The optical rotations were obtained using a JASCO J-815 spectrometer (JASCO International Co. Ltd., Tokyo, Japan). 1D and 2D NMR spectra were measured using a Bruker AVⅢHD spectrometer (Bruker Co. Ltd., Bremen, Germany) at 600 MHz for ¹H and 150 MHz for ¹³C in DMSO-d ₆ (δ _H 2.49 and δ _C 34.9). High-resolution mass data and advanced Marfey’s analysis were measured using a Waters Xevo G2-XS QTof mass spectrometer (Waters Co., Manchester, United Kingdom). Column chromatography was performed on macroporous adsorption resins (Diaion 4006; Nankai Chemical Co. Ltd., China) and the CombiFlash^® Rf system fitted with an ODS flash column (RediSep Rf C₁₈ flash column, 130 g). Semi-preparative HPLC was performed using an Agilent 1200 series liquid chromatography system (quaternary pump, autosampler, and diode array detector) using a Reprosil-Pur Basic-C₁₈ column (5 μm, 250 × 10 mm).

Nocardia sp. XZ19_369 was isolated from the nodules of sea buckthorn collected from Tibet, China. Their draft genomes were sequenced on an Illumina HiSeq platform (Illumina, San Diego, CA, United States), assembled using SPAdes V3.13.1, and submitted to the National Center for Biotechnology Information (NCBI) database (accession no. JACVED000000000 for XZ19_369). The phylogenetic tree (Supplementary Figure S1) was constructed using autoMLST (Alanjary et al., 2019).

Nocardia sp. XZ19_369 grown on the ISP2 medium (0.4% glucose, 0.4% yeast extract, 1% malt extract, and 2% agar) was inoculated into 100 mL of the ISP2 liquid medium contained in a 500-mL Erlenmeyer flask and cultured at 28°C for 2 days as the seed culture. For large-scale fermentation, 50 mL of the seed culture was inoculated into 10 Erlenmeyer flasks (5 L), each containing 1 L of the sterile YMS medium (0.4% yeast extract, 1% malt extract and 0.4% starch soluble), and then cultivated at 28°C for 10 days.

The culture broth (20 L) was centrifuged at 4,000 rpm for 30 min to obtain the mycelium and supernatant; the supernatant was analyzed by LC–MS using an ACQUITY UPLC^® CSH™ C₁₈ column (Waters, 1.7 μm, 2.1 × 100 mm, at 30°C) eluted with the gradient method (from 10% to 100% MeCN/H₂O, containing 0.1% formic acid at a flow rate of 0.3 mL/min for 10 min) (Supplementary Figure S2) and chromatographed over a macroporous adsorption resin column using the EtOH/H₂O gradient elution (0%, 30%, 50%, 80%, and 100%; each 5 L) to afford five corresponding fractions (Fr.1: 2350.1 mg, Fr.2: 1576.5 mg, Fr.3: 923.5 mg, Fr.4: 323.6 mg, and Fr.5: 375.2 mg).

Fr.3 (923.5 mg) was then fractionated using an ODS flash column eluted with a gradient MeCN−H₂O solution (1–10 min, 5%→25% MeCN; 10–50 min, 25%→50% MeCN; 50–70 min, 50%→100% MeCN; 70–80 min, and 100% MeCN; 15 mL/min) to yield four fractions (Fr.3–1−Fr.3–4). Fr.3–3 was further purified by reversed-phase semipreparative HPLC (Reprosil-Pur Basic-C₁₈ column; 5 μm, 250 × 10 mm, 2.5 mL/min, and 28% MeCN−H₂O in 0.1% trifluoroacetic acid) to yield 1 (5.2 mg) and 2 (3.0 mg).

