The DNA extraction protocol followed Visser et al. (2021). Briefly, DNA was extracted from the filters using the DNeasy Blood and Tissue Kit (Qiagen) with modified volumes. The DNA extractions were processed in sterile conditions under a clean bench to minimize contamination. After cleaning the outside of every filter with bleach, 720 μl of ATL-buffer and 80 μl of Proteinase K were added directly into the filter and incubated at 56°C for 2 h with agitation. After incubation, 600 μl of the buffer mix from the filter was transferred to a 2.0 ml Eppendorf tube and mixed with 600 μl AL-buffer and 600 μl 99% high grade ethanol. After this step, the normal DNeasy Blood and Tissue protocol was followed. DNA was eluted in 2 × 30 μl AE-buffer from the kit.

Rigorous laboratory measures were followed to reduce contamination. Only single-use consumables and single-capped PCR-tubes were used. DNA extractions, pre- and post-PCR were physically separated and sampling devices as well as laboratory equipment and surfaces cleaned with 50% bleach or RNAse AWAY (Carl Roth) containing bleach. To control for possible contaminations, negative controls were included at the filtration (MilliQ instead of seawater), DNA extraction (a blank filter was extracted and treated the same way as eDNA filter samples) and PCR (PCR-grade water instead of DNA sample) stages. As indicated below, positive controls were also included at the PCR stage.

A universal cephalopod primer targeting the nuclear 18S rRNA gene (Ceph18S_forward: CGCGGCGCTACATATTAGAC, Ceph18S_reverse: GCACTTAACCGACCGTCGAC, amplicon length = 140 – 190 bp, de Jonge et al., 2021) was used. The mitochondrial 16S rRNA locus (Jarman et al., 2006) was amplified as well following Visser et al. (2021). However, the positive controls on this sequencing run failed, so this data was excluded from further analysis.

For PCR amplification of the 18S rRNA gene, a 1-Step PCR protocol was used with the Illumina linker, Illumina index, Illumina adapter and a spacer for greater variability directly added to the primer sequence resulting in primer lengths of between 86 and 97 bp. All samples were amplified in duplicate resulting in six PCR products per sampling depth and site (three biological replicates × two PCR replicates) with individual Illumina tags for every replicate. Three positive controls (tissue derived DNA from Filippovia knipovitchi and Rossia palpebrosa, two cephalopod species that do not occur in the eastern Atlantic and one mock community with ∼10 ng/μl of each positive control species) and one negative control (PCR-grade water instead of DNA extract in the PCR reaction) were run in duplicate with individual taqs, and included on every PCR run.

The PCR reaction had a total volume of 25 μl and included 10 μl TaqMan Environmental Master Mix 2.0 (Applied Biosystems), 9 μl PCR-grade H₂O, 0.5 μl forward primer (10 μM), 0.5 μl reverse primer (10 μM) and 5 μl DNA extract. A touchdown PCR program was applied starting with an initial denaturation step at 95°C for 5 min, 8 cycles of 94°C for 30 s, 70°C for 30 s (decreasing this temperature by 1°C after every cycle) and 72°C for 1 min. Subsequently, 32 additional cycles were run with 94°C for 30 s, 62°C for 30 s, and 72°C for 1 min. The program was terminated by a final extension step of 72°C for 5 min.

DNA fragment sizes were verified on a 1.5% agarose gel stained with GelRed (Biotum). The PCR products were measured using a Qubit fluorometer and then pooled in equimolar concentrations, resulting in one library. The fragment size of the library was validated on a 2% agarose gel stained with GelRed (Biotum). Subsequently, the correct band was cut out (band size between 300 and 380 bp) and purified using the Zymoclean Gel DNA Recovery Kit (Zymo Research) following the manufacturer’s protocol. The library pool was quantified with the Qubit dsDNA HS Assay Kit (Molecular Probes Life Technologies). The insert size distribution was determined with the Tapestation4200: D5000 ScreenTape (Agilent). The working solution was diluted to 2 nM and the loading solution was prepared according to the Agilent protocol. The library pools were loaded with 8 pM. To increase diversity, 20% PhiX was spiked in the library. Sequencing was done on an Illumina MiSeq with the MiSeq Reagent Kit v3, 600 cycles (PE), 2x 300 bp (Illumina) at the Institute of Clinical Molecular Biology (IKMB), Kiel, Germany.

