Phg biosynthetic genes are co-transcribed as a multi-gene operon
As described above, the L-Phg biosynthetic genes (pglA-E) are organized in an operon-like structure (lpg) within the pristinamycin biosynthetic gene region (Fig. 1a) (Mast et al. 2011b). lpg is localized between the genes snbDE and snaD, which encode PI- and PII-specific peptide synthetases, respectively (Mast et al. 2011a). The gene mbtY is embedded in the lpg region and encodes a MbtH-like protein, which is suggested to interact with SnbDE but is not directly involved in Phg biosynthesis (Mast et al. 2011b). In order to determine if the pgl genes are co-transcribed and to ensure a successful transcription of the lpg operon in the heterologous expression studies later on, RT-PCR experiments have been conducted with RNA isolated from the S. pristinaespiralis wild type and primers that anneal to overlapping regions of the pgl genes (Fig. 1a). With these experiments, amplicons were obtained, which are specific for the overlapping regions between snbDE and pglA (A′), pglA and pglB (B′ ), pglB and pglC (C′), pglC and pglD (D′ ), and pglD and pglE (E′ ), respectively, revealing that all pgl genes are transcribed as one polycistronic mRNA and form an operon together with the Phg-specific NRPS gene snbDE (Fig. 1b). Since snbDE is located directly downstream of snbC with overlapping stop and start codons, respectively, and snbC has been shown to be regulated by PapR2, it can be estimated that snbC, snbDE, and the pgl genes together form a multi-gene operon, which is under regulatory control of the pristinamycin pathway-specific activator PapR2 (Fig. 1a).
Expression of L- and D-Phg operons in suitable host strains
To obtain constructs for the fermentative production of L-Phg, the native ~ 6-kb lpg operon from S. pristinaespiralis was cloned into the integrative vector pRM4 under control of the constitutive ermE* promoter, resulting in the expression construct pYM/lpg (Fig. S1). For production of the D-Phg enantiomer, an artificial D-Phg operon (dpg) was generated on the basis of the native lpg operon from S. pristinaespiralis: In a synthetic biology approach, the gene pglE, encoding the L-Phg aminotransferase in S. pristinaespiralis, was exchanged by the gene hpgAT from P. putida, which codes for a stereospecific D-Phg aminotransferase. This D-Phg aminotransferase is the only currently known L to D stereo-inverting aminotransferase (Walton et al. 2018). A recombinant PCR yielded the artificial dpg operon, which was cloned into pRM4, resulting in the expression construct pYM/dpg (Fig. S1). Both plasmids, pYM/lpg and pYM/dpg, were each transferred into different actinomycetes (S. pristinaespiralis Pr11, Streptomyces lividans T7, Streptomyces albus J1074, Amycolatopsis balhimycina, and Rhodococcus jostii RHA1) as homologous or heterologous host strains, respectively (–OE strains; Supplementary File). Strains with the empty pRM4 vector served as control (–C strains; Supplementary File). S. pristinaespiralis was used as expression strain because it is the natural producer of L-Phg, which is a building block for the biosynthesis of the streptogramin antibiotic PI. A. balhimycina was tested since it produces the structurally related non-proteinogenic amino acids hydroxy- and dihydroxy-phenylglycine, which are components of the glycopeptide antibiotic balhimycin (Pfeifer et al. 2001). S. lividans and S. albus are established heterologous expression strains (Nah et al. 2017) and R. jostii has a well-studied, intensive aromatic compound metabolism (Yam et al. 2011). All strains were grown in R5 medium in triplicate. After 30-h, supernatant samples were harvested and Phg amount (given in μg/L) was determined by HPLC-MS/MS analysis. Here, it should be noted that the applied method does not allow to distinguish between different Phg enantiomers. In order to determine enantiomerism of the produced Phg compounds, chiral HPLC analyses have been performed with various expression samples. However, Phg concentrations were too low to be detected (data not shown). HPLC-MS/MS analysis revealed that Phg amount was largest in samples from S. lividans (SL) and S. pristinaespiralis (SP) expression strains (> 1 μg/L), whereas only minor Phg amounts were measured for samples of S. albus (SA), A. balhimycina (AB), and R. jostii (RJ) expression strains (< 0.75 μg/L) (Fig. 3). No, or only trace amounts of Phg were detected in the respective pRM4 control samples (–C strains, data not shown). Interestingly, all D-Phg expression samples contained higher amounts of Phg than the respective L-Phg expression samples (Fig. 3). Altogether, from all tested strains, S. lividans and S. pristinaespiralis turned out to be the optimal hosts for fermentative Phg production.
