Metabarcoding is an emerging genetic tool to rapidly assess biodiversity in ecosystems. It involves high-throughput sequencing of a standard gene from an environmental sample and comparison to a reference database. However, no consensus has emerged regarding laboratory pipelines to screen species diversity and infer species abundances from environmental samples. In particular, the effect of primer bias and the detection limit for specimens with a low biomass has not been systematically examined, when processing samples in bulk. We developed and tested a DNA metabarcoding protocol that utilises the standard cytochrome c oxidase subunit I (COI) barcoding fragment to detect freshwater macroinvertebrate taxa. DNA was extracted in bulk, amplified in a single PCR step, and purified, and the libraries were directly sequenced in two independent MiSeq runs (300-bp paired-end reads). Specifically, we assessed the influence of specimen biomass on sequence read abundance by sequencing 31 specimens of a stonefly species with known haplotypes spanning three orders of magnitude in biomass (experiment I). Then, we tested the recovery of 52 different freshwater invertebrate taxa of similar biomass using the same standard barcoding primers (experiment II). Each experiment was replicated ten times to maximise statistical power. The results of both experiments were consistent across replicates. We found a distinct positive correlation between species biomass and resulting numbers of MiSeq reads. Furthermore, we reliably recovered 83% of the 52 taxa used to test primer bias. However, sequence abundance varied by four orders of magnitudes between taxa despite the use of similar amounts of biomass. Our metabarcoding approach yielded reliable results for high-throughput assessments. However, the results indicated that primer efficiency is highly species-specific, which would prevent straightforward assessments of species abundance and biomass in a sample. Thus, PCR-based metabarcoding assessments of biodiversity should rely on presence-absence metrics.
Citation: Elbrecht V, Leese F (2015) Can DNA-Based Ecosystem Assessments Quantify Species Abundance? Testing Primer Bias and Biomass—Sequence Relationships with an Innovative Metabarcoding Protocol. PLoS ONE 10(7): e0130324. https://doi.org/10.1371/journal.pone.0130324
Academic Editor: Mehrdad Hajibabaei, University of Guelph, CANADA
Received: January 22, 2015; Accepted: May 19, 2015; Published: July 8, 2015
Copyright: © 2015 Elbrecht, Leese. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Data Availability: The Illumina sequencing data are available via the Short Read Archive (accession numbers SRS731403 and SRS733820).
Funding: This work was supported by a grant of the Kurt Eberhard Bode Foundation to FL.
Competing interests: The authors have declared that no competing interests exist.
A minor proportion of all species on Earth are known . At the same time, anthropogenic impacts have initiated a mass extinction of species in the “Anthropocene” , with pervasive and often negative consequences for ecosystem functioning and human well-being [3,4]. To counteract biodiversity loss, fast and reliable tools are needed to assess and monitor biodiversity .
Stream biodiversity is particularity affected by anthropogenic degradation [6,7]. Therefore, large-scale monitoring and management programs have been established, for example, the European Union Water Framework Directive and the US Clean Water Act. In these biomonitoring programs, species lists, particularly of benthic invertebrate indicator species, are the central metric to assess the ecological status of freshwater ecosystems. For stream assessments, hundreds of benthic organisms are sampled in a standardised fashion, sorted, identified, and used in standardised analytical work flows (e.g. [8,9]). However, many benthic invertebrate larvae are difficult to identify at the species level, and thus the most practical taxonomic level for the identification of these organisms is often only the genus or family . This is a major concern, as different species within a genus or subfamily can have different ecological preferences and stress tolerances and belong to different functional feeding groups [11,12] see  for review. Even worse, frequent identification errors occur and many specimens are not detected in samples ; these limitations have direct consequences for the inferred ecosystem assessment metrics [10,14] and thus management decisions.
DNA barcoding allows for standardized and accurate species identification [15–18]. As this method is DNA based, it can be used to identify species reliably even when juvenile instars or fragments of organisms are available. For animals, a 658-bp standardized fragment of the mitochondrial gene COI (cytochrome c oxidase subunit 1) is typically used . DNA barcoding requires the establishment of an accurate reference database. For macroinvertebrates, this is best achieved by determining diagnostic characters (usually in male adult specimens [13,20,21]), sequencing the specimens, and depositing the COI sequences in a database such as the BOLD database . In times of declining taxonomic expertise [23,24], these curated and public barcode databases are indispensible to conserve taxonomic knowledge.
COI barcoding methods are well established for freshwater organisms [16,17,25] and initial studies have tested their potential for freshwater ecosystem assessments using classical Sanger-based sequencing [14,26]. Stein and co-authors showed that ten of 16 assessment metrics had higher statistical power using DNA barcoding than morphological assessment . However, Sanger sequencing requires that each specimen is processed individually in the laboratory, which is costly and extremely time-consuming for routine community assessments involving hundreds or thousands of specimens per sample.
This challenge can be overcome with the aid of next-generation sequencing, which enables the simultaneous analysis of millions of sequences. One next-generation sequencing technique termed metabarcoding (also called community barcoding) utilises the same principle as classical barcoding, yet with much higher throughput, allowing the simultaneous processing of hundreds of samples in a single analysis. When complete specimens are identified in bulk, it was suggested to use the term DNA metabarcoding to make a distinction to approaches using environmental DNA (eDNA) . However, as our findings largely apply to eDNA-based methods as well, we here refer to metabarcoding in a broad sense. Metabarcoding is currently being tested to address a wide range of biological problems, such as invasive species detection , gut content analysis , and assessment of microbial  and metazoan diversity, such as that of arthropods (e.g. [31,32]). Initial studies on benthic diatoms  and macroinvertebrates  show the potential of this method to revolutionise the way we monitor stream ecosystems. However, there are general challenges associated with the use of metabarcoding for ecosystem assessments. While preliminary bioinformatic pipelines for data analysis are available (e.g., Mothur , QIIME , UPARSE pipeline ), barcode reference databases are still incomplete. There are furthermore two problems of central importance that have not been addressed systematically. First, sampled organisms have vastly different biomasses, and thus small organisms may be lost owing to low number of sequence reads . Second, the amplification efficiency of the COI gene varies among species, and this might severely bias results [34,39] particularly in view of the variation in biomass. Precise estimates of biomass with respect to specimen recovery and primer bias have not been performed.
