Research Article |
Corresponding author: Ashrenee Govender ( agovender@ori.org.za ) Academic editor: Kelly D. Goodwin
© 2024 Ashrenee Govender, Sandi Willows-Munro, Sohana P. Singh, Johan C. Groeneveld.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Citation:
Govender A, Willows-Munro S, Singh SP, Groeneveld JC (2024) Net type, tow duration and day/night sampling effects on the composition of marine zooplankton derived from metabarcoding. Metabarcoding and Metagenomics 8: e119614. https://doi.org/10.3897/mbmg.8.119614
|
DNA metabarcoding requires only a single DNA fragment to detect a species in mixed zooplankton samples, compared to morphology-based methods that rely on the presence of intact specimens. However, metabarcoding protocols have not yet been fully standardised, thus hindering data comparability between studies. To converge on standardised metabarcoding protocols, we used an experimental field-sampling approach to compare the effects of sampling gear (ring-, Manta- and WP2 nets), day and night (DN) sampling and tow duration (5-, 10- and 15-minute tows) on marine zooplankton composition. High-throughput sequencing of the cytochrome c oxidase subunit I (COI) gene region with different primers and taxonomic assignment of amplicon sequence variants at 97% similarity to barcode records were used to identify species. Metabarcoding detected a total of 224 species, of which 92% matched prior occurrence records from the region. Malacostraca (crabs, hermit crabs, lobsters, prawns and shrimps) was the best-represented class (49%), followed by Copepoda (21%), Actinopterygii (ray-finned fishes; 21%), and Gastropoda (9%). Species counts ranged from 9–61 species per tow, with high proportions of unique species in replicate tows. Mean species counts did not differ significantly between net types, DN samples or tow durations, respectively. Proportionate representation amongst taxonomic classes remained within a narrow range, except when sampling deeper habitats with a smaller mesh size. DN samples showed no evidence of daily vertical migration of zooplankton. Consistent inferred species composition across net types, tow duration and DN sampling treatments reflects high detection sensitivity of individual-based sampling, allowing for greater flexibility in planning of zooplankton sampling regimes.
Experimental approach, gear-effects, high-throughput sequencing, species diversity, standardisation of field sampling
DNA metabarcoding provides a rapid and accurate method to record the species composition of mixed marine zooplankton samples (
Nevertheless, the direct integration of metabarcoding data into existing morphology-based time series remains challenging (
Furthermore, few standards exist for the collection of field data, laboratory approaches or analysis pipelines, thus hindering comparability of data collected by different laboratories (but see https://metazoogene.org/). Historical literature contains many studies on the effects of sampling methods and gear on quantitative estimates of zooplankton and species composition derived from morphological analysis. Examples are the performance of different plankton net types (
The standardisation of protocols for collecting metabarcoding data is a key aspect to consider when planning long-term biomonitoring surveys and for the integration of molecular information into existing quantitative time-series data. To propose standards, empirical evidence needs to be provided on best-practice approaches to field sampling. To do so, we used an experimental approach to compare the effects of sampling gear, day and night sampling and duration of tows on the metabarcoding-based composition of zooplankton samples collected in coastal waters off eastern South Africa. Specific aims were to: (1) generate a species list of zooplankton present in samples using metabarcoding; (2) validate identified species by cross-referencing with occurrence records; and for standardisation (3) evaluate the effects of different net types, day versus night sampling and tow duration on metabarcoding-based species composition.
Discrete sampling stations at 20 m, 100 m and 200 m depth soundings along a cross-shelf transect near Durban in the KwaZulu–Natal (KZN) Province of South Africa were sampled for zooplankton using plankton nets towed behind a boat (Suppl. material
Two field sampling protocols were followed to sample zooplankton: a net-type experiment to compare the effects of three different net types deployed during the day, using the same tow duration; and a day/night (DN) and tow duration experiment, using one net type (ring net) to compare the effects of day and night tows and of increasing tow durations, on zooplankton species composition derived from metabarcoding analysis.
For the net-type experiment, the ring- and Manta nets were towed horizontally just below the surface for 5 minutes at a time with a ground speed of 2–3 knots. The WP2 net was lowered to 10 m above the seafloor at each sampling station and then hauled vertically. All tows were undertaken at night, between sunset and sunrise. To provide more comprehensive data across a spatio-temporal range, samples were collected from stations at 100 and 200 m depth contours in September 2018 and August 2019. To avoid carry-over of DNA between successive tows, the ring net and cod-end were thoroughly rinsed with seawater between tows. To test for potential cross-contamination of samples (transfer of DNA and organisms between successive tows), three replicate ring net tows were performed at each sampling station. Even though the test cannot completely rule out cross-contamination, we hypothesised that high variability in the numbers of species identified in successive tows and a low level of shared species would indicate a low incidence of cross-contamination. A total of 20 samples were available for the net-type experiment, 12 from ring nets (four sets of three replicate tows) and four samples each from the Manta and WP2 nets.
