Overexpression of a type III PKS gene affording novel violapyrones with enhanced anti-influenza A virus activity

Background Type III polyketide synthases (PKSs) are simple homodimer ketosynthases that distribute across plants, fungi, and bacteria, catalyzing formation of pyrone- and resorcinol-types aromatic polyketides with various bioactivities. The broad substrate promiscuity displayed by type III PKSs makes them wonderful candidates for expanding chemical diversity of polyketides. Results Violapyrone B (VLP B, 10), an α-pyrone compound produced by deepsea-derived Streptomyces somaliensis SCSIO ZH66, is encoded by a type III PKS VioA. We overexpressed VioA in three different hosts, including Streptomyces coelicolor M1146, Streptomyces sanyensis FMA as well as the native producer S. somaliensis SCSIO ZH66, leading to accumulation of different violapyrone compounds. Among them, S. coelicolor M1146 served as the host producing the most abundant violapyrones, from which five new (2–4, 7 and 12) and nine known (1, 5, 6, 8–11, 13 and 14) compounds were identified. Anti-influenza A (H1N1) virus activity of these compounds was then evaluated using ribavirin as a positive control (IC50 = 112.9 μM), revealing that compounds 11–14 showed considerable activity with IC50 values of 112.7, 26.9, 106.7 and 28.8 μM, respectively, which are significantly improved as compared to that of VLP B (10) (IC50 > 200 μM). The productions of 10 and 13 were increased by adding P450 inhibitor metyrapone. In addition, site-directed mutagenesis experiment led to demonstration of the residue S242 to be essential for the activity of VioA. Conclusions Biological background of the expression hosts is an important factor impacting on the encoding products of type III PKSs. By using S. coelicolor M1146 as cell factory, we were able to generate fourteen VLPs compounds. Anti-H1N1 activity assay suggested that the lipophilic nature of the alkyl chains of VLPs plays an important role for the activity, providing valuable guidance for further structural optimization of VLPs. Electronic supplementary material The online version of this article (10.1186/s12934-018-0908-9) contains supplementary material, which is available to authorized users.

substrates, number of condensation as well as cyclization manner and thus have impacts on their product selectivity [11]. A lot of mutagenesis studies with the aim of broadening substrate specificity have been carried out in plant-derived type III PKSs. For example, the S338V variant of chalcone synthase (CHS) from Scutellaria baicalensis produced octaketides SEK4/SEK4b from eight molecules of malonyl-CoA instead of condensing 4-coumaroyl-CoA with three molecules of malonyl-CoA to generate naringenin chalcone as did the wild type CHS [12]. Substitution of N222 with Gly in octaketide synthase (OKS) from Aloe arborescens led to accumulation of a novel C 20 decaketide SEK15 in addition to the C21 heptaketide chalcone that is produced by the wild type OKS [13]. The L214I variant of Vitis vinifera stilbene synthase (VvSTS) bears an increased substrate binding pocket and a decreased cyclization pocket compared with those in the wild type enzyme, resulting in production of shortchain polyketides with improved efficiency but absence of long-chain polyketides; conversely, the sizes of both pockets in the T197A variant were increased, thus leading to generation of five new polyketides which are not produced by the wild type VvSTS [14].
In contrast, only a handful of mutagenesis studies have been reported for prokaryotic type III PKSs, among which the active-site cavity-forming residue Y224 in Sg-RppA from Streptomyces griseus [15] and Sc-RppA from S. coelicolor [16] has been studied the most. In both enzymes, Y224 was demonstrated to be important for starter substrate selection, but Sc-RppA showed a higher tolerance towards certain amino acid changes of Y224 than Sg-RppA. Mutants of Y224 were thus generated, which preferentially recognize unnatural acyl-CoA such as acetyl-CoA, acetoacetyl-CoA, hexanoyl-CoA and benzoyl-CoA instead of malonyl-CoA as starter substrate [15,16]. Another example is the mutagenesis of Gcs from S. coelicolor, which disclosed H261 and M274 are critical in controlling the substrate specificity and/or catalytic efficiency, as the H261A and M274A variants were capable of producing significantly increased amount of triketide pyrones in comparison to the wild-type Gcs [17].
Violapyrones (VLPs) are a group of α-pyrone compounds with antibacterial and anticancer activities [18][19][20]. Previously, we activated the VLP biosynthetic gene cluster via deletion of the global regulatory gene wblA so in deepsea-derived Streptomyces somaliensis SCSIO ZH66, leading to isolation of VLP B (10) (Fig. 1) [18]. This cluster is composed of a type III PKS gene vioA and a negative regulatory gene vioB. By inactivation of vioB, another four VLP compounds (VLP A, J, C and H) were obtained, which were subjected to evaluation of anti-MRSA (methicillin-resistant Staphylococcus aureus, MRSA) activity, demonstrating that the length of the alkyl side chains of VLPs played an essential role for the anti-MRSA activity [18].

