A Light‐Activated Acyl Carrier Protein “Trap” for Intermediate Capture in Type II Iterative Polyketide Biocatalysis

Abstract A discrete acyl carrier protein (ACP) bearing a photolabile nonhydrolysable carba(dethia) malonyl pantetheine cofactor was chemoenzymatically prepared and utilised for the trapping of biosynthetic polyketide intermediates following light activation. From the in vitro assembly of the polyketides SEK4 and SEK4b, by the type II actinorhodin “minimal” polyketide synthase (PKS), a range of putative ACP‐bound diketides, tetraketides, pentaketides and hexaketides were identified and characterised by FT‐ICR‐MS, providing direct insights on active site accessibility and substrate processing for this enzyme class.

Polyketides constitute ap rominent family of structurally and functionally diverse secondary metabolites, comprising renownedp harmaceuticals, agrochemicals and other productso f commercial interest. [1] Their biosynthesis proceedst hrough multiple decarboxylative Claisen condensation steps, involving acyl carrier protein (ACP) bound malonates and ketosynthase (KS) bounda cyl units (Figure 1a nd 2A). Ap olyketide carbon backbonei sassembled and modified, while remaining PKSbound, by auxiliary enzymes (ketoreductases, KRs;d ehydratases, DHs;a nd enoylreductases, ERs), until it is eventually released from the PKS (typically by thioesterase (TE) mediated hydrolysis/cyclisation)a nd further enzymatically elaborated to the matureb ioactive product. [2] PKSs are classifieda s" modular" or "iterative" andi nto different types according to their structural organizationand modus operandi. [3] Clinically important compounds such as the anticancer agents doxorubicin and daunorubicin are products of type II iterative polyketide synthase (iPKS) biocatalysis. [4] Iterative PKSs comprises ingle enzymes (type III), single multi-domainm odules (type I), or discrete enzymes (type II) which repetitively employt he same catalytic activities to assemble and modify polyketide carbon chains. In comparison to modularP KSs, for which the natureo fp olyketide products is mostly predictable on the basis of module number and composition and can be altered or evolved, [5] the investigation and the re-programming of iPKSs remainchallenginga nd underexploited.
Within iPKSs,t ype II systems are distinctive biomolecular factories:t hey are found prevalently in Gram positive Actinomycetes [4] and are made of discrete proteins acting in ac oncerted manner to ultimately generate complex aromatic metabolites, including tetracyclines, anthracyclines, benzoisochromanequinones,t etracenomycins, aureolic acids, angucyclines and pentagular polyphenols. Type II PKSs closelyr esemble type II fatty acid synthases (FASs) in their essential mechanisms of substrate processing, however they differ in terms of intermediate nature and substrate binding modes adopted by their essential ACP components. [6] Over time, the complex nature of proteinprotein and protein-substrate interactions,a sw ell as the fast kinetics of product assembly presented by these enzymes, have been the object of intense scrutiny. [7] For these studies model type II "minimal"P KS systemsh ave often been used. In this work, the actinorhodin minimal system has been our model system of choice.
At ype II "minimal"P KS is constituted by ah eterodimeric ketosynthase (KS)-chainl ength control( CLF) domain, [8] which catalyses and controls polyketide chain initiation and elongation;a nd by ad iscrete acyl carrier protein (ACP). This ACP delivers malonyl buildingb locks and intermediates to the KS-CLF Figure 1. SEK4/SEK4b biosynthesis by the type II actinorhodin (act)"minimal system":malonyl ACPdecarboxylative Claisen condensation, driven and controlled by act KS-CLF (Figure2A), generates an ACP-bound octaketide (2). In the absenceo ffurtherenzymatic processing(e.g. by aKRd omain), SEK4 and SEK4b (4 and 5)a re the mainproducts resulting from spontaneouso ctaketide cyclisation, dehydration and aromatisation.L egend:A CP = acyl carrier protein;KS-CLF = ketosynthase-chain length factor;K R= ketoreductase. complex to construct ap olyketonec hain via the 4'-phosphopantetheine (PPant) cofactor ( Figures 1a nd 2A). [9] Am alonyl Coenzyme A: ACP transacylase (MCAT) normally provides malonyl extender units to discrete type II ACPs, however this is not strictly required for am inimal system to functiona st ype II ACPs can self-malonylate. [10] In the absence of the ketoreductase actIII the postulated ACP-boundo ctaketide (2)s pontaneously folds to afford shunt products:i nt he case of the aromatic antibiotic actinorhodin, these are the octaketides SEK4 and SEK4b (4 and 5 respectively,F igure 