We describe an innovative and efficient strategy to analyse macroinvertebrate samples on an Illumina MiSeq sequencing platform. High sequence similarity in in amplicon sequencing can lead to decreased sequence quality on Illumina platforms . We deal with this issue by using uniquely tagged fusion primers targeting the standard barcoding region, which are simultaneously sequenced in forward and reverse sequencing direction to increase nucleotide diversity and thus improve read quality. With the new protocol, we performed two controlled experiments to address the two problems outlined above. First, we assessed the relationship between biomass and sequence abundance by sequencing genetically distinct specimens that differ widely in biomass, but belong to a single species. This allowed us to determine whether and when small specimens are lost owing to low read coverage. Second, we used equal amounts of tissue from 52 freshwater taxa to determine how well they are recovered given species-specific PCR amplification bias when extracting many species in bulk. All analyses were performed with ten replicates to improve statistical robustness.
Materials and Methods
Two experiments were performed (Fig 1). In experiment I, the influence of biomass on sequence abundance and the reproducibility of the method were tested using 31 stonefly specimens of the same species (Dinocras cephalotes), i.e., standardizing for a single species. In experiment II, species detection rates were tested using the standard barcoding primers LCO1490 and HCO2198  and controlling for tissue biomass.
Two MiSeq runs were used to increase the reproducibility and reliability of our novel metabarcoding protocol. A Experiment I: Dinocras cephalotes specimens with different COI barcodes were used to determine the reproducibility of the protocol and the influence of biomass on sequence abundance. B Experiment II: Ten sets of 52 aquatic taxa were homogenised, and DNA was extracted and amplified to determine which taxa could be recovered with MiSeq using the "ready to load" primers developed in this study.
Ethics statement: No protected species and areas were sampled for this study with the exception of the dragonfly larvae Cordulegaster sampled from the Deilbach (N51.3282, E7.1619). Here, special permissions were obtained beforehand from the Kreisverwaltung Ennepe-Ruhr and Mettmann. No further permissions were required for sampling all other non-protected species from the Felderbach (N51.3450, E7.1703) and Ruhr University Bochum pond (N51.4457, E7.2656).
Primer design and sequencing strategy
Fusion primers were designed that combined the standard COI LCO1490 (LCO) and HCO2198 (HCO)  with Illumina sequencing tags (S1 Fig). The advantage of fusion primers is that the COI barcoding fragments can be loaded directly onto the MiSeq sequencer after a single PCR and a purification step. COI amplicons are typically similar in base composition; therefore, three strategies were used to increase sequence diversity. First, 20% PhiX control was spiked into both MiSeq sequencing libraries. The PhiX library consists of fragments of a whole viral genome, which has a high nucleotide diversity. Second, the bases before the start of the Folmer primers were shifted by 0–4 bp increasing nucleotide diversity of amplicons at each read position (as described in ). Finally, a new approach was developed to increase diversity by simultaneously sequencing both LCO and HCO primers. A much higher per-site nucleotide diversity was observed using the consensus sequences of the 31 unique Dinocras cephalotes COI haplotypes (see experiment I below) (S2 Fig). Fewer peaks of low diversity (only up to 70% identical bases) were detected using both primers than using one primer at a time (100% identical bases). The parallel sequencing approach thus substantially reduced regions of low per-site diversity, improving read quality.
The adopted strategy of sequencing with LCO and HCO primers simultaneously as well as the 4-bp shifting strategy were used to differentially tag each of the ten replicates of both sequencing runs, allowing the removal of the Illumina tag in the adapters and the omission of the tag-reading step in the MiSeq runs.
Experiment I: Relationship between specimen biomass and sequence abundance
Samples and DNA extraction.
D. cephalotes larvae with known COI haplotypes were available from a previous study . All specimens were stored in 96% ethanol at -20°C. From the specimens, 31 samples with different biomasses that differed by at least 2 bp in the COI barcode from all other specimens were selected (for GenBank accession numbers see S1 Table). All specimens were photographed and one leg was removed as a backup. All specimens were dried overnight at room temperature, weighed with a Sartorius RC 210D scale (0.01 mg accuracy) by two scientists independently and mean values were used for subsequent analyses. For bulk DNA extraction, the 31 specimens (with a cumulative weight of 642.72 mg) were placed in a ceramic mortar and manually ground into a fine powder (20 min processing time) using liquid nitrogen (Fig 1A). One-fifth of the ground tissue was divided among 14 reaction tubes at ~9 mg each (9/642.72 = 1.4% of total tissue). DNA was extracted from the 14 aliquots using a modified salt extraction protocol . Extraction success was checked on an agarose gel. Then, 25 μL of DNA from each aliquot was treated with 0.55 μL of RNase (concentration 10 mg/mL, Thermo Scientific, Waltham, MA, USA) at 37°C for 30 min and cleaned up using the MinElute Reaction Clean up Kit to remove RNA (Qiagen, Hilden, Germany). The DNA concentration after cleanup was quantified using a Qubit 2.0 (Life Technologies, Carlsbad, CA, USA) with the Broad-Range (BR) Assay Kit.
COI amplification and sequencing.
