A comparative study at bioprocess and metabolite levels of superhost strain Streptomyces coelicolor M1152 and its derivative M1581 heterologously expressing chloramphenicol biosynthetic gene cluster

Microbial superhost strains should provide an ideal platform for the efficient homologous or heterologous phenotypic expression of biosynthetic gene clusters (BGCs) of new and novel bioactive molecules. Our aim in the current study was to perform a comparative study at the bioprocess and metabolite levels of the previously designed superhost strain Streptomyces coelicolor M1152 and its derivative strain S. coelicolor M1581 heterologously expressing chloramphenicol BGC. Parent strain M1152 was characterized by a higher specific growth rate, specific CO2 evolution rate, and a higher specific l‐glutamate consumption rate as compared with M1581. Intracellular primary central metabolites (nucleoside/sugar phosphates, amino acids, organic acids, and CoAs) were quantified using four targeted LC‐MS/MS‐based methods. The metabolite pathways in the nonantibiotic producing S. coelicolor host strain were flooded with carbon from both carbon sources, whereas in antibiotic‐producing strain, the carbon of l‐glutamate seems to be draining out through excreting synthesized antibiotic. The 13C‐isotope‐labeling experiments revealed the bidirectionality in the glycolytic pathway and reversibility in the non‐oxidative part of PPP even with continuous uptake of d‐glucose. The change in the primary metabolites due to the insertion of BGC disclosed a clear linkage between the primary and secondary metabolites.


| INTRODUCTION
The discovery of antibiotics resistance among a wide range of human pathogenic bacteria is increasing over time, causing a setback to our efforts to fight against diseases (Boucher et al., 2009;Rossolini et al., 2014). The World Health Organization recognizes antibiotic resistance as a global challenge to health security (Sood, 2016). The discovery of a new class of antibiotics has been becoming progressively more difficult due to biochemically and technically timeconsuming and costly processes that mostly end in futile efforts due to rediscovery of the same antibiotics (Baltz, 2008;Coze et al., 2013).
In this scenario, it is imperative to search for sustainable antibiotics production routes that are high throughput and cost-efficient.
Engineering the superhost strains using systems and synthetic biology approaches and heterologous expression of biosynthetic gene clusters (BGCs) of novel antibiotics into the superhost strains is one of the most promising approaches to discover and develop new antibiotics bioprocesses (Zhu et al., 2020). Streptomyces coelicolor A3(2) is the model species for different biotechnological purposes and is considered as a suitable species to create superhost strain because of the available knowledge of genome sequence, efficiency in precursors synthesis, and the presence of treasure of several secondary metabolite gene clusters (20 to 40 for a given Streptomyces species) with unknown functions (Bentley et al., 2002;Coze et al., 2013). These cryptic gene clusters may be a potential source of new medically useful antibiotics (Zerikly & Challis, 2009). The strategy of superhost strain creation is to simplify the overall metabolite profiles by deleting the native BGCs with low bioactivities and to increase precursor availability by removing competing sinks of carbon and nitrogen from unwanted pathways (Coze et al., 2013;Flinspach et al., 2010) S. coelicolor M1152 has been continuously being improvised for the discovery and analysis of new metabolites (Thanapipatsiri et al., 2015). For example, S. coelicolor M1581 strain was created from S. coelicolor M1152 by conjugating S. venezuelae cosmid pAH91 containing Chlm cluster (Lorena T. Fernández-Martínez et al., 2014).
Chloramphenicol is a bacteriostatic antibiotic, which is effective against both Gram-negative and Gram-positive bacteria, including many multiply drug-resistant strains such as Bacilli Lorena T. Fernández-Martínez et al., 2014;Sood, 2016). Several filamentous Gram-positive soil actinomycetes produce Chlm, but its production has been investigated primarily in S. venezuelae strain ATCC 10712 (Lorena T. Fernández-Martínez et al., 2014;Vining & Stuttard, 1995). Chlm biosynthesis utilizes the shikimate pathway by incorporating the primary metabolites phosphoenolpyruvate (PEP) from the glycolytic pathway and erythrose-4-phosphate (E4P) from the pentose phosphate pathway (PPP). Shikimate pathway produces chorismic acid as one of the intermediates, which acts as a precursor compound for the biosynthesis of aromatic amino acids (phenylalanine, tyrosine, and tryptophan) and the production of 4-amino-4deoxychorismate (ADC) Vitayakritsirikul et al., 2016).
ADC serves as a branch point for the pathway dedicated to Chlm and folic acid biosynthesis as shown in Figure 1 (Chang et al., 2001;Lorena T. Fernández-Martínez et al., 2014;Vining & Westlake, 1984).
The advancement in mass spectrometric analysis enables us to accurately quantify many metabolites in complex mixtures of biological samples. It can open a new window to improve our in-depth understanding of microbial physiology and the contribution of primary metabolites for secondary metabolite production (Coze et al., 2013;Rokem et al., 2007;. Recently, our group has reported optimized sample preparation workflow and five targeted LC-MS/ MS-based methods to generate high-resolution metabolite data to quantify intracellular central primary metabolite pools (amino acids, organic acids, NADs, CoAs, nucleosides, and sugar phosphates) (Kumar et al., 2021;Rost et al., 2020;Røst et al., 2020;Stafsnes et al., 2018). The application of our LC-MS/MS methods can reveal the differences between the superhost strain S. coelicolor M1152 and chloramphenicol producing strain S. coelicolor M1581 from the perspective of central primary metabolites. This can also reveal the interaction between genotype and phenotype and the linkage between primary metabolites and secondary metabolites.
In view of this, the present study focuses on the generation of comprehensive high-resolution intracellular primary metabolite profiles using advanced sampling and mass spectrometric methods at various growth stages of S. coelicolor M1581 in comparison to its parent and the superhost strain S. coelicolor M1152. The present work produces Chlm in bioreactors using S. coelicolor M1581 under controlled laboratory conditions and compares with the previous report with the same strain in shake flasks. The growth and the production kinetics were compared between both the strains and efforts were made to establish the linkage between phenotype and metabolomics. The 13 C-isotope-labeling experiments conducted in scaled-down bioreactors highlight the flow of carbons from L-glutamate and D-glucose in Chlm, its precursor compound, and other common primary metabolites in reference to the superhost strain. Our work will be an important contribution to the ongoing efforts of superhost strain improvement and heterologous BGC expression for the new and novel product biosynthesis.