For the DN/tow duration experiment, only the ring net was used, with tows conducted as above. Daytime samples were collected between noon and sunset and night samples from approximately 1 hour after sunset. Nets were towed for 5, 10 and 15 minutes at a time. Samples were collected from stations at 20 m and 100 m depth contours, in May and September 2022. A total of 24 samples were available for the DN/tow duration experiment, comprising of four samples in each of the six categories (day and night tows of 5, 10 and 15 minutes each).
All samples from both net-type and DN/tow duration experiments were washed from the tow net cod-ends into jars with 97% ethanol and stored at –20 °C before further processing in a genetics laboratory. Ethanol in sample jars were replaced after 24 hours to ensure optimal long-term storage conditions and to minimise the degradation of samples over time.
Individual tow net samples were homogenised in the 97% ethanol solution for 45 s using a consumer blender (Defy PB7354X, 350Wand 22,000 rpm) (
Polymerase chain reactions (PCRs) were performed in triplicate to minimise stochastic effects, bias and amplification errors (
The dada2 algorithm (
Species detected by metabarcoding were further validated by cross-referencing with occurrence records obtained from online databases such as the World Register of Marine Species (WoRMS; https://www.marinespecies.org), the Ocean Biodiversity Information System (OBIS; https://obis.org), the Global Biodiversity Information Facility (GBIF; https://www.gbif.org) and online literature.
Criteria used to compare metabarcoding-based species composition amongst experimental categories were individual species records, cumulative and mean (± standard deviation) species counts per experimental treatment and the proportional representation of Malacostraca, Actinopterygii, Copepoda (copepods) and Gastropoda (gastropods) in samples.
To examine variability at the level of individual sampling sites and test for potential cross-contamination, species collected by replicate ring net tows at individual sampling sites were compared. A null hypothesis of no difference in species counts was tested and Venn diagrams were constructed to compare unique and shared species within each set of three replicate tows. Where appropriate, Levene’s test was used to test for equality amongst variance of samples, followed by single and two-factor ANOVA to test for differences amongst sample means.
Water temperature (20.8–21.5 °C) and salinity (35.2–35.5 ppt) measured at 2 m below the sea surface remained within a narrow range across all sampling trips (Suppl. material
For the net-type experiment, the average flow volumes for 5-minute tows ranged between 19.1 and 39.8 kl per replicate set of three ring nets and between 36.0 and 37.6 kl per tow for individual Manta nets (Suppl. material
Sequencing was efficient with minimal filtering needed (for both forward and reverse reads) when merging the paired-end reads for all 44 zooplankton libraries (Table
Library | Depth (m) | Read count | Merged reads | Total amplicon sequence variants | Merged amplicon sequence variants assigned to species level (97%) | |
---|---|---|---|---|---|---|
Net type experiment | 2018 | 100 | 600746 | 21750 | 133 | 23 |
200 | 464824 | 33241 | 263 | 74 | ||
2019 | 100 | 619826 | 32631 | 292 | 65 | |
200 | 644274 | 48199 | 311 | 63 | ||
Total across sites | – | – | 2329670 | 135821 | 830 | 128 |
Day/night/duration experiment | Trip 1 | 20 | 639068 | 19180 | 228 | 48 |
100 | 641680 | 24187 | 456 | 107 | ||
Trip 2 | 20 | 602386 | 17125 | 335 | 67 | |
100 | 647686 | 17117 | 348 | 78 | ||
Total across sites | – | – | 2530820 | 77609 | 1047 | 178 |
For the 24 DN/tow duration libraries, a total of 2.5 million read counts were consolidated into 77,609 merged reads, of which a total of 1047 sequences were available for analysis across all groups amplified. The 1047 sequences were then collapsed into 178 ASVs that were matched to species level with a > 97% sequence similarity to sequences on BOLD or GenBank. Of the 1047 sequences, 589 (56%) sequences remained unassigned or could not be assigned to a species level at 97% similarity or above.
