Marina
Toplak
a,
Adelheid
Nagel
a,
Britta
Frensch
a,
Thorsten
Lechtenberg
a and
Robin
Teufel
*b
aFaculty of Biology, University of Freiburg, Schänzlestrasse 1, 79104 Freiburg, Germany
bPharmaceutical Biology, Department of Pharmaceutical Sciences, University of Basel, Klingelbergstrasse 50, 4056 Basel, Switzerland. E-mail: robin.teufel@unibas.ch
First published on 17th May 2022
The often complex control of bacterial natural product biosynthesis typically involves global and pathway-specific transcriptional regulators of gene expression, which often limits the yield of bioactive compounds under laboratory conditions. However, little is known about regulation mechanisms on the enzymatic level. Here, we report a novel regulatory principle for natural products involving a dedicated acetyltransferase, which modifies a redox-tailoring enzyme and thereby enables pathway furcation and alternating pharmacophore assembly in rubromycin polyketide biosynthesis. The rubromycins such as griseorhodin (grh) A are complex bioactive aromatic polyketides from Actinobacteria with a hallmark bisbenzannulated [5,6]-spiroketal pharmacophore that is mainly installed by two flavoprotein monooxygenases. First, GrhO5 converts the advanced precursor collinone into the [6,6]-spiroketal containing dihydrolenticulone, before GrhO6 effectuates a ring contraction to afford the [5,6]-spiroketal. Our results show that pharmacophore assembly in addition involves the acetyl-CoA-dependent acetyltransferase GrhJ that activates GrhO6 to allow the rapid generation and release of its labile product, which is subsequently sequestered by ketoreductase GrhO10 and converted into a stable intermediate. Consequently, the biosynthesis is directed to the generation of canonical rubromycins, while the alternative spontaneous [5,6]-spiroketal hydrolysis to a ring-opened pathway product is thwarted. Presumably, this allows the bacteria to rapidly adjust the biosynthesis of functionally distinct secondary metabolites depending on nutrient and precursor (i.e. acetyl-CoA) availability. Our study thus illustrates how natural product biosynthesis can be enzymatically regulated and provides new perspectives for the improvement of in vitro enzyme activities and natural product titers via biotechnological approaches.
The identification and investigation of the griseorhodin A biosynthetic gene cluster (BGC) in Streptomyces sp. JP95 revealed that early steps in the formation of rubromycin-type polyketides involve a minimal type II PKS, ketoreductases, cyclases/aromatases as well as diverse tailoring enzymes (Fig. 1).4,6 Notably, similar to many other pharmacophores of aromatic polyketides,10–14 formation of the hallmark bisbenzannulated spiroketal during late stage biosynthesis was attributed to distinct flavin-dependent oxidoreductases based on gene knock out experiments and metabolic studies.5 Recently, the biosynthesis of the spiroketal could be elucidated in more detail by in vitro studies with enzymes encoded by the 1 BGC (grh).15,16 Notably, these biosynthetic steps most likely proceed via ring A-reduced hydroquinonic intermediates (compounds introduced below denoted with a), while ortho-quinonic intermediates and final pathway products that are typically described in the literature are the result of autooxidation (compounds denoted as b).16 Accordingly, the advanced pentangular pathway intermediate dihydrocollinone (3a) serves as a precursor for the generation of the [6,6]-spiroketal intermediate dihydrolenticulone (4a), which is subsequently converted to dihydro-7,8-dideoxy-6-oxo-griseorhodin C (5a) that features the mature [5,6]-spiroketal motif. While 3a and 4a are moderately stable under aerobic conditions and can be detected by liquid chromatography (LC) methods, compounds featuring a mature pharmacophore such as 5a autooxidize too rapidly for detection. The formation of the spiroketal is facilitated by several flavoenzymes, which are known to be mechanistically versatile.10,11,13,17–25 Accordingly, generation of 4a mainly depends on the group A flavoprotein monooxygenase (FPMO) GrhO5,15 which is assisted by flavoprotein oxidase GrhO1.16 The subsequent conversion of the [6,6]-spiroketal-containing 4a into 5avia oxygenation and succeeding decarboxylative ring contraction depends on a second group A FPMO, GrhO6, while further downstream steps require further investigation and presumably proceed via dihydro-7,8-dideoxygriseorhodin C (6a) to the mature rubromycins such as 1 and 2.16 Both FPMOs were furthermore shown to moonlight as ring A reductases to increase substrate reactivity and likely enable salvaging of intracellular autooxidation products such as 3b and 4b.15,16 Notably, the GrhO6 product 5a/5b could not be directly observed previously but rather seco-7,8-dideoxy-6-oxo-griseorhodin C (7), which was shown to arise from the spontaneous hydrolysis of ring C of unstable 5a (presumably driven by the ring strain caused by the sp2-hybridized C6-ketone) and ring A autooxidation.
