Bioactivity and Metabolome Mining of Deep-Sea Sediment-Derived Microorganisms Reveal New Hybrid PKS-NRPS Macrolactone from Aspergillus versicolor PS108-62

Despite low temperatures, poor nutrient levels and high pressure, microorganisms thrive in deep-sea environments of polar regions. The adaptability to such extreme environments renders deep-sea microorganisms an encouraging source of novel, bioactive secondary metabolites. In this study, we isolated 77 microorganisms collected by a remotely operated vehicle from the seafloor in the Fram Strait, Arctic Ocean (depth of 2454 m). Thirty-two bacteria and six fungal strains that represented the phylogenetic diversity of the isolates were cultured using an One-Strain-Many-Compounds (OSMAC) approach. The crude EtOAc extracts were tested for antimicrobial and anticancer activities. While antibacterial activity against methicillin-resistant Staphylococcus aureus (MRSA) and Enterococcus faecium was common for many isolates, only two bacteria displayed anticancer activity, and two fungi inhibited the pathogenic yeast Candida albicans. Due to bioactivity against C. albicans and rich chemical diversity based on molecular network-based untargeted metabolomics, Aspergillus versicolor PS108-62 was selected for an in-depth chemical investigation. A chemical work-up of the SPE-fractions of its dichloromethane subextract led to the isolation of a new PKS-NRPS hybrid macrolactone, versicolide A (1), a new quinazoline (−)-isoversicomide A (3), as well as three known compounds, burnettramic acid A (2), cyclopenol (4) and cyclopenin (5). Their structures were elucidated by a combination of HRMS, NMR, [α]D, FT-IR spectroscopy and computational approaches. Due to the low amounts obtained, only compounds 2 and 4 could be tested for bioactivity, with 2 inhibiting the growth of C. albicans (IC50 7.2 µg/mL). These findings highlight, on the one hand, the vast potential of the genus Aspergillus to produce novel chemistry, particularly from underexplored ecological niches such as the Arctic deep sea, and on the other, the importance of untargeted metabolomics for selection of marine extracts for downstream chemical investigations.


Introduction
Marine sediments are formed by sinking planktonic matter, which are continually alimented from the water column and represent the largest source of organic carbon on earth [1][2][3]. The seafloor represents a heterogeneous surface, where microbial cell counts are Due to redundancy of several species based on the sequencing results (i.e., identical sequence), especially in the highly abundant bacterial genus Bacillus, we selected 32 bacteria and 6 fungi representing the taxonomic diversity for a small-scale cultivation campaign by an OSMAC approach and assessed their biotechnological potential. In order to expand the chemical diversity of the selected isolates, each fungal and bacterial strain was cultured in three different solid media that varied in carbon and nitrogen source(s) and Due to redundancy of several species based on the sequencing results (i.e., identical sequence), especially in the highly abundant bacterial genus Bacillus, we selected 32 bacteria and 6 fungi representing the taxonomic diversity for a small-scale cultivation campaign by an OSMAC approach and assessed their biotechnological potential. In order to expand the chemical diversity of the selected isolates, each fungal and bacterial strain was cultured in three different solid media that varied in carbon and nitrogen source(s) and salt concentrations. The fungi were grown on potato dextrose (PDA), rice (Rice), and glucose peptone yeast (GPY) agar, while the bacterial isolates were cultivated on marine agar (MA), starch casein KNO 3 (SCK) and glucose yeast malt (GYM) agar. Additionally, GPY (bacteria) and GYM (fungi) media were supplemented with potassium bromide, because bromide is a common salt ion in seawater, and many marine metabolites contain bromine substitution [20]. Therefore, we added potassium bromide (KBr) to the culture media to induce bromination of the biosynthesized compounds. As bromine can have toxic effects on microbes [21], the KBr addition was restricted to two media, GYM+Br and GPY+Br.
Of a total of 152 cultures (32 bacteria and 6 fungi each on 4 media), 136 cultures were successfully cultivated and subsequently extracted with EtOAc, while SCK medium did not support the growth of 14 bacterial isolates, and GPY+Br did not provide an adequate substrate for another two bacterial isolates. The EtOAc crude extracts were tested against (i) a set of infectious human pathogens: the ESKAPE panel that is composed of Enterococcus faecium, methicillin-resistant Staphylococcus aureus (MRSA), Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Escherichia coli; (ii) two opportunistic fungal pathogens, the yeast Candida albicans and the yeast-like Cryptococcus neoformans; and (iii) four human cancer cell lines (malignant melanoma cell line A375, the colon cancer cell line HCT116, the lung adenocarcinoma cell line A549, the breast adenocarcinoma line MDA-MB-231) plus the non-cancerous human keratinocyte cell line HaCaT. In total, high activity (defined as growth inhibition of ≥80% at test concentration of 100 µg/mL) was observed for 70 extracts in at least one assay. Activities against Gram-positive bacteria (MRSA and E. faecium) were most frequent ( Figure 1B,C); yet, none of the extracts inhibited the growth of Gram-negative bacteria. For the fungal isolates, the highest number of bioactive extracts was derived from GPY cultures; four extracts exhibited activity against MRSA and one against E. faecium ( Figure 1B, Table S2). High activity against the pathogenic yeast C. albicans was observed in PDA, Rice and GPY+Br extracts of Aspergillus versicolor PS108-62 and of Emericellopsis maritima PS108-61 only when cultured on PDA ( Figure 1B). As for the bacteria, the cultivation medium had again a clear impact on the bioactivity profile of the extracts ( Figure 1C). Cultures grown in SCK medium produced the least number of bioactive extracts with only six showing activity against E. faecium and seven against MRSA, while almost double that amount of MA extracts (13 against MRSA and 12 against E. faecium) showed activities against the Gram-positive pathogens ( Figure 1C). There was no difference in antibacterial activity of GYM and GYM+Br extracts; notably, only the GYM+Br extracts of two bacterial isolates showed high anticancer activities, namely, Rheinheimera perlucida PS108-96 (against A-375 cells) and Bacillus licheniformis PS108-67a (against all tested cell lines, Figure 1C). Supplementary Tables S2 and S3 outline the in vitro activity of all bacterial and fungal crude extracts.

