Molecular Diet Analysis of Adélie Penguins (Pygoscelis adeliae) in the Ross Sea Using Fecal DNA

Simple Summary The diet of Adélie penguins, Pygoscelis adeliae, in the Ross Sea was studied applying quantitative polymerase chain reaction (qPCR) protocols to their feces. Two krill species (Euphausia superba and Euphausia crystallorophias) and notothenioid fish (mainly Pleuragramma antarctica and Pagothenia borchgrevinki) were among the most abundant components of the diet. The composition of the two krill species and notothenioid fish was found to be strongly related to the geographic characteristics of the Ross Sea. Abstract The diet of Adélie penguins, Pygoscelis adeliae, is a useful indicator in understanding the ecological conditions of their habitats. The diets of Adélie penguins were studied using metabarcoding and quantitative PCR (qPCR) analyses of fecal DNA from seven habitats along the Ross Sea region. Using metabarcoding analysis with dual universal primers (18Sv9 and miniFish), the overall diet composition and detailed information about piscine prey were clearly elucidated. It was found that two krill species (Euphausia superba and Euphausia crystallorophias) and notothenioid fish were the most abundant in the diets of Adélie penguins. Among the notothenioid prey, Pleuragramma antarctica (56.50%) and Pagothenia borchgrevinki (18.21%) were the two most abundant species. qPCR analysis showed a significant geographic difference in the composition of main prey. Penguins inhabiting outbound parts of the Ross Sea (Capes Adare (CA) and Duke of York Island (DY)) mainly preyed on E. superba, without any significant changes in prey composition. By contrast, those inhabiting the inbound parts of the Ross Sea (Edmonson Point (EP) and Inexpressible Island (II)) preyed on E. crystallorophias and notothenioid fish rather than E. superba. Compared with the outbound habitats, prey compositions for penguins inhabiting the inbound regions were significantly different year to year, which was presumably due to the food availability based on the annual environmental and meteorological conditions of the coastal region along with the inbound parts of the Ross Sea.


Introduction
The Adélie penguin, Pygoscelis adeliae, is known to be the most abundant and widely spread penguin species in Antarctica, distributed along the entire coast of the continent [1]. As an omnivorous predator, it is regarded as a vital biomarker for ecological and environmental factors [2]. Thus, understanding the penguin's feeding ecology can provide basic information to establish models and plans for sustainability of the species in each habitat. Previous studies of the diets of Adélie penguins were conducted either directly via visual  (Figure 1 and Table 1). The size of Adélie penguin colonies in the area where the samples were collected varied from 1890 pairs at Edmonson Point to 227,000 pairs at Cape Adare (Table 1) [17,18]. Sample collections were conducted during the guard stage [19] in 3 consecutive breeding seasons, from December 2017 to January 2020. However, scat samples were collected only at CH and II for all 3 years, whereas 2-or 1-year samples were collected at the rest of the sites (Table 1). Except for samples collected at 3 sites (CA, CH, and II) in January 2020, all other scat samples were collected in December. At each sampling site, 11 scat samples were collected, with several exceptions: 5 at II in 2017, 7 at CW in 2018, and 10 at CH in 2017 (Table 1). Only freshly excreted unfrozen scat was collected around the nests, minimizing the impact on the penguins' incubation process and colonies.

Fecal DNA Isolation
Immediately after collection, the scat samples were put into 6 volumes of lysis buffer (50 mM Tris-HCl pH 8.0, 500 mM NaCl, and 50 mM EDTA) and stored frozen at −20 °C until delivery to the laboratory for further analysis. Fecal DNA was extracted using the AccuPrep genomic DNA extraction kit (Bioneer, Daejeon, Republic of Korea) following the manufacturer's instructions. The frozen samples in lysis buffer were homogenized using a FastPrep-24™ Classic homogenizer (MP Biomedicals) after adding sodium dodecyl sulfate (2%). The quantity and integrity of extracted DNA were analyzed using a NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Extracted fecal DNA was aliquoted and stored at −70 °C until being used for library preparation and qPCR analysis.

