Demonstration of starter unit inter-protein transfer from a fatty acid synthase to a multidomain, non-reducing polyketide synthase

Starter unit selection is a central “programming” event in the biosynthesis of polyketide natural products and an important contributor to their structural diversity.^[1] Among the iterative, non-reducing polyketide synthases (NR-PKSs) found in fungi this task is carried out by an N-terminal starter unit acyl transferase (SAT) domain. Generally these domains accept acetyl-CoA ubiquitously available in cells to initiate elongation of poly-β-ketide intermediates and account for the classical observation of a “starter unit effect.”^[2] Variations on this pattern are known, however, where a specialized fatty acid synthase (FAS)^{[2b, 3]} or a more highly-reducing PKS^[4] independently prepare a structurally more complex starter unit, which is selectively utilized to prime overall synthesis. An instance of the former is exemplified in the aflatoxin pathway.^{[3b, 5]} Here the FAS, HexS, comprised of yeast-like subunits HexA and HexB provides a C₆ fatty acyl starter unit 1 for the NR-PKS, PksA, to initiate biosynthesis of the aflatoxin (3) precursor norsolorinic acid anthrone (2, Figure 1). Not only does HexS generate a much shorter acyl chain than primary fungal FAS but also it appears to organize into an unusual α₂β₂γ₂ complex with PksA.^[3a] In mammals fatty acid synthase (FAS) is a seven domain α₂-homodimer whose thioesterase (TE) domain yields free fatty acid products. In fungi, FAS is an eight domain α₆β₆-heterododecamer lacking TE-domains that uses a malonyl/palmitoyl transferase (MPT) domain to eject products as fatty acyl-CoA thioesters. In vitro radiochemical studies of the HexS•PksA complex failed to show that the C₆ unit was released as free hexanoyl-CoA, thus implying, but not proving, that direct channeling to PksA had occurred. Multidomain protein dissection guided by the UMA algorithm^[6] has been shown to be a valuable avenue for deducing individual domain function by providing catalytically competent PKS fragments for in vitro reassembly.^{[2b, 7]} This “deconstruction” approach was employed to obtain the domain combinations necessary to reconstruct transacylation between HexS and PksA for detailed analysis.

Figure 1.

Initiation of aflatoxin biosynthesis by the fatty acid synthase HexA/B and type I iterative polyketide synthase PksA requires transfer of a hexyl starter unit from HexA to PksA.

The first step toward mimicking the transacylation chemistry was to generate acylated holo-ACP_HexA. Post-translational activation of apo-ACP_HexA is carried out in vivo by the 4’-phosphopantetheinyl transferase (PPT) domain encoded at the C-terminus of the HexA polypeptide.^[7b] The native reaction involves transfer of a phosphopantetheine arm from CoASH to the ACP. PPTases such as the type II bacterial PPTase Sfp have been extensively used to catalyze transfer of phosphopantetheine arms from CoA analogs regardless of the group attached to the cysteamine thiol.^[8] Under the assumption that the modified pantetheine arm should be inconsequential, 0.1 equivalent of PPT_HexA monodomain was used in vitro to activate ACP_HexA monodomain with hexanoyl-CoA. Surprisingly, ACP_HexA was post-translationally modified in relatively poor catalytic fashion as evidenced by MALDI-TOF analysis (Figure 2 and S1). Under the same reaction conditions, similar results were seen using CoASH, butanoyl-CoA, and octanoyl-CoA. Quantitative conversion of apo- to holo-ACP_HexA was accomplished only with a stoichiometric amount of PPT_HexA. The activation inefficiency may be a reflection of the chosen cut sites or a consequence of domain dissection itself where a normally in cis process is rendered in trans. Untethered, flexible linker sequence in the monodomain construct could inhibit proper protein–protein docking, a result of imperfect boundary prediction strategies. At least 50% of the protein sequence in the primary fungal FAS α and β subunits contributes not to catalysis but to structural integrity,^[9] so stripping away the intricate scaffolding may impair domain–domain recognition and reduce catalytic efficiency. It is worthwhile to note for comparison that the PPT and ACP monodomains from the S. cerevisiae primary metabolic FAS were shown to behave with typical catalytic efficiency for phosphopantetheinylation, requiring only 0.03 equivalents PPT for quantitative holo-ACP formation.^[10] PPT_FAS was cloned as a monodomain that is 8 residues shorter than PPT_HexA, and an ACP_FAS monodomain was 23 residues shorter at the N-terminus and 10 residues shorter at the C-terminus than the ACP_HexA monodomain. In particular, the ACP_FAS N-terminal cut site was guided by low complexity repeats that are not present in the HexA sequence. These details may imply a further functional departure that HexS exhibits when compared to fungal primary FAS.

