Study of bicyclomycin biosynthesis in Streptomyces cinnamoneus by genetic and biochemical approaches

The 2,5-Diketopiperazines (DKPs) constitute a large family of natural products with important biological activities. Bicyclomycin is a clinically-relevant DKP antibiotic that is the first and only member in a class known to target the bacterial transcription termination factor Rho. It derives from cyclo-(l-isoleucyl-l-leucyl) and has an unusual and highly oxidized bicyclic structure that is formed by an ether bridge between the hydroxylated terminal carbon atom of the isoleucine lateral chain and the alpha carbon of the leucine in the diketopiperazine ring. Here, we paired in vivo and in vitro studies to complete the characterization of the bicyclomycin biosynthetic gene cluster. The construction of in-frame deletion mutants in the biosynthetic gene cluster allowed for the accumulation and identification of biosynthetic intermediates. The identity of the intermediates, which were reproduced in vitro using purified enzymes, allowed us to characterize the pathway and corroborate previous reports. Finally, we show that the putative antibiotic transporter was dispensable for the producing strain.


1.1.
Kinetics of bicyclomycin production in MP5 medium (Suppl. In red, culture supernatant of the Δbcm::aphII/pJWe20 strain (mutant devoid of the bcm cluster, harbouring the bcmA gene under the control of its native promoter cloned into pOSV668). In green, culture supernatant of the Δbcm::aphII/pOSV668 (mutant devoid of the bcm cluster, harbouring the empty vector). Blue number corresponds to cIL (2) identified by mass spectrometry. It presents an m/z of 227 in positive mode. This compound has the same retention time, m/z ratio and fragmentation pattern as a cIL (2) authentic standard. Culture supernatants were analysed after 7 days. Various integrative plasmids containing bcm cluster intact or bearing an in-frame deletion of only one of each gene were introduced in this mutant strain. Strain and plasmid are indicated over each chromatogram. The deleted gene is in bold. Compounds 2 to 7 correspond to the major bicyclomycin-related products accumulating in the supernatant. These compounds were also detected in trace amounts by LC/MS analysis of the supernatant of the strain containing the intact bcm cluster (see Supplementary data sheet 1). Culture supernatants were analysed after 7 days.

Hypothetical bicyclomycin pathway deduced from the in vivo analysis of intermediates
After proposing an order for the action of the tailoring enzymes, we tried to exploit the fragmentation data to make hypotheses on the precise reaction that these enzymes catalysed.
Previous work 1 had shown that during the fragmentation of bicyclomycin (1), the first fragment to depart (m/z difference: 74) corresponds to the terminal diol of the leucyl moiety. The loss of this fragment is never observed for any of the products accumulated by the bcmC and bcmG deletion mutants. We can therefore propose that the two enzymes BcmC and BcmG add the two distal hydroxyl groups at the leucine lateral chain. However, there is no indication on the precise position modified by each of these two enzymes, hence the two possibilities proposed in Suppl. Fig. S4 (products 6a or 6b).
MS2, MS3 and MS4 fragmentation patterns of 7, the product of the action of BcmE on cIL, indicates that the DKP scaffold is unmodified in this product and that the hydroxylation occurs on the side chain of either the leucyl or the isoleucyl moiety, far away enough from the DKP moiety not to modify the fragmentation pattern in all daughter species. As BcmC and G are supposed to act on the extremity of the leucine lateral chain, the most probable position of BcmE-catalysed hydroxylation is hence on the side chain extremity of the isoleucyl moiety.
Concerning the last unattributed reactions (ether bridge formation and two hydroxylations), the two enzymes BcmB and BcmD should be involved. The deletion of bcmB led to the accumulation of product 5 (274 g/mol,

1.4.
In vitro characterisation of the cIL tailoring pathway 1.4.1 Suppl. Fig. S5 Supplementary Figure S5. HPLC (UV, 220 nm) chromatograms of in vitro enzymatic assays on cIL (2) (method B). (A) with BcmC, (B) with BcmE, (C) with BcmG. Chromatograms are stacked from bottom to top, the lowest corresponding to the initial state (t=0), then to reaction times of 1, 10 and 60 minutes. The position of the peaks corresponding to cIL (2) and product 7 is marked on the chromatogram; their identity was verified by ESI-MS.

NMR analyses
The proposed structures of the bicyclomycin biosynthetic pathway intermediates were confirmed by NMR analysis. The products 2, 4, 5, 6 and 7 were obtained in large quantity by the in vitro enzymatic reaction scale-up, purified on semi-preparative HPLC and analyzed by NMR. The 1 H, 13 C NMR assignments of the characterized compounds are given below. Products 3 and 1 were obtained in lower amounts and extracts could not be purified to homogeneity. However NMR signals of the products could be assigned using 2D correlation experiments.  (Table S6). The conformational and configurational analysis in the Cβ-Cγ fragment relied on the measurement of 2,3 JCH coupling constants involving Hβ protons and Cγ, Cδ1 and Cδ2 carbons, together with NOE analysis (Table S6). The NMR data were consistent with a predominant rotamer around Cβ-Cγ bond, with a χ2 angle (Cα-Cβ-Cγ-O) around +60°, and the configuration of Cγ atom was shown to be R.  (Table S7). Identical colour fillings are used for homologous genes. Several types of gene organisation are found in Actinobacteria, with bcmH upstream of bcmA, downstream of bcmG or absent. In Proteobacteria, bcmH is found downstream of bcmG and the order of the homologues of bcmE and bcmF is changed. Fig. S23

MCpseq1 TGGCACCCAGCCTGCGCGAG
Sequencing and verification of pMC1 to pMC15 (odd number)

MCpseq2 ATAAGCCCTACACAAATTGG
Sequencing and verification of pMC1 to pMC11 (odd number)

Sequencing and verification of pMC1
MCpseq4 TGCACCGCGTCGCCCGGCAG

Sequencing and verification of pMC3
MCpseq5 CCCGCCCGGACACGGGACAG

Sequencing and verification of pMC3
MCpseq6 CCGGAGTCCGCCGACCGCTT

Sequencing and verification of pMC5
MCpseq7 CCGGCCGGCTGTTCTCCGTG

Sequencing and verification of pMC7
MCpseq8 CGTTGTCGACGACCGTCGTC

Sequencing and verification of pMC7
MCpseq9 GACCGAACGCCTCGACCGAC

Sequencing and verification of pMC9
MCpseq10 AACCTGGGTGACGCGTTCCG Sequencing and verification of pMC11

Supplementary Table S5. Primers used in this study
Restriction sites added and used for cloning are underlined.