| Strains and cultivation conditions
We have used S. coelicolor M1152 and S. coelicolor M1581 for this study (Gomez-Escribano & Bibb, 2011;Lorena T. Fernández-Martínez et al., 2014). The protocols for spore preparation, spore pregermination, media preparation, and setting the experiments were adopted from the report . Both Dglucose (Sigma-Aldrich) and L-glutamate (Sigma-Aldrich) were supplied in the media as the carbon sources for the microorganisms. The cultivations were carried out in phosphate (SSMB-P) limited media in 3 L fermenters (Eppendorf Bioflo 320) under controlled conditions (30°C, pH 7.0) with an initial working volume of 1.8 L. Besides trace elements, SSMB-P media had carbon (D-glucose, 40 gL −1 ), carbon and nitrogen (Sodium L-glutamate monohydrate, 61.1 gL −1 ), phosphorous (Phosphate, 4.6 mM), and magnesium (Magnesium sulfate, 2.0 mM).
The pre-germinated spores were used as inoculum having~1 × 10 9 CFU L −1 . The pH was controlled at 7.0 by the automatic addition of 2 M HCl and 2 M NaOH. The agitation speed was set to 1000 rpm and aeration was provided at the flow rate of 0.3 vvm using filtered air (0.2 μm polytetrafluoroethylene filter). The dissolved oxygen (DO), pH, and temperature were measured with equipped probes and recorded online. Airflow, Oxygen (O 2 ), and carbon dioxide (CO 2 ) in the outlet gas were continuously measured and logged online using a Gas Analyzer (DASGIP GA4, BlueSens, Eppendorf).
The 13 C isotope labeling experiments were carried out in minifermenters with a working volume of 200 ml and under similar conditions as in 3-L fermenters. The agitation was provided using a magnetic bar and magnetic stirrer. The cultivation was carried out under the condition mentioned above in SSBM-P media, however, we have used D-glucose of which all six carbons were universally labeled with 13 C isotope (U −13 C6, 99%; CIL, ChemSupport AS) and its concentration was reduced to 20 g L −1 , instead of 40 g L −1 given in the original media. All carbons of L-glutamate (L-glutamic acid monosodium salt monohydrate, ≥98.0% (NT), 49621−1KG, Sigma-Aldrich) used in the experiment were naturally occurring and was not labeled with 13 C isotope.

