Could do better! A high school market survey of fish labelling in Sydney, Australia, using DNA barcodes

DNA extraction: optimization and final protocol

Initially, museum staff provided short training sessions in all laboratory techniques prior to commencement, including sterile technique, tissue sampling, use of pipettes, etc. Then, wet laboratory work (DNA extraction and PCR amplification) was carried out in the high school science classroom by the 14 participating students under the supervision of school and museum staff.

Approximately 20 mg of tissue was removed from each fish sample using a sterile scalpel blade and placed into 1.5 mL microcentrifuge tubes. To better fit with school scheduling, rapid DNA extraction processes were used. Initial trials on 12 samples from three fish products used a chelex-based DNA extraction protocol (Walsh, Metzger & Higuchi, 1991). As this protocol requires very accurate pipetting, it was successful for less than half the samples carried out by the students, thus we subsequently switched to the Bioline MyTaq™ Extract-PCR Kit (Bioline, Alexandria, Australia). The Bioline MyTaq™ Extract-PCR Kit appeared more robust to use by non-experts and was used for all subsequent samples following the manufacturer’s recommended protocols. However, we also compared the recommended protocol of diluting the final supernatant DNA solution 10-fold in water, vs. not diluting, finding that both methods were effective. Five successful samples from the optimization stage were included in the final data set (sample numbers SGS005–SGS011).

PCR amplification and sequencing

Polymerase chain reaction was performed on a Perkin Elmer 2400 PCR machine in the classroom. PCR used 40 cycles in 25 μL volumes following the manufacturer’s recommended protocols, with a final MgCl₂ concentration of 2.5 mM and annealing temperatures of 50–52 °C. Degenerate PCR primers were designed for this study based on a published primer cocktail. We took two published forwards primers, VF2_t1 and FishF2_t1 (Ivanova et al., 2007), which bind to exactly the same region of the COI gene but have different nucleotides in some positions, took the consensus sequence of these primers to make a degenerate sequence, and added a 5′-M13 tail to obtain our primer AMFishFm (5′-GTAAAACGACGGCCAGTCRACYAAYCAYAAAGAYATYGGCAC-3′). We repeated this procedure for the two reverse primers VF2_t1 and FishF2_t1 to obtain the degenerate reverse primer AMFishRm (5′-CAGGAAACAGCTATGACACYTCAGGGTGWCCGAARAAYCARAA-3′). The nucleotides in italics are 5′-M13 tails, used to facilitate sequencing with primers M13F (−21) and M13R (−27) (Messing, 1983). All primers are registered in the BOLD primer database. We also compared the performance of our newly designed primers with the primer pair FishF1/FishR1 (Ward et al., 2005).

Polymerase chain reaction products were visualised in the classroom using the E-Gels Precast Agarose Gel system (ThermoFisher Scientific, Waltham, MA, USA), and amplified PCR products were either sent to Macrogen (Seoul, South Korea) for Exo-SAP purification and DNA sequencing on an ABI 3730XL or were purified using ExoSAP-IT (Thermo Fisher Scientific, Scoresby, Victoria, Australia) in the Australian Centre for Wildlife Genomics, Australian Museum, and sequenced by the Australian Genome Research Facility (Sydney, Australia).

Sample replication

To test for consistency among students, we replicated sampling for 10 samples on different days and performed separate DNA extractions, PCRs and DNA sequencing. For 34 additional samples (sample IDs SGS201–SGS234) we performed PCR using both primer pair AMFishFm/AMFishRm and primer pair FishF1/FishR1. Sequencing results were compared. Only the highest quality forwards and reverse direction trace files were uploaded to BOLD for each sample.

Sequence upload and comparison

The BOLD student data portal (SDP) (Santschi et al., 2013) was used for initial analyses performed by the students. For samples with high quality sequence trace files in both forward and reverse directions, trace files were uploaded to the SDP, automatically assembled, and manually edited when necessary. The resulting consensus sequences were then entered into the SDP. All trace files were also edited and assembled using Geneious 9.1 (Kearse et al., 2012), and the resulting consensus sequences were used in BOLD searches on July 14, 2016, using the Species Level and Public Record databases on BOLD. Consensus sequences were uploaded to BOLD SDP if there was not already a consensus sequence in BOLD for that sample.

Misidentified samples

DNA barcode data revealed that four samples (indicated by asterisks in Table 1) were mislabelled. Each sample is discussed below.

Blue Warehou (Seriolella brama) was substituted with the closely related Silver Warehou (Spirodela punctata). The former has a reputation as a good eating fish, however, it has been overfished and its population has not rebounded after more than a decade, according to Australia’s Sustainable Seafood Guide (Australian Marine Conservation Society, 2014). The Silver Warehou has less appeal in food markets as it is perceived as lower quality for eating, but it is an abundant species in South East Australia. Blue Warehou, on the other hand, is listed as ‘conservation dependent’ under Australian Commonwealth environmental legislation, although that still permits targeted fishing of this species (Australian Fisheries Management Authority, 2016).

