DISCUSSION
Among the many docking events within a polyketide assembly line, those between KS’s and their upstream ACP’s are of high significance because they enable KS’s to gatekeep for properly processed intermediates and thus ensure the fidelity of the synthase18 . Although modern structural biology techniques, such as cryo-EM, have advanced our knowledge of the interactions between PKS domains, they are still resource-intensive and low throughput.7, 8 In contrast, modern in silico methods require few resources and are high throughput. Thus, we sought to identify the transacylation site using the most advanced folding/docking algorithms.
Alphafold-Multimer uses restraints derived from covariation analysis, both within and between domains, to guide folding/docking.15 In half of the 50 top solutions for the ACP+KS dimers, a structural feature is present at the interface that has not been previously observed - the TxLGDP motif of KS forms a type I β-turn rather than the α-helical turn present in reported structures and the other 25 top solutions. The β-conformation is correlated with a closer approach between the ACP and KS domains. The distance between the side-chain oxygen of the conserved ACP serine and the reactive KS cysteine is greater than 18.4 Å in 24 of the 25 top solutions with the α-conformation (20.8 Å, on average). The exception is AjuMod3 (15.8 Å), a module containing non-ribosomal peptide synthetase (NRPS) domains,36 in which an unusual phenylalanine in position 47’ makes contact with an unusual valine at position 275. The 17 top solutions that possess the β-conformation and a relatively rotated ACP possess distances between the phosphopantetheinylated serine and the reactive cysteine short enough for polyketides to transfer from the acyl-ACP to the KS (17.2 Å, on average).
The TxLGDP motif has not been experimentally observed in the β-conformation. However, these studies suggest that acyl chains are not readily transferred between ACP and KS when this motif is in the α-conformation. This raises the possibility that favorable interactions between the polyketide intermediate and the KS substrate tunnel facilitate the transition from the α-conformation to the β-conformation, which enables a closer association of the acyl-ACP and KS, productive interactions between KS and the phosphopantetheinyl arm (such as the coordination of H309 NεH and T311 OH with the distal carbonyl), and polyketide transfer through the approach of the thioester and the reactive cysteine. This may be part of the KS gatekeeping mechanism by which properly processed intermediates are selected for transfer to the reactive cysteine.18
During the reaction cycle of a module, ACP associates with the upstream AT and processing domains as well as the downstream KS (Figure 1). While it must be complementary to each of these cognate enzymes, how well ACP’s dock with noncognate domains introduced through natural recombination events or genetic engineering is unknown. As the domains of one module are able to evolutionarily co-migrate separately from the domains of other modules, their interfaces could diverge. If this is the case, engineering efforts should focus on module-swapping rather than domain-swapping. In a recent KS gatekeeping study, AmpKS15 was swapped for PikKS3 in PikMod3 ofP1 -P2 -P3 -P7 , a tetraketide synthase composed of PikMod1, PikMod2, PikMod3, and PikMod7.37 A triketide rather than the expected tetraketide was produced by the resulting synthase, indicating that the hybrid third module was completely skipped. An analysis of natural ACP/KS interfaces of AmpMod15 and PikMod3 reveals large differences in how their ACP’s interact with the KS surface residue in position 275 (asparagine in AmpKS15 and serine in PikKS3).
The docking solutions presented here show interfaces between KS’s and their upstream ACP’s that are quite distinct from one another (Figure 5, Table S2). The most dramatic example is the interface between AjuACP3 and AjuKS3, in which a valine occupies position 275 of AjuKS3 and AjuACP3 is rotated 17º relative to the consensus orientation. As mentioned above, KS’s containing a serine at position 275, such as PikKS3 and VstKS7, may not be recognized by most ACP’s since they associate with KS’s containing an asparagine at this position.
The transacylation site may be under selective pressure to diversify, whereas the extension site may be under selective pressure not to. The ordered assembly of PKS polypeptides could be jeopardized if the interfaces between KS’s and their upstream ACP’s were too similar. CDD’ and NDD’s do not make the only contact between PKS polypeptides, since ACP’s and KS’s also associate, at least while mediating the transacylation reaction and likely even when ACP is unacylated.27 The surface area for the transacylation site is larger than that observed for CDD/NDD interfaces (1338 Å2 for PikACP6 with PikKS6 vs. 1043 Å2 for one DEBS2 CDD helix with the dimeric DEBS3 NDD).26 If all the KS’s and ACP’s at the polypeptide termini equally docked with one another, the synthase could assemble incorrectly. In contrast, the evolution of polyketide assembly lines may benefit from the interface between KS’s and downstream ACP’s not diverging, such that a module still functions when moved to a new location in the same or a different synthase.7, 8 In the new location, the ACP and KS domains must respectively collaborate with a new upstream KS and downstream ACP to catalyze extension reactions.
The proposed transacylation site is consistent with the observed losses in activity from the 20 P1 -P6 -P7 point mutants (Figure 5b). That mutations to the conserved KS surface residues N275 and L315 resulted in the least active synthases agrees with the interactions between PikACP6 and PikKS6 proposed to enable the transacylation reaction. The AlphaFold-Multimer predictions show the sidechain of N275 making substantial contact with PikACP6, through Van der Waals contact with P80’ and hydrogen bond contact with both the C-terminal end of αIII’ and the N-terminal end of the one-turn helix upstream of αII’. The AlphaFold-Multimer predictions also show the invariant leucine in the TxLGDP motif at position 315 making significant hydrophobic contact with F77’ of PikACP6 when the motif is in the α-conformation and with a hydrophobic pocket on αII’ next to the phophospantheinylated serine when the motif is in the β-conformation. Each of the other mutations also resulted in losses of activity compared to unmutated P1 -P6 -P7 , usually moreso when located at or near the putative interface (e.g. , I319A, H281A, Q286A, D382A, D385A, Q330A). The relatively small decreases in activity observed for the R314A and Q322A point mutants might be surprising given the proximity of these residues to the interface; however, in the Alphafold-Multimer predictions the arginine and the glutamine sidechains make minimal Van der Waals contact with PikACP6. The low activity of the T140A point mutant control was expected from the importance of this residue at the PikKS6 extension site.7, 8
AlphaFold has apparently helped discover the transacylation site on the surface of KS and provided 50 examples of how upstream ACP’s dock at that interface. Such a feat through experimental structural biology techniques would require many more resources and much more time. AlphaFold will likely help predict other domain-domain interactions within PKS’s, although these should be experimentally tested within functional assembly lines. Our burgeoning understanding of how ACP domains dock with cognate enzymes is illuminating the dynamics and architectures of PKS’s and will help realize the potential of these assembly lines in producing designer molecules and medicines.