| Exometabolites quantification
Dry cell weight (DCW) was quantified by drying the washed biomass pellets at 110°C till constant weight was achieved (Kumar et al., 2021).
The residual glucose and the excreted metabolites in the fermentation broth were quantified using high-performance liquid chromatography (HPLC) equipped with a refractive index (RI) and a UV/VIS detector as described previously (Kumar et al., 2021). The stationary phase was a Hi-Plex column of dimension 300 × 7.7 mm, whereas the mobile phase was 0.05 M H 2 SO 4 in MQ-H 2 O at the flow rate of 0.8 ml min −1 . The residual L-glutamate in the fermentation broth was analyzed using LC-MS/MS following the method of intracellular amino acid analysis (described later), however using a shorter C18 column (1.7 µm, 2.1 × 50 mm) to reduce the analysis run time. 13 C L-glutamic acid (CIL, Inc.) was used as F I G U R E 1 Schematic representation of precursor and key intermediates involved in chloramphenicol, streptomycin, vancomycin, nystatin, and actinorhodin biosynthesis pathways Vitayakritsirikul et al., 2016). Primary metabolic pathways are shown with blue arrows and gray text while secondary metabolic pathways are shown with green arrows and text. ADC, 4-amino-4-deoxychorismate an internal standard. Phosphate concentration was followed by Phosphate test kit Quantofix 37210 (Sigma-Aldrich) paper strip.

| Chloramphenicol quantification
Chlm excreted in the fermentation broth was analyzed using UPLC-MS/MS instrumentation. The chromatographic column used in UPLC for the separation of Chlm was BEH C18 with a dimension 2.1 × 50 mm and pore size of 1.7 µm (ACQUITY UPLC ® , Waters), and the column was set to 50°C. The electrospray ionization source present in MS was operating in negative mode. Deuterated chloramphenicol, d5-Chlm (DLM-119-1.1, CIL, Inc.) having the monoisotopic mass of 327.14 was used as an internal standard. Mobile phase consisted of (A) water + 0.1% formic acid and (B) methanol + 0.1% formic acid, and flow rate was set at 0.250 mL/min. The mobile phase was run at 30% A for half a minute, then a linear gradient was programmed from 30% A to 0% A in 2.5 min, followed by 0% A kept for an additional 0.80 min period before the gradient was brought back to 30% A in 0.10 min. Finally, the column was equilibrated for 2.10 min before starting a new injection. The injection volume was 2 µl. Chlm was eluted at the retention time of 2.13 min. 13 C samples were analyzed by setting the MS method in MRM mode with similar mass by charge ratio (m/z) for parent and daughter ions, which was ranging from 321.14 (all 12C carbons) to 332.14 (all 13C carbons).
The cone and collision voltages were 20 and 4 V, respectively.

| Sampling and sample preparation
The online CO 2 and O 2 measurements were used to gauge the exponential and stationary phases. The time-series sampling for intracellular metabolite analysis was taken both in the exponential phase and the nutrient depletion phase. The sampling and sample preparation was carried out as described previously (Kumar et al., 2021). Data from one cultivation is presented (M1152 duplicate samples, M581 triplicate samples). Data has been verified using a second round of cultivation, that is, true biological replica, but is not presented as results from independent cultivations are hard to merge since there can be slight offsets on the time axis.
The cold extraction was used to extract CoAs from the biomass pellets (Bartosova et al., 2021). The CoAs were extracted using the extraction solvent having a composition of Acetonitrile: Buffer (50 mM NH 4 Ac, pH 5): CH 3 OH (7:2:1) The isotope dilution strategy was used in all methods of LC-MS/MS.