1). [8,10] In the presence of the ketoreductase actIII the postulated ACP-bound octaketide (2)i sc onvertedt om utactin (3,F igure1), [11] whereas the combined action of actIII-actVII (act ketoreductases,a romatase, cyclase and oxidases) ultimately convert 2 to actinorhodin (6). [12] For type II PKSs, malonyl-CoA remain the only elongating unit known to date, whereas av ariety of acyl buildingb locks (e.g.,a cetate, propionate,( iso)butyrate, benzoate…) can be used to prime these enzymes,a nd, together with post-PKS tailoring enzymes( e.g.,c yclases, oxidases, aromatases and methyltransferases), contributet ot he structural variation of type II PKS products. [4] In vitro reconstitution of enzyme activity [8,10] and in vivo genetic manipulation [13] have proved crucial in gathering the first insights into the determinantso fe nzyme priminga nd chain length control.M ore recent biophysical studies (e.g.,X -ray crystallography [14] and NMR [15] )o ft ype II PKS proteins have provided more in-depthk nowledge on protein/ substrate recognitiona nd productivec onformations. Someo f these investigationsh ave relied on the enzymatic loading of synthetically prepared intermediate mimics of ACPs, as the natural ACP-bound intermediates are intrinsically unstablea nd highly reactive. [15,16] The general inability of directly monitoring iterative intermediate formation and processing in real time constitutes as ignificant hurdle in gathering new knowledgeo f enzymek inetics and protein-substrate interactions required to devise novel syntheticb iology. [17] In our labs we have established ac hemical 'chain termination" methodology aimed at the captureo ft ransientp olyketide biosynthetic intermediates in vitro [18] and in vivo. [19][20] The method is based on the use of small molecule probest hat are nonhydrolysable mimics of ACP-bound malonate units. In competition with these last, the probesreact with enzyme-bound biosynthetic intermediates to off-load them in ar eadily availablea nd stable form for LC-MS characterisation. More recently we also reported the development of nonhydrolysable mimics of PCP (peptidyl carrierp rotein)-bound amino acids for the investigation of nonribosomal peptideassembly. [21] Herein we sought to extendt he scope of our methodology to develop protein-based tools aimed at intermediate capture in am ore stringent fashion. In particular, we envisaged that an acyl carrier protein, modified with an onhydrolysable mimic of malonyl pantetheine,may act as an intermediate "trap" for biosynthetic intermediates involvedi nt ype II PKS assembly,i n competition with the natural malonyl ACP 7 (Figure 2). In order to explore this and gather novel insights into the act type II PKS minimal system,aphotoactivatable acyl carrierp rotein 9 was chemoenzymatically prepared according to Schemes S1 and S2 in the Supporting Information. The 4,5-dimethoxy-2-nitrobenzyl (DMNB) photolabile group [22] was chosen as am eans to protect the pseudo-malonate moiety of carba(dethia) malonyl ACP 8, due to its relative ease of synthetici ncorporation and controllable removal at an on-protein damaging wavelength (365 nm). From commerciallya vailable d-pantothenic acidt he photolabile pantetheined erivative 10 was synthesised in 6steps (Scheme S1) and employed as substrate for the recombinant E. coli enzymes PanK, PPAT and DPCK, [23] to generate the corresponding coenzymeAphotolabile derivative (Scheme S2). Upon incubation of the latter with recombinant act apo-ACP and the phosphopantetheinyl transferase Sfp, [24] the desired photolabile ACP probe 9 was obtained ( Figure 2B). Light-activation of 9 was tested in the absence andi nt he presence of the act KS-CLF, employing either aK iloArc Broadband Arc Lamp or in an in-house built light box containing ac ircular 22 WU VA lamp [22] (Supporting Information). While isolated 9 could be deprotectedt o8 within a4hour irradiation period ( Figure S2), negligible deprotection of 9 took place in the presence of the act KS-CLF ( Figure S3). Enzymatic assays for the production of SEK4 and SEK4b by hexahistidine-tagged act ACP and KS-CLF were set up as previously reported, [10,25] adding the pre-photolysed and untagged ACP 8 at different times andv ariable concentrations( Ta ble S1). Enzymatic assay filtration throughN i-NTAa garose beads was carriedo ut in order to selectively isolate any species deriving from the "unnatural" ACP 8.T he samples recovered from this operation were concentrated and buffer-exchanged ahead of direct infusion into an FT-ICR-MS spectrometer (Supporting Information). The outcomeo ft hese experiments is illustrated in Figure 2 and detailed in the Supporting Information (Table S1).