Strain Selection for Large Scale Chemical Investigation
In order to select the most suitable strains for detailed chemical investigations, we employed a three-tier selection approach based on (i) unique biological activities, (ii) diverse chemical profiles and (iii) extract yields. For the first tier, activity against Gram-positive human pathogens was common among the deep-sea isolates ( Figure 1B,C). Inhibitory activity against C. albicans was unique to two fungi, and cancer cell lines were only inhibited by two bacterial isolates. Consequently, the bacterial strain B. licheniformis PS108-67a was pre-selected because two of its four crude extracts inhibited the growth of multiple cell lines. More precisely, the GYM+Br medium extract of B. licheniformis showed strong inhibition (≥80%) of all human cancer cell lines and the non-cancerous human keratinocyte cell line HaCaT (Table 1). Its GYM extract also showed moderate (46%) to high (94%) activity against all cell lines. Aspergillus versicolor PS108-62 was the second pre-selected (fungal) strain, mainly due to its strong inhibitory activity against C. albicans observed in its PDA, Rice and GPY+Br extracts, plus the moderate anticancer and anti-MRSA activities exhibited by its PDA extract (Table 1).  Next, the effects of culture media on the chemical space of the bioactive crude extracts of the two pre-selected strains (B. licheniformis PS108-67a and A. versicolor PS108-62) were investigated via a UPLC-QToF-MS/MS-based molecular networking (MN) approach [22]. No nodes in the networks could unambiguously be identified via automated annotation through the GNPS platform; however, manual dereplication (Tables S4 and S5) permitted the annotation of several molecular families (highlighted by eclipses) in the MNs shown in Figure 2.
The bacterial isolate B. licheniformis PS108-67a displayed a limited diversity of molecular families with only two clusters with more than two nodes in the MN (Figure 2A) to which the GYM and GYM+Br extracts contributed fairly evenly (GYM: 31 nodes; GYM+Br: 36 nodes). Three nodes belonging to the main cluster in the MN could be manually annotated to the known compound class of lichenysins [23] (Figure 2A). The large cluster consisted of nodes of the protonated and doubly charged ion species of the identified compounds, such as lichenysin G8a (C 54  ) was dereplicated from the raw UPLC-MS/MS data; however, the compound did not form a node in the MN. Lichenysins are well-known potent anionic cyclic lipoheptapeptide biosurfactants produced by B. licheniformis with reported bioactivities, including cytotoxicity to several cell lines [24], thus, they are most likely also responsible for the observed cytotoxicity including that against the non-cancerous HaCaT cell line. The remaining nodes in this cluster could not be dereplicated and thus could potentially represent new lichenysin analogues. The nodes found in the unannotated cluster A1 had molecular weights above 1000 Da, suggesting that these compounds are also of peptidic or polymeric nature. Due to several facts, i.e., our primary interest in drug-like small molecules (200-600 Da), the previous reports on cytotoxicity of lichenysins [24], and the herein observed general, non-selective toxicity of the B. licheniformis PS108-67a crude extract towards non-cancerous cell line HaCaT, this deep-sea strain was not prioritized for further chemical analyses. the larger network (left). The nodes of the remaining clusters B1-B4 and all other twonode clusters did not return any hits in multiple databases and, thus, may present putatively new compounds. As shown in Figure 2B, the unannotated cluster B4 was unique to the Rice extract, whereas the clusters B1-B3 shared nodes among all media extracts (Rice, PDA and GPY+Br). Notably, the crude extract of A. versicolor PS108-62 grown on GPY+Br medium was devoid of any compounds with the characteristic isotopic pattern of a bromine atom(s).  The MN for A. versicolor PS108-62 indicated a much broader chemical diversity with eight different molecular clusters containing more than two nodes ( Figure 2B). The PDA medium extract contributed the lowest number of nodes in the network (26 nodes), while 42 nodes were derived from the Rice extract and 45 from the GPY+Br extract. No nodes in the network could be identified via automated annotation, while manual dereplication associated the known compounds with seven nodes. The chemical structures of the annotated compounds and their MS/MS spectra are shown in Figures S6 and S7-S13, respectively. The single node (C 17 H 15 N 2 O 4 m/z 311.1033 [M + H] + ) from the GPY+Br extract was annotated as cyclopenol, a benzodiazepine alkaloid [25]. The two-node cerebroside cluster, including flavuside B (C 43 H 79 NO 9 Na m/z 776.5644 [M + Na] + ) [26,27], was expressed in all media extracts, i.e., Rice, PDA and GPY+Br. One of the three-node clusters was attributed to burnettramic acids that are complex linear PKS-NRPS hybrids. They were predominantly produced in the PDA medium, but they were also observed in the Rice medium extract. Burnettramic acid A (C 41 H 72 NO 12 m/z 770.5042 [M + H] + ) [28,29] was the only annotated compound in this cluster. The biosynthesis of the pyrroloindole alkaloids protubonines appeared to only be induced in the GPY+Br medium. This second small cluster with three nodes was represented by protubonine A (C 19 [30,31]. The quinazoline alkaloids produced by the Rice and GPY+Br medium extracts were clustered in two different, larger networks ( Figure 2B). Due to the clear differences observed in the experimental MS/MS spectra ( Figures S8-10 [32] were in the larger network (left). The nodes of the remaining clusters B1-B4 and all other two-node clusters did not return any hits in multiple databases and, thus, may present putatively new compounds. As shown in Figure 2B, the unannotated cluster B4 was unique to the Rice extract, whereas the clusters B1-B3 shared nodes among all media extracts (Rice, PDA and GPY+Br). Notably, the crude extract of A. versicolor PS108-62 grown on GPY+Br medium was devoid of any compounds with the characteristic isotopic pattern of a bromine atom(s).
Due to the observed bioactivity and chemical diversity discussed above, we decided to select A. versicolor PS108-62 for in-depth chemical studies. The yields of crude A. versicolor PS108-62 media extracts (each derived from 10 agar plates) varied significantly (Table 1). When grown on Rice medium, 102.8 mg of extract was recovered; this was four times higher than that obtained from the PDA medium (27.6 mg) and ten times higher than that from the GPY+Br medium extract (10.5 mg). Based on their bioactivity and overall chemical diversity, GYP+Br and Rice extracts both presented good candidates for a large-scale cultivation. However, considering the extraction yields and its contribution to unannotated MN clusters, Rice medium was selected for a large-scale cultivation experiment of A. versicolor PS108-62 for subsequent purification and chemical characterization of its metabolites.