| Computation and statistical analysis
The maximum specific growth rate, μ m was calculated in the exponential phase by the linear regression of the natural logarithm of biomass concentration and time. Similarly, the maximum specific growth rate, μ CO2 was calculated based on the natural logarithm of CO 2 evolution and time. The specific rates (specific D-glucose uptake rate, q gluc ; specific L-glutamate uptake rate, q glut ; specific CO 2 evolution rate, P CO2 ; specific chloramphenicol production rate, P Chlm ) were calculated by dividing the volumetric rate by the average biomass between the two points. The molecular weight of biomass was assumed as 27.0 g when calculating the carbon recovery.
The data from the UPLC-MS/MS instrumentation were acquired and quantified using MassLynx software (v4.2) and TargetLynx (Waters) application manager. All metabolite concentrations were calculated automatically based on a linear standards calibration curve, however, CoAs were quantified manually based on the slope of individual standards, prepared on biological matrix. The fermentation parameters were presented as an average of samples from two parallel cultivation with their standard deviation. All intracellular metabolites pools were reported as an average of two to three independent replicas and their standard deviation. The Omix software was used to visualize the change in the central carbon metabolites (Droste et al., 2011). Scores plot from principal component analysis (PCA) was run on a data set normalized to sum and autoscaled by the MetaboAnalyst online software (Xia et al., 2009).
The energy charge ratio (RCR) was calculated using the following formula: The relative summed fractional labeling (Rel SFL) of each metabolite was adapted from the previous report and calculated as shown below, where I n is the signal area of metabolite with n number of labeled carbon (Isotopologues) (Gombert et al., 2001). The scale of the Rel SFL number range from 0 to 100 and is an indicator of the 13 C label enrichment in the molecule. Control experiments were also conducted by using naturally labeled D-glucose and L-glutamate to verify the contribution of 13C carbon into naturally occurring 12C carbon .   I  I  I  I  n I  I  I  I  I  I n Rel SFL = (0 × + 1 × + 2 × + 3 × +… × ) ( + + + +… ) × 100 . and/or L-glutamate into CO 2 and H 2 O (Kumar et al., 2021). The RQ of around 0.6 during the L-glutamate depletion phase indicates that the TCA cycle was almost closed, and the basal respiratory activity of cells was only maintained by internally stored reduced compounds like lipids and proteins (Xiao et al., 2006). The consumption of internally stored compounds during L-glutamate depletion resulted in depletion of biomass concentration (Figure 2b). Contrary to S. coelicolor M1152, S. coelicolor M1581 had only a phosphate depletion phase, which started at around ̴ 54 h of batch cultivation (Figure 2c).
Comparing Figure 2a,c, a higher L-glutamate consumption rate in S.
coelicolor M1152 could be the reason for its early depletion and a corresponding L-glutamate profile of both strains further confirmed the same (Figure 2b,d). Interestingly, the uptake of D-glucose was inhibited when L-glutamate was completely depleted in the media ( Figure 2b). In S. coelicolor M1581, the cultivation media did not deplete with L-glutamate till the end of cultivation.
The different kinetic parameters of both strains are summarized in  Amino acids were more synthesized in the phosphate depletion phase of the S. coelicolor M1581 as compared with S. coelicolor M1152. The accumulation of amino acids could be due to less protein synthesis or just protein turnover that replenishes the amino acid pools. The metabolite profiling data is easier to interpret from a T A B L E 1 The different cultivation parameters such as specific rates, yields, and the carbon recovery for the Streptomyces coelicolor M1152 and S. coelicolor M1581 in the phosphate (SSBM-P) limited media F I G U R E 3 Heat map representation of the absolute concentration of intracellular metabolites in the time series sampling for the Streptomyces coelicolor M1152 and S. coelicolor M1581 in the phosphate (SSBM-P) limited media. White indicates that the metabolites were not analyzed/not included/trace amount (< 0.001 µmole/gDCW). Figure S1 shows the heat-map representation of relative standard deviation of intracellular metabolites concentration of two to three sample replicas metabolic network perspective. The Omix software was used to better visualize all the primary metabolic pathways (glycolysis, PPP, and TCA cycle) and all other linked metabolites (Droste et al., 2011).
The ratio of metabolites level after phosphate depletion to the exponential phase is shown in Figure 6a (Table S1). Previously, the ECR between 0.5 and 0.6 was reported and found to be influenced by the cell's mycelial live/dead ratio, sample processing, and other cultivation conditions . Among the CoAs, the downregulation of the important polyketide synthases (PKSs) precursor acetyl-CoA was observed to be a characteristic of phosphate depletion for both strains. This has no consequence for precursor supplies for the heterologous production of chloramphenicol in S. coelicolor, but will be important for polyketide synthase-based secondary metabolite production, e.g the S. coelicolor native actinorhodin, nystatin, and so forth (see Figure 1). Of course, sufficient supplies of acetyl-CoA being the substrate to enter TCA for complete oxidation and ATP production is necessary, but the present data set does not indicate a lower energy charge in the stationary phase.