Figure 2.

MALDI-TOF analysis of ACP_HexA post-translational modification via PPT_HexA. A) ACP_HexA activated with CoA and 0.1 equivalent (upper spectrum) or 1 equivalent (lower spectrum) PPT_HexA. Peak 1, apo-ACP (25024 Da); peak 2, holo-ACP (25364 Da); peak 3, apo-ACP + CoA adduct (25791 Da). B) ACP_HexA activated with hexanoyl-CoA and 0.1 equivalent (upper spectrum) or 1 equivalent (lower spectrum) PPT_HexA. Peak 1, apo-ACP; peak 2a, holo-ACP; peak 2b, hexanoyl-holo-ACP (25463 Da); peak 3, apo-ACP + hexanoyl-CoA adduct (25891). The non-covalent adducts with CoA and acyl-CoA^[15] can be attributed to the high CoA concentration relative to ACP. The smaller peak adjacent to each labeled peak is the α-N-d-gluconoyl-ACP species (+ 178 Da), a modification often seen in N-terminally His-tagged proteins expressed in E. coli.^[16]

PPT_HexA also appears to be biased against any acylated CoA, and requires high enzyme concentration and an extended reaction time to quantitatively convert apo-ACP_HexA to acyl-holo-ACP_HexA (Figure S2). From an efficiency viewpoint, discrimination against acyl-CoA ensures that an unmodified pantetheine arm is delivered to ACP_HexA, thus providing the activated enzyme form poised for hexanoyl synthesis. Under conditions that promote formation of acyl-holo-ACP_HexA, the acyl chain was found to hydrolyze from the phosphopantetheine arm to yield a mixture of holo-ACP_HexA and acyl-holo-ACP_HexA. Control experiments with ACP_PksA monodomain^[2b] and its presumed cognate PPTase, NpgA,^[7b] verify that acyl hydrolysis did not occur to acyl-CoAs spontaneously in solution during the reaction period (Figure S3). This behavior is thus a characteristic of the ACP_HexA/PPT_HexA system and may be a self-editing function in a megasynthase that does not contain a TE domain capable of this action.^[7c] NpgA was also tested as a potentially more efficient ACP_HexA monodomain activator based on the previous demonstration of NpgA–ACP_HexA transacylation using biotinylated CoA,^[7b] but NpgA performed even more poorly than PPT_HexA when assessed by MALDI-TOF (Figure S4).

Owing to the sensitivity of radiochemical assays, the small amount of hexanoyl-holo-ACP_HexA produced by in vitro activation was sufficient to monitor potential transacylation events. [1-¹⁴C]Hexanoyl-holo-ACP_HexA monodomain^[7b] was prepared as described above and mixed with SAT_PksA monodomain^[2b] to provide the minimal system expected to effect transthioesterification (Figure 3A). Because SAT_PksA can be acylated by free hexanoyl-CoA in solution,^[2b] unbound hexanoyl-CoA was removed from the hexanoyl-holo-ACP_HexA preparation by diafiltration. Domains with active site catalytic residue mutations (ACP-S179A and SAT-C117A) were also assayed to confirm that labeled hexanoyl chain bound to SAT_PksA originates from material delivered by ACP_HexA. Separation of the protein mixture by SDS-PAGE and subsequent autoradiography revealed rapid and efficient radiolabeled C₆ substrate transfer from ACP_HexA to SAT_PksA, consistent with the direct substrate channeling hypothesis. The equilibrium of this transfer appears to heavily favor SAT acylation, as a very small proportion of the hexyl chain remains bound to ACP_HexA following incubation with SAT_PksA.

Figure 3.

Monitoring ACP_HexA-SAT_PksA transacylation using [1-¹⁴C]hexanoyl-holo-ACP_HexA. A) Verification of ACP_HexA-SAT_PksA recognition. Reaction mixtures were separated by SDS-PAGE and visualized by autoradiography. Lane 1, ACP; lane 2, ACP-S179A; lane 3, ACP + SAT; lane 4, ACP-S179A + SAT; lane 5, ACP + SAT-C117A; lane 6, ACP-S179A + SAT-C117A. B) Autoradiograms of SDS-PAGE separations of transacylation time point assays between ACP_HexA and SAT or SAT-KS⁰-MAT. The proteins were denatured with SDS loading dye that did not include reducing agents, which causes SAT-KS⁰-MAT to resolve as three isoforms. Heat denaturation and reduction with thiol reagents were avoided to prevent substrate hydrolysis or transthioesterification. C) Quantitation of transacylations by band densitometry. Percent transfer to SAT was determined using the ratio of the SAT and ACP band densities.

