Long-distance Southern Ocean environmental DNA (eDNA) transect provides insights into spatial marine biota and invasion pathways for non-native species

The Southern Ocean surrounding Antarctica harbours some of the most pristine marine environments remaining, but is increasingly vulnerable to anthropogenic pressures, climate change, and invasion by non-native species. Monitoring biotic responses to cumulative impacts requires spatiotemporal baselines and ongoing monitoring - traditionally, this has been obtained by continuous plankton recorder (CPR) surveys. Here, we conduct a 3000 nautical mile environmental DNA (eDNA) transect from Hobart (Australia) to Davis Station (Antarctica). We evaluate eDNA sampling strategies for long-term open ocean biomonitoring by comparing two water volume and ﬁlter pore size combinations: large (12 L with 20 μ m) and small (2 L with 0.45 μ m). Employing a broad COI metabarcoding assay, we found the large sample/pore combination was better suited to open-ocean monitoring, detecting more target DNA and rare or low abundance species. Comparisons with four simultaneously conducted CPR transects revealed that eDNA detections were more diverse than CPR, with 7 (4 unique) and 4 (1 unique) phyla detections respectively. While both methods eﬀectively delineated biodiversity patterns across the Southern Ocean, eDNA enables surveys in the presence of sea-ice where CPR cannot be conducted. Accordingly, 16 species of concern were detected along the transect using eDNA, notably in the Antarctic region (south of 60 ° S). These were largely attributed to hull biofouling, a recognized pathway for marine introductions into Antarctica. In a warming Southern Ocean, continued biomonitoring is vital for conserving Antarctic ecosystems. We advocate for the long-term implementation of eDNA metabarcoding alongside CPR surveys to facilitate ecosystem-based management of these vulnerable environments.


Introduction
The Southern Ocean surrounding Antarctica harbours some of the largest, most pristine remaining marine environments (Brooks et al., 2022).Although the Southern Ocean covers approximately 10% of the world's oceans, it plays a disproportionately important role through the multiple ecosystem services it provides (Grant et al., 2013).It is particularly important for climate regulation through storing significant amounts of heat and carbon dioxide (Chen et al., 2019;Cavanagh et al., 2021).The high biological productivity is valuable in oxygen production, fuelling marine food webs and supporting important fisheries (Grant et al., 2013;Murphy et al., 2021).It is also an important home to a unique and diverse array of marine biota, boasting high levels of endemism (Rogers et al., 2012).However, Southern Ocean and Antarctic environments, and the valuable services they provide, are rapidly changing in response to increased anthropogenic pressures and climate change (Chown et al. 2015;Johnston et al. 2022).
The Southern Ocean comprises several biogeographic zones formed by distinct oceanographic fronts (Constable et al., 2014), in particular the boundary of the strong Antarctic Circumpolar Current (ACC) which encircles Antarctica and creates a sharp gradient between cold polar water and warmer waters further north.Bathymetry and sea ice coverage also influence the biotic habitat within the Southern Ocean (Constable et al., 2003;Massom & Stammerjohn, 2010).The environmental heterogeneity between and within these zones results in a variety of complex biological ecosystems characterised by distinct ecological communities that vary in species richness and diversity (Griffiths, 2010).These communities are highly susceptible to change in response to multiple local and global drivers (Morley et al. 2020a;Grant et al. 2021), exacerbated by anthropogenic impacts including climate change and resource exploitation (Rogers et al. 2020;Brooks et al. 2022).Moving forward, monitoring temporal and spatial changes in biodiversity, species distributions, and community structure will be essential in making ecosystem-based management decisions to conserve the unique biota of the Southern Ocean (Brooks et al. 2016;Kennicutt II et al. 2019) and to better understand the key role it plays in the dynamics of the oceans globally.
To date, biodiversity monitoring in the Southern Ocean has been inhibited by various logistical challenges (Griffiths, 2010), resulting in a paucity of comprehensive baseline data.Spatial coverage of biodiversity data has been limited by the remoteness of some parts of the Southern Ocean, which are rarely visited by scientific expeditions (Griffiths, 2010).Additionally, scientific expeditions are usually limited to the summer months in the polar regions, as sea ice extent and short days make winter expeditions challenging, resulting in a temporal bias of biodiversity data sets (Howell et al., 2021).Additional spatiotemporal datasets are urgently needed to provide more complete baselines and for comprehensively investigating biodiversity patterns in the face of growing anthropogenic impacts.
Continuous plankton recorder (CPR) surveys are routinely implemented to monitor Southern Ocean biotic diversity, particularly zooplankton communities (Hunt & Hosie, 2003;McLeod et al., 2010).The CPR is a device towed behind a ship at approximately 10 m depth, collecting plankton samples onto a continuously winding silk (Hunt & Hosie, 2003;Reid et al., 2003).CPR surveys have been conducted annually in the Southern Ocean since 1991, and have covered roughly 285, 000 nautical miles to date (Takahashi & Hosie, 2021), providing some of the longest running biological datasets (Hunt & Hosie, 2003;Richardson et al., 2006).However, the method requires detailed taxonomic identification of large numbers of specimens (McLeod et al., 2010), a time-consuming process that can be difficult when specimens are damaged, gelatinous or in non-adult life stages (Deagle et al., 2018).Furthermore, CPR surveys are designed to optimally target zooplankton communities and are therefore restricted in the taxonomic comprehensiveness of baseline data they can provide.
Environmental DNA (eDNA) biomonitoring presents a promising approach to complement and expand the current taxonomic scope of CPR surveys.eDNA metabarcoding describes biodiversity patterns through the amplification of trace amounts of DNA naturally shed into the environment by organisms.The technique typically targets DNA barcodes using broad-spectrum metabarcoding assays able to profile a wide-range of biota without the need for visual observation or identification of specimens.eDNA studies in various marine environments have demonstrated their ability to enrich temporal and spatial surveys by capturing biota across the tree of life (Berry et al., 2019(Berry et al., , 2023;;Minamoto et al., 2017).Notably, eDNA data can discriminate finescale spatial and habitat variation (Jeunen et al., 2019;West et al., 2021), and has demonstrated increased sensitivity to the presence of endangered or invasive species in comparison to conventional methods (Klymus et al., 2017;Nester et al., 2022).Early detection of non-native species is particularly important in the Antarctic region (south of 60°S) given the vulnerability of its unique biota to the negative impacts of nonnative settlement, and is consistent with the key principles of the Protocol on Environmental Protection to the Antarctic Treaty (1991).
Using ship-based surveys, we conduct one of the longest eDNA transects to date, spanning over 3000 nautical miles from Hobart, Tasmania (Australia) to Davis Station in East Antarctica.This study aimed to refine eDNA methodologies and determine their viability for surveying open ocean biotic communities by investigating biodiversity patterns in the Southern Ocean.To refine eDNA methodologies and determine their viability for surveying open ocean Animalia communities, two combinations of water volume and filter pore sizes were tested (12 L with 20μm, and 2 L with 0.45μm).Both water volume and pore size can influence the retention of eDNA from target organisms (animals) and non-target organisms (e.g., marine microorganisms), and optimal sampling strategies can differ depending on the sampled environment.We compared taxonomic resolution, diversity, and richness between these two approaches, and investigated their potential impacts on ecological inferences.We also investigated the capacity of eDNA surveys to identify distinct biodiversity patterns within expansive open ocean regions.Lastly, we compared the eDNA survey to concurrent CPR surveys, highlighting the strengths, limitations, and potential blind spots of each approach.Given the remoteness of the Southern Ocean and the difficulties associated with sampling the region, there is great potential to integrate eDNA sampling into voyages to Antarctica and use this data to inform decision-making, ocean-ecosystem models and conservation across the Southern Ocean.