| Cultivation and growth parameters
Mini fermenters (operating volume 50-200 ml) were designed to scale down the bioreactor cultivations to save expensive 13C-label substrate. The fermentation was carried out in a phosphate-limited media using both S. coelicolor M1152 and S. coelicolor M1581. In this study, experiments were designed to include 13 C labeled D-glucose at all carbon positions and 12 C labeled L-glutamate to minimize the cost without compromising the objectives of the study. Samples were taken at multiple time points in batch cultivation for the direct interpretation of dynamic 13 C labeling patterns (Buescher et al., 2015).
The cultivation results of mini fermenters containing the data of online monitoring and offline analysis are given for both strains ( Figure S2a,b). The D-glucose concentration in the SSBM-P media was reduced to 20 gL −1 from the original 40 gL −1 yet the D-glucose was never exhausted in the media till the time of the cultivation in both strains. The biomass concentration was reduced to less than half as observed in 3-L bioreactors. In these experiments also, it was observed a higher L-glutamate consumption rate as compared with Dglucose. In the stationary phase, the Chlm production was around 52 mg L −1 , similar to what was observed in 3-L bioreactors. It appeared that Chlm production was less affected by the scale-up parameters as also seen in the previous report with the same strain The carbon isotope labeling pattern of oxidative PPP intermediate, 6-phosphogluconate (6PG) was similar to that of upper glycolytic intermediates. The non-oxidative intermediates of PPP such as R5P, RL5P, and X5P were calculated combinedly. It was observed a significant relative composition of two 13 C labeled carbon (M+2) along with expected all 13 C labeled carbon (M+5) from D-glucose. This revealed that G3P synthesized from L-glutamate (M+0) feeds its carbon into PPP by the reactions catalyzed by transaldolase and transketolase (David L. Nelson, 2005). Previously it was also observed that when L-glutamate acts as a carbon source, the interaction of glycolytic intermediates to PPP is through the non-oxidative rearranging reactions catalyzed by transketolase and trans-aldolase F I G U R E 5 Heat map representation of intracellular metabolites for the Streptomyces coelicolor M1152 and S. coelicolor M1581 in the phosphate (SSBM-P) limited media. The ratio of the metabolite concertation at each sampling point to the average concentration across all sample points for each cultivation has been log2 transformed to better visualize the relative change during the cultivation. Red and green indicate a high and a low relative metabolite level, respectively. Gray indicates that the metabolites were not analyzed/not included/trace amount F I G U R E 6 The primary metabolic pathways and their associated metabolites were visualized using the Omix software (Droste et al., 2011). (a) Log2 fold change of central carbon metabolites (CCMs) corresponding to the stationary phase (72 h) relative to the exponential phase (36 h) for the Streptomyces coelicolor M1152. (b) Log2 fold change of CCMs corresponding to the stationary phase (60 h) relative to the exponential phase (30 h) for the S. coelicolor M1581. Both strains were cultivated in the phosphate (SSBM-P) limited media. Gray indicates that the metabolites were not analyzed/not included/trace amount (Aon & Cortassa, 2001). The isotopologue pattern of PEP and 3PG-2PG ( Figure S3) of the lower glycolytic pathway revealed a large fraction of these isotopologues were either 12 C (M+0) or 13 C (M+3) labeled carbon implying contributions of L-glutamate and D-glucose, respectively. Isotopologues of PEP and 3PG-2PG appeared to be a mixture of two isotopologues without undergoing any chemical reactions.