Compound Isolation and Structure Elucidation
Aspergillus versicolor PS108-62 cultivation was upscaled on 500 plates of rice medium for 21 days, at 22 • C, and subsequently extracted with EtOAc. The crude EtOAc extract was submitted to a modified Kupchan partition scheme [33] to yield three subextracts; aqueous MeOH (Km), DCM (Kd), and n-hexane (Kh). Further chemical work focused on the Kd subextract for the following reasons: The limited amount obtained for the Km subextract (34.9 mg) and its high salt content did not permit chemical investigation of this fraction. Additionally, no further work was performed on the Kh (6.355 g) subextract, due to its high lipid content, as indicated by its 1 H NMR spectrum ( Figure S14). The Kd subextract was fractionated on a C18-SPE cartridge to yield 12 fractions, which were tested against MRSA and C. albicans. The antimicrobial activity against C. albicans and MRSA was tracked to the later eluting fractions (Frs. 8-10). Fr. 8 was the most active SPE fraction with 100% and 93% inhibition at 50 µg/mL against MRSA and C. albicans, respectively.
The new compound 1, versicolide A ( Figure 3), was isolated from the bioactive SPE Fr.9 as a colorless film. The molecular formula C 29 H 49 NO 5 was deduced from the pseudomolecular ion at m/z 514.3510 [M + Na] + , requiring 6 degrees of unsaturation. The FT-IR spectrum of 1 contained absorption bands characteristic for amide (ν max 3388 cm −1 ), ester (ν max 1726 cm −1 ) and keto (ν max 1659 cm -1 ) functional groups, which were readily confirmed by the carbonyl resonances observed in the 13 C NMR spectrum at δ C 170.3, 172.1 and 212.8, respectively. Four additional olefinic carbon resonances (δ C 134.9, 134.7, 133.2 and 130.6) indicated the presence of two double bounds within 1 ( Table 2). Altogether these data account for five degrees of unsaturation; thus, 1 had to be monocyclic. The 1 H NMR and DEPT-HSQC spectra pinpointed to the presence of 10 methyl groups: one being primary (H 3 -19), seven secondary (H 3 (Table 2). subextract (34.9 mg) and its high salt content did not permit chemical investigation of this fraction. Additionally, no further work was performed on the Kh (6.355 g) subextract, due to its high lipid content, as indicated by its 1 H NMR spectrum ( Figure S14). The Kd subextract was fractionated on a C18-SPE cartridge to yield 12 fractions, which were tested against MRSA and C. albicans. The antimicrobial activity against C. albicans and MRSA was tracked to the later eluting fractions (Frs. 8-10). Fr. 8 was the most active SPE fraction with 100% and 93% inhibition at 50 μg/mL against MRSA and C. albicans, respectively.  Table 2). Altogether these data  Five spin systems (A-E) were identified in the 1 H-1 H COSY spectrum of 1 ( Figure 4A). The 1 H and 13   The COSY fragments (A-E) were linked together by key HMBC cross-peaks allowing the assembly of the 13-membered macrocyclic core structure ( Figure 4A). The segments A and B had to be connected based on the 1 H-13 C correlations of both H-2 and H-5 with the amide carbonyl (C-4). Fragments B and C were easily attached by the long-range couplings between H-5/C-6 and H-7/C-6. The connection of the segments C and D was based on the correlations between H-8/C10 and H3-26/C-10, and vice versa, by H-10/C-8 and H-10/C-26. The macrocyclic ring was fixed by the HMBC correlations of H-2, H2-20 (segment A) and H-12 (segment D) with C-1 carbonyl. Final HMBC correlations from H-12 to C-13, C-14 and C-28; from H3-28 to C-12, C-13 and C-14; and from H-14 to C-12, C-15, C-16 and C-28 readily attached the side chain (segment E) to macrocyclic core at C12, completing the gross structure of 1.
Configuration of versicolide A (1) was determined by DFT-NMR studies and comparison with synthetic model compounds from the literature. Versicolide A (1) contains eight stereocenters, giving as many as 128 possible pairs of enantiomers. Therefore, the problem was simplified by first studying the configuration of the 13-membered lactone ring. The simplified model compounds 1m ( Figure 5A) was used for these calculations. Compound 1m contains five stereocenters. The RR or SS relative configuration of the contiguous C-11 and C-12 was determined using NMR data, namely, the large coupling constant between H-11 and H-12, showing their anti relationship, and the NOESY correlations of H3-27 with H3-28 and H-14, showing the C-27 methyl group and the C-13/C-19 side chain to be gauche oriented (see also Figure 5B). The absolute configuration at this position was assumed to be 11R,12R on biosynthetic grounds, by analogy with closely related fungal metabolites thermolides [34] that are likely to have a related biosynthesis. Therefore, models of the eight possible diastereomers of 1m with the 11R,12R configuration (RRR- The COSY fragments (A-E) were linked together by key HMBC cross-peaks allowing the assembly of the 13-membered macrocyclic core structure ( Figure 4A). The segments A and B had to be connected based on the 1 H-13 C correlations of both H-2 and H-5 with the amide carbonyl (C-4). Fragments B and C were easily attached by the long-range couplings between H-5/C-6 and H-7/C-6. The connection of the segments C and D was based on the correlations between H-8/C10 and H 3 -26/C-10, and vice versa, by H-10/C-8 and H-10/C-26. The macrocyclic ring was fixed by the HMBC correlations of H-2, H 2 -20 (segment A) and H-12 (segment D) with C-1 carbonyl. Final HMBC correlations from H-12 to C-13, C-14 and C-28; from H 3 -28 to C-12, C-13 and C-14; and from H-14 to C-12, C-15, C-16 and C-28 readily attached the side chain (segment E) to macrocyclic core at C12, completing the gross structure of 1.
Configuration of versicolide A (1) was determined by DFT-NMR studies and comparison with synthetic model compounds from the literature. Versicolide A (1) contains eight stereocenters, giving as many as 128 possible pairs of enantiomers. Therefore, the problem was simplified by first studying the configuration of the 13-membered lactone ring. The simplified model compounds 1m ( Figure 5A) was used for these calculations. Compound 1m contains five stereocenters. The RR or SS relative configuration of the contiguous C-11 and C-12 was determined using NMR data, namely, the large coupling constant between H-11 and H-12, showing their anti relationship, and the NOESY correlations of H 3 -27 with H 3 -28 and H-14, showing the C-27 methyl group and the C-13/C-19 side chain to be gauche oriented (see also Figure 5B). The absolute configuration at this position was assumed to be 11R,12R on biosynthetic grounds, by analogy with closely related fungal metabolites thermolides [34] that are likely to have a related biosynthesis. Therefore, models of the eight possible diastereomers of 1m with the 11R,12R configuration (RRR-1m, RRS-1m, RSR-1m, and so on; see Figure 5A) were the object of DFT studies. Following our usual protocol [36], a conformational search on these model compounds was performed by molecular dynamics (see Section 4.7 for details), and for each diastereomer, 8 to 19 conformers with population >1% at 298 K were identified (Table S6). These were used for prediction of the 1 H and 13 C NMR chemical shifts. The isotropic shielding values, calculated at the mPW1PW91/6-311+G(2d,p)/PCM(MeOH) level, were converted to chemical shift using the linear regression parameters determined by Pierens [37] and available on the CHESHIRE website (http://cheshirenmr.info/). The results of these calculations, summarized in Table S6, univocally identified RSS-1m as the stereoisomer best-fitting experimental data, because it showed by far the best RMSD for 1 H and 13 C chemical shifts (0.091 ppm and 2.09 ppm, respectively) and showed a 99.98% probability in the DP4+ analysis [38] (Table S7). Computational results were fully supported by NOESY data. The eight lowest-energy conformers of RSS-1m, accounting for 96% of the conformer population, showed a very similar conformation of the macrocycle and only differed in the conformation of the isobutyl and 2-buten-2-yl groups linked at C-2 and C-12, respectively. All the cross-peaks observed in the NOESY spectrum were fully consistent with this conformation ( Figure 5B). These results defined the 2R,5S,7S,11R,12R configuration of the macrocyclic part compound 1.