As a reference for the obtained eDNA data, the 18S database described in Visser et al. (2021) was used. Briefly, cephalopod sequences from the SILVA 18S database were used to recursively search the NCBI Genbank database (accessed in June 2020) using a BLAST-based script from the EukRef project¹ until no further cephalopod sequences were found. This resulted in 169 sequences from 119 species. These cephalopod sequences were combined with all other eukaryotic 18S sequences from the SILVA database to prevent spurious assignments of non-cephalopod amplicons. Additionally, 18S rRNA sequences from 16 tissue extracts of voucher specimens collected during two research cruises (MSM49, WH383) in the eastern tropical Atlantic and North Atlantic were sequenced and added to this database (GenBank accession numbers OK663489: OK663503). DNA was extracted from tissue samples using the DNeasy Blood and Tissue kit (Qiagen) following the manufacturers protocol. DNA concentrations were measured with the NanoDrop Spectrophotometer (Thermo Fisher) and amplified with the same primer used for eDNA metabarcoding, Ceph18S. For PCR amplification, we used the same parameters as for the touchdown PCR program outlined above. All PCR products were Sanger sequenced on a 3730xl DNA analyzer (Applied Biosystems) at the Institute of Clinical Molecular Biology (IKMB) in Kiel, Germany. Forward and reverse strand sequencing of the PCR products was performed using the Sanger Sequencing Kit (Applied Biosystems). Primers and low-quality ends were trimmed from the sequences, checked manually, edited, and assembled using CodonCode Aligner (version 3.7.1). Information on the construction of the reference database can be found in Visser et al. (2021) and resulted in a total of 34,811 sequences of which 169 originate from cephalopods representing 107 species for the 18S rRNA gene.

The raw sequences from the eDNA metabarcoding were demultiplexed by the sequencing center (IKMB) and forward and reverse PCR primer sequences were removed with cutadapt version 1.18 (Martin, 2011). The subsequent bioinformatic analysis was conducted with the R package DADA2 version 1.15.0 (Callahan et al., 2016, p. 2) in RStudio version 1.1.463 (R Core Team, 2018; RStudio Team., 2020). Paired reads were merged and trimmed at a quality score ≤ 2. The maximum expected error rate that was accepted was set to three for forward reads and five for reverse reads. Expected errors (EE) are calculated from EE = sum(10^(-Q/10)) with Q being the quality score. Potential chimeras were removed and an amplicon sequence variant table created. Taxonomic assignment of unique reads was performed using IDTAXA with the R package DECIPHER version 2.6.0 (Murali et al., 2018) using the databases described above as training set. Taxonomic assignments occurring in the negative controls and blanks were removed from the corresponding plates or batch of samples, respectively. After taxonomic assignments by IDTAXA, sequences that were assigned to only genus or family level were rechecked with applying the naïve Bayesian classifier method used by BLASTn in GenBank (Altschul et al., 1990). A hit by BLASTn was accepted to species level with a percent identity of at least 99% and no closer hit than 90% to another species.

The universal Ceph18S primer targeting the 18S rRNA gene used in this study could not discriminate reliably to species level within the family Octopoteuthidae (de Jonge et al., 2021). To obtain this information, a novel species-specific primer targeting the cytochrome oxidase I (COI) region of the mitochondrial gene of Taningia danae and amplifying a 250 bp segment was designed using Primer3 (Untergasser et al., 2012) (Taningia_COI_forward: TCATGCAGGTCCCTCTGTTG, Taningia_COI_reverse: AGGTGTTGGTATAAGATGGGGT, amplicon length = ∼250 bp). Sequences for designing the primer were obtained from the GenBank database (NCBI, Accession: AY393902.1, Accession: MG591434.1, Accession: EU735402.1). The specificity of the primer was checked by carrying out an in-silico PCR with ecoPCR by OBITools (Ficetola et al., 2010). The in silico results showed amplification of T. danae, but also 15 other species, mostly belonging to Gonatidae, that do not occur in Cabo Verdean waters. Subsequently, the specificity of the primer was tested on DNA samples derived from tissue from seven cephalopod species (Bathypolypus sp., Megaleledone sp., Gonatus sp., Rossia sp., Brachioteuthis sp., Pareledone felix) and T. danae and sequenced using the Sanger Sequencing Kit (Applied Biosystems). The DNA was extracted with the DNeasy Blood and Tissue Kit (Qiagen) according to manufacturer’s protocol and diluted in 100 μl TE-Buffer.