eDNA sampling
Samples (n=138) were collected aboard the RSV Aurora Australis on a resupply voyage between Hobart, Tasmania (42°52'54.84"S147°20'29.76"E)and Davis Station, Antarctica (66°26'14.28"S77°28'24.6"E) in November 2019 (Figure 1; Table S1).Two combinations of water volume and filter pore sizes were tested at each sampling location: 12 L with 20 μm, and 2 L with 0.45 μm, herein referred to as "LargeVF" and "SmallVF" respectively.Water samples were collected and filtered approximately every 4 hours (4.52 ± 0.23) via the ship's uncontaminated seawater intake line (4 ± 2 m depth).SmallVF water samples (2 L, n=69) were filtered using a Sentino microbiology peristaltic pump (Pall Life Sciences) through 47 mm diameter, 0.45 μm pore size polyethersulfone filter membranes (Pall Life Sciences).Simultaneously, LargeVF water samples (12 L, n=69) were filtered using a Masterflex L/S console pump system (Cole-Parmer) through 25 mm diameter, 20 μm pore size nylon filter membranes (Merck).SmallVF filter membranes (47 mm diameter) were cut in half and immediately preserved at -80°C, with one half to be analysed and one to be stored as a reserve and a form of eDNA biobanking (Jarman et al., 2018).LargeVF filter membranes were not cut in half due to their smaller diameter (25 mm) and were stored whole at -80°C.Filtration equipment was rinsed with a 10% bleach solution and freshwater from the laboratory in between every sample, and soaked for 15 minutes with 10% bleach every tenth sample.Field controls consisted of 500 mL samples (n=10) of laboratory freshwater and the 10% bleach solution used for sterilisation.

CPR sampling
CPR sampling was conducted in parallel to the eDNA sampling using a Type II CPR (Mark V) and following standard Southern Ocean CPR methods (Hosie et al., 2003).The CPR was fitted with 270 μm nylon mesh and towed 100 m behind the ship at approximately 10 m depth.Four CPR transects were completed between latitudes 47°8'12.84"Sand 59°3'38.52"S,spanning a total of 1,858 nautical miles (Figure 1; Table S2).The instrument was not used in the presence of sea-ice, limiting sample collection at the Southern part of the voyage.CPR silk spools were fixed in 10% buffered formalin for morphological identification.

Laboratory processing
DNA was extracted from the filter membrane using a DNeasy Blood and Tissue Kit (Qiagen) in an automated QIAcube (Qiagen) DNA extraction system with the following modifications: 540 μl of ATL lysis buffer, 60 μl of Proteinase K, and a 3-hour digestion at 56°C.Extraction controls were processed in parallel with all samples to detect any laboratory or between sample contamination.Final DNA extracts were eluted in 100 μl of AE buffer.
DNA was amplified to target Animalia taxa using mitochondrial cytochrome c oxidase subunit I (COI) markers: m1COIintF (Leray et al., 2013) and jgHCO2198 (Geller et al., 2013), herein referred to as Leray-COI.Samples were serially diluted (1/5, 1/10 and 1/100) to optimise DNA input levels for quantitative PCR (qPCR) and remove potential PCR inhibitors.No-template controls were included on each qPCR plate.Metabarcoding was performed using fusion-tagged primers consisting of Illumina compatible sequencing adapters, a unique 6-8 bp multiple identifier tag (MID-tag), and the Leray-COI primer.Each sample and control were processed in duplicate using the same MID tag, to reduce stochasticity for species with low amounts of template DNA.qPCR reactions (25 μl) consisted of the following concentrations: 2 mM MgCl2, 1× AmpliTaq Gold PCR buffer, 1 U AmpliTaq Gold DNA polymerase (Applied Biosystems), 0.4 μM dNTPs (Astral Scientific), 0.1 mg BSA (Fisher Biotec), 0.6 μL of 5X SYBR Green dye (Life Technologies), 0.4 μM forward and reverse primer, 4 μL of eDNA template (at optimised dilution), made to volume with Ultrapure Distilled Water (Life Technologies).qPCR amplifications were performed using a StepOnePlus Real-Time PCR System (Applied Biosystems) in a single-step process using an adjusted touchdown thermocycler protocol with conditions: 94°C for 10 min, 16 cycles of 95°C for 10 s, 62°C (-1°C per cycle) for 30 s, and 72°C for 45 s, followed by 25 cycles of 46°C for 30 s, with a final extension of 72°C for 5 minutes.qPCRs were prepared in dedicated clean facilities within the TrEnD Laboratory (Curtin University).Amplicons were visualised on 1.5% agarose gels and duplicate reactions from the same eDNA template were combined and then pooled into a library at equimolar ratios.The resulting library was size selected using a Pippin Prep (Sage Science), purified using a Qiaquick PCR Purification Kit (Qiagen), and quantified using a Qubit 4.0 Fluorometer (Invitrogen) and a Qiaxcel Advanced System (Qiagen).The final library was diluted to 2 nM and sequenced on an Illumina MiSeq platform (Illumina) using a 500 cycle (2 x 250 bp) MiSeq V2 Reagent Kit for paired-end sequencing.

