Marine sediment samples were collected from the Amundsen Sea, Antarctica, at a depth range of 694–1249 m using a box corer during the RV ARAON research cruise between January and February of 2022 (ANA12C). At each location, approximately 50 g of muddy sediments were collected aseptically into sterile conical tubes. Some samples were treated on-board for actinomycete isolation, while others were maintained at a temperature of 4°C during a one-month transportation back to the laboratory in Korea.

The sediment samples were treated immediately after collection. They were pretreated using the stamping technique as follows: sediments were aseptically placed into a sterile petri dish, air-dried in a laminar flow hood, and then heated for 15 min at 57°C. The treated samples were pressed into a sterile form plug, and inoculated onto 1/10 MA (10% Marine broth (2216 Difco, US), 1.8% agar), 1/10 SYPA (0.1% soluble starch, 0.04% yeast extract, 0.02% peptone, and 1.8% agar) and SWA (1.8% agar) media plates by circular stamping, thereby inducing a serial dilution effect (Hameş-Kocabaş and Uzel, 2012). All media were prepared using 75% artificial seawater, which was prepared using commercial sea salt (Blue Treasure SPS Sea Salt, Qingdao Sea-Salt Aquarium Technology Co., Ltd., China). Two isolation methods were employed for the remaining samples moved to the laboratory: serial dilution and the stamping technique. The serial dilution process was carried out as follows: 1 g of sediment sample was added to a sterile 50 mL conical tube containing 9 mL of 1/2 MB (50% Marine broth (2216 Difco, US), 1% NaCl) and shaken at 180 rpm using an orbital shaker at 4°C for 7 days. From these tubes, 1 mL of an aliquot was transferred and mixed with another 9 mL of 1/2 MB to make a 10^-2 dilution factor. Following the dilution process, 0.1 mL of each sample was plated onto 1/10 MA, SYPA, and SWA media plates. Some media were supplemented with filter-sterilized cycloheximide (25 μg/mL) after autoclaving. The agar plates were incubated at room temperature (20–25°C) and observed for at least six months. Colonies with suspected actinomycetes morphology were subcultured on SYPA media and incubated at 28°C for 7 days. The pure cultures were preserved with 20% (v/v) glycerol and stored at -80°C.

The 16S rRNA gene was amplified by PCR for all strains using universal primers: 27F (5’-AGAGTTTGATCCTGGCTCAG-3’) and 1492R (5’-TACGGYTACCTTGTTACGACTT-3’) and sequenced by BioFact (Daejeon, Korea). PCR was performed in a Hushrun PCR cycler (BioFact, Deajeon, Korea) using an initial denaturation at 95°C for 15 min, followed by 30 cycles at 95°C for 20 sec, 56°C for 30 sec, and 72°C for 5 min. The BLAST program (https://blast.ncbi.nlm.nih.gov/Blast.cgi) (Altschul et al., 1990) and the web-based tool EZTaxon (https://www.ezbiocloud.net) (Chun et al., 2007) were used to compare the partial 16S rRNA gene sequences with the GenBank database. Sequences were aligned with closely related species, and the phylogenetic tree was constructed using maximum likelihood (ML) methods with a bootstrap analysis of 1,000 replicates using Molecular Evolutionary Genetics Analysis software version 12 (MEGA12) (Kumar et al., 2024).

All strains were precultured (28°C, 180 rpm for 7 days) in 15 mL SYP broth (1% soluble starch, 0.4% yeast extract, 0.2% peptone in 75% artificial seawater). The fermentation broth was inoculated into 1 L of SYP broth in a 2.5-L Erlenmeyer flask (1.5% v/v inoculation ratio). Following inoculation at 28°C with shaking at 180 rpm for 7 days, the culture was extracted with 1 L of ethyl acetate (EtOAc). The organic layer was then separated using a separation funnel. The organic phase was concentrated under a vacuum, yielding the crude extract. The crude extract was subjected to open column chromatography on silica gel 60 (Merck, Germany) and eluted with a step gradient of CH₂Cl₂-MeOH (99:1, 90:10, and 0:100) to yield three fractions. The extracts of most strains were tested for their antibacterial activity against methicillin-resistant S. aureus and Acinetobacter baumannii by determining their growth kinetics (Lee et al., 2025).