A subset of eDNA seawater samples from four stations (Santo Antão, Fogo, CVOO, Cyclone) were re-analyzed with the species-specific T. danae primer developed for this study, along with the above-mentioned universal cephalopod primer. All biological replicates were pooled and analyzed together with a positive (DNA extract of T. danae as a template) and a negative (PCR-grade water instead of DNA template) control, amplified via PCR in triplicate to increase the detection probability, which resulted in a total of 16 samples. Each PCR mixture contained 10 μl TaqMan Environmental Master Mix 2.0 (Applied Biosystems), 7 μl PCR-grade water, 0.5 μl dsDNase (ArcticZymes Technologies), 0.5 μl DTT Inactivation Buffer (ArcticZymes Technologies), 1 μl forward primer (10 μM), and 1 μl reverse primer (10 μM). The PCR-Mix was incubated for 20 min at 37°C to reduce contaminating DNA, followed by 20 min at 60°C for DNAse deactivation. After incubation, 5 μl DNA template was added to the PCR-mix and run with the following program: Initial denaturation step for 10 min at 95°C, 30 cycles with 94°C for 30 s, an annealing temperature of 60°C for 30 s and 72°C for 1 min, followed by a final elongation step of 72°C for 5 min.

All PCR products were examined on a 1.5% Agarose gel stained with GelRed (Biotum). Only samples showing a band on the gel were sequenced using the Sanger Sequencing Kit (Applied Biosystems) at IKMT. Sequences were processed with CodonCode Aligner v. 3.7.1.2. The forward and reverse primer sequences were cut off and forward and reverse reads assembled. All sequences were run against the NCBI database (22/01/2021) using BLASTn (Altschul et al., 1990). Only hits with a percent identity > 96% and e-value below 1e-80 were accepted as positive detections for Taningia danae (Supplementary Table 1).

A species accumulation curve (SAC) for the number of sampled sites was calculated only for eDNA using specaccum from the R package vegan (Oksanen et al., 2019). The method random was used calculating the mean SAC and its standard deviation from 100 random permutations of the data (Gotelli and Colwell, 2001). The Jaccard index, a measure of community dissimilarity, was calculated with the function vegdist (Oksanen et al., 2019) to measure similarity between the Canary Islands, Cabo Verde and Azores for all cephalopod taxa and for taxa belonging to the order Octopoda based on presence/absence data.

To compare cephalopod taxa occurring in the waters around Cabo Verde, the Canary Islands and the Azores, an extensive literature research was conducted. For the Cabo Verde islands, the species list generated in this study incorporated observations from three studies describing cephalopods in Cabo Verdean waters (Voss et al., 1998; Collins et al., 2001; Clarke, 2006) and our data. For the Azores, the species list from Visser et al. (2021) was used as it included all currently available data from both the literature and eDNA. For the Canary Islands, a recent review on cephalopod diversity off the Canary Islands was used, which included species from 48 published documents (Escánez et al., 2020) following the authors’ exclusion of species that were only hypothesized to occur off the Canary Islands due to their wide geographical distribution, but have not as yet been positively detected there.