Supplementary Table S6
Dominant rotamers for the leucyl chain of product 5 along with their relative configurations.

Supplementary Table S7
Dominant rotamers for the leucyl chain of product 4 along with their relative configurations.
NOE correlations are classified into strong (s), medium (m) and weak (w) intensities.

Deletion of the bcm cluster
The DNA region upstream of the bcm cluster was amplified by PCR with S. cinnamoneus genomic DNA as a template and oligonucleotides JWm23 and JWc10. The resulting PCR fragment was isolated as a HindIII/EcoRI fragment. The aphII kanamycin resistance gene was amplified using pOSV408 as template and oligonucleotides JWm19 and JWm20. The aphII gene was purified as an EcoRI fragment. The DNA region downstream of the bcm cluster was amplified by PCR with S.
cinnamoneus genomic DNA as template and oligonucleotides JWm24 and JWc25. This region was purified as an EcoRI/BcuI fragment. These three fragments were cloned together into HindIII/BcuIdigested suicide vector pOSV400 to create pJWm07. Effective cloning of the fragments in pJWm07 was controlled by sequencing.
Sequencing of pJWm07 revealed the presence of two mutations in the insert (Supplementary Figure S24). The first one was a point mutation G1021T in the intergenic region upstream of orf-2.
The second one was a deletion of nucleotides 6073 to 6374. This deletion included the last bp of orf+2, the intergenic region between orf+2 and orf+1 and the first 43 bp of orf+1. For unknown reasons, we were unable to obtain the PCR product of the downstream region without this deletion.
As these two mutations are in genes which are not involved in bicyclomycin biosynthesis, did not impair our bicyclomycin BCG characterization, we nevertheless used this mutated plasmid pJWm07 to construct the Δbcm strain.
pJWm07 was transferred to S. cinnamoneus by conjugation according to 13 . Exconjugants were selected for kanamycin resistance. Hygromycin-sensitive and kanamycin-resistant clones were then screened. Their genomic DNA was extracted and the replacement of the cluster by the kanamycin resistance cassette was verified by PCR using oligonucleotides JW-RT9 and JW-RT12.
Supplementary Figure S24. The genetic organization of the Δbcm mutant strain. Unwanted mutations induced during the strain construction are marked in red. The exact gradients were set as follows: For the purification of 7, the product of the reaction catalysed by BcmE with cIL (c 2) alone as a substrate, the linear gradient was set as to reach a 30% solvent B concentration in 30 minutes.
For the purification of compound 6, the product of the reaction catalysed by BcmE and BcmC with cIL (2) alone as a substrate, a first linear gradient was set to reach 10 % of solvent B in 5 minutes, followed by a second linear gradient to reach 20% solvent B in 20 minutes.
For the purification of compound 5, as the reaction mixture did not contain DMSO, it was loaded and directly washed with 100% solvent A during 5 minutes. Compound 5 was then eluted with a linear gradient set to reach 20% solvent B in 20 minutes.
The purification of 4, the product of the reaction catalyzed by BcmB with 5 as the substrate, was performed on an Hypercarb 150 x 10 (5 µM) column (Thermo Fisher Scientific). The reaction mixture was loaded and washed with 100 % solvent A for 5 minutes, then eluted with a linear gradient reaching 25% solvent B in 25 minutes. The collected peak contained a mixture of 4 and 5, was lyophilized, dissolved in water and purified on a LiChroCART 250 x 10 Purospher STAR RP-18e (5 µm) column (Merck) using the same loading and elution conditions as before.
The purification of 3, the product of the reaction catalysed by BcmD with 4 as the substrate, was performed on an ACE Excel 3 C18-PFP (150 x 4.6 mm) column (Advanced Chromatography Technologies). The flow was at 0.6 ml/min and the gradient as the same as for purification of 4.
All the products obtained were lyophilized and their identities were confirmed by NMR and mass spectrometry analyses.

NMR Analyses
Samples of compounds cIL (2) (10.6 mM), 7 (5 mM), 6 (20.9 mM), 5 (4.1 mM) and 4 (< 1 mM) were prepared in DMSO-d6 (Eurisotop, Saint-Aubin, France) in 3 or 5 mm NMR tubes (corresponding volumes 0.2 or 0.5 mL, respectively). NMR experiments were recorded on a 500 MHz Bruker Avance III spectrometer equipped with a 5-mm inverse TCI cryoprobe incorporating a Z-gradient coil. Spectra were recorded at 298.6 K. All data were processed and analyzed with Bruker TopSpin 3.2 program. 1 H and 13 C resonances were assigned via the analysis of one-dimensional 1 H, one-dimensional 13  coupling constants were also useful for the analysis of oxygen-substituted two-carbon fragments as they depend on the dihedral angle between the proton and 13 C-attached oxygen: 2 JC,H coupling constant is large (typically 4-7 Hz) when an oxygen substituent on a carbon atom is gauche to the