Global gap-analysis of amphipod barcode library

Introduction

Nature in the age of Anthropocene is facing numerous global changes and challenges. One of the drastic results of human associated activities is the acceleration of species extinctions, with one million species estimated to be presently critically endangered (IPBES, 2019). What is more, although the rate of species discovery grows, large numbers of species remain undescribed and it is believed many will not be recognized before they go extinct (Mora et al., 2011; Brix et al., 2020). This raises the challenge of efficient environmental monitoring, which is crucial for biodiversity recognition and preservation. Monitoring based on the taxonomic identification of organisms in samples is time-consuming and requires knowledge of the studied group. In the time of the taxonomic impediment (Ebach, Valdecasas & Wheeler, 2011), species identification methods offering an alternative to morphology-based methods are of great interest. Utilization of DNA-barcoding (identifying sequences of individual specimens), metabarcoding (high-throughput identification of bulk samples) and the use of environmental DNA (e-DNA, identifying DNA of taxa directly from water or soil sample, without collection of specimens) have been presented as promising methods in monitoring and ecological studies (e.g., Hajibabaei et al., 2012; Cristescu, 2014; Aylagas et al., 2018; Leese et al., 2018; Bush et al., 2019; Feio et al., 2020). The use of metabarcoding in assessing the status of ecosystems has already received the new term “Biomonitoring 2.0” (Bush et al., 2019). Such approaches require the existence of well-established barcode fragment libraries, which allow accurate recognition of organisms in the environment (Cristescu, 2014; Cowart et al., 2015; Oliveira et al., 2016; Múrria et al., 2020). Recent studies indicate that although the use of barcoding in biomonitoring has great advantages over morphological identification, the current gaps in barcode libraries may hinder their use (Weigand et al., 2019; Duarte, Vieira & Costa, 2020; Feio et al., 2020; Hestetun et al., 2020; Leite et al., 2020; Múrria et al., 2020; Vieira et al., 2021).

There are two main repositories where DNA sequences are deposited: NCBI GenBank (http://www.ncbi.nlm.nih.gov/genbank/, Sayers et al., 2020) and Barcode of Life Data System (BOLD, http://www.boldsystems.org, Ratnasingham & Hebert, 2007). In contrast to GenBank, which assembles nucleotide data of all genes, the primary aim of BOLD is to store data used for species barcoding, which in the case of Metazoa is the cytochrome c oxidase (COI) gene. The development of the BOLD database included the Barcode Index Number (BIN) system implementation (Ratnasingham & Hebert, 2013) that intends to help in biodiversity assessments by providing species-level taxonomic registry. Based on a molecular species delimitation method, each Molecular Operational Taxonomic Unit (MOTU) recognized by BOLD receives a unique alphanumeric code (BIN). Ideally, each BIN is associated with an accurate taxonomic (preferably species) identification and links to the voucher stored in a recognised institution. However, in practice this is not working well, and at the time of system implementation as many as 46% of BINs lacked species names (Ratnasingham & Hebert, 2013). This issue has arisen for a variety of reasons, which we investigate in this study using a particular faunal group, the Amphipoda, as a model.