Four randomly chosen DNA aliquots (1.4% of total tissue) were selected for COI amplification. In addition, 50 ng DNA from each of the 14 DNA aliquots was pooled to create a sample representing 20% of the total tissue. For each of these five samples, two PCR were run using one N5LCO and one N5HCO fusion primer, each uniquely tagged (S1 Fig). The same PCR master mix was used for all reactions to ensure identical PCR conditions for all replicates. The ten PCR replicates were then run simultaneously in a C100 Thermalcycler (BioRad, Hercules, CA, USA).
The COI fragment was amplified in a PCR reaction consisting of 1× PCR buffer (including 2.5 mM Mg2+), 0.2 mM dNTPs, 0.5 μM of each primer, 0.025 U/μL of HotMaster Taq (5Prime, Gaithersburg, MD, USA), 50 ng DNA, and HPLC H2O to a total volume of 50 μL. The PCR program was as follows: 94°C for 180 s, 30 cycles of 94°C for 30 s, 46°C for 30 s, and 65°C for 150 s, and 65°C for 5 min. PCR products were excised from a 1% TAE agarose gel and purified using the MinElute Gel Extraction Kit (Qiagen, Hilden, Germany). Concentrations were measured using the Qubit 2.0 BR Kit and the library for sequencing was prepared by pooling 12.3 ng of all ten amplicons. Then, paired-end sequencing was carried out by GATC Biotech (Constance, Germany) using the MiSeq with 300 bp paired-end sequencing.
S3A Fig includes a flow chart of the data processing steps. Sequences with a Phred score of >20 were demultiplexed using the base shift tags in both read directions using an R script (available on request). Primers were removed with cutadapt 1.4.2  and forward and reverse reads were concatenated to 540-bp fragments. Paired end sequencing generated 2*300 bp long fragments, which is not enough to recover the complete Folmer COI region, which is typically 658 bp in length. Furthermore, up to 30 bp of the reads are primer sequences, leading to 2*270 = 540 bp concatenated fragments. Sequences of each replicate were compared against a reference database using the blastn algorithm (blastn 2.2.29, ). Statistics and data subsetting were performed in R 3.1.2 . Hits shorter than 500 bp and those that matched two haplotypes equally well owing to sequencing errors and chimeras were removed from the hit table. To ensure reliable hits, only sequences that had a maximum of five mismatches and gaps were included in the analysis. The number of hits per haplotype was calculated and compared to the weight of the corresponding specimens.
Experiment II: Recovery of 52 different taxa
Samples and DNA extraction.
Freshwater macroinvertebrates were collected from various streams (and Daphnia from ponds) in western Germany and stored in 96% ethanol. Specimens were identified to the lowest taxonomic level possible based on morphology. Ten sets consisting of 52 unique taxa were photographed, and roughly equal amounts of tissue were dried overnight and weighed (Fig 1B, S2 Table). For Isoperla, Limoniidae, Tipulidae, and Cordulegaster boltonii, fewer than ten specimens were obtained; therefore, tissue from a single specimen was used in more than one replicate extraction. As each distinct morphotaxon was present only once in each of the ten replicates, barcoding prior to DNA extraction was not necessary. The 52 tissue samples per replicate were pooled for DNA extraction, and five replicates were ground in liquid nitrogen for 20 min, while the other five replicates were ground with a Qiagen TissueLyser LT (two times for 2 min at 50 Hz with a short centrifugation of the tubes in between). DNA was then extracted with the salt extraction protocol described in experiment I, and 10 μL of DNA for each of the 14 extraction tubes was pooled for each of the ten replicates. RNA was digested prior to PCR as described in experiment I.
COI amplification and sequencing.
The PCR conditions were identical to those used in experiment I, and all ten replicates were run simultaneously in a C100 Thermalcycler (BioRad, Hercules, CA, USA) using the same master mix (see above). One to five PCR aliquots were pooled for each of the ten replicates with the aim to test whether replication of PCR reduces stochastic effects (was not evaluated, due to small number of replicates). Amplicons were purified and size selected (500–1000 bp) using magnetic beads (SPRIselect, Beckman Coulter, Bread, CA, USA; ratio 0.55×/0.45×). PCR product concentrations were measured using the Qubit BR Kit and the library for sequencing was prepared by pooling 52 ng of all ten replicates. 300 bp paired-end sequencing on a MiSeq was performed by GATC Biotech.
S3B Fig includes a flow chart of the data processing steps. Reads were demultiplexed with a minimum Phred score of 25, primers were removed, and reads were concatenated as described in experiment I. Sequences from all ten replicates were pooled and dereplicated, and singletons were removed to find operational taxonomic units (OTUs) using the UPARSE pipeline (97% identity, ). Chimeras were removed from the OTUs using uchime_denovo. The remaining OTUs were identified using the BOLD barcoding database by querying against all barcode records. The ten replicates were dereplicated using derep_fulllength, but singletons were included in the data set. Sequences were matched against the OTUs with a minimum match of 97% using usearch_global. The hit tables were imported and the sequence numbers were normalised to the total sequence abundance and tissue weight for the various taxa.
Sequencing success and statistics
The MiSeq runs of experiments I and II yielded 9.63 and 14.07 Gb of read data, respectively. Both MiSeq runs showed good read quality (sequences with Q30 ≥ 85.3% and 83.3%). The complete MiSeq data from both runs are available online on NCBI with the accession numbers SRS731403 (experiment I) and SRS733820 (experiment II).
Sequences starting with the LCO primer for the first read were significantly overrepresented in experiment I (34% more frequent than sequences beginning with the HCO primer, t-test, p = 0.006, S4A Fig). In experiment II, we did not detect primer bias (t-test, p = 0.41, S4B Fig). However, each of the ten replicates had a unique tissue set, in contrast to experiment I, which used the same DNA pool derived from a single species (31 specimens) for all five aliquots.