The relative configuration of the side chain of versicolide A (1) was assigned as  [35] revealed that wrong values of the coupling constants were reported for 6c in the main paper; the correct values are reported here.
Following our usual protocol [36], a conformational search on these model compounds was performed by molecular dynamics (see Section 4.7 for details), and for each diastereomer, 8 to 19 conformers with population >1% at 298 K were identified (Table S6). These were used for prediction of the 1 H and 13 C NMR chemical shifts. The isotropic shielding values, calculated at the mPW1PW91/6-311+G(2d,p)/PCM(MeOH) level, were converted to chemical shift using the linear regression parameters determined by Pierens [37] and available on the CHESHIRE website (http://cheshirenmr.info/). The results of these calculations, summarized in Table S6, univocally identified RSS-1m as the stereoisomer best-fitting experimental data, because it showed by far the best RMSD for 1 H and 13 C chemical shifts (0.091 ppm and 2.09 ppm, respectively) and showed a 99.98% probability in the DP4+ analysis [38] (Table S7). Computational results were fully supported by NOESY data. The eight lowest-energy conformers of RSS-1m, accounting for 96% of the conformer population, showed a very similar conformation of the macrocycle and only differed in the conformation of the isobutyl and 2-buten-2-yl groups linked at C-2 and C-12, respectively. All the cross-peaks observed in the NOESY spectrum were fully consistent with this conformation ( Figure 5B). These results defined the 2R,5S,7S,11R,12R configuration of the macrocyclic part compound 1.
The relative configuration of the side chain of versicolide A (1) was assigned as (15R*,16S*,17R*) by comparison of NMR data reported for compounds 6a-6d, four diastereomeric synthetic analogues of khafrefungin [35] (Figure 5C): the chemical shift of the carbinol carbon atom C-16 (δ 79.5) of versicolide A (1) and the multiplicity of the corresponding proton H-16 (t, J = 5.6 Hz) matched very well the data reported for the diastereomer 6b, having the RSR configuration.
The relative configuration of the side chain to the macrocycle was studied by DFT prediction of NMR parameters of the whole molecule. The two possible stereoisomers (2R,5S,7S,11R,12R,15R,16S,17R, denoted as 1 in the following text, and 2R,5S,7S,11R,12R, 15S,16R,17S, denoted as tris-epi-1 in the following text) were examined. Rather than performing a conformational search of 1 and tris-epi-1 from scratch (which would be long and difficult because the size and flexibility of 1 would result in a huge number of possible conformers), we took advantage of existing conformational information. Therefore, 20 conformers were generated by linking in a combinatorial approach the 5 lowest energy conformers of RSS-1m (all conformers above 5%, accounting for 88.2% of population) with the four lowest-energy conformers of the RSR side chain (or their respective mirror images) generated in an independent calculation (all conformers above 5%, accounting for 90.1% of population). The two sets of 20 conformers were optimized by DFT at the B3LYP/6-31G(d,p)/SMD level, and no remarkable conformational change during optimization was observed for any of them, thus showing that the macrocyclic part of the molecule does not affect much the conformational space of the side chain, and vice versa, and therefore that the combinatorial approach chosen was appropriate. The 1 H and 13 C NMR chemical shifts of the two stereoisomers were calculated as described above (Table S8), and the results clearly pointed to compound 1 as the correct stereoisomer, in that it showed better RMSD ( 1 H: 0.092 ppm, 13 C: 1.88 ppm) than tris-epi-1 ( 1 H: 0.108 ppm, 13 C: 2.07 ppm), while the DP4+ probability of 1 was calculated as 99.93% (Table S9). Therefore, configuration of versicolide A (1) was fully determined as 2R,5S,7S,11R,12R,15R,16S,17R.
Burnettramic acid A (2) was the main constituent of the highly active SPE Fr.8. Its structure was unambiguously confirmed by comparison of its HRMS, NMR and specific rotation data with those reported in the literature [28,29]. Burnettramic acid A (2) has previously been isolated from a terrestrial Aspergillus sp., A. burnettii [28] and from the co-culture of A. versicolor with A. chevalier [29]. Due to low amounts, no compounds could be purified from the SPE Fr. 10.
The molecular network ( Figure 2) revealed further unannotated clusters with undescribed chemistry; therefore, whenever enough amounts were available, other SPE fractions of the Kd subextract were also considered for compound isolation and for confirmation of the initial dereplication efforts. Compound 3 was isolated as a colorless film from SPE Fr. 7 Table 2) were identical with those of versicomide A, a quinazoline-type metabolite previously reported from a crab-associated A. versicolor sampled from a hydrothermal vent [32], resulting in two possibilities, either compound 3 and versicomide A are the same compounds or they are enantiomers. The specific rotation of 3 ([α] Table 2). The COSY spectrum of 3 indicated the whole proton network to be composed of three spin systems (A-C) as shown in Figure 4B. As discussed above, fragment A was easily deduced from COSY spectrum and J-value analysis to belong to a 1,3,4-trisubstituted aromatic system, including the protons of H-7, H-8 and H-10. The spin system B was assigned to an isopropyl unit associated with a valine unit (H-14 to H 3 -17), while the spin system C was attributed to a sec-butyl moiety of an isoleucine (H-3 to H 3 -21). The three spin systems could be connected via HMBC correlations. The methoxy group H 3 -22 was attached at C-9 of the aromatic ring based on its correlation with C-9. The HMBC cross-peak between H-10/C-12 placed the first amide carbonyl at C-12 that correlated with H-14 of the valine moiety. H-14 further correlated with C-4 (δ C 149.7) and C-1, the second amide carbonyl (δ C 170.6). Final key HMBC correlations between H-3/C-4 concluded the presence of a Val-Ile-2,5-diketopiperazine unit in 3. So far, this structure accounted for 7 degrees of unsaturation. Based on the molecular weight and the remaining degrees of unsaturation, 3 had to contain a quinazolinone moiety concluding the planar structure of 3 that was identical to versicomide A [32]. However, the hypothesis that compound 3 is the enantiomer of (+)-versicomide A required H-3 and H-14 to be on the same side of the ring.