Primer Design and PCR Amplification of Prey Items
Metabarcoding analyses were conducted with two universal primers to amplify trace amounts of diet from fecal DNA. A eukaryotic-specific universal primer set targeting the V9 region of 18S rRNA (18Sv9) was adopted for the overall prey composition, with an amplicon size of approximately 115-140 bp [20]. For the fish taxa, we initially tested the MiFish primer set, which has been widely used for environmental DNA (eDNA) metabarcoding analysis of fish taxa [21]. However, it showed a high degree of cross-reactivity to host DNA, presenting little prey DNA in the metabarcoding results (data not shown). Thus, another fish-specific primer set targeting the same region was designed by multiple

Fecal DNA Isolation
Immediately after collection, the scat samples were put into 6 volumes of lysis buffer (50 mM Tris-HCl pH 8.0, 500 mM NaCl, and 50 mM EDTA) and stored frozen at −20 • C until delivery to the laboratory for further analysis. Fecal DNA was extracted using the AccuPrep genomic DNA extraction kit (Bioneer, Daejeon, Republic of Korea) following the manufacturer's instructions. The frozen samples in lysis buffer were homogenized using a FastPrep-24™ Classic homogenizer (MP Biomedicals) after adding sodium dodecyl sulfate (2%). The quantity and integrity of extracted DNA were analyzed using a NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Extracted fecal DNA was aliquoted and stored at −70 • C until being used for library preparation and qPCR analysis.

Primer Design and PCR Amplification of Prey Items
Metabarcoding analyses were conducted with two universal primers to amplify trace amounts of diet from fecal DNA. A eukaryotic-specific universal primer set targeting the V9 region of 18S rRNA (18Sv9) was adopted for the overall prey composition, with an amplicon size of approximately 115-140 bp [20]. For the fish taxa, we initially tested the MiFish primer set, which has been widely used for environmental DNA (eDNA) metabarcoding analysis of fish taxa [21]. However, it showed a high degree of cross-reactivity to host DNA, presenting little prey DNA in the metabarcoding results (data not shown). Thus, another fish-specific primer set targeting the same region was designed by multiple alignment of the currently reported 12S region of 122 Antarctic fish and Adélie penguins using Geneious software [22]. The barcode size amplified by the primer set, named "miniFish", was approximately 91 bp (Table 2).  [20] NGS18Sv9-R GTCTCGTGGGCTCGGAGATGTGTATAAGA GACAGCCTTCYGCAGGTTCACCTAC

Library Construction and NGS
Two PCR steps were used to construct the NGS library; 18Sv9 and miniFish primers were used for eukaryotic and fish taxa, respectively ( Table 2). The first PCR for the MiSeq platform was conducted with primers overhanging the linker sequence for the Nextera XT Index Kit (Illumina, San Diego, CA, USA). The first PCR mixtures (20 µL) contained 1 µL of a template (5-20 ng/µL conc.), 1 µL F/R primer (10 pmol) each, 2 µL dNTPs (each 2.5 mM), 2 µL 10× Ex Taq Buffer, 0.2 µL ExTaq HS DNA polymerase (Takara, Shiga, Japan), and DNase/RNase free water. PCR amplification was performed under the following cycling conditions: initial denaturation at 94 • C for 3 min, followed by 20 cycles of 94 • C for 30 s, then 60 • C for 30 s for 18Sv9 primers or 55 • C for 30 s for miniFish, 72 • C for 30 s, with a final extension of 3 min at 72 • C. The amplified PCR products were stained with loading STAR (Dynebio, Republic of Korea) after 1% agarose gel electrophoresis. The amplicons of the expected size on the gel (approximately 230 bp for 18Sv9 and 200 bp for miniFish) were purified using an AccuPrep Gel Purification Kit (Bioneer, Daejeon, Republic of Korea) for the second PCR.
The second PCR amplification was conducted using a Nextera XT Index Kit (Illumina, San Diego, CA, USA) following the manufacturer's instructions. The constructed libraries were identified by 1.5% agarose gel electrophoresis (approximately 270-300 bp) and further purified using the AccuPrep Gel Purification Kit (Bioneer, Daejeon, Republic of Korea). The integrity and quantity of the constructed libraries were measured using a Quantus fluorometer (Promega, Madison, WI, USA) and a BioAnalyzer 2100 (Agilent Technologies, Palo Alto, CA, USA). The indexed libraries were sequenced on the Illumina MiSeq platform using a MiSeq Reagent v3 600-cycle kit (Illumina, San Diego, CA, USA).