Bioinformatics and taxonomic assignments
Bioinformatics and taxonomic assignments of sequencing data were performed using eDNAFlow, a fully automated workflow that processes eDNA data from raw sequences to curated and non-curated zero-radius operational taxonomic units (ZOTUs) and their abundance tables (Mousavi-Derazmahalleh et al., 2021).Briefly, sequences were quality filtered, demultiplexed, denoised and erroneous sequences were removed using a combination of AdapterRemoval (Schubert et al., 2016), OBITools (Boyer et al., 2016), and Usearch (Edgar, 2010) via Zeus, an SGI Linux cluster based in the Pawsey Supercomputing Centre (Kensington, Western Australia).Resulting ZOTUs were queried against NCBI's GenBank nucleotide database using BLASTn.We used MEGAN (Huson et al., 2016) to assign taxonomy with lowest common ancestor parameters "min score": 300; "top percent": 5; "min support": 1; "min percent read to cover (query coverage)": 100.Taxonomic assignments of Animalia taxa were then manually curated and checked for additional entries using the BOLD identification engine (Ratnasingham & Hebert, 2007).Following the approach of Suter et al. (2021), species that hit at 100% within a ZOTU and could not be distinguished based on their COI sequence were merged e.g., 'Calanus propinquus/similimus '.All taxonomic assignments were further evaluated against knowledge of species distributions using the World Register of Marine Species (WoRMS Editorial Board, 2022) and Atlas of Living Australia (ALA -ala.org.au), and their status (invasive or pest) checked using the Global Invasive Species Database (GISD -iucngisd.org),the Global Register of Introduced Species (GRIIS -griis.org),and the CABI Invasive and Invasive Species Compendium (cabi.org/isc).Taxa classified as pest, invasive, or non-native are herein referred to as species of biosecurity concern.Taxa identified as hull-fouling organisms were removed from open ocean statistical analyses (Table S3).
ZOTUs assigned to contaminants, non-Animalia, and non-marine Animalia were removed and excluded from further analysis.These detections included humans (Homo sapiens ), dog (Canis lupus ), and groups such as insects (class Insecta) and birds (class Aves).Potential cross-contaminant ZOTUs were identified as marine Animalia ZOTUs with reads present in either field, extraction, or PCR controls.ZOTUs with more than 0.5% of reads originating from controls were entirely removed (Table S4 and S5).Below this, the number of reads present in controls or the percentage of the ZOTU comprising the control (whichever was higher) were removed from the samples.For example, a ZOTU assigned toClausocalanus brevipes had 21,300 assigned reads of which two (0.009%) were present in a field control.We therefore removed either two reads or 0.009% from each sample, whichever resulted in a higher number of reads.We opted for this approach over removing ZOTUs entirely as we found the total portion of control reads (23 reads) from these target taxa ZOTUs (57,405 reads) to be just 0.04%.A total of 17 ZOTUs were entirely removed and eight ZOTUs had reads removed.We believe our approach is conservative without removing indicator species or unnecessarily impacting the spatial trends of our data.Four eDNA samples (two LargeVF and two SmallVF) failed to sequence (less than 100 reads) and were removed from analysis, along with the corresponding or matching sample.

CPR morphological analysis
CPR silks (n=368) from the four tows were processed at the Australian Antarctic Division following standard methods (Hosie et al., 2003).In brief, silks were cut into sections representing five nautical miles and zooplankton were identified and counted under a dissecting microscope.Zooplankton were identified to the lowest taxonomic level possible and developmental stages of euphausiids and copepods were recorded.Taxa within some broad lineages were not differentiated (e.g., Chaetognatha), and damaged or unidentifiable specimens were grouped at coarser taxonomic levels where necessary (e.g., indeterminate Hyperiidea and indeterminate euphausiid furcilia).