LC-HR-MS/MS was conducted using a Thermo Fisher Scientific Vanquish UHPLC System integrated with a Thermo Scientific Q Exactive Plus mass spectrometer at the Core Research Support Center for Natural Products and Medicinal Materials (CRCNM). The fractions were prepared at a concentration of 1 mg/mL in methanol for analysis. The chromatographic separation was performed on an Acquity UPLC BEH C18 column (2.1 mm × 100 mm, 130Å, 1.7 μm). The column temperature was set to 25°C. The injection volume was set to 4 μL with a sample concentration of 1 mg/mL, corresponding to an injected amount of 4 μg. The flow rate was maintained at 0.3 mL/min throughout the analysis. The mobile phase consisted of two solvents: solvent A (water with 0.1% formic acid) and solvent B (acetonitrile with 0.1% formic acid). A gradient elution was performed starting with 5% solvent B, which was held constant for the first 2 min. The percentage of solvent B was then increased to 20% over the next 6 min and further increased to 100% over the following 6 min, reaching 100% at 14 min. This condition was maintained for 5 min. The system operated in positive ionization mode under the following acquisition parameters: an MS range of m/z 100–1500, an MS2 range of m/z 50–1500. The raw data of LC-HR-MS/MS were processed using MZmine version 4.3.0. The mass detection was performed by setting the noise levels at 2.0E2 for MS and 1.0E1 for MS2. Chromatograms were constructed with a minimum time span of 0.01 min, a minimum height of 3000, and an m/z tolerance of 0.0 (or 20.0 ppm). The chromatographic deconvolution was carried out using a baseline cutoff algorithm with the following parameters: a minimum peak height of 1500, a peak duration range of 0.01–3.00 min, a baseline level of 1000, an m/z range for MS2 scan pairing of 0.02 Da, and a retention time (RT) range for MS2 scan pairing of 0.1 min. Deisotoping was performed with an isotope peak grouper algorithm, using an m/z tolerance of 0.0 (or 20.0 ppm) and an RT tolerance of 0.1 min. Alignment was conducted using the join aligner module, with an m/z tolerance of 0.0 (or 20.0 ppm), an absolute RT tolerance of 0.1 min, a weight for m/z of 70, and a weight for RT of 25. Finally, peaks were gap-filled with a peak finder module, applying an intensity tolerance of 10.0%, an m/z tolerance of 0.0 (or 20.0 ppm), and an absolute RT tolerance of 0.2 min.

Feature-based molecular networking (FBMN) was conducted using the GNPS platform (https://gnps.ucsd.edu). The precursor ion and fragment ion tolerances were set to 0.02 Da. The molecular network was generated with a minimum cosine score of 0.7 and at least six matched fragment ions and peaks. MS/MS spectra were filtered to retain the top six peaks within a 50 Da window across each spectrum. Spectral data in the network were matched against the GNPS spectral library. The visualization of the molecular network was performed using Cytoscape version 3.10.2 (https://www.cytoscape.org/). The full FBMN workflow and results are accessible via the GNPS repository under Task ID eed2645813084a80a05f3936bbc76e69.

A total of 17 sediment samples were collected from the Amundsen Sea in Antarctica for the purpose of investigating the diversity of actinomycetes. This investigation employed a culture-dependent analytical approach. The latitude, longitude, and sediment depth were recorded ( Figure 1 ; Supplementary Table S1 ). A total of 24 actinomycete strains were isolated based on actinomycete-like morphology from the sediment samples. Strains were considered distinct if they differed in at least one of the following criteria: location, date of isolation, culture medium, isolation method, or morphological characteristics.