Combining the data from this study with records from the literature indicates that Cabo Verde is home to at least 102 cephalopod taxa of which 64 were identified to species level from 30 families. Nine of these taxa are rare and were detected in this region for the first time by this study. This includes three genera and six species detected with eDNA are known from the Atlantic Ocean, but had not been recorded around Cabo Verde before (Jereb and Roper, 2010). With 64 confirmed species, Cabo Verde can be considered a biodiversity hotspot for cephalopods comparable to the Canary Islands (Escánez et al., 2020) and Azores (see literature in Visser et al., 2021). This species richness is also comparable to those reported from other northern hemisphere open ocean systems identified as areas with high cephalopod species diversities such as the western North Atlantic (74 species), the North Sargasso Sea (68 species), the North African Subtropical Sea (72 species), and the Mediterranean outflow (70 species) (Rosa et al., 2008). The high species richness observed off Cabo Verde may be explained by the sampling sites being in vicinity of several seamounts (e.g., Senghor Seamount) and islands (Cabo Verde), topographical features that are known to enhance species diversity, creating various ecological niches for different species to thrive and evolve (Worm et al., 2003). Additionally, Cabo Verde is located at intermediate latitudes where temperate, subtropical and tropical species ranges overlap, leading to high diversities across trophic levels for predator and prey species (Worm et al., 2003). Many of the here detected cephalopod species belong to deep-sea families (Hoving et al., 2014) such as Cranchiidae, Histioteuthidae, Octopoteuthidae, Mastigoteuthida, and Chiroteuthida. This indicates that the identification of Cabo Verde as being a cephalopod diversity hotspot may be driven particularly by diversity in the deep sea. However, cephalopod diversity in shallow and coastal waters needs to be explored in more detail for a complete understanding of what is driving overall regional cephalopod community composition.

Two unexpected genera were detected with eDNA: Loligo and Alloteuthis. These genera are coastal and not known from Cabo Verde. They have not been detected by the demersal fisheries active around the islands prior to this study. The detections are unlikely to be false positives since they were absent from the corresponding negative controls. Another possibility may be the transport of eDNA by predators. Apex predators such as sperm whales, blue sharks and swordfish are known to feed on Loligo off the Azores and migrate to Cabo Verde (Clarke, 1956, Clarke et al., 1995, 1996), potentially releasing feces, and therefore eDNA, in Cabo Verdean waters. However, it can also not be ruled out that Loligo and Alloteuthis are undetected species in Cabo Verdean waters. Loligo and Alloteuthis are both distributed from tropical to temperate and subpolar coastal waters and occur along the northern and central west coast of Africa (Jereb and Roper, 2010). This indicates that species’ distributions may be broader than anticipated and eDNA detections of unknown and unexpected species must be considered carefully. In this context it is worth noting that in the South Atlantic Ocean the bathypelagic and benthopelagic squid Gonatopsis octopedatus has been dispersed more than 15,000 km from its region of original species distribution in the North Pacific, potentially by deep-sea currents (Arkhipkin et al., 2010).

This study was able to detect rare and elusive species with eDNA. One example is the bigfin squid Magnapinna sp. This squid is a bathypelagic species that is primarily known from underwater observations > 1800 m in the Pacific (Osterhage et al., 2020), Indian and Atlantic Ocean (Vecchione et al., 2001). Only a handful of damaged or juvenile specimens have been captured in the Pacific Ocean (Vecchione and Young, 1998). No observations or records have been made to date from the Cabo Verde region, although individuals have been captured 500 miles south of Cabo Verde from depths of between 998 and 1962 m in 2007 (Piatkowski, personal communication). This squid species has distinctive and relatively long, slim arms reaching lengths of over 1.5 m, compared to a mantle length of ∼15 cm (Osterhage et al., 2020). Magnapinna sp. is suggested to be globally distributed in bathyal and abyssal depths (Vecchione et al., 2001). This genus was solely detected with eDNA in this study and repeatedly in deep waters (1900 – 2500 m) off Cabo Verde, which is consistent with its known depth distribution in the Atlantic (Vecchione et al., 2001).

