New Rufomycins from Streptomyces atratus MJM3502 Expand Anti-Mycobacterium tuberculosis Structure Activity Relationships

Abstract

Four new rufomycins, 1–4, named rufomycins 56, 57, 58, and 61, respectively, exhibiting new skeletal features, were obtained from Streptomyces atratus strain MJM3502, and fully characterized. Compounds 1 and 2 possess a 4-imidazolidinone ring not previously encountered in this family of cyclopeptides, thereby resulting in a [5,17] bicyclic framework. The in vitro anti-Mtb potency of 3 and 4 is remarkable, with MIC values of 8.5 and 130 nM, respectively.

Graphical Abstract

Previous investigations of the Streptomyces atratus strain MJM3502, which was selected from ca. 7,000 actinomycete cultures via a high-throughput screening campaign, led to the re-identification of a series of rufomycins and expanded their potential as anti-tuberculosis (TB) leads.^1–6 In general, they possess a 21-membered heptapeptide ring, created via the non-ribosomal peptide synthetase (NRPS)^7,8, and the rufomycins 4–7 contain an additional in-chain cyclized 2-piperidinone.^5,6 A previous study expanded a series of rufomycins and preliminary anti-Mycobacterium tuberculosis (Mtb) structure-activity relationships (SAR).^5,6 Continued efforts to explore structural variants with possible anti-TB potential from MJM3502⁵ have now resulted in the isolation and characterization of four new albeit minor rufomycins, each possessing novel carbon skeletons in the rufomycin class [rufomycins 56, 57, 58, and 61 (1–4)].

The new skeletal variation in rufomycins 56 (1) and 57 (2) involving a 4-imidazolidinone moiety, leads to an unprecedented [5,17] bicyclic framework. Rufomycin 58 (3) is the first rufomycin with a 5-hydroxyl-5-methyl-2-pyrrolidinone moiety, whereas rufomycin 61 (4) contains a glycine residue instead of the alanine unit typical for all known rufomycins. All structures were established by detailed spectroscopic analysis, advanced Marfey’s analysis, and considered NRPS biosynthetic pathways. As 3 with an in-chain cyclic 2-pyrrolidinone showed a promising MIC value of 8.5 nM, its structural modification expands the possibility for pharmacophore modification of the rufomycin class of potential anti-TB leads.

Rufomycin 56 (1) was assigned a molecular formula of C₅₆H₇₉N₉O₁₂ from the (+)-HRESIMS ion at m/z 1070.5905 [M + H]⁺ and the ¹³C NMR data, requiring 22 double bond equivalents (DBEs). Its IR and NMR spectra indicate that 1 is a peptide, with similarity to the rufomycin class.⁵ Analysis of its ¹H and ¹³C NMR spectra (Table S1) showed three amide NH resonances (δ_H 7.80, 8.07, and 8.51) and one amine NH signal at δ_H 2.09, seven carbonyl signals resonating between δ_C 169 and 176 ppm, seven AA hydrogen/carbon resonances (δ_H 3.2~5.1/δ_C 46~65), and one ethoxyl group (OCH₂CH₃: δ_H 4.02–4.11, m, δ_C 61.6; OCH₂CH₃: δ_H 1.16, t, J = 7.1 Hz, δ_C 14.9). The ion at m/z 911.5 observed during MS/MS analysis (Figure S16), indicating loss of 159 amu fragment from the [M + H]⁺ molecular ion at m/z 1070.6, revealed the presence of an O-ethyl Leu AA residue. Three ions at m/z 840.4 (loss of 71 MW, Ala), 632.4 (loss of 208 MW, mNO₂Tyr), and 505.3 (loss of 127 MW, NMeLeu) that sequentially followed the precursor ion at m/z 911.5 indicated the presence of an Ala-mNO₂Tyr-NMeLeu motif. Characteristic resonances for epoxy retro-prenyl tryptophan (ErpTrp) and trans-crotonylated glycine (TrcGly) AA residues^2,5,8 in 1 were also observed from its NMR spectra. The information above revealed that 1 contained seven AA residues common to rufomycins, whereas the distinctly different ¹H and ¹³C NMR data as well as one new amine NH signal (δ_H 2.09) indicated that 1 is a very different rufomycin. Detailed analysis of the COSY/TOCSY cross-peaks (Chart S1) confirmed all seven AA skeletons via their scalar spin–spin coupling patterns. Presence of the AA sequence ErpTrp-NMeLeu3-mNO₂Tyr-Ala-NMeLeu2-OEtLeu1 could be assigned via the HMBC correlations from the α-hydrogens to the carbonyl carbons of the adjacent residues (Chart S1). The TrcGly (AA7) and NMeLeu2 (AA5) residues were linked by the amine NH, forming a long spin coupling fragment of AA5-H-2–H₂-3–H-4 (H₃-1′)–H-5–NH–AA7-H-2–H₂-3–H-4–H-5–H₃-6 based on the COSY/TOCSY correlations. This was confirmed by the HMBC correlations from AA5-H-5 (δ_H 4.76, brs) to AA7-C-1 (δ_C 175.8) and C-2 (δ_C 59.2). The 4-imidazolidinone motif was, thus, confirmed by connecting the TrcGly-NMeLeu2 motif with the ErpTrp residue through the peptidic nitrogen atom. This was further corroborated by the HMBC correlations from AA1-H-2 (δ_H 5.71, t, J = 7.7 Hz) to AA7-C-1 and AA5-C-5 (δ_C 71.5). Collectively, 1 was assigned as the first rufomycin with transannular cross-linkage, possessing a [5,17] bicyclic scaffold and named rufomycin 56 following the recently consolidated naming system.⁶