Identification of Southern Ocean fronts and zones
The Southern Ocean is divided into biophysical zones (see Figure 1) by several Antarctic Circumpolar Current (ACC) fronts, namely: the Subtropical Front (STF-S), the Subantarctic Front (SAF-N and SAF-S), the Polar Front (PF-N and PF-S), the Southern ACC Front (SACCF-N and SACCF-S), and the Southern Boundary (SBDY).Each of these zones exhibits relatively uniform water mass properties (Orsi et al., 1995;Sokolov & Rintoul, 2002), creating distinct habitats that support unique marine life (Bost et al., 2009).From north to south, these are: (1)Subtropical zone (STZ) extending north of the STF-S, (2) the Subantarctic zone (SAZ) extending from the STF-S to the SAF-S,(3) Polar Frontal zone (PFZ) covering the SAF-S to the PF-S,(4) Antarctic zone (AAZ) spanning the PF-S to the SBDY, and(5) South of the ACC zone (SACCZ) covers the area below the SBDY.These fronts and the zones they delimit each have unique environmental characteristics and form distinct habitats that support unique biota (Bost et al., 2009).As such, we aimed to integrate detected Animalia biodiversity with the biophysical properties of Southern Ocean zones.
While the concept of fronts as water mass boundaries is universally accepted (Chapman, 2014), the delineation of Southern Ocean fronts and their defining characteristics vary between studies (Chapman et al., 2020).We opted to define the fronts based on local criteria, in consideration with criteria adopted by other researchers and knowledge of where fronts are likely to be located over the Austral summer (November to March; Table 1).Although precise identification of these fronts is typically achieved at greater depths (Orsi et al., 1995), it has been observed that variations in both sea-surface temperature (SST) and sea-surface salinity (SSS) coincide with average front positions (Sokolov & Rintoul, 2002).Consequently, we applied this methodology in our study to define front locations and subsequently demarcate Southern Ocean zones for the specific purpose of our investigation (Table 1).Raw eDNA taxa were classified as Animalia, non-Animalia, or unclassified post-curation and contamination removal (Table S6).Differences in overall read numbers and the proportion of Animalia taxa (for both reads and ZOTUs) between LargeVF and SmallVF samples was explored visually in R Studio (v4.2.1; R Core Team, 2022).To test the differences observed, a paired Wilcoxon t-test was conducted between the eDNA volumes.We did not rarefy for a specific read depth, as our goal was to understand the relationship between sequencing depth, the proportion of assigned Animalia reads, and the number of species detected between the two eDNA sampling methods.Rarefying the samples would have reduced the sequencing and diversity patterns we wanted to observe and test, leading to false inferences about choice of volume and filter size.To examine the taxonomic resolution of each sampling method (i.e., the ability to accurately distinguish taxa), the proportion of Animalia assignments matched to various taxonomic levels (i.e., 'unclassified', order, family, genus, and species) was calculated.Animalia community composition was assessed at species level across Southern Ocean zones for the eDNA samples.Species identified as hull fouling were removed from this form of analysis.Alpha diversity was calculated for the LargeVF and SmallVF samples using Hill numbers at q=0 (species richness), q=1 (exponential Shannon's entropy index), and q=2 (inverse Simpson's index) within each zone (Alberdi & Gilbert, 2019), and the differences assessed using either paired t-tests or Wilcoxon paired t-tests.An indicator species analysis was performed using the R package indicspecies (De Cáceres et al., 2016) to determine relationships and patterns between species detected and Southern Ocean zones.

Spatial patterns in eDNA and CPR comparisons
A PCO was conducted on a Jaccard resemblance matrix of the presence-absence eDNA data to determine if spatial patterns were similar between LargeVF and SmallVF samples.To explore Southern Ocean diversity, spatial patterns, and allow for comparisons with CPR data, SmallVF and LargeVF samples were merged to form a single eDNA sample at each sampling location.Differences in Southern Ocean zone composition were explored using PRIMER v7 with the PERMANOVA + add-on (Anderson et al., 2008).Merged eDNA data was transformed to presence-absence and converted to a Jaccard matrix.A permutational multivariate analysis of variance (PERMANOVA) was used to test community variation between Southern Ocean zones (STZ, SAZ, PFZ, AAZ, and SACCZ) for the eDNA data.In the presence of significant differences, pairwise comparisons were performed to determine where they occurred.Community composition was then visualised using non-metric multidimensional scaling (nMDS).To allow for comparisons between morphological CPR and eDNA data using PERMANOVA and nMDS analyses, merged eDNA data were subset to samples taken within the length of the CPR transect, and only CPR silk with coordinates matching eDNA samples were compared (Table S7).This resulted in 41 sampling locations, each with a LargeVF, SmallVF, and CPR sample (n=123).SIMPER analysis was used to identify prominent Animalia species contributing most to pairwise dissimilarity between the survey methods in each Southern Ocean zone An accompanying R markdown file including code for all analyses and figure generation, with accompanying raw data and sequencing files, is available from the Australian Antarctic Division's Data Centre.The DOI will be made available upon acceptance of the manuscript.Results of Primer v7 outputs are provided in Tables S8-S13.

eDNA sequencing statistics & depth
The Leray-COI metabarcoding assay generated a total of 2.43 million sequencing reads post-quality, denoising and chimera filtering.These were assigned to 1273 ZOTUs, of which 808 were assigned to Animalia taxa.Using the LCA algorithm, 619 of the 808 Animalia ZOTUs were taxonomically resolved to species level, representing 98.16% of the total Animalia reads.Two ZOTUs could not be resolved and were excluded from analysis, both were hits to 'zooplankton environmental sample' on GenBank.While a reasonably large number of ZOTUs were assigned to Animalia taxa, this represented just 46.86% of the overall reads (1.14 million sequencing reads).

Comparing eDNA sampling strategies
Overall sequencing read numbers were similar between SmallVF (1.27 million reads) and LargeVF samples (1.16 million reads).However, over the whole survey just 11.37% of reads within the SmallVF samples were assigned to Animalia taxa, compared to 85.66% of reads within the LargeVF samples (Figures S2 and  S3).Off-target reads of SmallVF samples were largely dominated by subkingdoms Chlorophyta (47.49%),Harosa (19.89%), and Hacrobia (15.09%).A Wilcoxon signed rank test indicated median LargeVF Animalia read numbers (11,376 reads) were significantly higher than in paired SmallVF samples (696 reads; V=2077, p<0.001).These differences are reflected spatially within the data, with the proportion of Animalia reads within SmallVF samples consistently lower than LargeVF samples across all Southern Ocean zones (Figure 2a).For LargeVF samples, Animalia read proportion was relatively high across all zones (>75%).
A total of 68 marine Animalia species were detected using both eDNA sampling methods, representing 54 genera, 43 families and 24 orders.Predominant families were Calanidae (7 species), Euphausiidae (5 species), Clausocalanidae (5 species), and Myctophidae (5 species).Taxonomic resolution was comparable between the LargeVF and SmallVF samples, with over 76% of ZOTUs assigned to species level for both volumes (78.07%and 76.95% respectively).The overlap of detected species was large, with 51 species detected with both sampling methods.However, LargeVF detected 14 additional species (Figure 2b), primarily arthropod (8 species) and chordate species (4 species; Figure 2c).Accordingly, mean species richness (q=0) was significantly higher in the LargeVF samples (10.03 ± 0.44) compared to SmallVF samples (8.4 ± 0.44; t=4.825, df=64, p<0.001) across the entire transect.Although following similar trends across Southern Ocean zones (Figure S1), mean species richness (q=0) was significantly higher in LargeVF samples across all zones except the SACCZ (Table S14).Diversity (Hill numbers) became comparable between eDNA sampling methods when read abundance was taken into account, with no significant differences at q=1 or q=2 (Table S14).This trend suggests that LargeVF samples may be detecting more rare or low-abundant species, but a similar number of abundant species in comparison to SmallVF samples.A PCO of the eDNA data (Jaccard resemblance of the presence-absence data) revealed complete overlap of the two methods (Figures S4 and  S5).

