The Structural Relationship between Iterative and Modular PKSs

Abstract

Recent work has characterized the architecture of a highly-reducing iterative polyketide synthase (PKS), the mycocerosic acid synthase (MAS) (Herbst et al., 2016). Beyond sharing a structural similarity with the mammalian fatty acid synthase (FAS), the authors argue that the MAS architecture is representative of some PKS modules.

If ever there is a competition for the most elaborate enzymology performed, polyketide synthases (PKSs), the molecular machines responsible for the biosynthesis of diverse natural products, will be strong contenders for the gold medal. What makes PKSs fascinating is that they are not a single enzyme with a single activity, but rather an assembly of well-orchestrated enzymatic domains. Groups of enzymes termed “modules” biosynthesize polyketides not unlike a factory assembly line, handing over the product of one module to the next in line. These synthases are referred to as modular PKSs (modPKSs). Another class of PKS, referred to as iterative PKSs (iPKSs), utilize the equivalent of one module iteratively to build molecular complexity. For both modPKSs and iPKSs, we still lack the level of structural and functional understanding needed to harness their catalytic potential by reprogamming them to produce designer polyketides.

Over the last decade, structural biology has greatly contributed to our understanding of PKSs by providing molecular views of their components. In a recent report, Timm Maier and colleagues add to this growing collection by presenting crystal structures of mycocerosic acid synthase (MAS) from Mycobacterium smegmatis (Herbst et al., 2016). MAS is a highly reducing iPKS that is responsible for biosynthesis of long fatty acids, the mycocerosic acids, important components of the mycobacterial cell envelope. Herbst et al. employ a divide-and-conquer strategy often used in structural biology in which a large protein (MAS is 2,111 residues) is subdivided into smaller constructs that are more manageable and more likely to produce well-diffracting crystals. In this fashion, the authors determine the structures of the condensing region, which includes ketosynthase (KS) and acyltransferase (AT) domains, and the modifying region, which includes the ketoreductase (KR), dehydratase (DH), and enoylreductase (ER) domains.

The crystal structures clearly reveal that each of the KS, AT, DH, KR, and ER domains of MAS are folded and organized similar not only to those of iPKSs such as the mammalian FAS (Maier et al., 2008), but also to those of modPKSs such as the erythromycin synthase (Keatinge-Clay, 2012). The KS, AT, and the structured linker connecting them are equivalent to all published crystal structures of modPKS condensing regions. The DH displays the same dimerization interface as the DHs of modPKSs, the KR is monomeric as observed in structures of KRs from modPKSs, and the ER dimerizes as shown for at least the nucleotide-binding subdomain of an ER from the phthiocerol modPKS (PDB: 1PQW) (Gokhale et al., 2007).

The authors suggest that the architecture of the modifying region of MAS is representative of the modifying regions of some PKS modules. Here, it is important to note that the sequences within and surrounding the ERs of modPKSs are not all equivalent, and that the corresponding modifying regions may also diverge on a structural level (Figure 1). While an ER characterized from the phthiocerol synthase is apparently dimeric (PDB: 1PQW), ERs characterized from the spinosyn, curacin, and jamaicamide synthases are apparently monomeric (Zheng et al., 2012; Khare et al., 2015). This monomeric state might be due to the fact that the motif employed by dimeric ERs to dimerize is typically shorter in modPKS ERs and therefore insufficient to mediate dimerization. Furthermore, the ER/KR linkers of modPKSs (usually < 8 residues) are shorter than those of iPKSs (> 17 residues), as observed in the KR+ER didomain structure from the second module of the spinosyn synthase (Zheng et al., 2012). Thus, we can expect the ER-containing modifying regions of modPKSs to have at least two different architectures available to them depending on whether their ER domains and the ER/KR linkers enable dimerization or not. While the much more common architecture is apparently that of monomeric ERs, dimeric ERs likely reside within the epothilone, pellasoren, ajudazol, chondrochlorens, and phthiocerol modPKSs.

Figure 1. Sequence Alignment of ERs Indicate Whether They Are Dimeric or Monomeric.