When normalising the loss of sequences in each data processing step of experiments I and II, we found significant differences between the LCO and HCO primers with respect to the number of reads with no hits in experiment I (t-test, p = 0.01, S3A Fig) and final hits in experiment II (t-test, p = 0.01, S3B Fig). The primers (LCO and HCO) had small effects on final sequence abundance, with differences of 2.7% and 4.4% for experiment I and II. We observed a similarly small effect on sequences abundance for the individual specimens in experiment I; there were significant differences in sequence abundance between LCO and HCO for 17 of the 31 stonefly specimens (t-test, p<0.05, S5 Fig).
Amount of extracted tissue and species recovery.
S1 Table gives an overview of the weights of the 31 specimens. The two independent weight measurements differed in mean by 0.1 mg (SD = 0.03). In all 10 replicates, we recovered all 31 D. cephalotes specimens based on their unique haplotypes, including the smallest specimens, which only made up 0.023% (0.145 mg) of the total specimen biomass (Fig 2A). For two specimens, we recovered sequence artefacts (see S6 Fig); these did not affect further analyses. We did not observe significant differences in sequence abundance among the replicates using different amounts of extracted tissue as a template. However, there was slightly more variation in sequence abundance for the replicates in which DNA was extracted from only 1.4% of the total tissue volume than from 20% (S7A Fig). We observed a strong negative relationship between specimen weight and variation in sequence abundance, as shown in S5 and S7B Figs. Specimens with a low biomass tend to show relatively high variation in sequence abundance among the ten replicates.
A Specimen weight (y-axis) is plotted against read abundance per specimen (x-axis) for all ten replicates. The linear regression (red line) was highly significant for all replicates with p<0.001. B Ratio of N5LCO/N5HCO sequence abundance for the mixed and four single DNA aliquots. The red line indicates the mean ratio.
Influence of biomass on sequence abundance.
We found a highly significant positive linear correlation between specimen biomass and sequence abundance in all ten replicates (Fig 2A). The mean normalised sequence abundances had low standard deviations and the linear model fit well (p<0.001, R2 = 0.65, S8 Fig).
Reproducibility of sequencing results.
The sequencing results for the ten replicates were highly reproducible. Even when comparing absolute sequence numbers, the patterns were concordant (Fig 2A). We detected few outliers and low standard deviations for the ratio of LCO- to HCO-based haplotype read abundance for each of the five DNA extractions (Fig 2B and S8 Fig).
Recovery of different taxa with similar biomass.
S2 Table gives an overview of specimen weights of the 52 tissues parts, used in each of the ten replicates. We were able to reliably recover 83% (43) of the 52 taxa included in experiment II. We recovered many of the typical bio-indicator taxa such Ephemeroptera, Plecoptera, Trichoptera, and Diptera (Table 1). 34 taxa were recovered in all ten replicates (Fig 3). From the DNA extractions performed with the TissueLyser LT, six more specimens (2.31%) were recovered than when DNA was extracted with liquid nitrogen. Furthermore, we did not observe substantial differences in recovery rates when different numbers of PCR products were pooled.
Sequence abundances for the 52 morphologically identified taxa is shown in rows and the ten replicates used in the experiment in columns. Sequence abundance was normalised across the ten replicates and the amount of tissue used in each extraction. Sequence abundance of each specimens (morphotaxon) of the ten replicates is visualised by different shades of blue. If a field is i.e. half filled (50%) with the mid blue shade (= 1% of total sequences), the respective specimen represented 0.5% (50% of 1%) of the total sequences in that replicate. When no sequences or only a few sequences (below 0.003% of total abundance per replicate) were found for a specimen, it was scored as "No Hit," as indicated by an orange asterisk. On the right, K2P-corrected neighbour-joining (NJ) trees for each taxon, based on the most abundant sequence obtained for each specimen (calculated with MEGA6.06), are shown. MOTUs are defined by a 2% sequence difference based on the NJ tree.
On average, 99.52% of the hits could be assigned to the specimens used in each extraction (Fig 3 and S9A Fig). Sequences that did not match the target species were similar to sequences derived from a variety of benthic organisms, but also fungi and plants. Of all recovered 213 OTUs, 31 could not be identified using the BOLD database (S9B Fig).
Higher taxonomic resolution.
We were able to reliably assign OTUs to all 52 taxa included in the extraction. In 19 of the cases (37%), the morphologically identified taxa were assigned to more than one MOTU (K2P distance of detected haplotypes > 2%, Fig 3), indicating that the ten morphologically identified specimens per morphotaxon included several species. For example, we found two distinct MOTUs in the caddisfly genus Silo, three in the mayfly genus Rhitrogena, and four in the blackflies (Simuliidae) (Fig 3). This was expected, as morphotaxa could often only be identified at family or order level, and several species per morphotaxon may occur in the sampled aquatic habitats even at the same habitat patch. However, most morphotaxa represented a single MOTU (i.e. distinct biological species). Identification by COI barcode did not perform worse than identification by morphology. In fact, in 50% of the cases, barcoding identified specimens at a finer taxonomic level, and 11 morphotaxa included multiple species using COI data from the BOLD database (S9A Fig).
Variation of sequence abundance for different taxa.
Although many taxa were recovered, the number of sequences per taxon varied by four orders of magnitude despite the similarity in biomass used for extraction (Fig 3). In most cases, the sequence abundance obtained for a certain morphotaxon was consistent across replicates after normalising for slight differences in the amount of tissue biomass used. Exceptions were Arhynchobdellida, Ceratopogonidae, Scirtidae and Glossosomatidae (S3 Table), where efficiencies differed substantially between the MOTUs. There was no correlation between taxon biomass and the number of recovered sequences (using the means of ten replicates for each of the 52 taxa).