The BLAST analysis of the 16S rRNA gene sequences (1378–1471 bp) from the strains showed 98.57% or higher similarity value with sequences of previously described species. Based on the phylogenetic tree in Figure 2 , the actinomycete strains isolated from the Amundsen Sea were classified into five taxonomic orders: Micrococcales (four isolates), Kitasatosporales (three isolates), Propionibacteriales (three isolates), Mycobacteriales (eight isolates), and Geodermatophilales (six isolates). In the order Micrococcales, three isolates (LMA519, LMA520, and LMA521) and one isolate (LMA743) respectively formed different phyletic lineages within the genera Microbacterium and Micrococcus. Three isolates (LMA499, LMA744, and LMA746) formed phyletic lines encompassed by the genus Streptomyces. Three isolates (LMA649, LMA688, and LMA915) were identified as Nocardioides salarius J112 by BLAST analysis with 99.06 to 99.14% similarity. Eight isolates (LMA660, LMA661, LMA648, LMA686, LMA741, LMA742, LMA745, and LMA916) formed diverse phylogenetic lines within the evolutionary radiation encompassed by the two families Corynebacteriaceae and Nocardiaceae. The isolates LMA660 and LMA661 showed a low level of similarity to the known species Corynebacterium pilbarense IMMIB WACC-658 with similarity percentages of 98.63% and 98.70%, respectively. In the order Geodermatophilales, six isolates (LMA501, LMA503, LMA687, LMA689, LMA690, and LMA917) formed different phyletic lineages within the genera Blastococcus and Modestobacter. The strains LMA503, LMA687, and LMA917 were assigned to the genus Blastococcus. LMA503 showed a low level of similarity to the known species Blastococcus deserti SYSU D8006, with a similarity percentage of 98.57%. LMA687 and LMA917 exhibited a 98.77% and 98.71% similarity, respectively, to the known species Blastococcus litoris GP-S2-8 and Blastococcus aggregatus DSM 4725. The isolates LMA501, LMA689, and LMA690 were identified as belonging to the genus Modestobacter (Modestobacter marinus 42H12-1). All partial 16S rRNA gene sequences were submitted to GenBank. The accession numbers obtained are listed in Table 1 .

Fifteen strains (designated with red names in Figure 2 ) were selected for metabolite analysis based on their phylogenetic relationships. The selected strains belonged to the genera Rhodococcus, Corynebacterium, Nocardioides, Blastococcus, Modestobacter, Micrococcus, and Streptomyces. The extraction yields from 1 L cultures varied significantly among different genera, ranging from 31.2 mg to 73.9 mg. Approximately 70 mg of extract was obtained from strains belonging to the genus Streptomyces, 50 mg from Rhodococcus and Blastococcus strains, and around 30 mg from Modestobacter, Corynebacterium, Nocardioides, and Micrococcus strains following EtOAc extraction. The fermentative extracts were tested for bioactivity against both Gram-positive and Gram-negative bacterial pathogens. Nevertheless, the results obtained did not reveal any significant antibiotic activity (data not shown).

The antibacterial activity test was conducted at the preliminary screening stage using unpurified crude extracts and only two representative strains. To overcome these limitations and further explore the metabolic potential of the isolates, a molecular networking analysis was performed which yielded 3,702 parent ions (nodes) when the MS/MS data were analyzed by bacterial genus, as shown in Figure 3 . The production of parent ions, with the exclusion of media-derived components, showed a mix of genus-specific and shared patterns ( Figure 4A ). This specificity highlights the potential for discovering bioactive metabolites, as each strain contributes a unique chemical diversity. Within the molecular network, 1,414 parent ions (38.20%) were identified as unique to an individual genus. Remarkably, 203 metabolites (5.48% of the total ions) were specific to Rhodococcus, 104 metabolites (3.81%) to Corynebacterium, 133 metabolites (3.59%) to Nocardioides, 55 metabolites (1.49%) to Blastococcus, 52 metabolites (1.40%) to Modestobacter, 56 metabolites (1.51%) to Micrococcus, and 811 metabolites (21.91%) to Streptomyces ( Figure 4A ). Interestingly, the majority of detected ions (2,288 ions, 61.80%) were shared across isolates from multiple genera, indicating the presence of broadly conserved or widely distributed metabolic features.