Open questions remain, inter alia, regarding the reduction of the C6-ketone of 5a/5b to 6a/6ben route to the mature rubromycins that may be catalyzed by the putative ketoreductase GrhO10 and the corresponding functional homologs such as HyalO10 from 2 biosynthesis.5 In addition, the role of GrhJ/HyalJ, which belong to the acetyltransferase subfamily of GCN5-related N-acetyltransferases (GNATs), is particularly intriguing. Previously, investigation of the ΔgrhJ and ΔgrhO6 deletion mutants of the heterologous producer S. albus J1074 showed that 1 formation was abolished in both strains, which instead accumulated 4b as one of the major compounds. Accordingly, it was suggested that both enzymes play a role in [5,6]-spiroketal biosynthesis and the conversion of 4b into 5b. However, more recent studies showed that GrhO6 alone seemed sufficient for the in vitro formation of 5a/5b, although only 7 could be directly detected in adequate amounts (Fig. 1),16 raising the question as to how the rubromycin-producing bacteria prevent formation of this shunt product in vivo. In this work, we investigated the late stage redox tailoring of rubromycin polyketides and interrogated the cryptic role of the involved acetyltransferases, which revealed an unanticipated novel principle for controlling the metabolic flux in polyketide biosynthesis via direct acetylation and activation of key tailoring enzyme GrhO6.
Hence, these data confirm that GrhO10 represents a ketoreductase (KR) and acts on 5a/5b rather than 7. A closer inspection of GrhO10 as well as the functional homologues HyalO10 and RubG (rubromycin biosynthesis) revealed high similarity of these enzymes to the structurally characterized C17/C19 ketoreductase ARX21 from arixanthomycin biosynthesis (60% identity, 100% coverage).26 Multiple sequence alignment and homology modeling indicated that all important catalytic and substrate binding residues identified in ARX21 (S144, Y157, K161 and Y210, R154, respectively) are conserved in GrhO10 and its homologs (Fig. S5 and S6†), which is in line with the fact that they also act as C17 KRs on substrates with comparable chemical structure.
![]() | ||
Fig. 3 HPLC-chromatograms showing the time course for the formation of 5b in a reaction cascade containing 3b, NADPH, GrhO5, GrhO1, GrhO6 ± HyalJ. (A) In the reaction with HyalJ, after the initial reduction of 3b (ctrl, pink line – 10.5 min) to 3a (yellow line – 12.5 min) by GrhO5, 4a (pale yellow line – 14.7 min) is formed by the combined action of GrhO5 and GrhO1. Compound 4a is then further converted into 5a/5b (dark red line – 13.3 min) by GrhO6 together with HyalJ. At the same time, the hydrolysis product 7 (light red line – 10.7 min) is generated from 5a/5b. (B) In the reaction without HyalJ, the produced 4a is further converted into 7 (light red line – 10.7 min) by GrhO6 without evident accumulation of 5a/5b. In this case, 5a/5b (standard, dark red line – 13.3 min) could not be detected. Note, the small peak appearing at around 13.1 min after about 10 min reaction time based on the UV-visible absorption characteristics does not match 6a/6b. Instead, it is a minor side product mostly observed in enzyme assays lacking GrhO6 (see Fig. S11†). |
These results confirmed a cryptic role for the GNATs in promoting 5a/5b production and/or in counteracting spontaneous ring C hydrolysis to 7. Compound 7 was only recently identified and shown to be produced to some extent in vivo in different rubromycin polyketide producer strains (Fig. S13†).16 However, based on our results, an increased formation of 7 would be expected in the ΔgrhJ strain knockout out strain S. albus J1074 KR42,5 which was originally reported to accumulate 4a(b) as major product.16 Gratifyingly, the cultivation and re-analysis of this strain indeed showed that 7 is the most dominant polyketide-related compound in