The depth ranges stated in this study are a combination of general presence/absence of taxa. Since the cephalopod data presented here from the net catches is imbalanced with regard to day/night sampling, we do not differentiate between day and night distributions. In the Atlantic open ocean, cephalopod species diversity is known to be highest in the upper 200–1000 m and decreases with depth (Rosa et al., 2008) resulting in a bell-shaped diversity pattern with respect to depth. The same trend has been observed in pelagic fishes and invertebrates (Haedrich et al., 1980; Moranta et al., 1998; Smith and Brown, 2002; D’Onghia et al., 2004; Cartes et al., 2011; Papiol et al., 2012; Gaertner et al., 2013; Farré et al., 2016). High mesopelagic diversities coincide with enhanced primary productivity and optimal oxygen concentrations in overlying water layers (Levin et al., 2001) resulting in sinking of organic matter and high food supply at intermediate depths. Additionally, in this depth zone, predation pressure from top predators is intermediate and balanced by high or moderate production by low trophic levels providing prey for cephalopods and therefore facilitating their occurrence and coexistence (Gage and Tyler, 1991; Roxburgh et al., 2004). Cephalopod taxa community composition around the Cabo Verde islands of Santo Antão and Fogo, the two stations for which eDNA sampling was balanced with respect to depth, followed the same above-mentioned bell-shaped diversity pattern, with the highest number of species detected in the upper 1000 m, peaking at 600 m. This peak in cephalopod community composition correlates with species and biomass aggregations of myctophid fishes at 400–700 m in the central equatorial Atlantic (Kinzer and Schulz, 1985), which are, together with other fish species, primary prey of several cephalopods (Hoving et al., 2014; Merten et al., 2017; Villanueva et al., 2017). Overall cephalopod community composition decreased below 1000 m in comparison to the upper 1000 m, however, two peaks in cephalopod community composition were present at 1600 and 2200 m. The distribution of cephalopod species occurring in bathyal and abyssal depths is less well known than the distribution of epi- and mesopelagic species, partially due to difficulties in sampling the deep sea. Therefore, the species community composition is likely underestimated in these depths. The water layer immediately above the seafloor, the benthic boundary layer (BBL), is expected to support increased biomasses and species diversity (Brandt et al., 2007), which may explain the here observed peak in cephalopod species community composition at 2200 m depths. A previous study on eDNA of cephalopods in the hunting zones of Cuvier’s beaked whales (Ziphius cavirostris) showed that some predators specifically hunt in the BBL (Visser et al., 2021). Additionally, the high abundances of e.g., fish and crustaceans reported in the BBL (Dauvin and Vallet, 2006) suggests that BBL assemblages are of great importance for deep-sea food webs and nutrient cycling.

Due to the repeated observations of Taningia danae during video transects, but their absence in nets, we were particularly interested in the vertical distribution of this large squid, and hence developed and applied a species-specific primer on the eDNA samples. This species is recognized as one of the largest mesopelagic squid in tropical and subtropical oceans worldwide (Clarke, 1966) growing up to 150 kg and 1.7 m mantle length. Knowledge of this species is based on small juvenile individuals caught with mid-water nets (Roper and Vecchione, 1993), records generated from sperm whale stomach content analysis (Clarke, 1967; Okutani and Satake, 1978), specimens from fishing nets (González et al., 2003; Quetglas et al., 2006) and in situ observations with midwater camera systems (Kubodera et al., 2007; Gomes-Pereira and Tojeira, 2014; Hoving et al., 2019a). The presence of hundreds of beaks of T. danae in sperm whale stomachs suggests that this species is a very abundant deep-sea squid (Clarke, 1996b). However, adults are rarely caught with nets, potentially due to their occurrence at considerable depths and their strong swimming capabilities. The relatively frequent detections of T. danae during the video transects may be the result of T. danae being attracted to the light of the camera system (Hoving et al., 2019a) as seen in other studies applying lured camera systems in midwater (Kubodera et al., 2007; Gomes-Pereira and Tojeira, 2014). Juvenile and young T. danae undergo ontogenetic migration from the surface to 200 – 300 m water depth (Roper and Vecchione, 1993) and larger specimens perform short distance diel vertical migration from depths of 600 – 900 m during the day to shallower depths of 240 – 500 m during the night (Kubodera et al., 2007). Considering the hunting grounds of sperm whales and deep-sea sharks, adult T. danae were thought to occur below 1000 m (Clarke and Merrett, 1972). This deep occurrence is supported by the eDNA and video data of this study. The video surveys observed T. danae between 400 and 1400 m, whereas eDNA detected T. danae between 100 and 2500 m, with the majority of detections at 1600 m. The fact that T. danae was detected frequently in the eDNA samples highlights the potential importance and abundance of this elusive top-predator in deep-sea ecosystems, and shows the ability of eDNA to detect species that are rarely captured in nets.