Several studies show that DNA-based assessments are superior to morphological assessments in freshwater ecosystems (e.g., [14,25,33]); yet, reliable and standardised laboratory protocols need to be established prior to integrating metabarcoding assessments into existing monitoring programs. Here, we developed a new laboratory workflow and generated highly replicated next-generation sequencing data using the traditional 658-bp Folmer fragment. We used this data to systematically test, for the first time, for a relationship between specimen biomass and sequence reads in a standardised single-species setting (experiment I), and then assessed the impact of primer bias for multi-species samples using standardised biomass pools (experiment II).
Experiment I: Relationship between specimen biomass and sequence abundance
Benthic indicator organisms vary in biomass, and this variation depends on taxonomic group and life stage. Therefore, it is crucial to determine the relationship between biomass and sequence abundance to i) estimate taxon biomass in samples from read data, and ii) identify critical detection limits for a given sequence coverage and sample size. Piñol et al.  proposed a relationship between biomass and sequence reads, but did not systematically examine this hypothesis. Using 31 specimens of a single stonefly species (see ) that differed in biomass, we demonstrated a highly significant correlation between sequence abundance and specimen biomass. Irrespective of the tissue volume used for extraction, we recovered all 31 specimens. Weak outliers might be caused by differences in tissue conservation or mismatches in the primer binding sites of individual specimens . Variation in sequence abundance for each specimen was slightly higher among replicates that had less tissue as starting material. This trend was stronger for specimens with a smaller biomass. This result is expected because stochastic effects increase with reduced specimen biomass. However, our results indicate that the amount of tissue used in the DNA extraction was sufficient and did not lead to the systematic exclusion of small specimens as long as tissues are well ground and specimens have similar amplification efficiencies.
Furthermore, the LCO and HCO PCR replicates for DNA extraction yielded highly concordant results, emphasising the overall reliability of our protocol. We detected a slightly different sequencing efficiency between forward and reverse primers, but this only affected total sequence abundance and did not systematically alter the inferences from the data.
Experiment II: Primer amplification bias between species
Amplification of diverse multitemplate mixtures using universal primers can lead to highly unequal amplification efficiencies among products [39,49]. It has therefore been suggested that several group-specific primers are necessary for species monitoring . Here, we quantified the effect of primer bias using similar tissue biomass and different species (using ten replicates) to determine whether reliable species detection is possible.
The number of sequences per replicate was not biased by primer type (LCO or HCO), indicating that sequencing direction has a negligible effect when independent replicates are sequenced. Furthermore, we recovered a majority of specimens (83%) using one universal primer pair. However, using the Folmer primers, the number of sequences obtained varied among taxa by several orders of magnitude, probably because there were mismatches in the primer binding regions.
While amplification efficiencies were consistent among replicates of a morphotaxon, in particular when determined to species or genus level, some taxa that could only be determined at higher taxonomic level, contained different MOTUs with different amplification efficiencies (e.g. the order Arhynchobdellida). Presumably, these morphotaxa contained several taxonomically distant species, which are unequally well amplified with the Folmer primers. This is consistent with the primer bias which we already observed between morphotaxa and support the overall findings of strong taxon-specific amplification bias.
Implications for large-scale monitoring and future challenges
In this study, we established a quick and reliable protocol to assess the macrozoobenthic communities of stream ecosystems. We used a highly replicated and standardised approach for species detection using DNA metabarcoding and show that several technical and logistic problems have to be overcome before this protocol can be used for large-scale monitoring.
The results of experiment I show that it is possible to reliably estimate the biomass of a single species, but not its abundance because many small organisms generate the same number of sequence reads as a few large organisms. However, the results of experiment II show that primer efficiencies across different taxa greatly hinder species abundance assessments using PCR-based approaches, which is consistent with the findings of Piñol et al. . Thus, it is not possible to accurately estimate species biomass or even abundance in diverse environmental samples using amplification-based sequencing protocols. For accurate estimates of biomass, or even rough estimates, a PCR-free approach is needed; however, this requires further development . Currently, the monitoring of freshwater ecosystems is based on abundance metrics, which cannot be generated using the metabarcoding solutions currently available. Thus, for now, monitoring indices should use genetic data for presence-absence assessments. Initial studies on marine benthic taxa show that presence-absence data is sufficient for precise assessment indices [51,52], especially considering the additional information gained by species-level identification. Thus, the availability of such highly reliable data on the presence of species (even cryptic) can be very important for community descriptions. Furthermore, we used a universal COI primer with a broad target range , and only 17% of taxa went undetected. While this is already better than the error rates of several morphology-based studies (see  for a discussion), higher detection rates are desirable. This could be achieved if several group-specific primers or even more degenerate primers are used . Our protocol uses ten tagged fusion primers, and the Folmer primers can easily be supplemented with group-specific primers. To ensure that small specimens are detected, samples can additionally be fractionated into various size categories; extractions can be performed independently for each category, and template DNA amounts can be adjusted according to specimen size prior to amplification and sequencing.
The methods developed in this study can easily be adapted to assess the communities of other ecosystems. Our parallel sequencing strategy leads to an increase in per-site sequence diversity and read quality. The approach can be easily integrated into any other protocol for the MiSeq, HiSeq, or NextSeq protocols. While the use of full COI barcodes targeting the classical primer regions might give the highest taxonomic resolution, mini barcodes might be sufficient to detect most species and are popular in environmental DNA barcoding . However degraded DNA and contamination as in eDNA studies with amplicons is not a concern for organisms collected directly from streams. An approximately 400-bp barcode lying within the standard Folmer region would be optimal for both strategies, and plenty of sequence information is available to develop group-specific primers [29,32]. The use of a single universal primer pair that amplifies conserved ribosomal mitochondrial gene regions (e.g., 16S and 12S) could be effective . However, while this approach could have comparable taxonomic resolution as the COI barcode, it is currently limited by the lack of reference databases .