Diversity comparison
Diversity was compared between paired CPR and merged eDNA samples, i.e., eDNA water samples and CPR silks taken at the same coordinates.For this subset of samples, taxonomic resolution was greater using eDNA methods with almost 76.6% of ZOTUs assigned to species level, compared to the 58.84% of specimens detected across CPR silks identified to species level.At the phylum level, eDNA detected 7 phyla, of which 4 were exclusive to the survey method (Bryozoa, Cnidaria, Echinodermata, and Mollusca), while CPR detected 4 phyla of which Annelida was exclusive to the survey method (Figure 3a).Merged eDNA samples detected more species (37) than CPR (18), with only 12 species shared between the methodologies (Figure 3b).

eDNA
Animalia diversity was compared across the Southern Ocean Zones for all merged eDNA samples (i.e., LargeVF merged with SmallVF.Species richness varied across the zones, but was highest on average in the AAZ (10.46 ± 0.36) and the PFZ (9.67 ± 0.43) and lowest in the SACCZ (5.33 ± 0.84).PERMANOVA analysis of the presence-absence matrix confirmed community composition significantly differed between Southern Ocean zones (F=13.771,df=4, p<0.001), with pairwise testing indicating all zones were distinct in composition (Table S11).Taxonomic composition was seen to transition across the transect, with samples clustering into their respective Southern Ocean zones in the nMDS with slight overlapping (Figure 4a).Indicator species analyses identified common copepod taxa Calanus propinquus/similimus ,Calanoides acutus , and Neocalanus tonsus as contributing most to similarities between the zones.These taxa were among the most abundant in terms of reads and number of occurrences (Table S15).

eDNA & CPR
When comparing CPR data to merged eDNA data, significant interactions were observed between Southern Ocean zones and sampling method in the presence-absence matrix (Pseudo-F2=6.77,df=2, p<0.001), with pairwise testing indicating the methods significantly differed within each Southern Ocean zone.These interactions are illustrated in the nMDS (Figure 4b), with both sampling method (nMDS1) and Southern Ocean zone (nMDS2) having strong effects on community composition.Despite the effects of sampling method, patterns in composition were similar between methods, with a transition from SAZ to AAZ across both sampling methods (Figure 4b).Overlapping between the PFZ and AAZ was seen across all methods.
Indicator analyses revealed similarities were driven byLimacina retroversa (Mollusca), and copepod taxa C. propinquus/similimus , Metridia lucens, and Rhinacalanus gigas in the eDNA dataset, and copepod taxa C. similimus , R. gigasand Thysanoessa macrura in the CPR dataset.These taxa are common cosmopolitan species abundant in the Southern Ocean (Johnston et al. 2022) Diversity and composition differed between the zones depending on the chosen survey method, but was highest in the PFZ and AAZ for both methods (Figure S6).While all zones were dominated by arthropod taxa across methods, each method comprised a unique assemblage (Figure S6).SIMPER analysis was used to identify prominent Animalia species contributing most to pairwise dissimilarity between the survey methods in each Southern Ocean zone (Table S16-17).Differences were largely driven by occurrences of copepod taxa C. citer , Coryne eximia and C. propinquus/similimus in eDNA samples, andClausocalanus brevipes in CPR samples.

eDNA profiles post sea-ice contact: biosecurity concerns
Across the eDNA samples, we detected the presence of 16 species of biosecurity concern in eDNA samples using the criteria specified in 2.3.2 (Table S3; Figure 5).These species are either classified as invasive to Tasmania, non-native to Tasmania, native to Tasmania but detected close to the continent of Antarctica (south of 60°S), or cryptogenic (i.e., of uncertain origin; Table S18).The species comprised five phyla: Cnidaria (6, class Hydrozoa), Bryozoa (4, class Gymnolaemata), Arthropoda (2, classes Malacostraca and Thecostraca), Echinodermata (2, class Asteroidea), and Mollusca (2,class Bivalvia).Majority of species (87.5%) were detected at the beginning of the transect in the Derwent Estuary, Hobart (Tasmania), where they are known to occur.However, a large number of these species were detected much further south (13 of 16; Figure 5).Detections along the transect were sparse in the open ocean region, except for consistent detections of bryozoan species Bougainvillia muscus , Bugulina flabellata , and hydrozoan species Coryne eximia .At ˜59°S detections increased for multiple species, coinciding with the onset of sea-ice at this part of the transect.Total read numbers were relatively low (<100) for most of these detections (Table S18), reflecting that these species comprise a small component of the wider communities detected using a broad-spectrum Animalia assay.

Discussion
This study assesses the potential of integrating eDNA analysis alongside conventional CPR surveys to enhance biodiversity monitoring in the Southern Ocean, a region of exceptional ecological importance facing mounting anthropogenic pressures, climate change impacts, and threats of non-native species invasions.We assessed two eDNA sampling methods and directly compared these to conventional CPR to evaluate the scope of biodiversity detections and the spatial patterns inferred by each approach.Our findings illustrate the feasibility of eDNA in expanding existing CPR biodiversity assessments to encompass sea-ice regions, and highlight its potential as a valuable tool for the early detection of potential invasive species in Antarctica.This combined morphological and molecular approach, although specific to the studied open ocean environment, offers valuable insights for future monitoring endeavours in similarly vulnerable ecosystems.