The three archipelagos Cabo Verde, the Canary Islands and Azores investigated in this study share overall cephalopod community composition. However, the data indicates a difference in octopus community composition between Cabo Verde and the Azores as well as Canary Islands, suggesting a separation of Cabo Verde when focusing specifically on benthic octopuses. These findings are in line with studies proposing a biogeographical separation of Cabo Verde from Macaronesia based on the dispersal of benthic and coastal metazoan species (Freitas et al., 2019). Biogeographical separation and endemism are tightly linked to the larval dispersal ability of organisms as dispersal directly influences gene flow within and among populations. Populations may become either genetically more similar when subject to frequent mixing or they may become more isolated when mixing and dispersal is reduced or absent. There are strong inter-specific differences in cephalopod dispersal abilities. Cephalopod species that hatch planktonic paralarvae tend to be smaller and reach broader distributional ranges than species with larger, benthic hatchlings (Villanueva et al., 2016). Additionally, species with offshore larval distribution have been shown to have larger mean geographic ranges than species with inshore larval distribution (Macpherson and Raventos, 2006), a trend also observed in fish (Macpherson and Raventos, 2006), highlighting the importance of oceanographic currents as powerful dispersal mechanisms for pelagic species (Rowell et al., 1985; Saito and Kubodera, 1993; Bower et al., 1999; Roberts and van den Berg, 2005; Martins et al., 2014; Downey-Breedt et al., 2016).

Due to their muscular anatomy, strong swimming capabilities and highly mobile life style, some squid species are able to actively migrate hundreds of kilometers. For instance, the ommastrephid squid Sthenoteuthis pteropus spawns in the eastern equatorial Atlantic (Zuyev et al., 2002) and its paralarvae are quickly dispersed in the equatorial zone. Adult females are known to migrate 2500 km from Cabo Verde to Madeira and back (Zuev and Nikolsky, 1993; Zuyev et al., 2002). Other oceanic squids occurring off the Cabo Verde islands also occur off the Azores and Canary Islands, such as Taningia danae and Todarodes sagittatus. Due to their mobility and ability to cope with different oceanographic conditions, many squid species are able to sustain broad horizontal and vertical distribution patterns (Childress and Seibel, 1998; Zuyev et al., 2002; Rosa and Seibel, 2010; Hoving and Robison, 2012; Rodhouse et al., 2014; Doubleday et al., 2016). For species with limited swimming capacities, pelagic egg masses may facilitate dispersal (Young and Harman, 1985; Roberts et al., 2011).

The majority of octopus species have a short planktonic larval stage followed by a benthic adult lifestyle, and as such have a limited dispersal potential due to e.g., larger hatchling sizes, a characteristic linked with limited dispersal (Villanueva et al., 2016) and association with certain substrates. For example, the benthic octopus Pareledone turqueti is unable to disperse between sites that are separated by ocean depths > 1000 m, which presents a major physical barrier for the species (Allcock et al., 1997). This limited dispersal ability and therefore reduced gene flow in octopuses is also applicable to the common octopus Octopus vulgaris (Sampaio et al., 2018). The common octopus is the only benthic octopus occurring at all three archipelagos, the Canary Islands, Azores and Cabo Verde. However, O. vulgaris populations off Cabo Verde potentially show a developing endemism of this species (Sampaio et al., 2018). The other octopus species from the current study to occur across the three archipelagos were all pelagic octopuses that brood their eggs in midwater or have pelagic paralarvae (Japetella diaphana, Pteroctopus tetracirrhus, Vitreledonella richardi, and Tremoctopus violaceus). Therefore with increased dispersal abilities potentially comparable with those of oceanic squids. The oceanographic features of Cabo Verde form natural barriers to the north of the islands and toward the west coast of Africa, whereas the Canary Islands and Azores are connected by the Azores Current and the Madeira current, forming a seaway, with several shallow seamounts functioning as stepping stones for marine organisms (Freitas et al., 2019). As a result, benthic or benthopelagic organisms with reduced mobility, in contrast with the high mobility of squids, may become separated from species of the other archipelagos, leading to a separation of Cabo Verde from Macaronesia. This is indicated in octopus species composition in this study. To fully understand the biogeography of Macaronesia, not only benthic and coastal organisms should be taken into account, but also the pelagic, open ocean fauna should be considered.