The first part of the alignment shows the segment utilized by dimeric ERs to mediate dimerization. The shorter segment in SpnMod2-like modifying regions apparently does not mediate dimerization like the segment in MAS-like modifying regions. Crystal structures exist for the ERs of MAS, PpsC (the nucleotide-binding portion), JamMod2, CurMod5, and SpnMod2 (PDBs: 5BP4, 1PQW, 5DOZ, 5DP1, 3SLK). The second part of the alignment displays the ER/KR linker. Herbst et al. report similar SAXS profiles for MAS, EryAMod2, and msPksMod2. MAS, mycocerosic acid synthase, Mycobacterium smegmatis, A0R1E8; ssFAS, porcine FAS, Sus scrofa, A5YV76; PpsC, third module of the phthiocerol synthase, Mycobacterium tuberculosis, P96202; Aju, ajudazol synthase, Chondromyces crocatus, B1GYF7 and B1GYF9; Cnd, chondrochlorens synthase, Chondromyces crocatus, B9ZUJ6; Pel, pellasoren synthase, Sorangium cellulosum, I0J6Y3 and I0J6Y5; Epo, epothilone synthase, Sorangium cellulosum, Q9KIZ7; Rap, rapamycin synthase, Streptomyces hygroscopicus, Q54297, Q54296, and Q54299; Pks12, Pks12 synthase, Mycobacterium tuberculosis, I6XD69; Chl, chlorothricin synthase, Streptomyces antibioticus, Q0R4M5 and Q0R4M7; Spn, spinosyn synthase, Saccharopolyspora spinosa, Q9ALM5; Sor, soraphen synthase, Sorangium cellulosum, Q9ADL6 and Q53840; Jam, jamaicamide synthase, Lyngbya majuscula, Q6E7K0; Cur, curacin synthase, Lyngbya majuscula, Q6DNE2; Sti, stigmatellin synthase, Stigmatella aurantiaca, Q8RJY1; EryA, “erythronolide synthase,” gamma proteobacterium HdN1, E1VID6; msPks, pks, Mycobacterium smegmatis, Q3L885.

One domain not present in the MAS structures is the acyl carrier protein (ACP), which is known to be flexibly tethered and was not included in the constructs due to the possibility of it impeding crystallization. Nevertheless, the model of the complete MAS generated by Herbst et al. can be used to map out the range of motion available to the ACPs of highly reducing iPKSs as well as the ACPs of PKS modules containing dimeric ERs (Figure 2). In full-length MAS, ACP is covalently tethered to the KR domain through a flexible ~15 residue linker. Thus, this linker enables it to dock to each of the five enzymes—AT, KS, KR, DH, and ER—during the elongation of fatty acids into mycocerosic acids, a process that requires the addition of four methylmalonyl extender units. The ACPs of mammalian FASs enjoy similar freedom, with the caveat that their C-terminal end is bound to a mobile, monomeric thioesterase (TE) domain. The ACPs of ER-containing modules are often restrained by a C-terminal linker (on average 18 residues) covalently connected to downstream KS dimer. Such modules do not possess the signature for containing a dimeric ER, presumably because the ACPs would be too restrained to dock to the active sites of each enzyme. Modules containing dimeric ERs are located at the C-terminal ends of modPKS polypeptides, so their ACPs are not covalently restrained by a C-terminal linker.

Figure 2. Dimeric ERs in PKS Modules Can Cause Trouble on the Twofold.

If the architecture of the MAS modifying region (DH+ER+KR) were present in a PKS module covalently connected to a downstream module, as modeled in this stereodiagram, each ACP would not have adequate range of motion to access each of its cognate enzymes (AT, KS, KR, DH, ER, and the subsequent KS). As illustrated, a 15 residue linker connecting the MAS ACP to the last observable residue in the crystal structure of the MAS modifying region provides it sufficient freedom to dock to each cognate enzyme within MAS (represented by the blue components). The ACPs of iPKSs may enjoy similar ranges of motion. Equivalent architectures are also suspected within modPKSs such as the epothilone, pellasoren, ajudazol, chondrochlorens, and phthiocerol synthases; however, these modules are located at the C termini of PKS polypeptides, such that the ACPs are not covalently tethered to a downstream KS dimer. When an ER-containing module is covalently tethered to a downstream module, the linkers that join the ACPs to the KS dimer (an average-length linker of 18 residues is shown extending from an ACP to its anchor point on the KS domain) would not provide sufficient freedom for ACPs to access their cognate enzymes. Thus, ERs may have been evolutionarily relocated off the twofold axis of the synthase as suggested by the monomeric KR+ER structure from the second module of the spinosyn synthase.

The architectures of iPKSs and modPKSs are only starting to be elucidated. Perhaps the only synthase for which we have a complete architectural understanding is the fungal FAS (Jenni et al., 2007). This molecular machine possesses six equivalent reaction chambers, in which each condensing and modifying enzyme is firmly oriented with respect to one another and ACP is the only mobile element. The organization of the bacillaene synthase into a dense, ~100 MDa megacomplex may offer a glimpse into the higher order architectures of many PKSs (Gay et al., 2016). With these precedents, I believe that modPKSs, comprised by as many as 26 modules, are unlikely to exist as loose strands of domains with degrees of freedom between each of the condensing and modifying regions. As the Maier lab, our lab, and others in the community determine the atomic resolution structures of the components of large iPKS and modPKS machines, electron microscopy should be concomitantly and judiciously employed to help establish their relative orientations within their native, intact synthases.