For all 44 tows combined, metabarcoding detected zooplankton belonging to 27 orders, 89 families, 160 genera and 224 species (Suppl. material
The number of species identified per individual tow ranged from 9 to 61, with the mean number of species (± SD) for each replicate set of three tows being 29.7 ± 27.2; 29.3 ± 11.0; 23.0 ± 5.0; and 13.3 ± 3.8 species. The hypothesis of no difference in the number of species caught in replicate ring net tows was rejected in two of the four cases (χ2 = 8.205 and 49.92; df = 2; p < 0.05 in both cases), but accepted in the other two (χ2 = 2.151 and 2.174; df = 2; p > 0.05 in both cases). Venn diagrams indicated high ratios of unique species as a proportion of all species identified within each replicate set of three tows (Fig.
In the net-type experiment, metabarcoding identified 106 species from ring net samples (n = 12 tows), 47 from WP2 net samples (n = 4 tows) and 34 from Manta net samples (n = 4 tows). Mean species counts (±SD) for each net type were 23.8 ± 14.5 for ring nets, declining to 20.5 ± 10.4 for WP2 nets and 13.5 ± 8.7 for Manta nets (Fig.
Mean number of species per tow (± standard deviation) identified from samples per net type.
On average, Malacostraca contributed the largest proportion of all species identified from ring-(50.4%) and Manta net samples (48.2%), but in WP2 nets, there were more Copepoda (39.0%) than Malacostraca (34.2%) (Fig.
In the DN/tow duration experiment, metabarcoding could identify 125 species from ring net samples collected during the daytime and 116 species from night samples, but the difference was not significant (χ2 = 0.336; df = 1; p = 0.562) (Fig.
Number of species identified per taxonomic class as a proportion of all species for (A) day and night sampling and (B) tow duration (5, 10, 15 minutes) during day and night sampling (e.g. Day 5 = 5-minute tow during daytime).
Mean species counts per tow (all taxa combined) did not differ significantly between day (27.8 ± 7.7) and night samples (25.4 ± 11.7) (Students t-test; p = 0.56) and nor did counts differ between 5-minute (26.0 ± 10.9), 10-minute (27.1 ± 6.0) and 15-minute tows (26.6 ± 12.6) (ANOVA, F = 0.0244, df = 2, p = 0.976). The simultaneous analysis of the effects of DN and tow duration using a two-factor ANOVA with equal replication accepted all null hypotheses of no difference between DN (F = 0.289; p = 0.598) and tow-duration (F = 0.0224; p = 0.978) combinations, with no evidence for interaction between factors (F = 0.521; p = 0.602).
Malacostraca made up the bulk of species in both day (45.6% of species identified) and night samples (44.3%), followed by Copepoda (23.7% and 33.4%), Actinopterygii (17.1% and 14.4%) and Gastropoda (13.5% and 7.9%) (Fig.
The purpose of this study was to examine the effects of field sampling strategy (net-type, tow duration, day or night sampling) on the species composition of marine zooplankton derived from metabarcoding. Field sampling effects on the outputs of metabarcoding studies in the marine environment are rarely discussed in literature, as most methodological discussions focus on the molecular aspects of analyses. Our study contributes to the standardisation of methods used in metabarcoding studies of zooplankton, with a longer-term objective of comparability amongst datasets and integrating metabarcoding data into existing morphology-based time-series trends.
A total of 224 zooplankton species with > 97% similarity to barcode reference sequences were identified during the metabarcoding analysis. This is an underestimate of the true species richness in the samples. Some 40% of the sequences for the net-type experiment and 56% of the DN/tow experiment sequences could not be assigned to a species level because they did not match records on BOLD or GenBank with > 97% similarity. Furthermore, the mini-barcode primers used in our study were taxon-specific for Malacostraca and Actinopterygii (
Seventeen of 224 species (8%) identified from barcode reference data appeared to be false positives, with > 97% sequence similarity, but no validatory occurrence records from South Africa or the WIO region. False positive identifications can result from inaccuracies on BOLD and GenBank reference databases (
Selecting similar conditions for sampling across experiments (e.g. common sampling sites, replicate sampling, completing the full array of experimental treatments during the same trip) was aimed towards reducing natural variability so that the effects of net type, DN and tow duration on metabarcoding-based composition would be enhanced. A low level of shared species and high variability were observed between replicate ring net tows, suggesting a minimal occurrence of cross-contamination at sea. No significant differences were found in the mean numbers of species identified from sampling with various net types (ring, WP2 and Manta nets). Ring- and Manta nets were towed at depths < 5 m and, therefore, sampled similar epipelagic habitats during the same sampling trip. These two net-types caught similar proportions of malacostraca, copepods, ray-finned fishes and gastropods. WP2 nets were hauled vertically, from a maximum depth of 190 m and caught a greater proportion of copepods and fewer malacostraca than the ring and Manta nets, a difference attributed to deeper habitats sampled and/or a smaller mesh size used (200 µm) in the WP2 net.