Optimising open ocean eDNA sampling
Applications of eDNA biomonitoring in open ocean environments are expanding (Canals et al., 2021;Shelton et al., 2022;Suter et al., 2021Suter et al., , 2023;;Takeuchi et al., 2019), and refinement of these methodologies is required to determine their viability for long-term monitoring.Contributing to this, we evaluated diversity and richness patterns between two combinations of water volume and filter pore sizes, 12 L with 20 μm (LargeVF), and 2 L with 0.45 μm (SmallVF).Our findings revealed lower assigned Animalia reads in SmallVF samples, likely due to smaller filter pore sizes retaining non-Animalia micro-organisms that can co-amplify with the COI assay (largely micro-algae), and the smaller water volume containing less Animalia template.
Interestingly, recent research employing these combinations found that larger volumes and pore sizes detected more fragmented eDNA (Suter et al., 2023), indicating that coarse filters may retain short eDNA fragments, potentially bound to organic matter (Nagler et al., 2022), while capturing non-Animalia microorganisms less frequently than smaller pore sizes.Given the COI assay targets Animalia taxa, our results suggest that pore size and water volume can considerably impact assay performance and efficacy.As our study was conducted in the open ocean with relatively low numbers of Animalia taxa (Suter et al., 2021), these findings may not be applicable to coastal environments with high numbers of diverse Animalia species.Additionally, these non-target assignments are usually excluded from analyses or not reported in studies, particularly when established assays are used.However, reporting the assignment and proportion of non-target reads can assist in refining eDNA metabarcoding as a monitoring tool that is effective across a wide variety of environments.We therefore encourage studies to report the assignment and proportion of non-target reads, particularly when established 'universal' assays are used (Collins et al., 2019).
Animalia taxonomic resolution within eDNA methodologies was comparable, with high proportions of assignments to species level for both methods.However, we found LargeVF samples yielded more taxonomic assignments, and detected more unique species along the transect than SmallVF samples.These unique species comprised a small portion of the total LargeVF reads, indicating that more rare and low abundant species were detected, a trend that was reflected in Hill number analyses.This is again most likely attributed to the large pore size increasing the proportion of Animalia template in LargeVF samples, and therefore increasing the likelihood of detecting rare or low abundant taxa.The recommendation of pore sizes varies in the literature (Takahashi et al., 2023), with many studies suggesting the choice relies heavily on the characteristics of the sampling environment (Barnes et al., 2021;Budd et al., 2021;Cooper et al., 2022;Kumar et al., 2021).We recognise the difference between our pore sizes is large, and the use of different volumes introduces bias.Nevertheless, for open ocean environments, we suggest pore sizes larger than 0.45μm will maximise detected Animalia diversity and richness, and recommend further testing of large volumes (12 L or greater) across commonly used larger pore sizes (1.2μm, 5μm, 10μm, and 20μm) would allow further optimization of eDNA sampling for these environments.

eDNA derived biodiversity patterns of the Southern Ocean
Using eDNA metabarcoding, we successfully identified 68 unique Animalia species from 10 different phyla across the Southern Ocean.Community composition between Southern Ocean zones displayed significant differences in the species detected, reflecting a distinct latitudinal trend in species assemblages (see Figure 4).The spatial patterns detected were largely congruent with previously established Southern Ocean diversity profiles (Hunt & Hosie, 2003;McLeod et al., 2010;Pinkerton et al., 2020;Takahashi & Hosie, 2021), and followed similar spatial patterns for both eDNA methodologies.Similarities between open ocean zones driven by dominant Southern Ocean copepod species C. citer , C. propinquus/similimus , and C. acutus . These findings underscore the effectiveness of eDNA surveys in complex and dynamic open ocean environments, capable of detecting changes in communities.The significant changes in community composition between Southern Ocean zones not only accentuates the diverse ecological niches within this vast expanse, but highlights the ability of eDNA to elucidate large-scale ecological patterns.While the potential influence of water movement on eDNA dispersion between samples was considered, growing evidence suggests that eDNA signals tend to degrade or dilute rapidly over short spatial scales (Alexander et al., 2020;Jeunen et al., 2019;Koziol et al., 2019;Nester et al., 2020;West et al., 2020), thus mitigating the likelihood of significant cross-contamination or signal crossover between sampling locations.

Morphological CPR comparison
We successfully identified 87 unique species from 10 different phyla across the Southern Ocean using a combination of eDNA metabarcoding (LargeVF and SmallVF) and conventional CPR transects.When comparing to paired CPR samples (silk segments) taken at the same coordinates, eDNA detected more species across a wider range of phyla than CPR samples.eDNA detected seven phyla of which four were unique, while CPR detected four phyla of which Annelida was unique.Nonetheless, each survey method detected numerous unique species, and the overlap of species detected between approaches was small.Unique CPR detections were primarily copepod species, which can be attributed to its targeted approach for surveying plankton taxa.However, widespread and abundant copepod taxaClausocalanus brevipes and Thysanoessa macrura had noticeably fewer detections in eDNA samples.In this study, we used a broad 'universal' COI assay with a large amplicon length and primer degeneracies to expand the diversity of taxa detected.In such assays, common reads with fewer mismatches will often dominate PCR amplification, and rare templates may be underrepresented, resulting in PCR biases.Additionally, the relatively long amplicon length of the COI assay may not be effective at detecting degraded DNA fragments (Zhang et al, 2018).Targeted groupspecific assays with relatively short amplicon lengths may provide a more contemporary insight into these assemblages through increased detection rates (Jo et al., 2017;Wu et al., 2019).However, this approach comes with an increase in both time and financial costs, and should be considered with the specific project aims in mind.
Southern Ocean CPR surveys are among the most temporally and spatially extensive biological datasets available, providing invaluable baseline data for plankton taxa in a rapidly changing environment.Our data highlights the familiar strengths and limitations of the approach.Its targeted design (e.g., small mouth opening) may inadvertently omit species detected with eDNA metabarcoding, or damage specimens beyond identification.Additionally, certain life stages and morphology types (e.g., gelatinous) present challenges for morphological identification with CPR, and may only be broadly identified (e.g., egg, larvae).In contrast, eDNA metabarcoding can identify taxa regardless of life stage or specimen quality, but relies on often incomplete reference databases for accurate species assignments.In addition, eDNA survey locations are more versatile than CPR surveys -eDNA surveys can be conducted into the sea-ice zone.In doing so, we not only expanded the diversity and richness detected, but were able to detect several non-native species of concern in the Antarctic region that got scraped off the hull in the sea-ice zone.Surveys can also be expanded to include the whole water column (Canals et al., 2021;Feng et al., 2022), extending the current footprint and providing a method to detect these species of concern if settlement occurs i.e., at depth.Ultimately, each approach has its benefits, and the strengths of a combined genetic and conventional monitoring approach are well documented (Deagle et al., 2018;Nester et al., 2022;Alexander et al., 2022;West et al., 2022).We advocate for the integration of an eDNA survey to be conducted in parallel with CPR surveys, the intensity of sampling with these methods will ultimately depend on the question being addressed.