Bioinformatics Analysis
After adapter/index sequences were trimmed from the raw MiSeq reads and reads with a QV below 20, fewer than 100 nucleotides were discarded using CLC Genomics Workbench v.8.0 (CLC Bio, Cambridge, MA, USA). The trimmed raw reads were further merged using mothur (v.1.43.0) software [23]. The merged reads were screened with criteria of 7 bp overlap, zero mismatches, trimmed primer sequences, and expected size (115-140 bp for eukaryotes and 91 bp for fish taxa). The merged reads were clustered into operational taxonomic units (OTUs) at 99.6% of sequence identity. OTU clustering and the removal of chimeric sequences were conducted using UCHIME [24]. The OTUs were annotated with BLASTN (v.2.7.1) against the NCBI NT sequence database. Species names were assigned for each OTU with a sequence identity higher than or equal to 99%. Those whose identity ranged from 90 to 99% in the 18Sv9 metabarcoding and from 95 to 99% in the miniFish metabarcoding received genus names. OTUs with lower identities were considered "unknown". Nonmetric multidimensional scaling (NMDS) analysis was performed based on the Bray-Curtis distance of phylum and genus proportions from the 18Sv9 and miniFish metabarcoding, respectively.

qPCR Analysis
qPCR was performed using 3 designed specific primers to estimate the abundance of fish taxa and 2 krill species (E. superba and E. crystallorophias) in the fecal DNA sample sets ( Table 2). A standard curve was constructed for each primer set using the serially diluted plasmid harboring each target sequence, as described in a previous study [25]. qPCR was conducted using a Mic qPCR cycler system (BioMolecular Systems, Upper Coomera, Australia). The qPCR mixture in a 20 µL volume included 2 µL genomic DNA, 1 µL each of forward and reverse primer (10 pmol), 6 µL DNase/RNase free water, and 10 µL 2× Luna Universal qPCR master mix (New England Biolabs, USA). PCR conditions for fish taxa included initial denaturation at 95 • C for 5 min, followed by 40 cycles at 95 • C for 5 s, 55 • C, for 20 s, and 72 • C for 20 s, and a final extension at 72 • C for 3 min, while those for the 2 krill consisted of initial denaturation at 94 • C for 5 min, followed by 40 cycles at 94 • C for 30 s, 64.9 • C for 15 s, and 72 • C for 15 s, and a final extension at 72 • C for 5 min. A melting curve was generated after the final extension step of each assay in a 65 to 95 • C melting gradient at a rate of 0.3 • C/s. The fidelity of each qPCR assay was verified with corresponding fish and krill genomic DNA as positive control and water as negative control. Additionally, the specificity of the 2 krill-specific primer sets was confirmed by negative amplification for nontarget krill species, whereas the specificity of the miniFish primers was established using metabarcoding analysis. The calculation of copy numbers was based on Ct values of samples in each standard curve. Each copy number was normalized by dividing the 18S rRNA gene copy number as an endogenous control.