All assessment protocols rely on reference catalogues against which inventory data of a species from an ecosystem are compared. In particular, changes in species traits in a community (e.g. functional feeding groups, tolerance against pollution) are used as indicator values to evaluate the biological significance of inferred community changes. The efforts of large national and international barcoding consortia (BOLD, iBOL, and GBOL) have contributed to a substantial increase in both the size and quality of reference databases , which has provided a basis for species-level assessments. The protocol developed in this study enables the identification of nearly all macrozoobenthic species in an environmental sample. However, DNA-based assessments cannot assign biological traits to species. Therefore, to take full advantage of metabarcoding, acquiring ecological trait data at the species or even population level is the next crucial step. A combination of both data types, i.e., DNA species barcodes and ecological traits, will maximise the power of metabarcoding for the reliable assessment of ecosystem responses under stress and for biomonitoring.
Although more technical developments are necessary, we are confident that metabarcoding will widely replace present biomonitoring methods over time because 1) it has a higher taxonomic resolution , 2) it is cost efficient and fast and, most importantly, 3) it reduces human bias enabling comparisons among studies . Using our protocol, it is possible to assess community compositions within a week, from sampling to species identification.
We provide a highly reproducible laboratory protocol for processing macroinvertebrate samples in bulk and identified species using metabarcoding with the standard COI region. The technical accuracy of this method was supported by comparisons among many replicates. However, we also showed that the taxon abundance of diverse environmental samples can not be reliably assessed. Therefore, we suggest focusing on reliable presence-absence data obtained from replicated analyses. We are confident that the here presented protocol could be a useful resource to monitor a wide range of ecosystems in the next years.
S1 Fig. Fusion COI Primers developed in this study.
Fusion primer can be directly loaded onto the MiSeq system and universal primers modified or replaced.
S2 Fig. Increase of diversity by parallel sequencing.
By sequencing forward and reverse primers together, sequence diversity and thus read quality is increased.
S3 Fig. Number of reads excluded in data processing steps.
Includes flow charts of the bioinformatics processing of experiment I (A) and experiment II (B).
S4 Fig. Reads in each replicate after demultiplexing.
Data from experiment I (A) and experiment II (B).
S5 Fig. Experiment I: sequences per specimen.
Normalised sequence abundance for each stonefly.
S6 Fig. Experiment I: sequencing artefacts.
Sequence matches are shown for three individual specimens, including h28 and h13 that are affected by sequencing artefacts.
S7 Fig. Experiment I: Variability in sequence abundance.
Variability in sequence abundance between the ten replicates as well as dependence on specimen biomass.
S8 Fig. Experiment I: Sequence abundance depended on specimen biomass.
Mean normalised sequence abundance of all ten replicates, including standard errors.
S9 Fig. Experiment II: OTUs assigned to taxa.
Detailed overview of all 213 OTUs and their taxonomic identification using the BOLD database.
S1 Table. Information on Dinocras cephalotes specimen weights (in milligram) for experiment I
S2 Table. Information on specimen weights (in milligram) for experiment II.
We thank Janis Neumann, Hannah Schweyen, and Martina Weiss for help with the collection and identification of specimens. Alexander Weigand kindly determined the small Pisidium specimens. We also thank Ralph Tollrian for continuous support and helpful discussion. We further thank the EcoEvo Journal Club and Florian Altermatt for helpful suggestions that improved this manuscript.
Conceived and designed the experiments: VE FL. Performed the experiments: VE. Analyzed the data: VE. Contributed reagents/materials/analysis tools: FL VE. Wrote the paper: VE FL.
- 1. Blaxter M. Molecular systematics: Counting angels with DNA. Nature. 2003:122–124. https://doi.org/10.1038/421122a
- 2. Rockström J, Steffen W, Noone K, Persson A, Chapin FS, Lambin EF, et al. A safe operating space for humanity. Nature. 2009;461: 472–475. pmid:19779433
- 3. Millennium Ecosystem Assessment (2004) Ecosystems and Human Well-being: Synthesis. Island Press, Washington, D.C.
- 4. Cardinale BJ, Duffy JE, Gonzalez A, Hooper DU, Perrings C, Venail P, et al. Biodiversity loss and its impact on humanity. Nature. 2012;486: 59–67. pmid:22678280
- 5. Pfrender ME, Hawkins CP, Bagley M, Courtney GW, Creutzburg BR, Epler JH, et al. Assessing macroinvertebrate biodiversity in freshwater ecosystems: advances and challenges in DNA-based approaches. Q Rev Biol. 2010;85: 319–340. pmid:20919633
- 6. Vörösmarty CJ, McIntyre PB, Gessner MO, Dudgeon D, Prusevich A, Green P, et al. Global threats to human water security and river biodiversity. Nature. 2010;467: 555–561. pmid:20882010
- 7. Dudgeon D, Arthrington AH, Gessner MO, Kawabata Z-I, Knowler DJ, Lévêque C, et al. Freshwater biodiversity: importance, threats, status and conservation challenges. Biol Rev. 2005;81: 163. pmid:16336747
- 8. Meier C, Haase P, Rolauffs P, Schindehütte K, Schöll F, Sundermann A, et al. Methodisches Handbuch Fließgewässerbewertung. 2006;: 1–110.