eDNA profiles post sea-ice contact: biosecurity concerns
We detected the eDNA of 16 species of Antarctic biosecurity concern, with many detections in the southern part of the transect.This list included species classified as invasive or non-native to Tasmania, native to Tasmania but detected close to the continent of Antarctica (south of 60°S), or cryptogenic (i.e., of uncertain origin).Hull biofouling and ballast water constitute the two main anthropogenic pathways for marine species to Antarctica (Hughes & Ashton, 2017).Ballast water introductions into the region are considered a minimal threat as vessels generally take on ballast water in the region following unloading of cargo (Lee & Chown, 2009;Lewis et al., 2003), and discharge of ballast water is prevented by the Antarctic Treaty (ATCM (Antarctic Treaty Consultative Meeting), 2006).Comparatively, hull biofouling has been described as the greatest potential pathway for marine introductions (Lee & Chown, 2009).Six of the species recorded have previously been recorded on the hull of the RSV Aurora Australis (Lewis et al., 2003(Lewis et al., , 2004(Lewis et al., , 2005)), and the remaining are either known biofouling (GISD -iucngisd.org,GRIIS -griis.org)or possible biofouling organisms (hull fouling recorded at genus levels, (Lee & Chown, 2009;Lewis et al., 2004Lewis et al., , 2005)).We therefore believe the presence of these species as biofouling on the RSV Aurora Australis is the most likely transport explanation for their DNA detection.With exception of the most northerly collections near Tasmania, it is unlikely these species are established at the locations where their eDNA was recovered.
Detections of these taxa were relatively sparse in open ocean zones except for two bryozoan (B.muscus and B. flabellata ) and one hydrozoan species (C.eximia ) that were detected almost consistently along the transect.While it is plausible that the frequency of these detections could be the result of large populations on the hull of the ship, it is also possible that these species are present in the ships uncontaminated seawater line due to their small size and colonial growth forms (ALA -ala.org.au).This raises general issues for underway seawater sampling as sampling over the side of the vessel is largely impossible given the harsh conditions of the Southern Ocean and other methods (such as collection using Niskin bottles) are only feasible on voyages with dedicated marine science time.The potential growth of species within the uncontaminated seawater line must be assessed in future eDNA surveys using this water source.
An increase in detection frequency was observed at 59degS for the remaining taxa and coincided with the onset of sea-ice.The abrasion of sea-ice on vessel hulls is thought to act as a natural hull cleaner, removing fouling organisms with the sheer force associated with pushing through sea-ice (Hughes & Ashton, 2017;Lee & Chown, 2009;Lewis et al., 2004).The sudden reappearance and increase in detections indicate that these biofouling species may have been scraped off the hull into the water column, potentially breaking apart encrusting or colonial organisms and increasing the likelihood of eDNA detection.Although abrasion by seaice is considered a significant factor reducing the number of marine biofouling species from being introduced to the region (Hughes & Ashton, 2017), it cannot be relied upon to prevent introduction or settlement.Ships may not encounter sea-ice due to temporal variation in sea-ice extent and thickness (McCarthy et al., 2019), as well as an overall decline in response to climate change (Eayrs et al., 2021;Ludescher et al., 2019;Raphael & Handcock, 2022).Furthermore, certain areas of ships may not be subject to ice abrasion, highlighted by Lee and Chown (2007) who revealed the invasive musselMytilus galloprovincialis had survived ship transportation throughout the broader Antarctic region within two sea chests of theSA Agulhas .
Our study detected DNA assigned to the Northern Pacific sea star (Asterias amurensis ), a species previously identified as a high-risk potential invader to Antarctic ecosystems (Clarke et al., 2022;Holland et al., 2021) The Northern Pacific sea star is an aggressive global invader due its high fecundity (Aguera & Byrne, 2018;Ling et al., 2012) and broad thermal tolerance (Byrne et al., 2016), and the detrimental effects it has on native species and community composition are well documented (Byrne et al., 1997;Parry & Hirst, 2016).It was likely introduced through ballast water and is now abundant in several southern Australian ports, including Hobart (Tasmania; Byrne et al., 1997).The species displays extensive phenotypic plasticity, is able to alter its spawning times to align with local conditions, and has an introduced range in cold climate locations including Canada and Alaska (Buttermore et al., 1994;Byrne et al., 1997Byrne et al., , 2016;;Ling et al., 2012).Concerningly, recent environmental modelling indicated the Northern Pacific sea star could survive at all Australian Antarctic stations under current or future climate scenarios (Holland et al., 2021).
It is crucial to recognise that we cannot determine the source or developmental stage of these eDNA detections, and we therefore cannot rule out the possibility that these detections represent transport of non-viable DNA, larvae, or dead organisms.Additionally, while these species were not detected in any of our field or laboratory controls implemented, overall read numbers were relatively low, most likely due to the broadspectrum of Animalia species detected with the assay used.To increase the sensitivity of these detections and provide a comprehensive overview of their potential distribution, we recommend the implementation of a suite of species-specific assays targeting these species of concern.The reporting of invasive marine species (IMS) in Antarctic regions is infrequent (Cardenas et al., 2020;Hughes & Ashton, 2017;Tavares & De Melo, 2004;Thatje & Fuentes, 2003), and while no established IMS populations are currently reported, the rapid warming of the region may see an increase in the number and range of organisms capable of surviving (McCarthy et al., 2019).It is widely recognised that preventing IMS establishment is a more successful and cost-effective approach than eradication (Finnoff et al., 2007;Hanley & Roberts, 2019), a particularly difficult task in remote locations (Rout et al., 2011).Currently, there are no systematic IMS surveillance programs in the Antarctic region (Holland et al., 2021).The implementation of eDNA surveys presents a unique opportunity to monitor these otherwise hard to detect threats.