Fecal DNA Metabarcoding of Pygoscelis Adeliae in the Ross Sea Region Using 18Sv9 Universal Primers
In total, 143 fecal DNA from seven sample sites were prepared during the surveys from December 2017 to January 2020. Metabarcoding analysis was conducted with 14 pooled samples from different habitats and years (Table 1). To know the overall diet composition of Adélie penguins, the first metabarcoding analysis was performed using the 18Sv9 universal primer set. On average, 204,657 raw reads were generated from each pooled sample based on MiSeq sequencing. The lowest and highest raw read numbers were obtained at II in 2018 (65,911) and MC in 2017 (413,458) ( Table 3). After trimming, more than 80% of the raw reads for each site were used for further analysis as the reliable "merged reads," supporting the quality of NGS reads (Table 3). Despite their specificity for eukaryotic taxa, the merged reads generated by the 18Sv9 universal primer set still presented a high number of bacterial taxa, whose proportion in pooled samples ranged from 24.72 to 68.17%. In addition to bacterial sequences, the ratio of penguin DNA was high, ranging from 17.54 to 50.07%. After eliminating the bacterial and penguin DNA, only tiny proportions (0.07-31.15%) were obtained as putative prey items of Adélie penguins (Table 3). We failed to identify the species level, mainly because of low sequence variability in the 18Sv9 marker. Hence, prey items in each pooled sample were further classified into five phyla: Annelida, Arthropoda, Chordata, Mollusca, and Unknown ( Figure 2). Among the five phyla, arthropods were most abundant, making up 61.33% on average, followed by unknown (26.20%) and chordates (11.66%). Two phyla, Annelida (0.39%) and Mollusca (0.43%), were negligible ( Figure 2). classified into five phyla: Annelida, Arthropoda, Chordata, Mollusca, and Unknown ure 2). Among the five phyla, arthropods were most abundant, making up 61.33% o erage, followed by unknown (26.20%) and chordates (11.66%). Two phyla, Ann (0.39%) and Mollusca (0.43%), were negligible ( Figure 2).  The arthropod OTUs belonged to either E. superba or E. crystallorophias, the av proportion of which among arthropod OTUs was 75.64%. E. crystallorophias (49.88% approximately two-fold higher than E. superba (25.14%; Figure 2). The copepod O placed second, with an average proportion in arthropod OTUs of 16.92%. Among copepod genera, Calanus (7.78%) was the most abundant, followed by Metridia (4 and Paraeuchaeta (1.90%) ( Figure 2). All OTUs in the phylum Chordata showed the hi identity to notothenioid fish, which is the most abundant fish taxon in the Southern Only two OTUs in Annelida have identified either Lumbrineris sp. or Tomopteris sp., w average proportion was only 0.40% among the five taxa. One OTU in Mollusca sh The arthropod OTUs belonged to either E. superba or E. crystallorophias, the average proportion of which among arthropod OTUs was 75.64%. E. crystallorophias (49.88%) was approximately two-fold higher than E. superba (25.14%; Figure 2). The copepod OTUs placed second, with an average proportion in arthropod OTUs of 16.92%. Among seven copepod genera, Calanus (7.78%) was the most abundant, followed by Metridia (4.60%) and Paraeuchaeta (1.90%) ( Figure 2). All OTUs in the phylum Chordata showed the highest identity to notothenioid fish, which is the most abundant fish taxon in the Southern Sea. Only two OTUs in Annelida have identified either Lumbrineris sp. or Tomopteris sp., whose average proportion was only 0.40% among the five taxa. One OTU in Mollusca showed the highest identity to Limacina helicina, a tiny swimming planktonic sea snail, which was identified only one time at CA in 2019. Finally, an unknown OTU showed the highest identity to Eimeria furonis, suggesting putative gastrointestinal helminths. Although Parorchites zederi and the nematode Stegophorus macronectes are among the parasites commonly found in penguins in Antarctica [26], we failed to identify the exact species, mainly because of the lack of reference sequences in the database.
Nonmetric multidimensional scaling (NMDS) analysis was conducted to compare community structures obtained by metabarcoding (Figure 3). Three clusters were observed via NMDS in the 18Sv9 metabarcoding analysis ( Figure 3A). The first group (CW in 2018 and CH in 2017) was characterized by a high number of unknown parasites, mainly for those starved. The second group (CA in 2019 and II in 2017) showed a high number of Chordata taxa, mostly fish, whereas the others (group 3) exhibited an arthropod-dominant group, most of them belonging to krill. These data support the suggestion that krill are major prey of Adélie penguins in the Ross Sea, with a few exceptional cases. the highest identity to Limacina helicina, a tiny swimming planktonic sea snail, which was identified only one time at CA in 2019. Finally, an unknown OTU showed the highest identity to Eimeria furonis, suggesting putative gastrointestinal helminths. Although Parorchites zederi and the nematode Stegophorus macronectes are among the parasites commonly found in penguins in Antarctica [26], we failed to identify the exact species, mainly because of the lack of reference sequences in the database. Nonmetric multidimensional scaling (NMDS) analysis was conducted to compare community structures obtained by metabarcoding (Figure 3). Three clusters were observed via NMDS in the 18Sv9 metabarcoding analysis ( Figure 3A). The first group (CW in 2018 and CH in 2017) was characterized by a high number of unknown parasites, mainly for those starved. The second group (CA in 2019 and II in 2017) showed a high number of Chordata taxa, mostly fish, whereas the others (group 3) exhibited an arthropod-dominant group, most of them belonging to krill. These data support the suggestion that krill are major prey of Adélie penguins in the Ross Sea, with a few exceptional cases.