Each compound 1–2 (50 μg) was dissolved in a 100 μL of 6 N HCl and heated at 110°C for 24 h. After heating, the hydrolysates were divided into two parts and dried under the N₂ flow. The dried hydrolysates and each standard amino acid (L-Ala, L-Val, L-Ser, and L-Asp) were dissolved in 30 μL of the 0.1 M NaHCO₃ solution. To each reaction vial, 30 μL of L-FDAA (N _α-(2,4-dinitro-5-fluorophenyl)-L-alaninamide, 1% solution in acetone) was added and heated at 40°C for 1 h (Kiyonaga Fujii, 1997a; Kiyonaga Fujii, 1997b). After cooling at room temperature, 30 μL of 0.1 M HCl was added to each vial. The mixtures were diluted with 500 μL of MeOH. The other part of hydrolysates and standard amino acids were derived with D-FDAA in the same manner. The aforementioned derivatives were analyzed with LC–MS using an ACQUITY UPLC^® CSH™ C₁₈ column (Waters, 1.7 μm, 2.1 × 100 mm, at 30°C) eluted with the gradient method (from 15% to 60% MeCN/H₂O with a 5% isocratic MeOH containing 1% formic acid at a flow rate of 0.3 mL/min for 30 min) and isocratic elution method (from 12% MeCN/H₂O with a 5% isocratic MeOH containing 1% formic acid at a flow rate of 0.3 mL/min for 30 min). The configurations of amino acids were confirmed by comparison with the authentic standards. The L-FDAA/D-FDAA derivatized authentic amino acids provided retention times (t _R, min): L-Ala-L-FDAA and L-Ala-D-FDAA (7.80 and 9.22 min, m/z 342), L-Val-L-FDAA and L-Val-D-FDAA (11.52 and 14.18 min, m/z 370), L-Asp-L-FDAA and L-Asp-D-FDAA (6.22 and 6.76 min, m/z 386), and L-Ser-L-FDAA and L-Ser-D-FDAA (9.99 and 11.32 min, m/z 358). The presence of Ala, Val, Asp (Asn), and Ser residues in 1 and 2 was also assigned to the L-configuration.

In order to simplify the calculation, the bismethylated guanidino fatty acid (MGFA) was replaced with acetic acid, which was not impacted with the accuracy of the result (Daranas and Sarotti, 2021). Conformational analysis was performed using OpenBabel (O'Boyle et al., 2011) with a genetic algorithm at the MMFF94 force field, with energies within the 3.0 kcal/mol energy threshold (Spartan14. Wavefunction, 2019). The conformers were optimized using Gaussian 16 (Gaussian Inc.) (M. J. Frisch, 2010) at the M062X/6-311+G (d, p) level in vacuum. At the same level, the frequencies were calculated to provide the relative thermal free energies (ΔG), which are used to calculate the equilibrium populations. NMR chemical shift calculations for those optimized conformers within their Boltzmann distribution (>1%) were performed using the GIAO method at the mPW1PW91/6-311+G (d, p) level in DMSO with the PCM model. The calculated chemical shifts of conformers for 1–2 were averaged in terms of their relative Gibbs free energy and the Boltzmann distribution theory. Finally, the calculated NMR chemical shift values were averaged according to Boltzmann distribution for each conformer and fit to the experimental values by linear regression. DP4+ probability analysis was performed according to the reported methods (Grimblat et al., 2015).

Antibacterial assays were conducted in flat bottom, sterile 96-well plates (Corning, America) in triplicate, using a broth microdilution protocol (Cockerill, 2012). Candida albicans (ATCC 10231), Staphylococcus aureus (ATCC 29213), Enterococcus faecium (ATCC 35667), Candida tropicalis (ATCC 1369), Escherichia coli (ATCC 25922), Klebsiella pneumoniae (ATCC 700603), and Pseudomonas aeruginosa (ATCC 27853) were used as test strains. Each bacterial culture (100 μL) containing ca. 5 × 10⁴ CFU was added to each well of 96-well plates. Nocaviogua A and B (1 and 2) were dissolved in DMSO. A measure of 1 μL of each work solution of compounds 1–2 and the corresponding positive drugs (1.28–0.00125 mg/mL) were added to each well and incubated at 30°C for 24 h for determining MIC values.