H-3 lacked an NOE cross-peak with H-14, but exhibited a strong NOE with H-15. The NOE between H-3 and H-15 is diagnostic for a cis relationship between H-3 and the isopropyl group of the valine unit at C-14, placing H-3 and H-14 on the opposite sides of the ring.
This means that compound 3 must be a diastereomer of (+)-versicomide A, in which the two amino acids have opposite configuration at the α carbon.
This prompted us to re-examine the original (+)-versicomide A literature [32]. In this paper, the configuration of the chiral centers relied on chiral phase HPLC analyses of the acid hydrolysate of the compound. Based on the retention time comparison, the two chiral amino acid units were identified as D-Val and D-allo-Ile. However, the reported evidence does not seem entirely solid because, as shown in Supplementary Figure S73 of reference [32], the retention times of the individual amino acids are close, and additional peaks that could match different stereoisomers are present in the chromatogram. Curiously, then, the published structure of (+)-versicomide A [32] contains L-Val and L-Ile (see our Figure S30) adding further discrepancy to the configuration of the stereocenters in this compound. Unfortunately, the authors did not report any NOESY data for versicomide A, to allow us to draw a comparison with our data. The opposite (+) specific rotation value to that of versicomide A, and the NOE data indicated that 3 is definitely a new compound, which we called (−)-isoversicomide A. The identical 1D NMR data of 3 and versicomide A in the same solvent (Table S10) are not impossible in principle, even if they are diastereomers, but it is also possible that the relative configuration of (+)-versicomide A also needs to be revised and that they are in fact enantiomers. We were unable to perform additional experiments to identify the configuration of 3, due to minute amounts of compound isolated, and hence, we left the stereocenters of 3 unassigned, as shown in Figure 3.
The SPE Fr.4 and Fr.5 were also subjected to RP-HPLC separation and yielded the known benzodiazepine alkaloids cyclopenol (4) and cyclopenin (5), respectively. Interestingly cyclopenin, which is highly related to cylopenol, was not detected in MN, possibly because of poor ionization. The structures of both compounds were confirmed by comparison of their HRMS and NMR data to the literature data [25,[39][40][41] . These compounds were originally reported from a terrestrial Penicillium sp. [41,42], but later isolated from marine Aspergillus spp. [32,43].