Fecal DNA Metabarcoding of Pygoscelis Adeliae in the Ross Sea Region Using miniFish Primers
Piscine prey items of P. adeliae were analyzed using the miniFish primer set (Figure 4). Approximately 1.3 million to 6.1 million raw reads were generated at each site. After trimming, the average proportion of qualified merged reads was 45.93%, ranging from 3.22 to 76.97% (Table 4). Interestingly, extremely low numbers of qualified merged reads were obtained at CA in 2018 (3.22%) and 2019 (8.11%). The proportion of non-fish reads ranged from 4.10 to 15.73%, indicating high specificity for piscine taxa.

Fecal DNA Metabarcoding of Pygoscelis Adeliae in the Ross Sea Region Using miniFish Primers
Piscine prey items of P. adeliae were analyzed using the miniFish primer set (Figure 4). Approximately 1.3 million to 6.1 million raw reads were generated at each site. After trimming, the average proportion of qualified merged reads was 45.93%, ranging from 3.22 to 76.97% (Table 4). Interestingly, extremely low numbers of qualified merged reads were obtained at CA in 2018 (3.22%) and 2019 (8.11%). The proportion of non-fish reads ranged from 4.10 to 15.73%, indicating high specificity for piscine taxa.
The NMDS result showed two clusters but failed to identify any patterns by years and habitats ( Figure 3B). The NMDS analysis showed that two clusters were determined by the proportions of the two main fish species. The first cluster included the nine sites where P. antarctica was the most abundant. At the same time, P. borchgrevinki was equivalent to or higher than P. antarctica at the other five sites (CH, DY, and CW in 2018 and CH and EP in 2019), indicating no spatiotemporal pattern ( Figure 3B).

qPCR Analysis of Main Prey Items
On the basis of the metabarcoding analysis of fecal DNA using two primer sets, two krill species (E. superba and E. crystallorophias) and notothenioid fish, especially P. antarctica and P. borchgrevinki, were identified among the most abundant prey items of Adélie penguins (Figure 4). To conduct a more accurate quantitative analysis of those main prey items, qPCR was adopted for individual samples of fecal DNA. Three primer sets, including two krill-specific primer sets (E. superba and E. crystallorophias) and one fish-specific (miniFish), were used for the quantitative analysis. The proportions of the main prey items showed clear geographic patterns. Higher proportions of E. superba and E. crystallorophias were respectively identified in the outbound and inbound habitats in the Ross Sea ( Figure 5 and Table 6). In fact, the proportions of E. crystallorophias showed a clear opposite pattern to those of E. superba, whose highest proportions were identified in the most inbound habitats, EP and II. The proportion of E. superba at those two habitats was less than 1% ( Figure 5). By contrast, the proportion of E. superba was highest in the most outbound habitats, including CA and DY. The proportion of piscine diets showed a similar pattern with E. crystallorophias, in which the two most inbound habitats, EP and II, showed their highest proportions, while those in CA and DY were lower than 10% ( Figure 5).    Annual proportional changes of the main prey items (Notothenioid fish, E. superba, and E. crystallorophias) were also analyzed at two sites (CH and II), where all 3-year scat samples were collected in the Ross Sea ( Figure 6). Proportions of the main prey were similar for 3 years at the outbound habitat (CA), among which E. superba was the largest, followed by notothenioid fish and E. crystallorophias. By contrast, their proportions were highly variable year to year in the inbound habitat (II). In 2017 and 2019, a higher proportion of fish prey was identified, while a predominant proportion of E. crystallorophias was found in 2018 ( Figure 6). Even between 2017 and 2019, we were able to identify differences in krill composition. Compared to 2017, much higher proportions of E. superba and E. crystallorophias were found in 2019. However, E. crystallorophias was the main krill prey of Adélie penguins in the inbound habitat, II ( Figure 6).