Optimal production media for Phg production
In order to define the best Phg production conditions, the optimal producer strains S. pristinaespiralis (SPlpg-OE, SPdpg-OE) and S. lividans (SLlpg-OE, SLdpg-OE) were grown in two different culture media—the complex medium R5 and the pristinamycin production medium HT7T. Samples were taken at different time points (24, 48, 72, and 96 h) and Phg amount was determined by HPLC-MS/MS. Phg was detected in all S. pristinaespiralis (SPlpg-OE, SPdpg-OE) and S. lividans (SLlpg-OE, SLdpg-OE) expression samples, whereas only trace amounts of Phg were measured in the respective pRM4 control samples (Fig. 4a–d). Overall, Phg production was generally higher (even if statistically significant only for L-Phg expression samples as shown in Fig. S2) and more consistent in HT7T medium than in R5 (Fig. 4b, d vs a, c). Interestingly, Phg concentrations decreased in nearly all media and all expression hosts after reaching the maximal level, which suggests a degradation or metabolization of the expression product. An exception was found for S. lividans expression strains in HT7T medium, where Phg production steadily increased to cultivation time point 96 h (Fig. 4b). Thus, Phg metabolization in S. lividans seems to be medium dependent. For S. pristinaespiralis samples, Phg decrease in the pristinamycin production medium HT7T might also be explained by a subsequent incorporation of Phg into PI. Furthermore, it was observed that Phg concentrations in general were higher in D-Phg expression strains than in L-Phg expression strains, which was consistent with the data obtained from the Phg expression studies in different host strains (Fig. 4a–d vs Fig. 3). D-amino acids are known for their poor metabolic usability (Elmadfa and Leitzmann 2015). Hence, the higher Phg amount in the D-Phg expression strains might be explained by a rather poor metabolization of the unnatural D-Phg enantiomer. Due to the observation that overall Phg production was more stable and consistent in HT7T and with regard to subsequent genetic engineering approaches targeting pristinamycin-specific genes in S. pristinaespiralis host strains (see below), the pristinamycin production medium HT7T was used as Phg production medium for further analyses.
Deletion of a gene of the phenylacetyl-CoA degradation pathway significantly improves Phg production in S. pristinaespiralis but not in S. lividans
In order to increase Phg production in the optimal producer strains S. lividans and S. pristinaespiralis, we aimed to genetically manipulate key steps within primary metabolism involved in precursor supply to direct the metabolic flux towards Phg production. As a target of manipulation, we chose the phenylacetyl-CoA degradation pathway since phenylacetyl-CoA is a suggested precursor for the biosynthesis of Phg (Mast et al. 2011a; Osipenkov et al. 2018) (Fig. 2). In a previous study from Zhao et al. (2015), it has been reported that the paaABCDE (paa) operon from S. pristinaespiralis encodes a putative phenylacetyl-CoA epoxidase multicomponent enzyme system, which is responsible for the degradation of phenylacetyl-CoA (Zhao et al. 2015). It was suggested that derepression of the paa operon in S. pristinaespiralis leads to a higher flux of phenylacetyl-CoA towards the phenylacetic acid catabolic pathway and thus to less precursor supply for L-Phg biosynthesis (Zhao et al. 2015). By contrast, it can be assumed that an inactivation of the paa genes in S. pristinaespiralis drives the phenylacetyl-CoA flux towards Phg biosynthesis. Thus, we aimed to inactivate the paa operon in S. pristinaespiralis—but also S. lividans, since a homologous paa operon is present in the S. lividans genome (Supplementary File)—and overexpress the Phg operons in the engineered mutant strains in order to increase production yields. For this purpose, the gene region paaA-E in S. pristinaespiralis and S. lividans, respectively, was