An active pathway of vertical carbon distribution by cephalopods is achieved by vertical migration. Many cephalopods spend the paralarval and juvenile phase of their lifecycle in upper surface waters, benefiting from primary production (Hoving et al., 2014). As they grow and mature, they descend to deeper layers to reproduce, transporting carbon from shallow to deep waters via ontogenetic migration. The second active pathway of carbon distribution is diel vertical migration, a common trait of many deep-sea cephalopods (Jereb and Roper, 2010; Hoving et al., 2014; Judkins and Vecchione, 2020). These cephalopods migrate close to the surface at night for feeding and descend to deeper layers during the day, to potentially avoid visual predators or to maintain their energy reserves in the colder, deeper water (Sutton, 2013). Off Cabo Verde, 31% of the confirmed cephalopod taxa are known to undergo ontogenetic migration such as Sthenoteuthis pteropus, Taningia danae and most of the cranchiids (Jereb and Roper, 2010; Judkins and Vecchione, 2020). Thirty-five are also known to perform diel vertical migration (e.g., Sthenoteuthis pteropus, Taningia danae as well as many enoploteuthids) (Jereb and Roper, 2010).

Carbon can also be distributed passively when cephalopods die after reproduction, either sinking to the seafloor or floating to the surface, depending on the species (Fields, 1965; Stockton and DeLaca, 1982; Roper and Vecchione, 1996; Martin and Christiansen, 1997; Nesis et al., 1998; Boyle and Rodhouse, 2005; Hoving et al., 2017). To date, there are no records of cephalopod carcasses on the seafloor in the Cabo Verde region. However, stomach content analysis of deep-sea scavengers off the Cabo Verde islands have found ommastrephid squid in the stomach contents of the cusk eel (Neobythites sp.) and the velvet belly lanternshark (Etmopterus spinax) (Clarke and Merrett, 1972). Both scavengers are benthic and likely fed on carcasses rather than actively hunted epi- and mesopelagic squid.

Species in the Cabo Verde area which may form substantial foodfalls when they die (settling to the seafloor) are T. danae and S. pteropus. Taningia danae is suggested to be associated with bottom waters when spawning (Nesis, 1987), indicating that this species may be consumed by benthic organisms after spawning and death. Given the estimated high abundance and large size (∼1.70 m mantle length), biomass from this species may contribute significantly to the regional carbon cycle of the deep sea. The orangeback flying squid S. pteropus is another abundant and dominant squid species, in terms of numbers of individuals and biomass (instantaneous biomass: 4–6.5 million tons, annual total biomass production = 34–52 million tons) in the eastern tropical Atlantic (Zuyev et al., 2002; Jereb and Roper, 2010). Sthenoteuthis pteropus can reach maximum mantle lengths of 65 cm and has a broad distribution from Madeira (36°N) in the North to South Africa (32°S) in the South. This species’ distribution spans the whole of Macaronesia and individuals are known to migrate thousands of kilometers from Cabo Verde to Madeira and back. Spawning of S. pteropus is not linked to bottom waters, but occurs in the epipelagic zone. This species is not neutrally buoyant and individuals sink to the seafloor when they stop swimming. Tissues from this species have high protein concentrations (17%) (Zuyev et al., 2002), rendering them as a valuable food source. Sthenoteuthis pteropus undergoes ontogenetic and diel vertical migration down to 1200 m water depth (Moiseev, 1991), resulting in active transport of carbon from the surface to the mesopelagic ocean. Due to its high abundance, wide distribution and migration ability, good nutritional value, large size, spawning style and negative buoyancy, S. pteropus may transport carbon from epi- and mesopelagic zones to the deep-sea benthos.