Discussion
We analyzed the diets of Adélie penguins in the Ross Sea region by metabarcoding of fecal DNA with multiple markers (18Sv9 and miniFish). The barcode sizes amplified by the two primer sets were approximately 91-140 bp, which were much smaller than regular barcodes. The quality and quantity of dietary DNA remaining in the scat are often much lower compared to other environmental samples because of the vigorous digestion process [27]. Therefore, amplifying of short sequences, named "minibarcodes", is used for samples containing degraded DNA such as processed food, water, soil, or scat [28,29]. Although it is one of the most widely used methods [30][31][32], mitochondrial COI minibarcodes show a high degree of cross-reactivity with bacterial DNA presenting most of their sequences, which outnumbered the trace amounts of degraded dietary DNA. We alternatively applied the nuclear 18Sv9 universal primer set, which was explicitly designed for eukaryotic taxa [33,34]. However, it also amplified a large amount of predator and bacterial DNA, accounting for up to 80% of total metabarcode reads (Table 3). Blocking primers are often used to suppress the host DNA in metabarcoding results [35,36]. We also tried them, but this attempt failed because they suppressed the DNA of penguins and dietary items (data not shown). Despite the short amplicon size and low sequence variability, metabarcoding analysis using the 18Sv9 primer set successfully showed the overall