Summarizing, the presence of 12 C originating from L-Glutamate into glycolytic pathway revealed that there was bidirectionality in the glycolytic pathway even with continuous uptake of D-glucose. Reversibility in the non-oxidative part of PPP which synthesizes F6P could be another possibility especially when more 12 C carbons were found in X5P/R5P/RL5P than in G6P.
Citrate in the TCA cycle is formed by the reaction between OAA and acetyl-CoA. It was observed an abundance of isotopologues having two 13 C labeled carbon (M+2), one 13 C labeled carbon ( Chlm are shown in Figure 8. The 13 C labeling patterns were the same in ATP and GTP (Purine nucleoside phosphates) but different in CDP (Pyrimidine nucleoside phosphate), which could be justified because of the same precursors (two carbon from glycine, two carbon from formate, one carbon from bicarbonate) for ATP and GTP but different precursors (three-carbon from aspartate and one carbon from bicarbonate) for CDP. The isotopologues of ATP, GTP, and CDP were more uniform for S. coelicolor M1152 as compared with S. coelicolor M1581, which could again indicate a higher influx of L-glutamate carbon into the system and consistent with the above studies. Interestingly, the Rel SFL of all metabolites in both strains were decreased in the exponential phase, consistent with the higher Lglutamate uptake rate. The Rel SFL in these metabolites was nearly stable in the stationary phase. However, the Rel SFL of the intermediates of the TCA cycle, especially the metabolites downstream of αKG were increased in the stationary phase due to the decreased Lglutamate consumption rate enabling the inflow of 13 C labeled glycolytic intermediates into the TCA cycle. PEP (lower glycolysis) and E4P (  The average Rel SFL (%) of all sampling points is plotted using the Omix software for the (a) Streptomyces coelicolor M1152 and the (b) S. coelicolor M1581. The ratio of Rel SFL (%) of S. coelicolor M1581 to S. coelicolor M1152 was log 2 transformed and then visualized (c) using Omix software (Droste et al., 2011). For (a) and (b), the Rel SFL varies from 0% (green) to 100% (red), whereas for (c), the color code varies from minimum to maximum ratio KUMAR AND BRUHEIM | 159 metabolites) is not critical for chloramphenicol biosynthesis having PEP and E4P as precursors, but needs close monitoring for heterologous expression of other BGCs, for example, polyketides (see Figure 1).

| CONCLUSIONS
The superhost strain S. coelicolor M1152 and its heterologously expressing chloramphenicol BGC strain S. coelicolor M1581 were comparatively studied at the bioprocess and metabolite levels. The

CONFLICT OF INTERESTS
The authors declare that there are no conflict of interests.

AUTHOR CONTRIBUTIONS
Conceptualization, methodology, software, validation, formal analysis, investigation, data curation, writing-original draft preparation, writingreview and editing, visualization, supervision: Kanhaiya Kumar and Per Bruheim. Resources, project administration, funding acquisition: Per Bruheim. All authors have read and agreed to the published version of the manuscript.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study may be available from the corresponding author, Per Bruheim, upon reasonable request.