Phylogenetic analysis of VioA and distribution of vio cluster
To better understand the function of VioA, phylogenetic analysis was performed to with other characterized bacterial type III PKSs. As shown in Fig. 2, VioA belongs to the B2-2 clade [2], which preferentially uses short-and medium-chain (C 2 -C 12 ) acyl-CoA as starter. VioA is closest to Cpz6 from the caprazamycin biosynthetic gene cluster [21], and they are clustered with DpyA and Gcs, which are proposed to recognize both CoA-and ACP-tethered β-keto acids from branchedchain or straight-chain fatty acid metabolism as starters, and to generate pyrones by lactonization of a linear polyketide intermediate [2]. In contrast, the members of the other subclade in B2-2 use malonyl CoA as both starter and extender unit to give scaffolds of pyrones and resorcinols by lactonization or Claisen-, aldol-type cyclization, respectively. We further mined the vioAB locus from other Streptomyces genomes in Genbank, and found another 21 Streptomyces genomes harboring vioAB homologous loci (Additional file 1: Table S3). Notably, most of them are located in a linear plasmid, probably contributing to their horizontal gene transfer during evolution.

Overexpression of vioA in different Streptomyces strains
Type III PKSs capture acyl-CoA substrates from primary metabolism. Considering the variety of the acyl-CoA pools in different biological backgrounds, the VLP gene cluster was overexpressed in three different hosts, including the general heterologous expression host S. coelicolor M1146 [22], the marine-derived Streptomyces sanyensis FMA [23] as well as the native producer S. somaliensis SCSIO ZH66 [18,24]. To get rid of the negative regulatory function of vioB [18], the vioA gene was put under the control of the constitutive promoter P gapDH followed by introduction into different hosts as described in the materials and methods section. HPLC analysis of the fermentation broths showed that in addition to VLP B (10), several other VLPs compounds were also accumulated in M1146/pWLI807 (Fig. 3i) and ZH66/pWLI807 (Fig. 3iii), conversely, only compound 1 was accumulated in FMA/ pWLI807 (Fig. 3v), indicating the significant influence of the expression hosts on the products. Their relative yields in each host were indicated in Additional file 1: Figure S1.

Identification of the accumulated VLPs in the overexpression strains
From the large scale fermentations of the overexpression strain M1146/pWLI807, compounds 1-14 were isolated and identified via detailed NMR spectroscopic analysis.

Anti-H1N1 activity of VLPs
Before evaluating the anti-H1N1 activity of compounds 1-14, the cytotoxicity of compounds 1-14 in MDCK cell was evaluated by MTT assay [26]. The results in Table 3 showed that compounds 1-14 exhibited no significant cytotoxicity and CC 50 value for compounds 1-14 were more than 1400 μM. Compounds 1-14 were evaluated for their anti-H1N1 activity by using CPE inhibition assay [27]. As shown in Table 3, compounds 11 and 13 showed moderate anti-H1N1 activities with IC 50 values of 112.7 and 106.7 μM, respectively, which is comparable to that of the positive control ribavirin (IC 50 = 112.9 μM); delightedly, compounds 12 and 14 exhibited stronger anti-H1N1 activities than ribavirin up to fourfold, with the IC 50 values of 26.9 and 28.8 μM, respectively. In contrast, compounds 1-10 were inactive against H1N1 virus up to the concentration of 200 μM. Comparing the structures and bioactivities of these compounds, we proposed that the length and polarity of the alkyl side chains at C-3 and C-6 play essential roles for the antiviral activity, in which the one containing longer alkyl side chain with lower polarity gives better activity.