This radiochemical method for tracking transacylation was extended to determine the effect that domain dissection exerts on system kinetics. A time course transacylation assay compared the relative efficiency of transfer to SAT_PksA monodomain, SAT–KS⁰–MAT tridomain^[7a] and SAT–KS⁰–MAT + PT–ACP⁰–TE/CLC tridomains (Figure 3B). The KS and ACP_PksA active sites were mutated in the tridomains to ensure that the hexanoyl chain remained bound to SAT. To limit potential hydrolysis of the labile bound thioester, transacylation reactions were quenched with SDS loading dye that lacked a reducing agent. SAT-KS⁰-MAT was purified to a single species under reducing (β-mercaptoethanol) SDS-PAGE conditions, but resolves as three isoforms in the absence of reducing agent (Figure S5). Thus, the three bands present in the autoradiograph of labeled SAT-KS⁰-MAT are isoforms present under nonreducing conditions. Transacylation to SAT–KS⁰–MAT at 25 °C was too rapid to reliably measure early time points, and the hexanoyl chain was noticeably hydrolyzed from SAT–KS⁰–MAT at later time points (data not shown), so reactions were carried out at 4 °C to slow the transfer for accurate quantitation. The relative rate of transfer to the SAT–KS⁰–MAT tridomain (0.10 ± 0.007 s⁻¹) was three-fold faster than transfer to the SAT_PksA monodomain (0.029 ± 0.001 s⁻¹) at 4 °C as measured by band densitometry. The presence of the KS and MAT domains presumably enhances the structural integrity of SAT in the more native-like tridomain construct for docking of ACP_HexA to SAT_PksA. A similar three-fold kinetic effect was observed in simpler transacylation experiments comparing PksA SAT or SAT–KS⁰–MAT⁰ using hexanoyl-CoA.^[7c]

The dissected system was next evaluated for the potential influence other PksA domains (PT, ACP_PksA, and TE/CLC) may exert on FAS-PKS association. The rate of transacylation was unchanged when ACP_HexA and SAT–KS⁰–MAT were incubated in the presence of the PT–ACP⁰–TE/CLC tridomain (0.10 ± 0.004 s⁻¹), indicating that PT–ACP⁰–TE/CLC is not involved in the ACP_HexA–SAT interaction. This observation supports the proposed spatial domain organization model of PksA,^[11] based on the α₂-mammalian FAS structure and the dimeric double hot-dog fold of the PT domain of PksA, which is shared with the dehydratase domain in animal FAS. In this model, SAT, KS, and MAT are situated proximal to each other, analogous to the AT and KS domains of mammalian FAS, while PT, ACP, and TE/CLC are expected to reside distal to the SAT–KS–MAT domain cluster.

These results suggest that within the HexS complex ACP_HexA must be more externally accessible than its fungal ACP_FAS counterpart. In the crystal structure of fungal primary FAS,^{[9, 12]} ACP domains are flexibly tethered inside a dome structure formed by three β-subunits that sit on either side of a 6 α-subunit wheel structure. This essentially traps the ACP domains inside the active complex, as the openings in its walls are sized to permit only small molecule diffusion (~25 Å) and would not accommodate an ACP exit or SAT entrance. Because HexS is believed to assume a simpler α₂β₂ heterodimeric structure, it is not as likely to sequester ACP inside such an elaborate architecture—a key structural change that is functionally consistent with the FAS–PKS pair being able to pass the hexyl starter unit by FAS ACP–PKS SAT interaction.