vivo alongside 4b, fully consistent with the in vitro results (Fig. S14†). However, although 5a/5b was faster formed and clearly accumulated in the enzyme assays, its spontaneous hydrolysis also proceeded at significantly higher rates compared to pure 5b in the same buffer without enzymes (Fig. 3, S4 and S10†). Further scrutiny of the assays revealed that reduced 5a is substantially more susceptible to undergo hydrolysis than its oxidized counterpart 5b. Therefore, GrhO5 and GrhO6 are likely responsible for the accelerated 5a/5b decomposition due to their ring A reductase activities.15,16
Because GrhJ and HyalJ are predicted GNATs, acetyl-CoA (AcCoA) was next added to assay mixtures containing 3b, NADPH, GrhO5, GrhO1, GrhO6 together with HyalJ. Interestingly, supplementation of AcCoA resulted in an increase in 5a/5b formation in vitro (Fig. S15 and Table S1†), which, surprisingly, became more pronounced with each freezing–thawing cycle that HyalJ had undergone before (Fig. S16†). Therefore, we speculated that HyalJ is purified from E. coli as holo-enzyme with bound AcCoA, but over time loses its ability to boost 5a/5b formation due to the degradation of AcCoA. To test this hypothesis, the enzyme was freshly produced, purified and immediately sacrificed by the addition of 1 eq. (v/v) EtOAc + 10% FA. Then, the organic phase was removed, and the aqueous layer analyzed by UPLC-HRMS, which clearly confirmed the presence of AcCoA (Fig. S17 and S18†). Interestingly though, UPLC-MS analysis revealed that HyalJ is not only co-purified with AcCoA, but also with almost identical amounts of propionyl-CoA and methylmalonyl-CoA as well as minor amounts of malonyl-CoA (Fig. S19–S22†). To find out, whether binding of these acyl-CoA derivatives also promotes 5a/5b formation, enzymatic assays with apo HyalJ (which has virtually no effect on 5a/5b formation in the absence of free AcCoA) were repeated in the presence of supplemented propionyl-CoA, malonyl-CoA and methylmalonyl-CoA. While propionyl-CoA and methylmalonyl-CoA had an equal effect to AcCoA, the addition of malonyl-CoA to the reaction mixture did not result in increased 5a/5b formation, consistent with the minuscule amounts of malonyl-CoA co-purified with HyalJ compared to the other CoA-derivatives (Fig. S19 and S23†).
To further investigate whether K5 is the site of acetylation, freshly purified GrhO6 (as control) and GrhO6 after turnover of 4b in the presence of AcCoA and HyalJ were in-gel digested with an N-Asp protease and analyzed by UPLC-HRMS. Despite low signal intensity of the peptide of interest due to unfavorable protease cleavage sites and poor ionization, analysis of the extracted protein fragments indicated that reacted GrhO6 is acetylated at K5 (Fig. S28†), fully in line with the bioinformatic predictions. Accordingly, a [M + 2H]2+-fragment with 332.661 m/z corresponding to DTK(Ac)GTT was detectable that was absent in the control. At the same time, the unmodified counterpart ([M + 2H]2+, 311.656 m/z) of this fragment was exclusively detected in the control. These results corroborate that the activation of GrhO6, which boosts 5a/5b formation, depends on the covalent modification of K5 by the ac(et)yltransferase HyalJ. GNATs typically have a strictly conserved fold, ideally suited for homology modeling.27,33,34 Accordingly, the predicted structure of HyalJ revealed the putative AcCoA binding site as well as catalytic residues. Based on its position, His81 may thus function as critical residue of the oxyanion hole for stabilization of the negative charge in the transition state following the attack of GrhO6's K5 side chain on the thioester bond of the CoA-ester substrates. The active site pocket itself is sizeable, likely explaining why HyalJ can also bind and activate bulkier CoA-esters such as propionyl-CoA and methylmalonyl-CoA (Fig. 4 and S29†).