All seven AA residues in 1 were assigned as l-configured by using the advanced Marfey’s analysis (Figure S17) and based on the reasoning that they share a common biosynthetic origin with the known rufomycins.^5,9–11 The ROESY correlations (Figure 2) of AA5-H-5/AA1-H₂-3 and AA5-H₃-1′/AA7-H-2 revealed that AA7-H-2 and AA5-H-5 were trans-oriented, in combination with the assigned l-configured TrcGly residue (AA7-C-2S), indicated the absolute configuration of AA5-C-5 as R. The Newman projection formulae along the AA5-C-4–C-5 bond in 1 (Figure 2) showed that the two bulky groups were placed with an anti-relationship. The ~0 Hz value of J_4,5 indicated that H-4 and H-5 are in a gauche relationship, as a large coupling constant would be expected between two hydrogens with an anti-relationship.^12,13 Accordingly, AA5-C-4 was assigned an R absolute configuration. The AA1-C-2′ S absolute configuration of the chiral carbon in the epoxide ring was tentatively assigned based on biosynthetic consideration.^4,5,7,8 Collectively, the structure of 1 was assigned as shown.

Figure 2.

Newman projection formulae along the AA5-C-4–C-5 bond for partial structure of 1.

Rufomycin 57 (2) had a molecular formula of C₅₆H₇₇N₉O₁₂ based on the (+)-HRESIMS data and ¹³C resonance count. Near identical NMR (Table S1) and MS/MS data (Figure S26) of 2 compared to 1 suggest a very close structural relationship between them. Next, the ¹H and ¹³C NMR resonances of all seven AA residues were assigned and then connections between the underlying nuclei established via COSY, TOCSY, and HMBC correlations (Figures S20, 21, and 23). This established the [5,17] bicyclic skeleton analogous to that of 1. The characteristic signal resonances for AA5-CH-5 at δ_C 72.7 and δ_H 4.95 (brs) confirmed the presence of the 4-imidazolidinone in 2. The only difference between 1 and 2 was the presence of a methoxyl group in 2 instead of the ethoxyl group in 1. The HMBC correlation from OCH₃ (δ_H 3.72, s) to AA6-C-1 (δ_C 174.5) proved that the methoxyl group was attached to this carbonyl, which was consistent with the observed subtle substituent chemical shift effects at AA6-C-1 (Δδ_C 482 ppb) between 1 and 2, while the other carbons exhibited minor Δδ_C values <150 ppb (Table S2). The relative configurations of all stereocenters of 2 were, thus, shown to be identical with those of 1 based on their similar NMR and J value patterns (Tables S1 and S2).

Scheme 1 delineates a possible biosynthetic pathway for 1 and 2, with the co-isolated rufomycins 6/7^5,6 as their common precursor. Their acid-promoted dehydration can yield intermediate a, which produces b via nucleophilic attack of the AA7 nitrogen towards the iminium function. Intermediate b can then undertake an acid-triggered rearrangement, and the AA6–AA7 peptide bond hydrolyses to yield c. This intermediate can then readily transform into d through a nucleophilic attack of the AA1 nitrogen onto the iminium function at AA5-C-5. Esterification of the AA6-C-1 carboxy group likely occurred during the isolation/purification process, eventually generating 1 and 2.