The results from the three methods employed in this study each indicated differing cephalopod families to be regionally dominant. Mastigoteuthidae followed by Ommastrephidae and Loliginidae were the most commonly detected families with eDNA, whereas with nets, the most commonly detected families were Cranchiidae followed by Pyroteuthidae and Enoploteuthidae. This is in line with net data from the literature from the eastern North Atlantic, with Cranchiidae (Liocranchia reinhardtii) being the most commonly detected family in rectangular midwater trawls caught in 1968 and 1972, followed by Onychoteuthidae (Onychotetuhis banksii) (Clarke, 2006). However, Pyroteuthidae and Enoploteuthidae only represented 5 and 2%, respectively, of the samples collected by the trawls described in Clarke (2006). This highlights the potential collection bias toward certain species, as a function of the trawling gear characteristics being used, and therefore that differences in counts can occur within methods (i.e., between different trawls) as well as between methods (i.e., between eDNA and trawl catch analysis). For example, in Clarke (2006), three different types of nets were used with a net opening of between 0.5 m² (Multinet) and 480 m² (pelagic trawl) and mesh sizes of between 335 μm and 4 mm. In the current study, smaller net openings and mesh sizes, such as those of the IKMT and Multinet, mainly caught paralarvae and small individuals between 4 and 190 mm mantle length, whereas the pelagic trawl captured larger individuals of between 14 and 240 mm mantle length. However, even the pelagic trawl used here was not able to catch very large squids. Video surveys conducted by different gears also resulted in differing sampling bias. The manned submersible JAGO used in this study is loud under water, potentially scaring organisms away. PELAGIOS on the other hand is attached to a research vessel via cable, causing vibrations in the water column along the cable length, which could also influence observations. Both gears emit light that may attract some species and repel others. The majority of observed cephalopods that could be identified to at least family level was within video data collected with PELAGIOS, whereas within JAGO video data, Cranchiids, which are often less mobile, were more often observed. Within JAGO data cephalopods were often observed in mid escape, moving swiftly and therefore could not be identified to high taxonomic resolution.

The video observation results were in general accordance with the observations from the nets, both detecting the Cranchiidae as the most abundant family. The second most commonly detected family were the Mastigoteuthidae which aligned with the eDNA results. This discrepancy in assessment of the most frequently observed families reported by these three methods may partially be the result of differing characteristics of the sampling methods. Cranchiids are abundant and when small often less mobile, therefore they are more easily caught with nets than some other families. Ommastrephids on the other hand are highly mobile, muscular and often larger squids that may be able to easily escape approaching nets. As a defense strategy, ommastrephids and mastigoteuthids also release large amounts of ink when being threatened. This ink, which contains DNA, may result in a higher abundance of eDNA in their environment than that associated with some of the other families, e.g., cranchiids. Cranchiids have been observed to eject ink inside their mantle cavity for camouflage, therefore not releasing it externally (Hunt, 1996). This may reduce the amount of eDNA released by that cephalopod group. Information on eDNA shedding rates of cephalopods is needed, as the size of an individual does not necessarily reflect the amount of e.g., tissue being shed. Further, differing oceanographic conditions might influence the detectability of eDNA (Pinfield et al., 2019).

The three sampling approaches applied in this study complemented each other in genera and species detections. It seems on first examination of the data that net tows were the most effective sampling method for assessing cephalopod deep-sea diversity, since specimens for reference databases were also collected. However, when comparing the eDNA data with trawl data from all cruises individually, net tows were either more effective, comparable or less effective in finding taxa when compared to the eDNA approach. Between 9 and 20% of species were detected with both the eDNA and net approach, indicating a considerable contribution of 20 – 43% in species detections when combining eDNA with net trawls. However, direct and quantitative comparisons between and within the three sampling methods applied here was not possible and was not the priority of the current study, as the sampled sites, seasons, years, targeted depths and gear used within and between methods differed substantially, as well as the sample sizes. The volume of water that was sampled with CTD niskin bottles for the eDNA approach was several magnitudes less than that which was physically and optically sampled by nets and video surveys, respectively. As the rarefaction curve for eDNA did not reach the asymptote from the data collected, sampling more stations and depths would likely have resulted in additional detected taxa. Similarly, the addition of more targeted marker genes may have increased the number of species and taxa detected by the eDNA methodological approach. By increasing the sampling size for eDNA samples we can expect eDNA to become progressively more equivalent, or even superior, to the net trawl methodological approach in detecting taxa, as has been the case in other studies focusing on other organisms, such as fish (Thomsen et al., 2012; Boussarie et al., 2018; Govindarajan et al., 2021). However, reference specimens are required from a region to compare with obtained eDNA sequences.

This study highlights that the complementary use of different methods yields the most holistic view of deep-sea cephalopod communities. Nets remain indispensable since reference specimens are needed to detect, compare and validate sequences obtained with eDNA. However, both the eDNA approach and video surveys are non-invasive and appropriate for integration into the surveillance and management plans of marine ecosystems, but efforts should be increased to incorporate more primers into the analysis procedure and also the seafloor should be sampled for eDNA analysis.