The selectivity of zooplankton tow nets can, inter alia, be influenced by avoidance behaviour, clogging of the net mesh, escape and patchiness (
The DN samples showed no evidence of daily vertical migration (DVM) of zooplankton (reviewed by
The absence of a significant difference in metabarcoding-based species counts in tows with increasing duration can partially be explained by the effects of zooplankton patchiness in the water column. Zooplankton patchiness refers to the aggregation of zooplankton in specific areas (or patches) at horizontal and vertical planes, on scales ranging from centimetres to kilometres (
In conclusion, the detection of 224 zooplankton species, of which 92% matched prior distribution records, emphasises the robustness of metabarcoding in characterising zooplankton communities. Mean species counts obtained from metabarcoding analysis did not differ significantly between net types, DN samples or tow duration, respectively. The proportionate representation amongst taxonomic classes remained within a narrow range across experimental treatments, except when sampling deeper habitats with a smaller mesh size. The consistency of metabarcoding-based species composition across experimental treatments reflects high detection sensitivity of individual-based sampling, allowing for greater flexibility in planning of zooplankton sampling regimes. Thus, we contribute important empirical evidence for standardising field sampling methods when collecting marine zooplankton to be analysed with metabarcoding.
Opinions expressed and conclusions arrived at are those of the authors and are not necessarily to be attributed to the National Research Foundation of South Africa. We thank the South African Institute for Aquatic Biodiversity (SAIAB) for the use of the RV Phakisa and its crew for sampling at sea, especially the skipper, Thor Eriksen.
The authors have declared that no competing interests exist.
No ethical statement was reported.
We acknowledge funding support from the DSI/NRF/ACEP Captor Project (Grant 110763) for field and laboratory work. The first author (A.G.) received post-doctoral funding support through the Professional Development Programme of the National Research Foundation (NRF).
Conceptualization: AG, JCG, SPS, SWM. Data curation: AG, JCG. Formal analysis: JCG, AG. Funding acquisition: JCG. Investigation: AG, JCG. Methodology: SWM, AG, JCG, SPS. Project administration: JCG. Resources: JCG. Software: AG. Supervision: JCG, SPS, SWM. Validation: AG, JCG. Visualization: JCG, AG. Writing - original draft: JCG, AG. Writing - review and editing: SPS, SWM.
Ashrenee Govender https://orcid.org/0000-0002-2860-4610
Sandi Willows-Munro https://orcid.org/0000-0003-0572-369X
Sohana P. Singh https://orcid.org/0000-0003-3484-7800
Johan C. Groeneveld https://orcid.org/0000-0002-9831-9073
All raw sequence reads and codes used to perform analyses are available from Figshare (https://doi.org/10.6084/m9.figshare.25577301.v2); additionally, the raw sequence data have been uploaded to NCBI (https://www.ncbi.nlm.nih.gov/sra/PRJNA1115141).
Supplementary information
Data type: docx
Explanation note: table S1. Latitudinal and longitudinal data points for sampling stations at 20 m, 100 m and 200 m depth soundings along a cross-shelf transect near Durban in the KwaZulu–Natal (KZN) Province of South Africa. table S2. The six primer cocktails used in this DNA metabarcoding study (first round PCR): each of the COI primer cocktails amplify different fragments of the COI-5P gene region. Illumina adapter target sequences (indicated in bold and underlined) were used in accordance with the workflow from the Illumina 16S Metagenomics protocol. These adapter targets allow Nextera indexing and Illumina adapter addition through PCR. table S3. Environmental data measured with the Sea-Bird SB19plus Profiler CTD at approximately 2 m below the sea surface. Data are unavailable for the 1st part of the DN/tow duration experiment. table S4. Water volume (kilolitres, kl) passing through ring nets (0.5 m2 net opening) and manta nets (0.075 m2 net opening) based on flow meter data. Data are unavailable for the 1st part of the DN/tow duration experiment. table S5. Species detected by metabarcoding of zooplankton collected over the continental shelf of Durban, South Africa and verification of adult distribution ranges.