The cytotoxic effects of compounds 1–2 were evaluated against uveal melanoma cells 92-1 (UM92-1), non-small cell lung cancer (NCI-H2170), and three breast cancer cell lines (SK-BR-3, MDA-MB-231, and MDA-MB-453) by the 3-(4, 5-dimethyl-2-thiazolyl)-2, 5-diphenyl-2H tetrazolium bromide (MTT) method. After cells (ca. 3×10³ cells/200 μL/well) were seeded in a 96-well plate and cultured in a 5% CO₂ incubator at 37°C for 24 h, 1 and 2 were added to each well for 48 h incubation. Then, the medium was removed, and the MTT solution was subsequently added to each well and maintained for 4 h. After removing the supernatant, 150 μL DMSO was added to dissolve purple crystals. Ultimately, the absorbance value was read at 570 nm using a microplate reader (Elx800, BioTek Instruments, Inc., United States). The assays were performed four times, and bleomycin was used as the positive control.

Nocaviogua A (1) was purified as white amorphous powder. HRESIMS of 1 disclosed a molecular formula of C₄₁H₆₄N₁₁O₁₁S using a quasi-molecular ion at m/z 918.4514 {[M + H]⁺, calcd 918.4507, with 16 double-bond equivalents (DBEs)}. The ¹H NMR and ¹³C NMR data (Table 1) recorded in DMSO-d ₆ revealed nine amide and/or ester carbonyl carbons (δ _C 165.1–172.4), nine alkenyl carbons (δ _C 98.7–155.2 and δ _H 7.16–5.28), and eleven NH (δ _H 6.83–8.95), accounting for 14 DBEs and requiring that 1 incorporates two rings. The existence of one alanine (Ala) moiety in 1 was confirmed due to the presence of the spin system NH-CH-CH₃ revealed by the ¹H-¹H COSY sequential correlations, together with the HMBC correlations from Ala₁-2 (δ _H 4.20) and Ala₁-3 (δ _H 1.16) to Ala₁-CO (δ _C 171.5). One valine (Val) unit in 1 was inferred by the presence of a spin system of -NH-CH-CH(CH₃)₂, which is indicated by the ¹H-¹H COSY correlations, and then confirmed by the HMBC correlations from Val₃-2 (δ _H 4.16) and Val₃-3 (δ _H 1.97) to Val₃-CO (δ _C 171.3). The existence of one serine (Ser) residue was identified based on the spin system of -NH-CH-CH₂ deduced from ¹H-¹H COSY correlations, as well as the key HMBC correlations from Ser₄-2 (δ _H 3.84) and Ser₄-3 (δ _H 3.50) to Ser₄-CO (δ _C 170.3) (Figure 1). One asparagine (Asn) or aspartic acid (Asp) moiety was inferred by the presence of the spin system of -NH-CH-CH₂, which is indicated by the ¹H-¹H COSY correlations, and then confirmed by the HMBC correlations from Asp₆/Asn₆-3 (δ _H 2.75, 2.88) to Asp₆/Asn₆-CO₁ (δ _C 170.3) and Asp₆/Asn₆-CO₄ (δ _C 172.4), and Asp₆/Asn₆-2 (δ _H 4.84) to Asp₆/Asn₆-CO₄ (δ _C 172.4). One glycine (Gly) residue was discovered due to the presence of the spin system of -NH-CH₂ revealed by the ¹H-¹H COSY correlation, together with the HMBC correlations from Gly₇-2 (δ _H 3.49, 4.07) to Gly₇-CO (δ _C 167.6). The existence of one avionin, which includes three amino acid residues, namely, Avi₁, Avi₂, and Avi₃ moieties, can be revealed by the spin system of NH-CH-CH₂ in Avi₂ and NH-CH = CH in Avi₈, together with the key HMBC correlations from Avi₂-2 (δ _H 4.17) to Avi₂-CO (δ _C 171.2), from Avi₅-3 (δ _H 2.74, 3.86) to Avi₅-CO (δ _C 169.1) and Avi₅-C₂ (δ _C 60.9), and from Avi₅-NH (δ _H 6.83) to Avi₅-CO (δ _C 169.1) and Avi₅-C₂ (δ _C 60.9). The double-bond geometry of the Avi₈ moiety was established as Z based on the small coupling constant (7.2 Hz) and the ROESY correlation between Avi₈-(CH)₁ and Avi₈-(CH)₂. Additionally, an N-terminal MGFA was confirmed by the spin system of -CH-CH-CH-CH-CH₂-CH₂-CH-CH-CH₂-CH₂-CH₂-CH₂-NH and two -NH-CH₃ units, and key HMBC correlations from MGFA-2 (δ _H 5.99) to MGFA-CO (δ _C 165.1), and from MGFA-15 (δ _H 2.72), MGFA-16 (δ _H 2.72), and MGFA-13-NH (δ _H 7.32) to MGFA-C₁₄ (δ _C 155.2). The alkenyls in MGFA were all E arrangements based on the large coupling constant (equal or greater than 15.0 Hz).