A sample sold as Bream (SGS209), which is a generic term referring to members of the family Sparidae, was identified as Jackass Morwong, Nemadactylus macropterus, family Cheilodactylidae. This species is listed as overfished by the NSW Department of Primary Industries but as sustainable by the Australian Fisheries Management Authority (2016).

A sample of Latchet (SGS115), which is a type of Gurnard, Pterygotrigla polyommata, was identified as either Red Gurnard, Chelidonichthys kumu, or Spiny Red Gurnard, Chelidonichthys spinosus. These fishes look similar and this is likely a genuine case of mistaken identity rather than intentional mislabelling.

A single sample of Flathead (SGS117) was also identified as either Red Gurnard, Chelidonichthys kumu, or Spiny Red Gurnard, Chelidonichthys spinosus, although this sample was sequenced only once and cross-contamination or mislabelling of samples cannot be ruled out.

Species which cannot be identified to species level using DNA barcodes

Striped Marlin (K. audax) and White Marlin (K. albida) cannot be distinguished using DNA barcodes. Hanner et al. (2011) presented both COI data and data from a nuclear gene, rhodopsin, from both species with the same result, and suggested that the taxonomy of these species may need revision. Regardless, the species can be separated by incorporating information on their geographic distribution as the former species is found in the Indian and Pacific Oceans while the latter species is found in the Atlantic Ocean.

Two samples of Big-Eye Ocean Perch, Helicolenus barathri, and one of Ocean Perch (Helicolenus barathri or Helicolenus percoides) were identified by BOLD as one of three species, Helicolenus barathri, Helicolenus lahillei, or Helicolenus dactylopterus. It is unclear whether DNA barcodes lack the resolving power to distinguish these species or whether there are some misidentified specimens on BOLD. However, we note that the latter two species occur in the Atlantic Ocean while Helicolenus barathri occurs in the Southwest Pacific, so they may be distinguished by their place of origin, if that information is available. Furthermore, the other Southwest Pacific species, Helicolenus percoides, is easily distinguished from the other three species by DNA barcodes.

Two samples of Blue Grenadier, M. novaezelandiae, and two samples of Hoki, either M. novaezelandiae or M. magellanicus, were identified by BOLD IDE as either of these two species, that is, DNA barcode data cannot distinguish the two species. However, the online Catalog of Fishes (Eschmeyer et al., 2018) lists M. magellanicus as a junior synonym of M. novaezelandiae.

For Sand Whiting, Sillago ciliata, BOLD gave 100% matches to two species, Sillago ciliata and Stenoria analis. This lack of resolution was investigated more thoroughly by Krück et al. (2013) using both mitochondrial and nuclear genes. The authors found that a single site in the DNA barcode region of COI was usually diagnostic for these two species, however, that was not the case for three of their 60 samples (5%), which showed discordance between nuclear and mitochondrial gene phylogenies, evidence of a past hybridization event. Nuclear gene data are therefore required to distinguish these two species. The ranges of these two commercial species overlap in Queensland so it may prove important to distinguish these species in the Queensland fishery.

Samples not identified to species level by BOLD IDE, but identifiable with other methods

Tuna (Thunnus) species cannot be accurately identified by the BOLD IDE because it relies on threshold distances and neighbour-joining trees, and there is insufficient COI sequence divergence between species for these methods to accurately distinguish species. However, Lowenstein, Amato & Kolokotronis (2009) demonstrated that DNA barcode data can be used to distinguish Thunnus species if one uses a character-based approach or if one simply takes the closest match in a BLAST search (or on BOLD). We used the latter approach to refine our identification of tuna samples, and concluded that both were correctly identified as Yellowfin Tuna, Thunnus albacares.

Similarly, the BOLD IDE identified our two samples of Yellowtail Scad, Trachurus novaezelandiae, as one of three species, Trachurus novaezelandiae, Trachurus declivis, or Trissolcus japonicus, however the closest matches were to Trachurus novaezelandiae. Furthermore, using the BOLD Published database, NJ trees showed clear separation of Trachurus novaezelandiae from the other two species.

Eastern Sea Garfish, Hyporhamphus australis, shows a similar pattern on BOLD, with very low divergences among three species, however a BOLD NJ tree separates Hyporhamphus australis from the other Hyporhamphus species, Hyporhamphus melanochir and Hyporhamphus ihi.

Samples not identified to species level by BOLD IDE, possibly due to misidentified reference samples on BOLD

For Hyporhamphus australis, a sequence labelled as Arrhamphus sclerolepis is also clustered with Hyporhamphus melanochir sequences. However, that A. sclerolepis sequence is well separated in the NJ tree from conspecific sequences and would appear to be misidentified.