- 9. Kenney MA, Sutton-Grier AE, Smith RF, Gresens SE. Benthic macroinvertebrates as indicators of water quality: The intersection of science and policy. Terr Arthropod Rev. 2009;2: 99–128.
- 10. Haase P, Pauls SU, Schindehütte K, Sundermann A. First audit of macroinvertebrate samples from an EU Water Framework Directive monitoring program: human error greatly lowers precision of assessment results. J North Am Benthol Soc. 2010;29: 1279–1291.
- 11. Merritt RW, Berg MB, Cummins KW. An Introduction to the Aquatic Insects of North America. 2009.
- 12. Pauls SU, Graf W, Haase P, Lumbsch HT, Waringer J. Grazers, shredders and filtering carnivores—The evolution of feeding ecology in Drusinae (Trichoptera: Limnephilidae): Insights from a molecular phylogeny. Mol Phylogenet Evol. 2008;46: 776–791. pmid:18171625
- 13. Jones FC. Taxonomic sufficiency: The influence of taxonomic resolution on freshwater bioassessments using benthic macroinvertebrates. Environ Rev. 2008;16: 45–69.
- 14. Stein ED, Martinez MC, Stiles S, Miller PE, Zakharov EV. Is DNA Barcoding Actually Cheaper and Faster than Traditional Morphological Methods: Results from a Survey of Freshwater Bioassessment Efforts in the United States? Casiraghi M, editor. PLoS ONE. 2014;9: e95525. pmid:24755838
- 15. Hebert PDN, Ratnasingham S, de Waard JR. Barcoding animal life: cytochrome c oxidase subunit 1 divergences among closely related species. Proc R Soc Lond B Biol Sci. 2003;270: S96–S99.
- 16. Zhou X, Adamowicz SJ, Jacobus LM, DeWalt RE, Hebert PD. Towards a comprehensive barcode library for arctic life—Ephemeroptera, Plecoptera, and Trichoptera of Churchill, Manitoba, Canada. Front Zool. 2009;6: 30. pmid:20003245
- 17. Zhou X, Jacobus LM, DeWalt RE, Adamowicz SJ, Hebert PDN. Ephemeroptera, Plecoptera, and Trichoptera fauna of Churchill (Manitoba, Canada): insights into biodiversity patterns from DNA barcoding. J North Am Benthol Soc. 2010;29: 814–837.
- 18. Sweeney BW, Battle JM, Jackson JK, Dapkey T. Can DNA barcodes of stream macroinvertebrates improve descriptions of community structure and water quality? J North Am Benthol Soc. 2011;30: 195–216.
- 19. Hebert PDN, Cywinska A, Ball SL, deWaard JR. Biological identifications through DNA barcodes. Proc R Soc Lond B Biol Sci. The Royal Society; 2003;270: 313–321.
- 20. Zhou X, Kjer KM, Morse JC. Associating larvae and adults of Chinese Hydropsychidae caddisflies (Insecta:Trichoptera) using DNA sequences. J North Am Benthol Soc. 2007;26: 719–742.
- 21. Ekrem T, Stur E, Hebert PDN. Females do count: Documenting Chironomidae (Diptera) species diversity using DNA barcoding. Org Divers Evol. 2010;10: 397–408.
- 22. Ratnasingham S, Hebert P. BOLD: The Barcode of Life Data System (http://www.barcodinglife.org). Mol Ecol Notes. 2007;7: 355–364. pmid:18784790
- 23. Hopkins GW, Freckleton RP. Declines in the numbers of amateur and professional taxonomists: implications for conservation. Anim Conserv. 2002;5: 245–249.
- 24. Agnarsson I, Kuntner M. Taxonomy in a changing world: seeking solutions for a science in crisis. Syst Biol. 2007;56: 531–539. pmid:17562477
- 25. Jackson JK, Battle JM, White BP, Pilgrim EM. Cryptic biodiversity in streams: a comparison of macroinvertebrate communities based on morphological and DNA barcode identifications. Freshw Sci. 2014;33: 312–324. pmid:24309212
- 26. Mächler E, Deiner K, Steinmann P, Altermatt F. Utility of environmental DNA for monitoring rare and indicator macroinvertebrate species. Freshw Sci. 2014. https://doi.org/10.1086/678128
- 27. Taberlet P, Coissac E, Hajibabaei M, Rieseberg LH. Environmental DNA. Mol Ecol. 2012;21: 1789–1793. pmid:22486819
- 28. Bohmann K, Evans A, Gilbert MTP, Carvalho GR, Creer S, Knapp M, et al. Environmental DNA for wildlife biology and biodiversity monitoring. Trends Ecol Evol (Amst). 2014;29: 358–367. pmid:24821515
- 29. Pompanon F, Deagle BE, Symondson WOC, Brown DS, Jarman SN, Taberlet P. Who is eating what: diet assessment using next generation sequencing. Mol Ecol. 2012;21: 1931–1950. pmid:22171763
- 30. Poisot T, Péquin B, Gravel D. High-Throughput Sequencing: A roadmap toward Community ecology. Ecol Evol. 2013;3: 1125–1139. pmid:23610649
- 31. Yu DW, Ji Y, Emerson BC, Wang X, Ye C, Yang C, et al. Biodiversity soup: metabarcoding of arthropods for rapid biodiversity assessment and biomonitoring. Methods Ecol Evol. 2012;3: 613–623.