Some SAT domains have been shown to be nonessential for PKS priming if a suitably activated starter unit is provided to the system.^{[4, 7c, 13]} In these situations, the KS is thought to be directly loaded from acyl-CoA, or to a lesser extent in some cases from an acyl-SNAc ester, when the SAT domain contains an active site mutation or natural motif that would prevent covalent substrate linkage. Acyl-CoA is significantly preferred over acyl-SNAc in these examples, presumably because CoA offers a complete pantetheine arm reminiscent of substrate delivery by an ACP domain. We sought to determine if ACP_HexA-delivered substrate in the HexS•PksA system could function by a similar bypass strategy. Using a radiochemical assay analogous to those mentioned previously, hexanoyl-holo-ACP_HexA failed to acylate SAT⁰–KS–MAT, a PksA tridomain with an inactive SAT domain (C117A) (Figure S5). This observation confirms that SAT_PksA is an essential gatekeeper for substrate selection and cannot be bypassed to directly acylate the KS with substrate delivered by ACP_HexA, consistent with previous observations of reconstituted activity in PksA.^{[2b, 7c]}

The success of direct radiolabeled substrate transfer between ACP_HexA and SAT suggests that, at the very least, a weak or transient protein-protein interaction is occurring in the system. To obtain a more direct measure of ACP_HexA•SAT_PksA affinity, pull-down assays using Ni-resin-immobilized SAT monodomain or tridomain as bait for ACP_HexA were carried out (Figure 4). Following washes with low ionic strength buffer to minimize disruption of noncovalent interaction, yet to also remove unbound species, column-bound bait and associated prey were eluted by competition with imidazole. Based on visual comparison of Coomassie-stained band intensities on SDS-PAGE, SAT-KS⁰-MAT has a greater relative affinity than SAT_PksA for ACP_HexA, a characteristic consistent with the relative transacylation rates derived from the radiochemical experiments above. The increased affinity of SAT-KS⁰-MAT and ACP_HexA further supports the faster observed transacylation rate and suggests that SAT adopts more native-like organization in the tridomain structure for efficient ACP_HexA docking.

Figure 4.

SAT_PksA and SAT-KS⁰-MAT pull-down of untagged ACP_HexA. His-tagged SAT_PksA (A) or SAT-KS⁰-MAT (B) was immobilized on Ni-NTA resin and used as bait for ACP_HexA[His-tag]. M, molecular weight marker; lane 1, column flow-through after binding steps; lane 2, wash step 1; lane 3, wash step 4; lane 4, eluate after application of 250 mM imidazole buffer to displace the His-tagged complex

Other biosynthetic systems that employ a fatty acid starter unit to prime elongation by a type III PKS are known and provide both parallel and contrasting models for the work described here.^[14] In the case most similar to HexS•PksA, a bacterial type I FAS in Azotobacter vinelandii was shown to make C₂₂-C₂₆ fatty acids from [¹⁴C]malonyl-CoA and transfer them to two different type III PKSs for phenolic lipid biosynthesis.^[14c] Transacylation is presumed to occur between the FAS ACP and the PKS because the FAS lacks a TE or MPT domain for acyl chain hydrolysis or CoA ester formation, respectively, and mutation of the PKS active site cysteine prevented localization of radiolabeled starter unit onto the protein. An alternative strategy for fatty acyl starter unit incorporation is seen in the biosynthesis of alkylresorcylic acids by Myxococcus xanthus, in which fatty acids from primary metabolism are activated by a fatty acyl-AMP ligase and covalently bound to a type II ACP, which then primes a type III PKS.^[14d] Neurospora crassa employs a more direct approach to pentaketide alkylresorcylic acids by loading the type III PKS with acyl-CoAs taken from primary fungal metabolism.^[14b] Finally, Dictyostelium discoideum biosynthesis of acylphloroglucinol differentiation-inducing factors occurs via Steely, a genetically fused type I FAS–type III PKS hybrid protein that presents the PKS with an ACP-bound hexanoyl unit.^[14a] In each of these examples an ACP, loaded in different ways, directly delivers the acyl “starter” to a type III PKS to initiate synthesis.

In the present studies minimal domain reconstitution has facilitated close radiochemical analysis of a FAS–NR-PKS partnership occurring in a fungal secondary metabolic pathway. The ACP_HexA-bound hexyl starter unit is transferred directly to the PksA SAT domain to initiate noranthrone biosynthesis. The three-fold difference in relative transacylation rates for SAT monodomain and SAT–KS⁰–MAT tridomain, coupled with enhanced relative affinity of ACP_HexA for the tridomain, suggest ACP_HexA•SAT_PksA contact is optimized when SAT is further stabilized within a more complete tridomain environment. A mechanism for selective FAS ACP activation with CoA is also intimated by the apparent aversion PPT_HexA exhibits for acylated CoA. Through the interposition of a SAT domain in NR-PKSs, HexS•PksA interprotein transacylation represents a distinct control point in the common goal to marry fatty acids and polyketides for biosynthetic purposes.