The connection sequence of these nine groups was established by HMBC correlations from Ala₁-NH (δ _H 8.06) to MGFA-CO (δ _C 165.1), from Avi₂-NH (δ _H 7.91) to Ala₁-CO (δ _C 171.5), from Val₃-NH (δ _H 7.49) to Avi₂-CO (δ _C 171.2), from Ser₄-NH (δ _H 7.40) to Val₃-CO (δ _C 171.3), from Avi₅-NH (δ _H 6.83) to Ser₄-CO (δ _C 170.3), from Asp₆/Asn₆-NH (δ _H 8.50) to Avi₅-CO (δ _C 169.1), from Gly₇-NH (δ _H 8.82) to Asp₆/Asn₆-CO₁ (δ _C 170.3), and from Avi₈-NH (δ _H 8.95) to Gly₇-CO (δ _C 167.6), suggesting that nine units were connected to yield a sequence of MGFA-Ala₁-Avi₂-Val₃-Ser₄-Avi₅-Asp₆/Asn₆-Gly₇-Avi₈. The two DBEs were accounted for realizing the dicyclic structure of 1. The key HMBC correlations from Avi₂-3 (δ _H 3.01, 2.11) to Avi₅-2 (δ _C 60.9) established the connectivity of Avi₂ and Avi₅. Additionally, considering the chemical shift of Avi₈-1 (δ _C 98.7) and Avi₅-3 (δ _C 41.0), and the requirement of unsaturation, the connection of (Avi₅-3)-S-(Avi₈-1) was deduced as found in nocavionin and microvionin (Wiebach et al., 2018). Based on the combination of previous information and the molecular formula, one Asp unit was assigned in 1. Thus, the planar structure of 1 was determined, as shown in Figure 1. Notably, 1 contains an odd number of nitrogen atoms and an even number of hydrogen atoms, which suggests that the MGFA unit with the positively charged functional group is found in goadvionins (Kozakai et al., 2020).

The configurations of the amino acid residues in 1 were further determined by the advanced Marfey’s method. The acid hydrolysates of 1 were derivatized with L-FDAA and D-FDAA. By comparing the retention times with standards (Supplementary Figures S5–S8; Supplementary Table S1), the absolute configurations of Ala₁, Val₃, Ser₄, and Asp₆ were all established as S-configurations, leaving two chiralities in Avi₂ and Avi₅ residues to be clarified.

The remaining two configurations in Avi₂ and Avi₅ are proving to be a challenge to resolve. However, unlike the NMR spectra of nocavionin and microvionin in D₂O and H₂O, respectively, which showed the presence of conformer isomers (Wiebach et al., 2018), the NMR spectra of nocaviogua A (1) in DMSO-d ₆ did not exhibit any obvious signals indicating different conformers. This allows us to use the GIAO NMR calculation to determine their stereochemistry. In order to solve the structure of complex, large, and highly flexible molecules using readily available computational resources, the MGFA was replaced with acetic acid and the four possible diastereomers 1a–1d (Figure 2) were evaluated. Conformational searches were performed at OpenBabel with the MMFF94 force field and an energy cutoff value of 3.00 kcal/mol. Subsequently, the obtained conformers were optimized by the DFT method at the M062X/6-311+G (d, p) level in vacuum, and the NMR chemical shifts were calculated at the PCM/mPW1PW91/6-311+G (d, p) level. Accordingly, the calculated carbon chemical shifts for 1a showed the highest similarity toward the experimental values with the smallest corrected mean absolute error (CMAE, 1a: 1.5113 vs. 1b–1d: 2.0405, 1.9820, and 1.8977 in ¹³C NMR data and 1a: 0.1397 vs. 1b–1d: 0.2068, 0.1813, and 0.2073 in ¹H NMR data, respectively) values (Supplementary Tables S2–S3), particularly for Avi₂ and Avi₅ moieties (Figure 3). In addition, DP4+ analysis based on NMR data provided 100% probability for the isomer 1a (Supplementary Table S4), indicative of the S, S configuration for Avi₂ and Avi₅ moieties in 1. Thus, the whole structure of compound 1 was fully determined and designated as nocaviogua A (1).