The Order Amphipoda are peracarid crustaceans belonging to the class Malacostraca. They are very diverse components of aquatic environments. According to the World Amphipoda Database (WAD, Horton et al., 2020, accessed on 17-07-2020) there are 10,235 accepted amphipod species, the majority of which (78%) inhabit the marine realm, around 20% are freshwater species and just 2% are terrestrial taxa (Horton et al., 2020; Väinölä et al., 2008). The discovery rate of new species has grown steadily since the first amphipod species description and has particularly accelerated in the last six decades (Horton et al., 2020) with mean number of over 100 taxa annually described since the 1960s (Coleman, 2015). If the trend from the last sixty years persists, we may expect to have ca. 8,000 new species described by 2100. More conservative estimates predict that 6,100 new species will be described by that date (Arfianti, Wilson & Costello, 2018). The use of molecular methods in the studies of Amphipoda has revealed very high species diversity (e.g., Knox et al., 2012; Verheye, Backeljau & d’Udekem d’Acoz, 2016; Tempestini, Rysgaard & Dufresne, 2018; Jażdżewska & Mamos, 2019) and revealed the existence of cryptic species complexes within widely distributed taxa (Witt, Threloff & Hebert, 2006; Mamos et al., 2014; Wysocka et al., 2014; Havermans, 2016). Amphipoda are not only a species-rich group, but they also often dominate the crustacean assemblages in which they occur (e.g., Corkum, 1989; Humphries, Davies & Mulcahy, 1996; Vinogradov, Volkov & Semenova, 1996; Jazdzewski et al., 2001; Väinölä et al., 2008; Frutos, Brandt & Sorbe, 2017; Brix et al., 2018; Havermans & Smetacek, 2018). They can be found in both the benthos and the pelagic realm, presenting a variety of states of mobility (from epibenthic clingers to fully mobile swimmers) and, as a result, possess a wide variety of feeding habits including herbivory, detritivory, necrophagy, omnivory, predation and ectoparasitism (Barnard & Karaman, 1991; Vinogradov, Volkov & Semenova, 1996; Dauby, Scailteur & De Broyer, 2001; Väinölä et al., 2008). Being diverse and abundant they are important prey items for other invertebrates and vertebrates, including fish, birds and mammals (e.g., Dalpadado et al., 2001; Dauby, Nyssen & De Broyer, 2003). Certain species of Amphipoda are used in laboratory ecotoxicological studies (Hyne & Everett, 1998; Bundschuh et al., 2013; Major et al., 2013). Some amphipod species are well-adapted to anthropogenic environments such as artificial structures used in coastal protection or are part of fouling communities, and have shown a high invasion potential worldwide (e.g., Bij de Vaate et al., 2002; Kelly et al., 2006; Cabezas et al., 2014; Rewicz et al., 2015; Beermann et al., 2020; Sedano et al., 2020).

The combined factors of high diversity and the important role played by amphipods in the aquatic ecosystem highlight the need for accurate species identifications which are required for biological monitoring programs. The use of DNA-barcoding may speed up the identification process, but it will only succeed if the barcode library is well-established and robust. Recent gap-analyses of the barcode libraries in aquatic European environments showed very large differences in the coverage between different taxonomic groups and geographic regions (Weigand et al., 2019; Feio et al., 2020; Hestetun et al., 2020; Leite et al., 2020; Vieira et al., 2021). These studies used species lists restricted to particular geographic regions or chosen taxonomic groups. Basic summaries concerning the extent of amphipod data in BOLD identified problems with lack of taxonomic identification or detailed geographic information as well as contamination with human or bacterial DNA and provided recommendations to improve the data (Radulovici & Coleman, 2017; Coleman & Radulovici, 2020). However, to date there are no detailed analyses that have been conducted on a single taxon on a global scale.

In this study we have conducted a gap-analysis of the barcode library of a single crustacean order, the Amphipoda, on a global basis. In producing an up-to-date picture of the current state of knowledge, we will provide researchers with a detailed understanding of the both the strengths and the potential limitations of the use of DNA barcodes for identifications. We also propose recommendations for future initiatives that involve molecular data and produce new barcodes to fill the gaps in our knowledge of this taxon.

Material and Methods

Data for the present study were retrieved from BOLD by searching the “Public Data Portal” using the keyword “Amphipoda”. A combined dataset of all records was downloaded as an .xml file on June 24th 2020.

All records of the barcoding fragment of the cytochrome c oxidase I (COI-5P in BOLD) were extracted (29,016 records). This extracted dataset was used for all further analyses conducted by using various filtering options in an Excel spreadsheet. 2,579 records, represented by sequences shorter than 500 bp or having more than 1% ambiguous nucleotides for which BINs were not ascribed, were removed from dataset. Continued analysis of the dataset revealed some duplicate records (1,468 records, 734 cases, File S1). These derived from data harvested by BOLD from GenBank and seemed to be associated with an update of the records in GenBank. In the dataset, these records had an identical sample ID that referred to a GenBank Accession Number but with an additional ‘.1’ appended (e.g., KP713892 and KP713892.1) and with an identical identification provided. The differences were often linked with more detailed geographical information in the case of one record from the pair. Only the more detailed entry was retained for continued analysis. One sequence of Niphargus novomestanus S. Karaman, 1952 (KR858496, BOLD:ADD1128) was removed from the dataset because it was deleted from GenBank by its submitter (“This record was removed at the submitter’s request because the source organism cannot be confirmed.” GenBank website). The resulting dataset contained 25,702 records (Fig. 1, File S2).