In the FBMN analysis, a total of 3,702 nodes were detected, among which 269 compounds were successfully annotated in the GNPS library ( Figures 4B–D ). The remaining 3,433 nodes (92.73%) were unannotated, highlighting the extensive presence of yet-to-be-identified metabolites. The analysis revealed that several bacterial genera share common classes of compounds, highlighting conserved biosynthetic pathways across diverse species. For instance, cyclic dipeptides (diketopiperazines) such as cyclo(l-Val-l-Pro) and cyclo(l-Leu-l-Pro) were found in Rhodococcus, Corynebacterium, Nocardioides, Blastococcus, Modestobacter, Micrococcus, and Streptomyces. These compounds are known for their bioactive properties, such as antifungal effects against Fusarium sp. and Aspergillus sp (Soldatou et al., 2021), antibacterial activity against S. aureus and Bacillus subtilis (El-Gendy and Rateb, 2015), and vancomycin-resistant Enterococcus (Rhee, 2002), as well as quorum-sensing modulation (Holden et al., 1999). Their occurrence across the genus suggests a fundamental role in bacterial metabolism. Similarly, fatty acid amides, including palmitamide and 9-octadecenamide, were detected in Rhodococcus, Corynebacterium, Nocardioides, Blastococcus, Modestobacter, Micrococcus, and Streptomyces. Indole derivatives (particularly indole-3-acetic acid and indoxyl sulfate) and adenine derivatives were found to be prevalent in genera such as Rhodococcus, Corynebacterium, Nocardioides, and Blastococcus. Fatty acids and their esters, including linoleic acid and sorbitane monopalmitate, were also commonly detected, indicating the role of lipid metabolism across these bacterial genera.

In addition to the common compound classes, we identified several genus-specific compounds that further differentiate the metabolic profiles of each bacterial genus. For example, Rhodococcus exhibited a wide range of diketopiperazines, indole derivatives, and polyphenols, along with the unique presence of β-caryophyllene oxide and monactin. The detection of β-caryophyllene oxide suggests its potential role in biotechnological applications, highlighting its importance in microbial secondary metabolism. The monactins, known family of macrocyclic antibiotics, were found in a Figure 4B cluster, which suggests that Rhodococcus species may possess antimicrobial properties (Phillies et al., 1975).

Corynebacterium strains shared many compound classes with Rhodococcus, but they were distinguished by the presence of biotin methyl ester, phenacetin, and deoxycholate, which may serve specific functions in metabolism or stress responses. Additionally, fatty acid derivatives such as palmitamide, 9Z-octadecenamide, 13Z-docosenamide, and linoleic acid further corroborate the unique fatty acid synthase system in Corynebacterium. This system produces coenzyme A (CoA)-linked fatty acids instead of acyl carrier protein (ACP)-linked intermediates (Takeno et al., 2013). The presence of biotin methyl ester and lumichrome indicates the potential for vitamin biosynthesis, consistent with findings from metabolic engineering studies (Feierabend et al., 2021).