Bioactivity Results of Purified Compounds
Due to low isolation yields, only compound 2 and 4 could be tested in vitro for their antimicrobial activity against C. albicans and MRSA. Compound 2 was inactive against MRSA but inhibited the growth of C. albicans with an IC 50 value of 7.2 µg/mL (positive control nystatin had an IC 50 value of 2.2 µg/mL). This result is coherent with the previously reported activity of 2 towards C. albicans (IC 50 value 0.5 µg/mL, MIC value 1 µg/mL) in the literature [28,29]. No antimicrobial activity was detected for compound 4 at the test concentration of 100 µg/mL. Considering the initial anti-Candida activity of the crude fungal extract, it is possible that burnettramic acid A (2) contributes to the anti-yeast activity of A. versicolor PS108-62.

Discussion
The rapid development of resistance towards currently available antibiotics and antifungals urges the search for novel drugs in new ecological niches [32]. Due to access difficulties, the Arctic deep sea has remained largely underexplored for biodiscovery efforts in comparison to other environments [44]. Yet, several studies on psychrophiles and psychrotrophs have shown great potential for biodiscovery. A total of 352 new metabolites have been reported from organisms living in cold environments, between 2006 and 2016 [9]. In this study, we set out to (i) cultivate microorganisms from Arctic deep-sea sediments, (ii) assess their bioactivity and metabolomes and (iii) to isolate new, bioactive compounds.
The culturable microbial fraction associated with the sampled Arctic deep-sea sediments consisted of 70 bacterial isolates, of which over 50% belonged to the genus Bacillus. Microbial communities associated with deep-sea sediments vary considerably in the available literature. For instance, Jroundi et al. [45] also found Bacillus to be the most abundant genus in Mediterranean deep-sea sediments by both culture-dependent and independent techniques, while several others reported high abundance of Proteobacteria in deep-sea sediments [46]. In the present cultivation-based study, Proteobacteria made up less than 15% of the bacterial isolates. Far fewer studies have looked at the culturable diversity of deep-sea fungi. In most cases, the majority of fungal isolates belonged to genera known from terrestrial environments, i.e., being facultative marine fungi. Here, we were able to recover six isolates belonging to the phylum Ascomycota and one isolate was a basidiomycetous yeast. To the best of our knowledge, we provide the first accounts of cultured Protomyces and Kondoa from deep-sea sediments, while Emericellopsis have previously been isolated from deep-sea sediments of the northern Porcupine Bank in the Atlantic Ocean [47]. Aspergillus spp. are commonly reported in deep-sea sediments all over the globe [48], where they have been described as facultative marine fungi that were originally washed in from the land, but have adapted to thrive under extreme conditions [49,50]. Laboratory experiments demonstrated that spores of A. sydowii can germinate and grow under high hydrostatic pressures that were equivalent to deep-sea pressure of up to 5000 m and temperatures of 5 • C [51].
Our Arctic deep-sea microbial isolates depicted great potential to produce antimicrobial compounds, and were particularly active against Gram-positive human pathogens. Almost 85% of the isolates showed moderate to high anti-MRSA activity, and 71% showed activity against E. faecium. Anti-yeast activity was rare, with only two isolates showing high activity against C. albicans and five isolates exhibiting moderate activity against C. neoformans. Culture media had a considerable effect on the observed bioactivities. The results obtained from the applied OSMAC approach suggest that the nutrient-rich media GYM/GYM+Br as well as GPY/GPY+Br are favoring the production of bioactive metabolites, as crude extracts originating from these media showed the highest antibiotic activities against MRSA and E. faecium. Antibiotic production has frequently been reported in marine microbes when glucose was used as the main carbon source (e.g., [52][53][54]). With regard to the addition of bromide, we did not observe the production of additional brominated compounds in the UPLC-MS profiles of the bioactive extracts through the enrichment of the GYM medium with KBr. However, amongst these extracts were the only ones that showed anticancer and cytotoxic activities. Several studies have previously reported an induction of anticancer compounds after media substitution with bromine salt, as reviewed by Romano et al. [55]; for instance, Aspergillus flavipes PJ03-11 produced two new anticancer compounds, flavichalasine N and O, in a rice medium substituted with NaBr [56]. Hence, also in our study, bromine supplementation may have triggered the production of cytotoxic compounds in some isolates, i.e., Bacillus PS108-67a and Rheinheimera PS108-96. Additionally, the selected fungal isolate A. versicolor PS108-62 displayed differential bioactivities after addition of bromine. While the GPY extracts were devoid of any biological activities, the GPY+KBr extract showed potent antifungal activity against C. albicans. Through molecular networking, we were able to visualize the effect of the culture media on the strain's chemical space and demonstrated that the biosynthesis of specific molecular families was media-dependent. The initial dereplication enabled annotation of a few of these molecular families, but a large portion of the metabolome remained unidentified. Still, the MN-based untargeted metabolomics approach proved to be a powerful approach complementary to bioactivity data for selection and work-up of microbial extracts for their new, bioactive chemical constituents.
Genes coding for type I and type III polyketide synthases (PKSs), nonribosomal peptide synthetases (NRPSs), and terpene synthases (TPSs) are common in fungi, including deep-sea fungi [57,58], being responsible for biosynthesis of a vast diversity of metabolites [17,59]. From deep-sea Aspergillus spp., numerous polyketides, peptides and terpenoids have been reported [59][60][61]. The fusion of different biosynthetic pathways results in enhanced structural diversity, which is often associated with potent bioactive properties [62]. A range of fungal PKS-NRPS hybrids have been reported that are of linear nature (e.g., burnettramic acid A) [28,63], while macrocyclic PKS-NRPS hybrids, resulting from intramolecular Diels-Alder reactions, are less common [62]. The new versicolide A (1) is a unique macrocyclic PKS-NRPS hybrid formed by a polyketide chain linked via leucine. Only very few examples of similar hybrid PKS-NRPS macrolactones have been reported. These include thermolides from Talaromyces thermophilus [34] from the fungal class Eurotiomycetes that also includes the genus Aspergillus; georatusin from Geomyces auratus (class: Leotiomycetes) [64]; metacridamides in Metarhizium acridum (class: Sordariomycetes) [65]; and tricholides from cyanobacteria [66]. In contrast to classical PKS-NRPS hybrids that are linked via C-C bonds, these mixed biosynthesis macrolactones have an ester bond linkage between the polyketide and the non-ribosomal peptide units [34]. Due to integration of different amino acids and post-translational modifications, these molecular families can be very diverse; for instance, 13 different thermolide analogues have been reported with varying amino acid substitutions [67]. A retrospective MN analysis indicated that versicolide A (1) is part of the unannotated cluster B1, which is composed of five nodes. This suggests that there are more versicolide analogues to be described from A. versicolor PS108-62. In the future, we will optimize the fermentation process to pursue isolation of further analogues and determine their bioactivities.
In conclusion, the isolated Arctic deep-sea sediment culture collection provides a great resource for biodiscovery of antimicrobial metabolites, in particular against the Grampositive pathogens, MRSA or E. faecium, while activity against the fungal pathogens was less common among the 77 isolates and only observed in two deep-sea fungi. Furthermore, this study demonstrates that the biotechnological potential of the well-studied fungal genus Aspergillus is far from being exhausted, and in particular, isolates from extreme environments, such as the Arctic deep sea, are a promising source of new antimicrobial metabolites. The MN-based untargeted metabolomics and initial bioactivity results allowed us to prioritize this prolific fungus for a chemical work-up, resulting in the isolation of the new PKS-NRPS macrolide versicolide A (1). During the isolation process, we also isolated and characterized (−)-isoversicomide A (3), a new member of the rare versicomide family. Versicomides are a small group of quinazolines reported from one marine and one terrestrial Aspergillus sp. [32,68]. The present study may suggest a higher chemical and stereochemical diversity and broader distribution of this small, but intriguing class of quinazolines in the fungal kingdom.

General Procedures
Specific rotations of the compounds were measured on a monochromatic light source in MeOH at 20 • C on a Jasco P-2000 polarimeter (Jasco, Pfungstadt, Germany). The NMR spectra were acquired on a Bruker AV 600 spectrometer (600 and 150 MHz for 1 H-NMR and 13
PCR amplification was performed using primers ITS1F and ITS4R for fungi [71] and primers Eub27f and 1387r for bacteria [72], in a total reaction volume of 25 µL, consisting of 1 µL of template DNA, 1 µL of each primer (concentration: 10 µM), 12.5 µL of Dream Taq Master mix (ThermoFisherScientific, Schwerte, Germany), and 9.5 µL DNA free water (ThermoFisher Scientific, Schwerte, Germany). The following PCR protocol was used for amplification of fungal DNA: initial denaturation for 85 s, at 94 • C, followed by 30 cycles of denaturation of DNA, at 95 • C, for 35 s, primer annealing, at 55 • C, for 55 s, and elongation for 3 min, at 72 • C. A final elongation step for 10 min, at 72 • C, completed the PCR. For bacterial DNA, PCR was conducted as follows: 30 cycles of each 30 s of DNA denaturation, at 92 • C, primer annealing, at 55 • C, and elongation, at 72 • C. A final elongation step for 5 min, at 72 • C, completed the PCR. The correct length of PCR products was checked by gel electrophoresis on a 1% agarose gel run for 20 min at a voltage of 120 V in 1× TBE buffer, taking 100 bp plus DNA ladder (Thermo Scientific™ GeneRuler™, Sunnyvale, CA, USA) as length standard. PCR products were submitted for Sanger sequencing to the company LGC Genomics GmbH (Berlin, Germany), applying the same primers as used for PCR. The obtained sequences were checked for quality and trimmed using ChromasPro V1.33 (Technelysium Pty Ltd., South Brisbane, Australia). The sequences were then compared to the NCBI Genbank (https://www.ncbi.nlm.nih.gov/ genbank/, accessed on 11 November 2022) using the Nucleotide BLAST function. Obtained isolate sequences were deposited in GenBank, assigned accession numbers are given in Table S1. All isolates were cryopreserved, at −80 • C, using the Microbank TM system (PRO-LAB Diagnostics, Richmond Hill, Canada).