Long-term conservation impacts
The extreme nature of the Southern Ocean requires a range of techniques to monitor the effects of climate change and increasing anthropogenic pressures (Gutt et al., 2018;Howell et al., 2021).Multiple resupply voyages occurring over the Austral summer presents a compelling opportunity to develop spatial and temporal baseline data for a broad range of Southern Ocean taxa.The simplicity of the sampling protocol could facilitate the beginning of a 'community science' program, as suggested by Howell et al. (2021), with nonspecialists collecting samples annually on resupply voyages.While sampling quality and reproducibility will need to be carefully managed (Lukyanenko et al., 2016), the ability to generate large amounts of spatiotemporal data is invaluable, and the importance of developing taxonomically diverse baseline data cannot be understated.
To reduce the logistical challenges of eDNA sample collection and to increase the extent of potential survey areas, automated eDNA samplers have been designed and are being increasingly implemented in field surveys (Breier et al., 2020;Formel et al., 2021;Govindarajan et al., 2022;Hendricks et al., 2022;Herfort et al., 2016;Ribeiro et al., 2019).Automated eDNA sampling can be either active through the use of autonomous underwater vehicles (Truelove et al., 2022) or drones (Aucone et al., 2023), or passive by leaving filter membranes submerged for extended periods of time (Bessey et al., 2021).These automated approaches could allow samples to be taken at great depths and inaccessible areas, such as under ice shelfs, with the potential to dramatically increasing the survey area beyond any current biological survey method.Moreover, the automation of laboratory processing, bioinformatic pipelines and analyses is simultaneously addressing issues of reproducibility and standardisation and increasing the speed at which eDNA data can be generated (Mousavi-Derazmahalleh et al., 2021).

Conclusion
Monitoring the growing effects of anthropogenic pressures and climate change in the Southern Ocean requires a baseline understanding of the distribution and composition of it taxonomically diverse marine biota.Here, we have demonstrated the potential of eDNA metabarcoding as a valuable long-term biomonitoring tool for open ocean environments, expanding the scope of existing CPR surveys.Our evaluation of eDNA methodologies found larger volumes (12 L) with larger pore sizes (20 μm) were better suited to open ocean surveys as they detected greater proportions of Animalia taxa, and a higher number of rare or low abundant species.The detection of several non-native species using eDNA metabarcoding was a significant finding of the study, underscoring the potential risk of shipping as a vector for the introduction of non-native species and further confirming the risks of monitoring remote environments using a singular approach.
Given the vulnerability of Antarctic environments to the introduction of non-native species, future work should optimise a suite of targeted group or species-specific assays to enhance sensitivity and increase the likelihood of detecting these species in a timely manner, particularly in near-shore Antarctic environments.By combining eDNA detections with CPR and ongoing collection of biophysical data, the number of species detected increased more than two-fold.We also found that although CPR surveys inadvertently omit several phyla, both methods detected high numbers of unique taxa.We advocate for the incorporation of eDNA metabarcoding to support long-term CPR surveys of the Southern Ocean, noting that the ability to overlay more metabarcoding data (across the tree-of-life) and biobanking of samples (Jarman et al., 2018) has not been explored here.
As the allocation of monitoring resources and establishment of potential mitigation strategies heavily relies on ecological interpretations of long-term datasets, the timely implementation of these surveys cannot be understated.With multiple annual resupply voyages across the Southern Ocean occurring over the Austral Summer, a unique opportunity exists to monitor long-term spatial and temporal patterns in the region.We anticipate that combined morphological and molecular approaches will have profound implications for the conservation and ecosystem-based management of the Southern Ocean and the Antarctic region.

Figure 2 :
Figure 2: (a) Proportion of reads classified as 'Animalia', 'non-Animalia', or 'not assigned' across LargeVF (12 L and 20 μm filter pore size) and SmallVF (2 L and 0.45 μm filter pore size) samples within Southern Ocean zones.(b) Venn diagram of the number of unique species detected by LargeVF and SmallVF eDNA metabarcoding, and the number of shared species detections using these approaches.(c) The number of unique species detections exclusive to and shared between LargeVF and SmallVF eDNA metabarcoding by Phyla.

Figure 3 :
Figure 3: Comparison of Animalia species composition detected by eDNA metabarcoding using the Leray COI assay, and CPR surveys in the Southern Ocean.(a) Number of unique and shared species per Animalia phyla for each method.(b) Proportional Euler diagram of unique and shared species between methods.(b) Number of unique and shared species per Animalia phyla for each method.

Figure 4 :
Figure 4: (a) Nonmetric multidimensional scaling (nMDS) ordination plot of Animalia species composition across Southern Ocean zones using eDNA metabarcoding.(b) nMDS ordination plot of Animalia species composition using eDNA metabarcoding (triangle) and CPR (circle) across Southern Ocean Zones.Southern Ocean zones are distinguished by colour, with latitude lines overlayed.nMDS ordination plots were conducted on a Jaccard resemblance matrix of presence-absence data.

Figure 5 :
Figure 5: Read proportion of species of biosecurity concern detected using eDNA methodologies.Reads are plotted on the x axis by latitude in the southerly direction of the transect (Hobart, Tasmania to Davis Station, Antarctica.Proportions are calculated by total reads per individual species, and not by the overall total reads, so that detection trends of less abundant reads are not overlooked due to the presence of more dominant/abundant reads.The status of each species is indicated by the different colours (see legend) as well as the onset of sea-ice (blue box).

Table 1 :
Location of Southern Ocean zones in this study, and the average sea surface temperature (SST degC) and sea surface salinity (SSS).References included for the defining characteristics of each front.