Site-directed mutagenesis of VioA
The broad substrate promiscuity of VioA makes itself an excellent candidate for further enzyme engineering to generate diverse VLPs. We next set out to further investigate the biosynthetic potentials of VioA in vitro. However, no soluble VioA was obtained after exploring different conditions (data not shown). Therefore, we turned to probe its function by expressing different versions of sitemutated vioA in S. coelicolor M1146. Firstly, we did multiple sequence alignment of VioA with selected reported type III PKSs (Additional file 1: Figure S16). With the purpose to probe the substrate promiscuity and/or to improve catalytic efficiency of VioA, I174 and L190 (corresponding to T197 and I214 in VvSTS, respectively), as well as Y229 and S242 (corresponding to H261 and M274 in Gcs, respectively) were substituted with Ala, Ile, Ala and Ala, respectively. Structure modeling was simultaneously performed as described in the Materials and Methods section to help understanding the underlying mechanism. As shown in Additional file 1: Figure S17A, both I174A (ii) and Y229A (iv) displayed severe defection on VLPs production, supporting their important roles in substrate binding (I174) and cyclization (Y229), respectively (Additional file 1: Figure S17B). No change was observed for L190I (Additional file 1: Figure S17A, iii), indicating this substitution probably had no influence on the cyclization pocket (Additional file 1: Figure S17B). Conversely, VLPs production was completely abolished in S242A (Additional file 1: Figure S17A, v), demonstrating S242 to be essential for the activity of VioA, which was consistent with its position being close to the cyclization pocket in the structural model (Additional file 1: Figure S17B).

P450 inhibitor increased the yields of anti-H1N1 VLPs by blocking side-chain oxidation
The above bioassay results indicated that the presence of hydroxyl-(1-3) or keto-group (5) in the alkyl side chain at C-6 has negative impact on the bioactivity. The introduction of hydroxyl-or keto-group might happen before (as an oxidized starter unit) or after the assembly of the pyrone ring (as a tailoring step). To test if they are assembled by cytochrome P450 monooxygenases, 2 mM of P450 inhibitor metyrapone was added into the fermentation medium. The results (Fig. 5) showed that  the production of compounds 1-3 and 5 were decreased by ~ 2.3-, ~ 1.4-, ~ 3-, and ~ 7.8-fold, respectively, and simultaneously, the yields of 10 and 13 were increased by ~twofold and ~fourfold, respectively. Conversely, no obvious changes of the other compounds were observed.
This result indicated that the presence of hydroxyl-or keto-group is assembled by an unknown P450 located in the genome of the heterologous host as proposed in Fig. 6.

Discussion
Heterologous expression serves as a proven effective approach for activating silent secondary metabolites gene clusters [28][29][30][31]. As type III PKSs are simple homodimer ketosynthases, they are especially convenient to be manipulated. In this study, the type III PKS gene vioA from deepsea-derived S. somaliensis SCSIO ZH66 was put under the control of the constitutive promoter P gapDH followed by introduction into three different expression hosts. The accumulation of VLP products with different profiles in these three hosts (Fig. 3) supported the importance of precursor availability as well as genetic backgrounds of expression hosts. The structural diversity of the VLPs compounds accumulated in S. coelicolor M1146/pWLI807 indicated that VioA can condense CoA-or ACP-tethered β-keto acids of different chain length with both ethylmalonyl CoA and methylmalonyl CoA, similar to that of Gcs [32]. However, compared to the data reported so far, VioA might recognize more diverse CoA-or ACP-tethered β-keto acids from fatty acid metabolism than Gcs. In the present study, although we tried to broaden the substrate promiscuity and/or improve catalytic efficiency of VioA via mutagenesis based on sequence alignment as well as previously mutagenesis results [14,17], it is not surprising that no variants with expected properties were obtained. The VvSTS variants T197A and L214I were able to produce polyketides with different profiles than those of the wild-type VvSTS by changing the sizes of the substrate binding pocket and the cyclization pocket [14]; however, the corresponding substitutions in VioA led to severely impaired activity (I174A, Additional file 1: Figure S17A, ii) and no impact at all (L190I, Additional file 1: Figure  S17A, iii). In Gcs, replacements of H261 and M274 with Ala significantly increased the yields of triketide pyrones [17]; on the contrary, the corresponding variants Y229A and S242A of VioA displayed severely impaired (Additional file 1: Figure S17A, iv) and totally abolished activity (Additional file 1: Figure S17A, v), respectively. The structure model of VioA supported the important roles of these mutated residues (Additional file 1: Figure S17B). Crystallographic studies would be contribute to disclose the underlying molecular basis for the substrate promiscuity of VioA and provide reliable guidance for further optimization.
Herein, for the first time, VLPs were shown to display anti-H1N1 virus activity (Table 3). Except for compound 12 harboring a 3-ethyl-4-hydroxy-α-pyrone ring,  hydroxy-α-pyrone backbone but with diverse side chains at C-6. The differences in their activity can be ascribed to the influence of the substituent at C-3 and the alkyl side chain at C-6. The anti-H1N1 activity increased with decrease in the polarity of the compounds, suggesting that the lipophilic nature of the alkyl chain plays an important role for the activity, which is consistent with their anti-MRSA assay results [18]. These findings indicated prospective directions for improving anti-H1N1 activity of VLPs.