inactivated by replacing it against a thiostrepton resistance cassette (thioR) (Supplementary File, Fig. S3). This resulted in the mutants SPpaa::thio and SLpaa::thio, respectively, in which the Phg expression constructs pYM/lpg and pYM/dpg, as well as the empty vector as a control, were each transferred to. The paa control strains, SPpaa::thio-C and SLpaa::thio-C and the host strains SPpaa::thio lpg-OE, SPpaa::thio dpg-OE, SLpaa::thio lpg-OE, and SLpaa::thio dpg-OE were grown in HT7T medium and supernatant samples at different time points were used for Phg production analysis. HPLC-MS/MS measurements of the samples from the engineered host strains revealed that Phg production in the S. lividans paa expression samples was almost the same as in the wild-type-derived expression samples (Fig. 5a vs Fig. 4b): maximal Phg production at 96 h was measured for SLpaa::thio lpg-OE at 1.00 μg/L compared with 0.94 μg/L for SLlpg-OE and 0.95 μg/L for SLpaa::thio dpg-OE compared with 1.2 μg/L for SLdpg-OE. In contrast, Phg production was strongly improved for S. pristinaespiralis paa-derived expression samples (Fig. 5b vs Fig. 4d): Already after 24 h, Phg amount in SPpaa::thio lpg-OE (1.57 μg/L) was 5-fold higher than in SPlpg-OE (0.31 μg/L) and remained high until 96 h. Here, the production decline at 72 h might be an artifact since standard deviations for the SPpaa::thio lpg-OE samples in general were quite high. Phg production was also significantly improved for SPpaa::thio dpg-OE strains, where the maximal Phg production at 96 h (1.30 μg/L) was 3.7-fold higher than in non-engineered SPdpg-OE strains (0.35 μg/L). Overall, the significant improvement of Phg production in S. pristinaespiralis paa host strains most likely results from the directed flux of the phenylacetyl-CoA precursor towards the Phg biosynthetic pathway. The fact that Phg production was improved for SPpaa::thio lpg-OE compared with SPpaa::thio dpg-OE might be explained by the different enzyme kinetics of the two aminotransferases. D-amino acid transaminases, such as HpgAT (encoded in the dpg operon), are commonly known to have a very low transamination activity towards D-Phg (Soda and Esaki 1994). Thus, PglE may convert the accruing phenylglyoxylate precursor more efficiently to L-Phg than HpgAT can convert it to D-Phg.
Deletion of Phg aminotransferase gene pglE slightly improves Phg production in S. pristinaespiralis
In a recent study, we showed that the L-Phg aminotransferase PglE is responsible for the conversion of phenylglyoxylate to L-Phg in S. pristinaespiralis (Osipenkov et al. 2018) (Fig. 2). Deletion of pglE leads to an accumulation of phenylglyoxylate (Osipenkov et al. 2018). Due to this increased basal precursor availability, we were interested how the Phg operon expression in the S. pristinaespiralis pglE mutant (MpglE) would influence production performance. Besides that, inactivation of the native pglE gene could deliver a genetic background for the production of enantiopure Phgs in S. pristinaespiralis. Thus, the MpglE mutant was used as parental strain for the expression of the Phg operons. Strain denomination is similar as reported above and samples were treated as outlined before. HPLC-MS/MS analysis revealed that Phg production in MpglE host strains (MpglE lpg-OE and MpglE dpg-OE) was overall slightly higher than in S. pristinaespiralis wild-type-derived strains (Fig. 6 vs Fig. 4d): An improvement was observed for the MpglE lpg-OE samples, where a maximal production of 0.56 μg/L Phg at 48 h was measured, which is an increase of 1.8-fold compared with the maximal value of 0.31 μg/L Phg at 24 h in the SPlpg-OE sample. For MpglE dpg-OE expression samples, no tremendous Phg production improvement was observed (Fig. 6). Therefore, one could speculate that the slightly increased Phg rates in MpglE lpg-OE may result from a somehow favorable basal phenylglyoxylate precursor supply.