Extraction and Compound Isolation
After 21 days of growth, the rice medium cultures (500 plates) of A. versicolor PS108-62 were cut into pieces, homogenized by an Ultra-Turrax at 19000 rpm and extracted with EtOAc (2 × 400 mL per 20 plates). The extracts were collected in a separatory funnel and washed twice with 200 mL of Milli-Q ® water to remove salts and water-soluble compounds. The organic phases were combined and evaporated under reduced pressure (200 bar, 150 rpm, at 40 • C). The crude extract (6.51 g) was subjected to a modified Kupchan partition scheme [33] to yield three subextracts, namely, n-hexane (Kh, 6.355 g), DCM (Kd, 217.9 mg) and aqueous MeOH (Km, 34.9 mg). A portion of Kd subextract (87.9 mg) was fractionated by C18 SPE cartridge (45 mL/5000mg); each fraction of the SPE was eluted with 50 mL of solvent starting from 100% water (Fr.0) to 100% methanol (Fr.10) with 10% increments of methanol in water. The last fraction (Fr.11) was eluted with a MeOH/DCM mixture (1:9). Fr.9 was subjected to RP-HPLC on a Luna 5µ C18 column (250 × 4.6 mm) with a gradient of MeCN in water from 30:70 to 20:80 in 8 min then to 0:100 in 6 minutes at a flow rate of 1 mL/min to yield an impure fraction. This fraction was re-chromatographed on a X select HSS T3 column (High-Strength Silica C18, 2.5 µm, 150 × 4.6mm) to yield the new compound versicolide A (1, 0.5 mg, t R

UPLC-QToF-MS Analysis
Chromatograms were acquired with a UPLC I-Class System coupled to a Xevo G2-XS QToF Mass Spectrometer (Waters®, Milford, MA, USA) using a modified protocol described previously [73]. Samples were injected at a concentration of 0.1 mg/mL with an injection volume of 1 µL. The separation was performed using a HSS T3 column, at 40 • C, and at a flow rate of 0.6 mL/min. A gradient of H 2 O + 0.1% formic acid (FA) for the mobile phase A, and MeCN + 0.1% FA, for the mobile phase B. The solvent system was as follows: from 99:1 to 0:100 in 11.5 min followed by washing and reconditioning of the column over 3.5 min. The mass range was set from m/z 190 to 1200, the capillary voltage was 3.0 kV, the cone and the desolvation gas flow were at 50 and 1200 L/h, respectively. The source temperature was 150 • C and the desolvation temperature 550 • C, and the sampling cone and source offset were set at 40 V and 80 V, respectively. Data-dependent acquisition (DDA) mode was used to select the five most intense precursor ions of each MS spectrum (excluding the peaks corresponding to the isotopic contribution of the 13 C). Those ions were then fragmented using a collision energy (CE) ramp: low CE from 15-25 eV to high CE of 40-60 eV. Solvent (MeOH) and non-inoculated media extracts were injected under the same conditions and used as controls.

Molecular Network
The raw data were converted to mzXML file format using MSConvert (version 3.0.19059, Vanderbilt University, Nashville, TN, USA). For classical molecular networking of the small-scale culture extracts (strain B. licheniformis PS108-67a: https://gnps.ucsd.edu/ ProteoSAFe/status.jsp?task=a81c111412ec4f6482acb3f72b49139b and strain A. versicolor PS108-62: https://gnps.ucsd.edu/ProteoSAFe/status.jsp?task=6a874cb754af46bfba3cae3 d80aa6de2 the data was filtered by removing all MS/MS fragment ions within +/− 17 Da of the precursor m/z. MS/MS spectra were window filtered by choosing only the top 6 fragment ions in the +/− 50 Da window throughout the spectrum. The precursor ion mass tolerance was set to 0.5 Da and a MS/MS fragment ion tolerance of 0.5 Da. A network was then created where edges were filtered to have a cosine score above 0.7 and more than 5 matched peaks. Further, edges between two nodes were kept in the network if and only if each of the nodes appeared in each other's respective top 10 most similar nodes. Finally, the maximum size of a molecular family was set to 100, and the lowest scoring edges were removed from molecular families until the molecular family size was below this threshold. The spectra in the network were then searched against GNPS' spectral libraries. The library spectra were filtered in the same manner as the input data. All matches kept between network spectra and library spectra were required to have a score above 0.7 and at least 5 matched peaks. Solvent and media blanks in group G6 were filtered before networking. The networks were visualized in Cytoscape ® (version 3.5.1, Institute for SystemsBiology, Seattle, WA, USA).