Each record in the dataset was then further refined by sorting into categories according to the level of taxonomic identification. The following categories were used: order, family, subfamily, genus and species. Where records were provided with a temporary species identification, i.e., they are recognised as separate morphospecies but are not determined to correspond to a known taxon—they were treated as a separate category. In the whole dataset ca. 2.5% of records (596 individuals, 145 BINs) had uncertain identification with “cf.” or “aff”. Because the majority of them (417 records, 101 BINs) were associated with five species of one genus (Gammarus) for simplification all such records were treated as final species identifications. However, it is understood that the use of open nomenclature, when applied to identifications, provides an indication of the level of uncertainty, and may be intended to indicate the presence of new species or species complexes.

The data in BOLD come from wide variety of projects, some of which involve detailed taxonomic study by specialists, others are focused on monitoring or other topics in which taxonomic specialists are not involved. For the purposes of our analyses it was assumed that the identification accuracy was equal throughout the whole dataset, regardless of its origin. In several cases identification of the specimens within a single BIN varied strongly, with some records remaining at order level while others were determined to the species level. BINs aim to represent a putative species, so in the above example, the most detailed taxonomic information was applied to all records within the single BIN. Sometimes multiple (most often two) species or genus names were associated with a single BIN (87 cases). Each of these cases was checked individually. Sometimes it was an obvious misidentification of a single individual within a large group - if this was noted the misidentified record was added as an additional element to the records identified to the lowest congruent level (e.g., if the genus name matched the BIN genus, the misidentified taxon was added as an additional record identified to the genus level, if the lowest congruent level was family it was added to the family records); and the taxon identification of the majority of records was applied as correct. When it was impossible to judge which name was correct, the name of the identifier was checked and identifications carried out by taxonomists specializing in Amphipoda were prioritised over those provided by a non-specialist study. Where this process did not give a satisfactory conclusion, the BIN was allocated an identification at a rank that was congruent for the different records. The list of taxa with incongruent identifications together with an explanation of the final decision is presented in File S3.

Based on the taxonomic identification of the records the associated BINs were divided into the following environmental categories:

a) marine

b) freshwater

c) terrestrial.

Taxa that inhabit both marine realm and brackish environments were allocated to the marine category. Taxa from freshwater also occurring in brackish waters were allocated to the freshwater category. All representatives of the family Talitridae were treated as terrestrial taxa. Where taxonomic information was not detailed enough to provide environmental information about the particular BIN, the geographic data (coordinates and/or locality description) of the associated records were used to ascribe a particular BIN to one of the above categories. In some cases, this necessitated checking the original publication. A small number of unallocated BINs (18) and associated records (44) were used only in the first general summary of amphipod barcodes, but they were removed from further analyses (File S4).

In order to verify the correct environmental allocation of BINs, all BINs with records possessing coordinates were plotted on a map using the software QGIS2.16.1 (QGIS Development Team, 2018). Cases where incongruence between the ascribed environment and the geographic position appeared were checked individually. For those records without detailed geographic information the country of origin was taken from either BOLD or the associated publication.

In order to verify the barcode coverage within the studied group a list of BINs associated with a species name was compared with the list of accepted amphipod species names available in the World Amphipoda Database (WAD, Horton et al., 2020), accessed on 17-07-2020). A barcode quality assessment of the species represented in BOLD, based on the grading system proposed by Oliveira et al. (2016) and slightly modified by Fontes et al. (2020) was applied. This system consists of five grades: A – consolidated concordance (>10 sequences of a single morphospecies grouped in a single BIN), B– basal concordance (same as grade A but between three and 10 sequences available in the library), C– multiple BINs (one morphospecies assigned to more than one BIN), D– insufficient data (single species is assigned to single BIN but it is represented by less than three sequences in the barcode library), E– discordant species assignment (more than one species assigned to a single BIN). Fontes et al. (2020) provide an R-based application (Barcode, Audit & Grad System – BAGS), and uses only those records possessing species names. Since our aim was to focus on all available barcode records (including sequences identified only to higher ranks), the assessment was carried out manually. Additionally, as a result of initial treatment of the dataset, misidentified species records or BINs with unclear species identification, were already removed, so category E (discordant species assignment; Oliveira et al., 2016; Fontes et al., 2020) was not recorded. For the purpose of the present study Lysianassoidea incertae sedis was treated as an additional family. The amphipod families were divided into four categories depending on the number of species in each: low species rich families (up to 10 species), moderately species rich families (from 11 to 30 species), species rich families (31–100 species), very species rich families (more than 100 species). This division allowed verification of pattern between the species richness of the family and its representation in BOLD.