Deletion of the PI-NRPS gene snbDE significantly improves Phg production in S. pristinaespiralis
As suggested above, the decrease of Phg in SPlpg-OE samples may be due to an incorporation of L-Phg into PI (Fig. 4d). Thus, a strategy to increase Phg production in S. pristinaespiralis is to block PI biosynthesis. In order to do that, we inactivated the gene snbDE in S. pristinaespiralis (Supplementary File), which encodes the PI-specific NRPS module SnbDE that uses L-Phg as a building block for PI biosynthesis (Mast et al. 2011b). The respective mutant MsnbDE::thio was used as expression host for the Phg operon expression. The derived host strains MsnbDE::thio lpg-OE and MsnbDE::thio dpg-OE, as well as the control MsnbDE::thio-C, were grown in HT7T and samples were analyzed for Phg production by HPLC-MS/MS. HPLC-MS/MS analysis revealed a maximal Phg production in samples MsnbDE::thio lpg-OE (0.87 μg/L) and MsnbDE::thio dpg-OE (1.27 μg/L) at 96 h, which was an increase of ~ 3-fold compared with maximal production values in wild-type-derived samples SPlpg-OE and SPdpg-OE (0.30 μg/L and 0.35 μg/L), respectively (Fig. 7 vs Fig. 4d). Furthermore, it was found that Phg concentration in the MsnbDE::thio-derived strains increased continuously, whereas a decrease was observed in the wild-type-derived samples at later time points. Actually, the Phg production profile of the MsnbDE::thio-derived strains more resembled the production profile of the S. lividans host strains (Fig. 7 vs Fig. 4b). Thus, it can be assumed that Phg production in the MsnbDE::thio-derived strains is steadily increasing because Phg is not utilized for PI biosynthesis and thus accumulates, which may also happen in S. lividans because this strain does not produce pristinamycin.
Deletion of the pristinamycin TetR–like regulatory gene papR5 significantly improves Phg production in S. pristinaespiralis
As we had incident that Phg production performance in S. pristinaespiralis depends on the pristinamycin biosynthesis capability (see above for MpglE, MsnbDE samples), we aimed to further enhance Phg production by using a pristinamycin superproducer as expression host. In a previous study, we showed that the S. pristinaespiralis repressor mutant papR5::apra produces up to ~ 300% more pristinamycin than the wild-type strain (Mast et al. 2015). Due to this high pristinamycin production capability, we used papR5::apra as a host for Phg operon expression. The derived host strains papR5::apra lpg-OE and papR5::apra dpg-OE, as well as the control strain papR5::apra-C, were grown in HT7T and samples were analyzed by HPLC-MS/MS for Phg production. HPLC-MS/MS data revealed that Phg production was significantly increased in papR5::apra-derived host strains compared with the wild-type-derived ones: papR5::apra lpg-OE and papR5::apra dpg-OE produced approximately 3.3-fold and 2-fold, respectively, more Phg than the wild-type-derived expression strains (papR5::apra lpg-OE: 1 μg/L; papR5::apra dpg-OE: 0.72 μg/L Phg) (Fig. 8 vs Fig. 4d). Phg production was increased especially in the papR5::apra lpg-OE host strain, which was also observed for the other Phg precursor–engineered host strains (SLpaa::thio, SPpaa::thio, and MpglE). Thus, Phg-related precursor engineering seems to affect more L-Phg than D-Phg biosynthesis. This might be explained by the less favorable enzymatic properties of HpgAT, as mentioned before. The reason why Phg concentration is stable or even increasing in these expression hosts might be because PI production is oversaturated with Phg precursor and thus Phg would accumulate. Overall, the improvement of Phg production in the papR5::apra-derived host strains most likely results from the elevated levels of precursor supply in the course of an increased pristinamycin biosynthesis.