Scheme 1.

Plausible Biosynthetic Pathways for Compounds 1 and 2

An alternative biosynthetic pathway for 1 and 2 starting with the co-isolated rufomycin 10^5,6 can also be proposed (Scheme S1, see details in Supporting Information).

The molecular formula of rufomycin 58 (3) was determined as C₅₃H₇₃N₉O₁₃, based on the (+)-HRESIMS and ¹³C NMR data, and demanded 22 DBEs. The ¹H and ¹³C NMR spectra (Table S1) of 3 contained the typical signal patterns for Ala, mNO₂Tyr, TrcGly, and ErpTrp. Additionally, seven amide carbonyl groups and a characteristic methylene to lower frequencies at δ_H −0.36 and 1.46 were observed, suggesting 3 to be a rufomycin analogue.⁵ The functionalities mentioned above and the 21-membered ring of rufomycin accounted for 21 DBEs, indicating the presence of an additional ring. This was deduced as a 5-hydroxyl-5-methyl-2-pyrrolidinone, formed between the AA5 and the nitrogen of the adjacent AA6, as verified by the HMBC correlations (Scheme 2) from AA5-NCH₃ to C-2 (δ_C 58.3); from AA5-H-2 to C-1 (δ_C 172.0), C-4 (δ_C 88.4), and AA-4-C-1 (δ_C 174.3); from AA5-H₃-5 to C-3 and C-4; and from AA6-H-2 (δ_H 4.14) to AA5-C-2. The remaining AA sequence and the position of the heptacyclic ring connection were determined by the COSY/TOCSY data and the HMBC correlations from the α-hydrogens to the carbonyl carbons of the adjacent residues (Figures S30, 31, and 33). The AA5-HO-4 group was α-oriented based on the ROESY correlations between H₃-5 and H-2 (Figure S34). The absolute l-configurations of the AA residues were determined using advanced Marfey’s analysis (Figure S37). The assignment of AA1-C2′S absolute configuration at the epoxide ring followed biosynthetic assumptions.^4,5,7,8

Scheme 2.

Proposed Biosynthetic Pathway and HMBC Correlations (Red Arrows in c) of the 5-hydroxyl-5-methyl-2-pyrrolidinone moiety of 3

One remarkable property of 3 was evident from HPLC analysis, which exhibited three peaks, 3A–C (Figure 3I). While initially considered to be three different compounds, and isolated individually, their ¹H NMR and MS spectra (Figures S38 and 39) were nearly identical, suggesting that 3A–C are three species of 3 (conformers, hemiaminal anomers) that can interconvert and are sufficiently stable at RT for preparative separation. Indeed, re-injection of each material collected as a single peak gave identical three peak patterns (Figure 3II), indicating that the underlying three molecular species form a thermodynamic equilibrium. This was consistent with the three observed pairs of characteristic low frequency resonances (Figure S39).

Figure 3.

I) HPLC chromatogram of 3; II) HPLC chromatograms for 3A–C HPLC condition: 55% acetonitrile in H₂O; flow rate 2.8 mL/min; detector with 254 nm.

Compound 4 (rufomycin 61) shared the molecular formula C₅₃H₇₃N₉O₁₂ with 3, as gleaned from the (+)-HRESIMS and ¹³C NMR data. Analysis of the NMR spectra (Table S1) verified that 4 is closely related to rufomycins 4–6,^5,6 possessing the same 6-hydroxyl-5-methyl-2-piperidinone formed from the carbons of AA5 and the nitrogen of adjacent AA6 as shown by COSY and HMBC data (Figures S42 and S44). A pair of doublets (δ_H 3.79 and 4.47, both d, J = 17.5 Hz) assigned by HSQC to a methylene group (Figure S43), together with the absence of the methyl doublet in rufomycins 4–6, suggested that 4 possessed a Gly residue. The HMBC correlations (Figure S44) from AA4-H₂-2 (δ_H 3.79 and 4.47) to AA4-C-1 (δ_C 169.2) and AA3-C-1 (δ_C 173.5) corroborated the presence and position of a Gly residue. The absolute configurations of the other 6 AAs were determined as being l based on advanced Marfey’s analysis (Figure S48) and biosynthetic consideration. The α-configurations for both the hydroxyl and the methyl groups in the 6-hydroxyl-5-methyl-2-piperidinone ring were determined by the ROESY cross-peaks of AA5-H-2/H-4 and H-5/H-4 and H₃-1′ (Figure S45). Thus, the structure of 4 was assigned as shown and named rufomycin 61.