Strains and plasmids
All strains and plasmids used in this study are listed in Additional file 1: Table S1. Escherichia coli DH5α was used as the host for general subcloning [33]. E. coli ET12567/pUZ8002 [34] was used as the cosmid donor host for E. coli-Streptomyces intergeneric conjugation. The deepsea-derived S. somaliensis SCSIO ZH66 has been described previously [18,24]. S. coelicolor M1146 [22] and S. sanyensis FMA [23] were used as the host strains for heterologous expression. Plasmid extractions and DNA purifications were carried out using standardized commercial kits (OMEGA, Bio-Tek, Guangzhou, China). PCR reactions were carried out with primers listed in Additional file 1: Table S2 using Pfu DNA polymerase (TIANGEN, Beijing, China). Oligonucleotide synthesis and DNA sequencing were performed by Sunny Biotech company (Shanghai, China). Restriction endonucleases and T4 DNA ligase were purchased from Fermentas (Shenzhen, China).

Bioinformatic analysis
The evolutionary history was inferred using the Neighbor-Joining method [35]. The optimal tree with the sum of branch length = 10.17229024 is shown. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1000 replicates) are shown next to the branches. The tree is drawn to scale, with branch lengths in the same units as those of the evolutionary distances used to infer the phylogenetic tree. The evolutionary distances were computed using the Poisson correction method and are in the units of the number of amino acid substitutions per site. 29 protein sequences were used for analysis. All positions containing gaps and missing data were eliminated. There were a total of 301 positions in the final dataset. Evolutionary analyses were conducted in MEGA5 [36].

Overexpression of VioA
The vioA gene and the constitutive promoter P gapDH were amplified from the genome of S. somaliensis SCSIO ZH66 using primer pairs of vioAFP/vioARPBamHI and P gapDH FPEcoRI/P gapDH RP (Additional file 1:

Production and analyses of VLPs
Spores of Streptomyces strains were inoculated into 50 mL of medium in a 250 mL flask for production analysis or into 200 mL in a 1 L flask for isolation, and were incubated at 30 °C, 220 r.p.m for 7 days. The culture supernatants were extracted twice with an equal volume of EtOAc. The combined EtOAc extracts were concentrated in vacuo to afford a brown residue, which was dissolved in MeOH, filtered through a 0.2 μm filter, and subjected to HPLC analysis.

Site-directed mutation
For this work, the site-directed mutagenesis was created by overlapping primer mutagenesis [37]. To make each mutation, pairs of overlapping oligonucleotides, Additional file 1: Table S2, were synthesized. The first round of PCR was done using each of two mutagenic oligonucleotides and each of two (flanking) oligonucleotides complementary either to the 5′ or 3′ ends of the P gapDH ::vioA. The two resulting PCR products were mixed, annealed and extended by few PCR cycles. The resulting Gel-purified full-length PCR products were cloned into pMT3 and confirmed by DNA sequencing.

Protein structure modeling
The structural model of VioA was done by using I-TASSER server (http://zhang lab.ccmb.med.umich .edu/I-TASSE R) [38]. The C-score for the VioA model is 1.15, indicating a high degree of structural homology to the templates, which is additionally confirmed by the low RMSD of 4.1 ± 2.8 Å. COACH was then used for protein-ligand-binding site prediction [39,40].

Biological assays
The cytotoxicity of compounds was measured by the MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyl tetrazolium bromide; Sigma-Aldrich, USA) assay. Confluent MDCK cell cultures in 96-well plates were exposed to different concentrations of compounds in triplicate, using incubation conditions equivalent to those used in the antiviral assays. Next, 10

Additional file
Additional file 1: Table S1. Plasmids and strains used in this study. Table S2. Primer pairs used in this study. Table S3. Homologous locus of vioAB in different Streptomyces genomes. Figure S1. Relative yields for compounds 1-14 in different strains. Figure S2. Spectral data of 1. Figure  S3. Spectral data of 2. Figure S4. Spectral data of 3. Figure S5. Spectral data of 4. Figure S6. Spectral data of 5. Figure S7. Spectral data of 6. Figure S8. Spectral data of 7. Figure S9. Spectral data of 8. Figure S10. Spectral data of 9. Figure S11. Spectral data of 10. Figure S12. Spectral data of 11. Figure S13. Spectral data of 12. Figure S14. Spectral data of 13. Figure S15. Spectral data of 14. Figure S16. Multiple-sequence alignments of VioA with selected type III PKSs. Figure S17. Site-directed mutagenesis study of VioA.