Discussion
We analyzed the diets of Adélie penguins in the Ross Sea region by metabarcoding of fecal DNA with multiple markers (18Sv9 and miniFish). The barcode sizes amplified by the two primer sets were approximately 91-140 bp, which were much smaller than regular barcodes. The quality and quantity of dietary DNA remaining in the scat are often much lower compared to other environmental samples because of the vigorous digestion process [27]. Therefore, amplifying of short sequences, named "minibarcodes", is used for samples containing degraded DNA such as processed food, water, soil, or scat [28,29]. Although it is one of the most widely used methods [30][31][32], mitochondrial COI minibarcodes show a high degree of cross-reactivity with bacterial DNA presenting most of their sequences, which outnumbered the trace amounts of degraded dietary DNA. We alternatively applied the nuclear 18Sv9 universal primer set, which was explicitly designed for eukaryotic taxa [33,34]. However, it also amplified a large amount of predator and bacterial DNA, accounting for up to 80% of total metabarcode reads (Table 3). Blocking primers are often used to suppress the host DNA in metabarcoding results [35,36]. We also tried them, but this attempt failed because they suppressed the DNA of penguins and dietary items (data not shown). Despite the short amplicon size and low sequence variability, metabarcoding analysis using the 18Sv9 primer set successfully showed the overall prey items of Adélie penguins in the Ross Sea at the phylum level, in which krill and notothenioid fish were the main prey items (Figure 2).
In addition to the diet study, 18Sv9 rRNA metabarcoding results of the penguins' fecal DNA provided additional information. For instance, diverse putative parasites were identified, including Platyhelminthes (Diphyllobothrium pacificum, Parorchites zederi, Parorchites sp., Spirometra erinaceieuropaei, and Zygobothrium sp.), Chaetognatha (Pseudosagitta sp.), Tardigrada (Acutuncus antarcticus), Nematoda (Contracaecum osculatum), and Apicomplexa (Eimeria sp.). Many of them are among the previously reported parasites of Adélie penguins. P. zederi is one of the common parasites in Adélie penguins [26], and D. pacificum is one of the most widely identified parasites among the animals inhabiting Antarctica [37]. Eimeria sp. is also one of the three main Coccidian parasites (Apicomplexa) of penguins inhabiting Livingston Island and King George Island [38]. It is generally known that parasite composition is highly related to diet composition [39,40], and in particular, euphausiid species are known to be intermediate hosts for parasite species of Adélie penguins [26]. Further studies should be conducted to determine the relationship between diet composition and parasitic species, which may provide helpful information about their ecological implications in the Ross Sea. Moreover, parasite composition should be further monitored to estimate the health condition of Adélie penguins. Although Dibothriocephalus, Adenocephalus, and Diphyllobothrium are considered commensal parasites, the larva of some Spirometra can cause life-threatening diseases to host species [41]. Therefore, monitoring them with the use of molecular tools would be important not only in terms of their ecological implications but also for the sustainability of penguin colonies.
While the 18Sv9 primer set presented the overall prey items of Adélie penguins in the Ross Sea at the phylum level, its low sequence variability and short barcode size made it possible to analyze up to the species level. We obtained detailed information regarding the penguins' fish diet using a fish-specific universal primer, miniFish, targeting their mitochondrial 12S region. The newly designed miniFish primer set showed a high degree of specificity for fish taxa (Notothenioidei) in the Southern Sea without amplifying either penguin or bacterial DNA (Table 4). Its amplicon size was also short (91 bp), which is suitable for amplifying fish sequences from highly degraded DNA samples isolated from scat, soil, or water. The amplified region lies within the MiFish marker, which is applicable to the established reference database [21]. Most of the haplotypes obtained by miniFish showed a high degree of sequence identity, assigning each species name and supporting a high number of reference sequences for fish species inhabiting the Ross Sea. However, because of its short amplicon size, some species in three genera, Chionodraco, Pogonophryne, and Trematomus, were indiscriminate using miniFish metabarcoding (Table 5). Those weaknesses can be compensated by comparative analysis of regional fish assemblage data obtained via eDNA metabarcoding using a more extended barcode primer set such as MiFish [21]. The newly designed miniFish primer would be a valuable tool for diet studies of piscivorous predators, such as penguins and seals in Antarctica. miniFish metabarcoding showed that P. antarctica was the most abundant fish prey of Adélie penguins, making up 56.50% on average, followed by P. borchgrevinki (18.21%), Chionodraco spp. (9.61%), and Trematomus spp. (5.15%), which is consistent with previous studies [42,43]. Since Adélie penguins are omnivorous predators, this result is reasonable, knowing that the Antarctic silverfish, P. antarctica, is the most abundant pelagic fish inhabiting the Ross Sea [44,45]. However, there are different reports regarding Adélie penguin diets in previous studies. Krill and P. borchgrevinki were found to be the main diet according to underwater camera surveillance [46]. Another study, based on isotope 13C and 15N values, revealed that euphausiids are the main