The biosynthesis of both compounds 3 and 4 almost certainly involves A-domain promiscuity, as documented: “Relaxed specificity is not uncommon, particularly for hydrophobic amino acids: a number have been reported in Bacillus spp.”¹⁴ In the case of compound 3, the adenylation domain picks up norvaline instead of the normally encoded isoleucine (only a CH₃ difference) and inserts that in the heptapeptide. An example of this misreading occurs in polymyxin E1, where the 7^th amino acid is Leu, Ile, or Nva.¹⁵ Exactly which cytochrome P-450 variant is responsible for the oxidation of the C-4 methylene to the ketone is a matter of conjecture (e.g., propyl oxidation as in Scheme 2, a to b). Subsequent nucleophilic attack of the nitrogen atom in AA6 upon the AA5-C-4 carbonyl group produced the 5-hydroxyl-5-methyl-2-pyrrolidinone. Similarly, in compound 4, the A-domain picks up glycine instead of the normally encoded alanine and inserts that in the heptapeptide.

A systematic evaluation of the anti-mycobacterial activities of 1, 3, and 4 (Table 1) used previously reported methods^3,5,16 and included the following determinations: MIC versus M. tuberculosis H37Rv (ATCC 27294), M. abscessus (ATCC 19977), and M. avium (ATCC 15769); MBC against M. tuberculosis H37Rv, acquired on days 7 and 14; cytotoxicity to mammalian cells; and binding affinities (K_D) with both caseinolytic protein C1 (ClpC1) N-terminal domain (NTD) and full length protein (FL) by surface plasmon resonance (SPR). Interestingly, compound 1 failed to show anti-mycobacterial activity or strong binding at the concentrations tested. In contrast, 3 showed remarkable inhibitory activity against M. tuberculosis, M. abscessus, and M. avium with MIC values of 8.5, 1200, and 300 nM, respectively. Both 3 and 4 showed tight binding affinities against both ClpC1-NTD and FL as well as time-dependent bactericidal properties, consistent with previous rufomycin studies.⁵ Moreover, the cytotoxicity of 1, 3, and 4 was evaluated in a mammalian cell line (Vero cells; ATCC CRL-81), indicating a high selectivity of 3 and 4 against mycobacteria (3, SI = 10,200; 4, SI > 77).

Table 1.

Biological Data of 1, 3, and 4: in vitro Activity against Mycobacteria (μM) and Target Binding by SPR (K_D, μM)

Cpd	Mtb H37Rv MIC	Vero cell IC50	Mtb SIc	M. abscessus MIC	M. avium MIC	day 7 MBC	day 14 MBC	KD NTD	KD FL
1	>10	>100	NCf	>10	>10	NTd	NTd	NBe	2.1
3 g	0.0085	87	10,200	1.2	0.30	>10	0.058	0.17	0.21
4	0.13	>10	>77	0.59	4.3	>10	2.4	0.23	0.28
RMPa	0.028	>100	>3,570	>4	0.16	0.24	0.12	NAd	NAd
INHb	0.46	NTd	NCf	>8	>8	0.99	4.0	NAd	NAd

^aRifampin (RMP) as positive control.

^bIsoniazid (INH) as positive control.

^cSelectivity index (SI) = IC₅₀/MIC.

^dNT, not tested and/or NA, not applicable.

^eNB, no binding observed.

^fNC, not calculated.

^gContaining the three-species (3A-C) equilibrium at RT (see text).

In summary, this study established the structures of four new rufomycins covering three new skeletons. Compounds 1 and 2 with a [5,17] bicyclic skeleton failed to show anti-Mtb activity, suggesting that the unaltered macrocyclic skeleton of the rufomycins is essential for their anti-Mtb activity. The new in-chain cyclic 2-pyrrolidinone in 3 was associated with strong anti-Mtb activity, equal to or greater than that of the 2-piperidinone rufomycins that are known to exhibit already high in vitro potency. It is interesting that compound 4 was a single diastereomer in the 6-hydroxyl-5-methyl 2-piperidinone ring, corresponding to rufomycin 6 stereochemistry. The results shed further light on post-translational structural modifications in rufomycins and inform semi-synthetic efforts for SAR optimization of this class of potential anti-TB leads.

Figure 1.

Structures of 1–4.