Results

Of the 25,702 amphipod COI records, 46.5% (11,958 records) were freshwater, 43.5% (11,169 records) were from the marine realm, and 9.8% (2,531 records) were terrestrial taxa. Of the 3,835 recognized BINs in total, 45% (1,726 BINs) belonged to freshwater taxa, 50% (1,920 BINs) were marine, and 4.5% (171 BINs) were from terrestrial taxa. 44 records (0.2%) and their associated 18 BINs (0.5%) could not be ascribed to the above environmental categories and were not considered further (Figs. 2A, 2B).

More than half (57.5%) of the records available in BOLD possessed coordinates, and 20% had information about the country of origin. Geographic information about the remaining 22.5% was provided only in the original publication. Geographic information is more comprehensive for marine taxa, where 71% of records possessed coordinates (compared to 47% for freshwater, and 50% for terrestrial taxa). Molecular studies of freshwater Amphipoda are focused mainly in the Northern hemisphere (particularly European countries, Russia and United States) while in the Southern hemisphere, Australia, New Zealand and Argentina are well studied (Fig. 3A). There is a complete lack of records (amphipod sequences) from Brazil, equatorial America and vast areas of Africa. Similar patterns of data coverage were seen for marine amphipods, which have greater numbers of records along European, North American and East Asian coasts. In the Southern hemisphere, Australia, New Zealand and Antarctica had larger numbers of barcode records (Fig. 3B). However, vast areas of the deep sea and the Arctic Ocean remain undersampled. Terrestrial Amphipoda in Europe, North America, China, Australia and Chile were the best represented (Fig. 3C), but sampling gaps were seen in the continents of South America and Africa.

The majority of records (69.8%, 17,922 recs.) had a complete species-level identification. Of the remaining 30.2% of records, 5.6% (1,433 recs.) had received temporary names (open nomenclature), 11.3% (2,902 recs.) remained identified at the genus level, 0.2% (40 recs.) at subfamily, 5.0% (1,285 recs.) at family, and 8.1% (2,076 recs.) at the order level. Levels of identification varied according to the environment, with marine taxa having greater proportions of taxa identified only to higher taxonomic ranks (Fig. 4A). The majority of BINs (3,817) were associated with species names (55.7%, 2,126 BINs). These were followed by BINs identified to the order level (13.3%, 506 BINs), generic or family level (10.7%, 407 BINs each) and those with a temporary name (9.4%, 359). BINs with only a subfamily name constituted just 0.3% (12). Greater variations between environments were seen for the BINs, with 74% (1,284) of freshwater BINs having a species level identification, compared to only 39% (751) of marine BINs (Fig. 4B). More than 20% (444, 23%) of the BINs for marine taxa remained identified at the order level.

Regardless of the environmental origin, the majority of BINs were represented by a single sequence (Fig. 5). BINs represented by five or fewer sequences constituted around two thirds (67%, 114 terrestrial BINs to three quarters, 78%, 1,488 marine BINs) of BINs recorded in a particular environment. Freshwater taxa had 41 BINs (2.4%) represented by more than 50 sequences, compared to 28 (1.5%) for marine taxa, and eight (4.7%) for terrestrial taxa. When only those BINs with complete species-level identifications are considered, the proportion of sequences representing a particular MOTU does not change, with freshwater taxa having 78% of BINs (1,016) represented by five or fewer sequences. Almost three quarters of marine BINs (71%, 525 BINs) had five or fewer sequences in BOLD, while this proportion was 61% (56 BINs) for terrestrial taxa. Freshwater taxa had 35 BINs (3%) represented by more than 50 sequences, compared to 27 (4%) for marine taxa, and 6 (7%) for terrestrial taxa. The best represented BIN in BOLD (801 sequences) belonged to the terrestrial species Orchestoidea tuberculata Nicolet, 1849 (BOLD:ACQ3380), followed by the marine species Gammarus oceanicus Segestråle, 1947 (BOLD:AAA1262, 553 sequences), and the freshwater species Diporeia hoyi (S.I. Smith, 1874) (BOLD:AAA1473, 512 sequences). A further 26 BINs were represented by more than 100 sequences, including 17 freshwater, seven marine and two terrestrial BINs (File S5).