food source [47]. Those results may reflect regional characteristics, and metabarcoding analysis covered all cases. We identified a higher proportion of P. borchgrevinki than P. antarctica at several sites ( Figure 4). In addition, compared with krill, an equal or higher number of fish taxa were shown at II in 2017 and 2018 ( Figure 2). These proportions changed from year to year, and long-term metabarcoding analysis using a dual primer set would be one efficient strategy to determine the relationship between the proportions of prey and the environmental characteristics in the Ross Sea. Although P. antarctica was identified as the main fish prey of Adélie penguins, it is worth noting that 11 fish genera have been identified as prey items for these penguins in the Ross Sea, some of which are actually too large to be their prey ( Figure 4). Furthermore, two genera, Chionodraco and Trematomus, made up approximately 15% of fish prey on average. Different from the two small pelagic fish prey, these two are known as demersal fish species. Terra Nova Bay, a highly productive area in the Ross Sea, is an important spawning and nursery ground for several fish species [48]. Thus, it would also be worth investigating the relationship between the reproduction of notothenioids and the dynamics of Adélie penguins.
While Adélie penguins' main diet was found to be krill and notothenioid fish, which is consistent with previous studies [6,15], neither hydrozoans nor cephalopods were identified from fecal DNA in this study ( Figure 2 and Table 3). Recent studies showed that hydrozoans have increasingly grown as penguin prey, which is presumably because of climate change [6,15]. Cephalopod Psychroteuthis glacialis was also considered to be among the prey items of Adélie penguins [49]. However, these differences may come from the seasonal availability of non-fish taxa in some habitats, especially those located inbound of the Ross Sea. In previous studies, except for fish taxa, a high degree of seasonal changes in the penguins' diet was shown, which was presumably due to food availability by periodic changes in ice-covered regions along the inner coastal part of the Ross Sea such as polynyas [49,50]. Subsequently, more frequent and long-term surveys of Adélie penguins in the Ross Sea are required in order to understand the relationship between environmental changes and food availability in the region, and a long-term molecular survey of scat samples would be one of the most cost-effective strategies.
According to previous research, krill was found to be the most important prey of Adélie penguins in the Ross Sea. One of the most interesting findings by molecular analysis was a clear geographic difference of krill prey within the Ross Sea [51,52]. Higher proportions of E. superba were identified in the more outbound habitats, while the proportion of E. crystallorophias increased in the more inbound habitats ( Figure 5). Differences in the abundance of two krill, E. superba and E. crystallorophias, are due to their availability in the Ross Sea. E. superba is considered a pelagic species in offshore areas, whereas E. crystallorophias is limited to inshore waters under ice [53]. Therefore, outbound habitats such as CA and DY would be favorable for E. superba, while the coastal waters near two inbound habitats, EP and II, may provide favorable conditions for E. crystallorophias rather than E. superba. We also identified that the prey of Adélie penguins inhabiting outbound habitats showed a rather stable pattern; the main prey was E. superba, with small amounts of fish species (Figure 6). By contrast, those in inbound habitats of the Ross Sea face a high degree of variation in their prey according to environmental changes. One of the most important environmental changes is the dynamics of the ice pack along the Ross Sea. Since they prefer being underneath the ice pack, changes in the availability of E. crystallorophias or fish may be related to polynya. Previous studies showed that the population of Adélie penguins is directly linked to sea ice [54] and temperature fluctuations [55]. This ecological indicator's productivity has been linked to Antarctic krill [56], which is also sea ice-bound [57,58]. This could help researchers better understand the population dynamics of Adélie penguins, especially among the inbound habitats in the Ross Sea. Collectively, we established a pipeline to analyze the diets of Adélie penguins from scat samples in the Ross Sea using molecular techniques. We were able to obtain detailed information on the prey species of Adélie penguins. We also found that the penguins' diets were highly affected by the geographic characteristics of the Ross Sea. Frequent and long-term surveys using this methodology would provide detailed information regarding the feeding ecology of Adélie penguins inhabiting the Ross Sea Region Marine Protected Area, which would be of practical use for their scientific conservation and management.

Conclusions
The diets of Adélie penguins (P. adeliae) were analyzed using fecal DNA with NGS and qPCR techniques. The combination of molecular tools used in this study successfully elucidated that two krill species and notothenioid fish are the penguins' main diet species. However, diet compositions were highly variable according to the environmental conditions and geographic food availability. These data will help researchers expand our knowledge of the role of Adélie penguins in the ecosystem of the Ross Sea. Further studies should be conducted to understand the relationship between meteorological phenomena such as the occurrence of polynya and the diets of Adélie penguins in the Ross Sea.