Goldband Snapper, Pristipomoides multidens, was identified by BOLD IDE as either Pristipomoides multidens or Pomadasys maculatus. Yet, the single matching sample of Pomadasys maculatus has clearly been misidentified as it belongs to a different family and the other >50 representatives of this species on BOLD are distantly related and are not found in BOLD searches using Pristipomoides multidens sequences.

Yellowtail Kingfish, Seriola lalandi, was identified by BOLD IDE as Seriola lalandi or Schistura zonata. BOLD IDE NJ trees show Seriola lalandi as a single cluster of some 30 samples, with five samples of Schistura zonata from Brazil nested within this cluster. There is <1% distance among members of this cluster. However, there is a second cluster of Schistura zonata samples from the USA, placed >7% distant from Seriola lalandi, which suggests that the Brazil samples may have been misidentified (or the taxonomy is incorrect). In this case the Brazil ‘Schistura zonata’ sequences are in the Public Record Database while the USA sequences are unpublished.

For Centroberyx affinis, a search of the BOLD Public Record Database provided a 98.3% match to a single specimen of this species. However, searching the Species Level Record Database found a 100% match to a cluster of unpublished sequences of both Centroberyx affinis and Centroberyx australis. Supposed C. australis sequences group into three different barcode clusters. Thus, our sample is likely correctly identified, but no firm conclusions can be drawn at this stage.

Our single Shark sample was identified as Carcharhinus brevipinna. A single record of Carcharhinus limbatus was embedded within the cluster of Carcharhinus brevipinna on BOLD. That single record is well separated from the many other Carcharhinus limbatus on BOLD and thus likely represents a misidentified specimen. This is an example of the limitations of the BOLD database and highlights the importance of database validation and utility of museum voucher collections.

Finally, the Bluespotted Goatfish, Upeneichthys vlamingii, was identified by BOLD IDE using the Species Level Record Database as either Upeneichthys vlamingii or Uroplectes lineatus. However, the two Uroplectes lineatus records on BOLD are both unpublished and are separated by >4.5% COI distance, so it appears likely that one of the Uroplectes lineatus sequences has been misidentified.

Use of standard names

The AFNS AS SSA 5300-2015 (AFNS) specifies that fish sold to consumers (i.e. retail sales and restaurants) must be identified by their standard fish name. However, the AFNS is a voluntary standard and Food Standards Australia and New Zealand (FSANZ) does not mandate compliance with the standard (Senate Standing Committee on Rural and Regional Affairs and Transport, 2014). A total of 20 species of fish that we sampled, or 40% of our sample, did not precisely comply with the standard. However, if we exclude spelling errors and cases where vendors supplied more information than the standard requires (e.g. Rainbow Trout rather than just Trout) then the error rate drops to 22%, which is still substantial. Many published studies have not differentiated between mislabelling due to the use of unofficial names, and actual misidentification or misrepresentation of species.

Accuracy of product labels

DNA barcodes confirmed the retailers’ identifications of all but four of the 61 samples and 39 species sampled, that is, 7% of samples and 10% of species were mislabelled. While this is poorer than the only other published market survey in Australia that used DNA barcodes (Lamendin, Miller & Ward, 2015) it is at the lower end of the range of mislabelling rates reported from surveys conducted in other countries. In a review of 51 seafood authentication studies Pardo, Jiménez & Pérez-Villarreal (2016) reported that the average rate of mislabelling was 30%. Restaurants and takeaways generally recorded higher levels of mislabelling than fishmongers and supermarkets, but only comprised 10% of samples analysed in all studies. Our survey was restricted to two major retailers of fresh fish in the Sydney central business district, selling mostly Australian produced fish. It is possible that different results would be obtained if sampling imported fish, or other localities in Australia, or restaurants.

Adequacy of the Australian fish names standard

As a matter of principle, wild caught IUCN Red-Listed species should not be sold for food. At the very least, consumers should be empowered to make choices that support sustainable use of natural resources. This is impossible when retailers are permitted to group many species under one label, for example, ‘Snapper’ is an umbrella term that has found to be applied to more than 60 species in 16 fish families (Cawthorn, Baillie & Mariani, 2018). An example in the AFNS is Basa. Our single sample of Basa was identified as Pangasianodon hypophthalmus. The AFNS defines Basa as any member of the family Pangasiidae, therefore this name is compliant. However, the AFNS usage of this name is extremely broad, and is inconsistent with global standards. On the United Nations Food and Agriculture Organization fish names list, Pangasianodon hypophthalmus is called Swai, while Basa is reserved for Pangasius bocourti. Both species are farmed in Asia, but natural populations of Pangasianodon hypophthalmus are Red-listed as Endangered. Two other Pangasiidae, Pangasianodon gigas and Pangasius sanitwongsei, are Red-listed as Critically Endangered. DNA barcode data provides robust discrimination of these species, yet any of them can be called Basa in Australia. Ideally, the names standard would distinguish between these species and also between wild and farmed populations of these species. Furthermore, the industry should only allow imports of species in the Least Concern category.