Conversely, Nocardiodes strains were enriched in aromatic amines such as anthranilate and 4-hydroxyphenylacetamide. These compounds are often associated with secondary metabolic processes and may play a role in bacterial survival under certain environmental conditions. A range of fatty acid derivatives including linoleic acid, palmitamide, palmitoyl ethanolamide, 13Z-docosenamide, and 9Z-octadecenamide, have been identified, aligning with the known fatty acid composition of this genus (Yoon et al., 2005; Mitzscherling et al., 2022). The detection of dipeptides (Val-Leu, Phe-Pro, His-Pro, Phe-Glu) and cyclic dipeptides [e.g., cyclo(l-Val-l-Pro), cyclo(l-Leu-l-Pro)] supports the presence of peptidoglycan-derived amino acids, which are integral to the cell walls of Nocardioides species. Additionally, the presence of adenosine and 5’-deoxyadenosine aligns with the high G+C content characteristic of Nocardioides genomes (Yoon et al., 2005).

In Blastococcus strains, they showed a distinct profile dominated by histamine-based lipids and bioactive compounds such as Denatonium and Nerolidol, which suggests potential antimicrobial and bioactive properties (Hezbri et al., 2024; Urzì et al., 2004). The detection of fatty acids such as linoleic acid, oleic acid, and 12(13)-epoxy-9Z-octadecenoic acid further aligns with the fatty acid profile of Blastococcus, which predominantly includes isopalmitic acid (iso-C16:0), cis-palmitoleic acid (iso-C16:1 H), and cis-10-heptadecenoic acid (C17:1 ω8c) (Hezbri et al., 2024; Urzì et al., 2004). Furthermore, palmitamide and 9Z-octadecenamide are amide derivatives that are structurally related to the polar lipids in Blastococcus, such as phosphatidylcholine and phosphatidylethanolamine (Montero-Calasanz et al., 2022; Yang et al., 2019).

Modestobacter strains shared some similarities with Blastococcus, because of the production of indigo red and deoxycholate. Metabolite profiling of Modestobacter strains identified palmitamide and octadecanamide, both fatty acid amides, consistent with the presence of iso-C16:0 and cis-9-octadecenoic acid (C18:1 ω9c) fatty acids in Modestobacter species. Additionally, phospholipid-related metabolites such as lauryldiethanolamine and N,N-dimethyldodecylamine N-oxide were observed, which may correspond to the diverse polar lipids reported in Modestobacter species, including phosphatidylethanolamine and phosphatidylinositol (Jiang et al., 2021). The identification of adenosine and N6-(Δ2-isopentenyl)-adenine also supports the nucleotide-derived metabolic profile of Modestobacter, which has been implicated in menaquinone biosynthesis (Jiang et al., 2021).

Micrococcus strains produced unique compounds such as carbazole, a nitrogen-containing polycyclic aromatic compound, and tris(2-butoxyethyl) phosphate. These compounds may offer insights into the ecological adaptability of Micrococcus strains and their potential industrial applications. Metabolites such as cyclo(l-Val-l-Pro), Phe-Pro, cyclo(l-Leu-l-Pro), N-acetyl-tyramine, lumichrome, xanthine, and indole-3-acetate were identified. These compounds support the production of bioactive peptides and alkaloids in the genus Micrococcus, including thiazolyl peptide antibiotics (e.g., micrococcin and kocurin) and carotenoid-derived pigments with antimicrobial properties (Tizabi and Hill, 2023). The identification of pyrrolo[1,2-a]pyrazine-1,4-dione, an alkaloid-related compound, aligns with the genus’s ability to produce bioactive nitrogen-containing metabolites (Tizabi and Hill, 2023).

Finally, Streptomyces were found to produce a diverse array of secondary metabolites, including cyclo(l-Val-l-Pro), Phe-Pro, His-Pro, Val-Leu, and Leu-Gly, which are commonly associated with antimicrobial and signaling activities in actinomycetes. Furthermore, the presence of nerolidol, a sesquiterpene alcohol, suggests the potential for further antimicrobial activity in these strains (Krist et al., 2015). The detection of linoleic acid and 9,12-octadecadiynoic acid, both with cytotoxic and antimicrobial properties, further underscores the adaptability of Streptomyces in extreme environments (Pacios-Michelena et al., 2021).