Nocaviogua B (2), white powder, was determined as C₄₁H₆₆N₁₂O₁₁S, based on the HRESIMS peak at m/z 935.4766 [M + H]⁺ (calcd for C₄₁H₆₇N₁₂O₁₁S, 935.4773), implying 15 DBEs. The ¹H NMR and ¹³C NMR data (Table 1) for 2 that were recorded in DMSO-d ₆ revealed nine amide and/or ester carbonyl carbons (δ _C 165.5–172.7) and nine alkenyl carbons (δ _C 99.3–158.3 and δ _H 7.19–5.38), accounting for 14 DBEs and requiring only one ring incorporation into 2. The ¹H and ¹³C NMR (DMSO-d ₆) data for 2 were closely similar to those with 1 except for three signals of NH (Asn₆-4-NH₂: δ _H 7.33 and 6.89 and Ser₄-NH₂: δ _H 8.11) in 2, which is more than 1, indicating that the amide bond between Val₃ and Ser₄ was opened, and the Asp unit in 1 was replaced with the Asn residue in 2. This hypothesis was further confirmed by HMBC and ¹H-¹H COSY correlations (Figure 1). Based on the coupling constant and the ROESY correlation, the geometry of the double bond in the Avi₈ moiety was established as Z and the alkenyls in MGFA were all in the E arrangement, which were the same as in 1. Considering the number of nitrogen atoms and hydrogen atoms, the MGFA unit in 2 was also with the positively charged functional group as in 1.

The absolute configurations of Ala, Val, Ser, and Asn moieties were determined as L by the advanced Marfey’s method (Supplementary Figures S11–S14; Supplementary Table S1). The configurations in Avi₂ and Avi₅ units of compound 2 were also determined using GIAO NMR calculations by evaluating the four possible diastereomers 2a–2d (Supplementary Figure S27). Consequently, the calculated carbon chemical shifts of 2a showed the highest probability to the experimental values with the smallest CMAE (2a: 1.4284 vs. 2b–2d: 1.6495, 1.8486, and 1.6907 in ¹³C NMR data and 2a: 0.2186 vs. 2b–2d: 0.2303, 0.2732, and 0.2248 in ¹H NMR data, respectively) values (Supplementary Tables S13–S14). In addition, DP4+ analysis based on NMR data with 100% probability for the isomer 2a (Supplementary Table S15), assigned Avi₂ and Avi₅ units as S, S configurations, respectively, are identical to those determined in compound 1. Thus, the whole structure of compound 2 was fully determined and designated as nocaviogua B (2).

The two compounds were evaluated for their antibacterial activity on Candida albicans (ATCC 10231), Staphylococcus aureus (ATCC 29213), Enterococcus faecium (ATCC 35667), Candida tropicalis (ATCC 1369), Escherichia coli (ATCC 25922), Klebsiella pneumoniae (ATCC 700603), and Pseudomonas aeruginosa (ATCC 27853) by MIC values. Regrettably, both compounds 1 and 2 showed no antibacterial effect at concentrations up to 128 μg/mL.

The cytotoxicity of compounds 1-2 was assayed against uveal melanoma cells (UM92-1), non-small cell lung cancer (NCI-H2170), and three breast cancer cell lines (SK-BR-3, MDA-MB-231, and MDA-MB-453). Compound 1 exhibited weak cytotoxicity against UM92-1, NCI-H2170, and MDA-MB-231, exhibiting an inhibition rate of approximately 15%, 30.0%, and 31.1%, respectively, at a concentration of 50 μM (Figure 4). Compound 2 shows no obvious cytotoxic activity at the same concentration.