Out of the 3,817 studied BINs, just over half (55.7%, 2,126) were associated with a species-level identification, representing 1,001 species. Freshwater BINs with species identification reached 1,284, associated with 453 species, while 751 marine BINs were determined to 496 species. Of the 91 terrestrial BINs, 52 species were identified. Generally, a single morphological species was associated with each BIN (68%, 680 cases, 288 in freshwater, 359 marine, 33 terrestrial). 17% of the identified species were associated with two different BINs (72 freshwater, 82 marine and 14 terrestrial) (Fig. 6). There were however 19 cases when one single morphological species was represented by more than 10 BINs (17 freshwater, one marine and one terrestrial) (File S6). The greatest number of BINs was recorded for the freshwater species Gammarus balcanicus Schäferna, 1923 represented by 143 BINs (45 BINs were identified as “cf.” or “aff.”) followed by another freshwater taxon Hyallella azteca (Saussure, 1858) (62 BINs) and Gammarus fossarum Koch, 1836 (51 BINs; 19 BINs identified as “cf.” or “aff.”). Among terrestrial taxa the highest molecular variation (12 BINs) was recorded for Morinoia japonica (Tattersall, 1922) (present in BOLD under former generic name Platorchestia), while Apohyale stebbingi Chevreux, 1888 (with 11 BINs recognized) was the most diverse among marine species.

Of the 239 accepted families of Amphipoda (238 families and Lysianassoidea incertae sedis), 105 (44%) were represented by at least one species in BOLD (Table 1). The largest number of families had up to 20% of species barcoded, while only ten families had more than half of the known species barcoded (File S7). Thirteen families lacking barcoded species had at least one barcoded taxon identified at the genus level, a further five families had a taxon identified at the family level.

Just under ten percent (999 spp., 9.7%) of the 10,330 accepted species of Amphipoda (Horton et al., 2020) had barcodes. Of the nominal species possessing barcodes almost 500 (496 spp.) are marine, 451 spp. are freshwater and 52 spp. are terrestrial taxa. The data coverage of the majority of species, no matter their environmental origin, is not sufficient for the barcodes to be trusted according to the quality control system (Table 2) (Oliveira et al., 2016; Fontes et al., 2020). Additionally, a large group of taxa is represented by multiple BINs; only 10% of species represent consolidated concordance of available barcodes.

The breakdown of amphipod families according to the assigned categories of richness and their respective representation in BOLD can be seen in Table 3. Almost every one of the very species rich families had at least one species barcoded (31 families out of 32), and 22 of 30 species rich families are represented in BOLD. For both moderately low and low species rich families 26 possessed at least one representative in BOLD constituting respectively 48% and 21% of all families each (File S7). The mean coverage of barcodes for species in each of the above groups was around 10% with the highest observed for low species rich families (12%) and the lowest (8%) recorded for families grouping from 30 to 100 species. However, if the families without any molecular information were removed from the study these numbers considerably change. The low species rich families (1–10 spp.) had a barcode coverage at the level of 49%, moderately species rich families (11–30 spp.) reached 21% of coverage, while the rich and very rich amphipod families (more than 30 spp.) had only 9–10% of species studied.

A third of families (34) have at least one species characterized by consolidated concordance of available barcodes (category A of the quality grading system). Another third of families (38) do not have any species in categories A or B, indicating that the species already studied represent a potential cryptic diversity or the available data are insufficient (Table 4).

Within the very species rich families, the best representation in BOLD was recorded for Niphargidae (36.5% of known species represented with a barcode), Gammaridae (31%) and Crangonyctidae (16%). Only the family Stegocephalidae did not have any representative with species level identification (although barcodes belonging to this family but identified at genus level were present). The least studied families within this group (but having at least one species barcode) were: Phoxocephalidae (1% of the species with a barcode), Dexaminidae, Liljeborgiidae and Maeridae (ca. 2% of the species with a barcode). Among species rich families 41% of the species from Pseudoniphargidae had barcodes, while the Epimeriidae and Pontogammaridae had 20% and 19%, respectively. The best represented moderately species rich families were Metacrangonyctidae, Oxycephalidae and Hyperiidae with 55%, 50% and 48% of the associated species represented with a barcode respectively. Within low species rich families four (Baikalogammaridae, Crymostygidae, Cyllopodidae and Tryphanidae) had all known species represented with barcodes, but other than Cyllopodidae (two species) the families are monotypic (File S7).

Discussion

Supplemental Information