- 32. Gibson J, Shokralla S, Porter TM, King I, van Konynenburg S, Janzen DH, et al. Simultaneous assessment of the macrobiome and microbiome in a bulk sample of tropical arthropods through DNA metasystematics. Proc Natl Acad Sci. 2014;111: 8007–8012. pmid:24808136
- 33. Zimmermann J, Glöckner G, Jahn R, Enke N, Gemeinholzer B. Metabarcoding vs. morphological identification to assess diatom diversity in environmental studies. Mol Ecol Resour. 2014; 1–17. https://doi.org/10.1111/1755-0998.12336
- 34. Hajibabaei M, Shokralla S, Zhou X, Singer G, Baird DJ. Environmental barcoding: A next-generation sequencing approach for biomonitoring applications using river benthos. PLoS ONE. 2011;6: e17497. pmid:21533287
- 35. Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M, Hollister EB, et al. Introducing mothur: Open-Source, Platform-Independent, community-supported software for describing and comparing microbial communities. Appl Environ Microbiol. 2009;75: 7537–7541. pmid:19801464
- 36. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, Costello EK, et al. QIIME allows analysis of high-throughput community sequencing data. Nat Methods. 2010;7: 335–336. pmid:20383131
- 37. Edgar RC. UPARSE: highly accurate OTU sequences from microbial amplicon reads. Nat Methods. 2013;10: 996–998. pmid:23955772
- 38. Carew ME, Pettigrove VJ, Metzeling L, Hoffmann AA. Environmental monitoring using next generation sequencing: rapid identification of macroinvertebrate bioindicator species. Front Zool. 2013;10: 1–1. pmid:23356547
- 39. Piñol J, Mir G, Gomez-Polo P, Agustí N. Universal and blocking primer mismatches limit the use of high-throughput DNA sequencing for the quantitative metabarcoding of arthropods. Mol Ecol Resour. 2014; 1–12. https://doi.org/10.1111/1755-0998.12355
- 40. Krueger F, Andrews SR, Osborne CS. Large scale loss of data in low-diversity Illumina sequencing libraries can be recovered by deferred cluster calling. Preiss T, editor. PLoS ONE. 2011;6: e16607. pmid:21305042
- 41. Folmer O, Black M, Hoeh W, Lutz R, Vrijenhoek R. DNA primers for amplification of mitochondrial cytochrome c oxidase subunit I from diverse metazoan invertebrates. Mol Marine Biol Biotechnol. 1994;3: 294–299. pmid:7881515
- 42. Lundberg DS, Yourstone S, Mieczkowski P, Jones CD, Dangl JL. Practical innovations for high-throughput amplicon sequencing. Nat Methods. 2013;10: 999–1002. pmid:23995388
- 43. Elbrecht V, Feld CK, Gies M, Hering D, Sondermann M. Genetic diversity and dispersal potential of the stonefly Dinocras cephalotes in a central European low mountain range. Freshw Sci. 2014;33: 181–192. pmid:25000790
- 44. Sunnucks P, Hales DF. Numerous transposed sequences of mitochondrial cytochrome oxidase I-II in aphids of the genus Sitobion (Hemiptera: Aphididae). Mol Biol Evol. 1996;13: 510–524. pmid:8742640
- 45. Martin M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet journal. 2011;17: 10–12.
- 46. Camacho C, Coulouris G, Avagyan V, Ma N, Papadopoulos J, Bealer K, et al. BLAST+: architecture and applications. BMC Bioinformatics. 2009;10: 421. pmid:20003500
- 47. R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/. 2014.
- 48. Elbrecht V, Poettker L, John U, Leese F. The complete mitochondrial genome of the stonefly Dinocras cephalotes (Plecoptera, Perlidae). Mitochondrial DNA. 2013. https://doi.org/10.3109/19401736.2013.830301
- 49. Polz MF, Cavanaugh CM. Bias in template-to-product ratios in multitemplate PCR. Appl Environ Microbiol. 1998;64: 3724–3730. pmid:9758791
- 50. Zhou X, Li Y, Liu S, Yang Q, Su X, Zhou L, et al. Ultra-deep sequencing enables high-fidelity recovery of biodiversity for bulk arthropod samples without PCR amplification. Gigascience. 2013;2: 4. pmid:23587339
- 51. Ranasinghe JA, Stein ED, Miller PE, Weisberg SB. Performance of Two Southern California Benthic Community Condition Indices Using Species Abundance and Presence-Only Data: Relevance to DNA Barcoding. Crandall KA, editor. PLoS ONE. 2012;7: e40875. pmid:22879881
- 52. Aylagas E, Borja A, Rodríguez-Ezpeleta N. Environmental status assessment using DNA metabarcoding: towards a genetics based Marine Biotic Index (gAMBI). Consuegra S, editor. PLoS ONE. 2014;9: e90529. pmid:24603433
- 53. Meusnier I, Singer GA, Landry J-F, Hickey DA, Hebert PD, Hajibabaei M. A universal DNA mini-barcode for biodiversity analysis. BMC Genomics. 2008;9: 214. pmid:18474098
- 54. Deagle BE, Jarman SN, Coissac E, Pompanon F, Taberlet P. DNA metabarcoding and the cytochrome c oxidase subunit I marker: not a perfect match. Biol Lett. 2014;10: 20140562–20140562. pmid:25209199
- 55. Clarke LJ, Soubrier J, Weyrich LS, Cooper A. Environmental metabarcodes for insects: in silicoPCR reveals potential for taxonomic bias. Mol Ecol Resour. 2014;14: 1160–1170. pmid:24751203
- 56. Taberlet P, Coissac E, Pompanon F, Brochmann C, Willerslev E. Towards next-generation biodiversity assessment using DNA metabarcoding. Mol Ecol. 2012;21: 2045–2050. pmid:22486824
- 57. Ji Y, Ashton L, Pedley SM, Edwards DP, Tang Y, Nakamura A, et al. Reliable, verifiable and efficient monitoring of biodiversity via metabarcoding. Holyoak M, editor. Ecol Lett. 2013;16: 1245–1257. pmid:23910579