In selected samples, putative ACP-bound nonhydrolysable intermediates, including di, tetra,p enta and hexaketide species (including mono-and di-dehydros pecies) were identified and characterised by HR-MS analysiso fp rotein-charged states ranging from 11 + to 6 + ;t hese species were absent in control samples (Supporting Information). Also, in assays of KS-CLF with 13 C 3 -malonyl ACP (instead of 7)a nd 8,p utative ACPbound nonhydrolysable species bearing an even number of 13 Ca toms were observed( Figure 2C and Supporting Information), consistent with their expected polyketide nature. Overall, in the analyses of KS-CLF assays in the presence of 8,c arba(dethia) acetyl-ACP was the most abundants pecies, whereas the putative captured intermediates were presenti nl ow abundance. To characterise further these speciesw ithouta dditional sample manipulation, the 4'-phosphopantetheine( PPant)e jection assay [26] was attempted directly on the heterogeneous samples infusedi ntot he FT-ICR-MS, however this did not lead to small molecule detection. The 4'-PPante jection assay employs collisionally activated dissociation (CAD) to preferentially cleave the 4'PPant ion. However,w ec annot exclude CAD possibly fragmenting the enzyme-boundp olyketide speciesi nt he conditions employed to analyse such complex mixtures. In order to improvet he confidence with whichw ei dentified low abundance putative ACP-bound species, we used autocorrelation to identify periodic patterns in data. The result of the isotopic distribution of the ion at the right charges tate can indeed support the identification of signals close to the noise threshold, as demonstrated by Palmblad et al. [27] The resultso f this approach applied to our samples supported the assignment of the manually identified putative species( see Supporting Information).
Amongst the identified species, ACP-bound diketide and tetraketide species were mosta bundantlyd etected in samples deriving from the simultaneous addition of both 8 and 7, whereas ap utative di-dehydro hexaketide ( Figure S9) was the most commonly observed product, in av arietyo fc onditions (Table S1). Parallele xperiments conducted using photolabile Nacetylcysteamine-based chain termination probes, [22] in the in vitro assembly of SEK4 and SEK4b, did not lead to any offloaded putative intermediates (data not shown), leading us to postulate that the ACP probe interacts more efficiently with the minimal system.
The assembly of actinorhodin by the act PKS, as well as that of relatedp roducts and other typeIIP KS-derived metabolites, has been the object of intensive scrutiny and still holds a number of unresolved questions, including severalc oncerning intermediate sequestration and stabilisation and protein-protein interactions. [15] Herein we have shown that, through the use of ac hemoenzymatically generated malonyl ACP nonhydrolysable mimic (8), direct evidence of novel ACP-bound polyketones pecies involved in type II PKS assembly can be obtained. These species, in contrast to others previously reported, [28] are nonhydrolysable from the carrier protein, hence they should constitute useful chemical biologyt ools for mechanistic and structuralinvestigations.
The ability of 8 to intercept putative biosynthetic intermediates in vitro, such as those herein presented, and, conversely, the inabilityo fN-acetylcysteamine based probes to do so (data not shown), support that, for the type II act minimal system,t he KS-CLF active site is mostly accessible to ACPbound substrates rathert han free species. The varied nature of putative ACP-bound intermediates observed in our experiments, including in those where the unnaturalp seudo-malonyl ACP 8 was presenti nd efectt ot he natural malonyl ACP substrate 7,s uggests dynamic interactions between the KS-CLF and the ACP,w ith proteins thatc an interchange, [29] and protein-protein interactions efficiently guidingA CP-bounds ubstrates into the KS active site for processing. [30] The putative captured ACP-bound species detected in thesee xperiments may possibly reflect the kinetics of carbon chain assembly and folding, with specific steps, for example, diketide and tetraketide formation,r elativelys low in comparison to others. This would be in agreement with crystallographic studies of act KS-CLF "caught in action" with adiketide and atetraketide species bound to its cysteine active site. [14] Further work will be required to corroborate the preliminary insights gathered by our experiments. Nonetheless, the ACP probe 8 herein prepared and evaluated constitutes ar are example of protein "trap" for biosynthetic species [31] and the first for polyketide synthases. Its use in conjunction with advanced FT-ICR-MS analyses and data analysis tools represents an ew promising approachf or the study of challenging biosynthetic enzymes that make use of dynamic carrier proteins,i ncluding fatty acid synthases and nonribosomal peptidesynthetases.

Experimental Section
The chemoenzymatic preparation of 9,i ts photolysis to 8 and use of 8 in enzymatic assays generating SEK4 and SEK4b, as well as FT-ICR-MS analyses of putative captured enzyme-bound intermediates, are reported in the Supporting Information.