Computational Details
The eight diastereomeric model compounds 1m (Figure 5A) of the macrocyclic moiety of versicolide A (1) were generated using the Builder module in the INSIGHT II/Discover package (BIOVIA, 5005 Wateridge Vista Drive, San Diego, CA 92121, USA). The conformational space of each diastereomer was explored using high-temperature molecular dynamics (MD) [74]. A series of 10 ns MD runs at 600 K in the CFF91 force field were performed for each diastereomer, using the INSIGHT II/Discover package, and the coordinates were saved every 5 ps, and the resulting 2000 structures from each run were minimized in the same force field.
All the unique (non-duplicated) structures from MD within 5 kcal/mol from the lowest-energy conformer of each stereoisomer were used as input structures for density functional theory (DFT) calculations (Table S6). DFT calculations were performed using the program Gaussian 16 (Revision C.01, Gaussian Inc., Wallingford CT, USA). All conformers were optimized by DFT at the B3LYP/6-31G(d,p) level of theory, and the energy of the structures within 3.5 kcal/mol from the lowest-energy conformer were re-evaluated at the same level of theory using the SMD continuous model for the solvent, MeOH. All the conformers populated more than 1% (roughly those within 2.0 kcal/mol from the lowest-energy conformer) were used for prediction of the 1 H and 13 C NMR chemical shifts. The Cartesian coordinates of these conformers are available as Supplementary Information in a file called "compounds1m.txt".
The isotropic shielding values were calculated using the Gauge Invariant Atomic Orbitals (GIAO) method at the mPW1PW91/6-311+G(2d,p) level and the PCM model for the solvent (MeOH). Average isotropic shieldings for each stereoisomer were calculated using Boltzmann statistics (T = 300 K). Average isotropic shieldings were converted to chemical shift using the linear regression parameters available in the literature for this level of theory [37] (Table S7).
The 1 H and 13 C NMR chemical shifts of the two alternative 2R,5S,7S,11R,12R,15R,16S, 17R and 2R,5S,7S,11R,12R,15S,16R,17S stereoisomers of the full molecule of versicolide A (1) were predicted using the same DFT methods and level of theory as described above (Tables S8 and S9). The Cartesian coordinates of the conformers used for prediction are available as Supplementary Information in a file called "compounds1.txt".

Bacterial and Fungal Bioactivity Assays
The crude extracts were tested against methicillin-resistant Staphylococcus aureus DSM 18827, Escherichia coli DSM 1576, Pseudomonas aeruginosa DSM 1128, Klebsiella pneumoniae DSM 30104, Acinetobacter baumannii DSM 30007, Enterococcus faecium DSM 20477, the yeastlike fungus Cryptococcus neoformans DSM 6973 and the yeast Candida albicans DSM 1386. All bacterial test strains were cultivated in TSB medium (1.2% tryptic soy broth, 0.5% NaCl), except for E. faecium, which was cultivated in M92 medium (3% trypticase soy broth, 0.3% yeast extract, pH 7.0-7.2). M186 medium (1% glucose, 0.5% peptone from soybeans, 0.3% yeast extract, 0.3% malt extract) was used for cultivation of C. albicans and C. neoformans. Overnight cultures of the test organisms were prepared and diluted to an optical density (600 nm) of 0.01-0.03. To prepare the assay, the crude extracts and fractions (20 mg/mL DMSO stock solution) were dissolved in medium and transferred into a 96-well microtiter plate, and 200 µL of the cell suspension cultures was added to each well. The final test concentration was 100 µg/mL. The inoculated microplates were incubated for 5 h, at 37 • C and 200 rpm (E. faecium without shaking), and C. neoformans for 7 h, at 28 • C and 200 rpm. The detection of inhibitory effects was performed by adding 10 µL of a resazurin solution (0.3 mg/mL in phosphate-buffered saline) to each well and incubating again for 5-60 min before the fluorescence signal (560 nm/590 nm) was read by the microplate reader (Tecan Infinite M200) as described [75]. For E. faecium, the pH indicator bromocresol purple was used. For C. neoformans the absorbance at 600 nm was measured to detect cell growth inhibition. The resulting values were compared to a positive antibiotic control (chloramphenicol for S. aureus, E. coli, K. pneumoniae, ampicillin for E. faecium, polymyxin B for P. aeruginosa, doxycycline for A. baumannii and amphotericin B for C. neoformans) and a negative control (no compound) on the same plate. For determination of IC 50 values of pure compounds, a dilution series was prepared and tested as described for crude extracts and fractions. The IC 50 values were calculated using MS Excel (Microsoft Corporation) as the concentration showing 50% inhibition of the viability on the basis of a negative control (no compound) and compared with the positive control (nystatin).

Anticancer and Cytotoxicity Bioactivity Assays
Four human cancer cell lines, namely, malignant melanoma cell line A375 (CLS, Eppelheim, Germany), colon cancer cell line HCT-116 (DSMZ, Braunschweig, Germany), lung carcinoma cell line A549 (CLS, Eppelheim, Germany), breast cancer line MDA-MB-231 (CLS, Eppelheim, Germany) and the non-cancerous human keratinocyte line HaCaT (CLS, Eppelheim, Germany) were used to evaluate the activity of the crude extracts by monitoring the metabolic activity using the CellTiterBlue Cell Viability Assay (Promega, Mannheim, Germany). HaCaT cells were cultivated in RPMI medium, A549 and MDA-MB-231 cells in DMEM: Ham's F12 medium (1:1) supplemented with 15 mM HEPES and A375 and HCT116 cells in DMEM medium supplemented with 4.5 g/L of D-Glucose and 110 mg/L of sodium pyruvate. All media were supplemented with L-Glutamine, 10% fetal bovine serum, 100 U/mL of penicillin and 100 mg/mL of streptomycin. Cell cultures were maintained, at 37 • C, under a humidified atmosphere and 5% CO 2 . The cell lines were transferred every 3 or 4 days into fresh medium. For the experimental procedure, cells were seeded in 96-well plates at a concentration of 10.000 cells per well in RPMI. After 24 h incubation, the medium was removed, and 100 µL of the test sample, adjusted to final concentrations of 100 µg/mL by diluting in growth medium, was added to the cells. Doxorubicin, as a standard therapeutic anticancer drug, was used as a positive control. Following addition of the crude extracts, plates were incubated for 24 h, at 37 • C. Afterwards, the assay was performed according to the manufacturer's instructions (Promega, Mannheim, Germany) and fluorescence was measured using the microplate reader Tecan Infinite M200 at excitation 560 nm and emission of 590 nm.

Conclusions
The microbiota associated with Arctic deep-sea sediments still remain underexplored for their biotechnological potential. Here, we isolated and cultivated bacteria and fungi from the Arctic deep-sea floor (−2454 m) and investigated their bioactivity. The chemical composition of the two bioactive, pre-selected strains were profiled by molecular network based untargeted metabolomics, which led to the prioritization of A. versicolor PS108-62 for downstream chemical work. From this strain, we isolated a new hybrid PKS-NRPS macrolactone, versicolide A, a new quinazoline (−)-isoversicomide A, as well as three known compounds. Our investigation confirms, one more time, the vast potential of polar sediment associated microorganisms for marine biodiscovery efforts, especially when combined with modern methodology such as metabolomics and broad-panel bioactivity screening. This study also highlights the fact that the heavily studied fungal genus Aspergillus possesses an enormous talent for producing new compounds with intriguing chemical scaffolds and bioactivity. Our future studies will continue unlocking the biomedical potential of polar deep-sea microorganisms.