As malonyl-CoA, the main precursor for rubromycin biosynthesis, is directly formed from acetyl-CoA, it stands to reason that the GNATs monitor the availability of pathway precursors. Accordingly, a lack of acetyl-CoA (due to an arrest of the citric acid cycle), indicative of a phase of starvation, may prompt a rapid halt of rubromycin biosynthesis and thereby prevent stagnation or even cell death from nutrient depletion, while at the same time freeing resources, e.g., for sporulation.37 In contrast, during high-nutrient conditions, rubromycin polyketide biosynthesis is affordable and pushed forward by acetylated GrhO6. Notably, the metabolic profile observed for the ΔgrhJ knock-out strain of the heterologous producer most likely mimics that of the starved-out cell. It is therefore tempting to speculate that the respective final products in each scenario (4a/4b and 7 as opposed to mature rubromycins during starvation and high nutrient conditions, respectively) fulfill different biological functions and thus enable a broader metabolic adaptability and thus increased survivability. For example, 4a/4b has antibacterial as well as antiproliferative activity and was furthermore shown to act as inhibitor of the serine protease human leukocyte elastase,5 clearly highlighting the bioactivity of this compound. Evidently, the impaired functionality of non-acetylated GrhO6 leads to accumulation of 7 and prevents 6a/6b formation. Concurrently, 4a/4b can be partially transformed into 8 by the action of GrhO10/Hyal010, as verified by the in vitro assays. The potent inhibitor 8 then further impedes GrhO6 functionality, which may explain why the cells of the ΔgrhJ knock-out strain do not completely convert 4a/4b into 7 and instead accumulate both compounds (Fig. 5). These data, however, are also a reminder that metabolic profiles of bacteria strongly depend on the growth conditions, as nutrient-rich media in the laboratory typically do not reflect the natural environment, in which nutrient scarcity may often be the rule rather than the exception. So far, it remains unclear if GrhO6 remains activated for its entire lifetime or whether deacetylases exist that target the enzyme; obvious enzyme candidates are not encoded in the rubromycin BGCs or their vicinity. Further studies are required to address these open questions and the biological implications of the reported metabolic adjustments.
In the in vitro assays, conditions for shunt product 7 formation are favorable due to the presence of NADPH in combination with non-acetylated GrhO6. This leads to a boost in ring A reduction of autooxidized ortho-quinonic pathway intermediates and thus to the accumulation of 5a, which is evidently more susceptible to undergo spiroketal hydrolysis compared to 5b. Most likely, this is equally true in vivo under the reducing conditions present in the cell, which is supported by the observation that the ΔgrhJ knock-out strain, which can only produce non-acetylated GrhO6, accumulates major amounts of 7. However, the complete lack of 6a/6b formation in the coupled assays with non-acetylated GrhO6 even in presence of a large excess of GrhO10 is surprising and should at least enable the partial conversion of 5a/5b into stable 6a/6b. A plausible explanation for these observations is that product release by non-acetylated GrhO6 is significantly slowed down, resulting in the hydrolysis of 5a/5b into 7 within its active site and thereby depriving the ketoreductase of its substrate. Hence, it appears likely that GrhJ-mediated acetylation may affect specific interactions of GrhO6 with 5a/5b and/or conformational dynamics required for product release. This is substantiated by the fact that group A FMPOs are well known to undergo conformational changes and, e.g., feature mobile flavin cofactors whose movements are tightly controlled by the protein matrix and depend on the presence of substrates.10,15,19,20,38–40 Overall, these results also emphasize that the different redox states of (hydro)quinonic pathway intermediates and final products drastically affect the molecular properties and therefore need to be taken into account with respect to their biosynthesis, stability and activity.
In summary, we could show that the direct modification of a key redox tailoring enzyme by a dedicated N-acetyltransferase is required to enable the production of advanced on-pathway precursors en route to the rubromycins. Our in vitro reconstitution of the biosynthetic steps and their regulation combined with the in vivo observations from the ΔgrhJ knock-out strain clearly confirm that these acetyltransferases govern rubromycin polyketide biosynthesis. Accordingly, this novel principle of controlling the metabolic flux and effectuating pathway branching in polyketide biosynthesis enables the formation of distinct final products with alternate pharmacophores. The results presented herein thus provide deeper insights into the regulation and formation of bioactive secondary metabolites. We anticipate that similar strategies will be exposed in other biosynthetic pathways, which also brings new opportunities for the investigation of enzymes that thus far proved inactive in vitro and provides knowledge for future pathway bioengineering efforts as well as for chemoenzymatic approaches for natural product formation.
Footnote |
† Electronic supplementary information (ESI) available. See https://doi.org/10.1039/d2sc01952c |
This journal is © The Royal Society of Chemistry 2022 |