The Modules of trans-Acyltransferase Assembly Lines Redefined with a Central Acyl Carrier Protein

Abstract

Here, the term “module” is redefined for trans-acyltransferase (trans-AT) assembly lines to agree with how its domains cooperate and evolutionarily co-migrate. The key domain in both the polyketide synthase (PKS) and nonribosomal peptide synthetase (NRPS) modules of assembly lines is the acyl carrier protein (ACP). ACPs not only relay growing acyl chains through the assembly line but also collaborate with enzymes in modules, both in cis and in trans, to add a specific chemical moiety. A ketosynthase (KS) downstream of ACP often plays the role of gatekeeper, ensuring that only a single intermediate generated by the enzymes of a module is passed downstream. Bioinformatic analysis of 526 ACPs from 33 characterized trans-AT assembly lines reveals ACPs from the same module type generally clade together, reflective of the co-evolution of these domains with their cognate enzymes. While KSs downstream of ACPs from the same module type generally also clade together, KSs upstream of ACPs do not – in disagreement with the traditional definition of a module. Beyond nomenclature, the presented analysis impacts our understanding of module function, the evolution of assembly lines, pathway prediction, and assembly line engineering.

Introduction

Acyl carrier proteins (ACPs) form transient domain-domain interactions with cognate enzymes in their biosynthetic pathways^1–2. These interactions enable specific transformations to occur on the acyl chain covalently shuttled by that ACP. This holds for both Type II systems in which each domain is encoded on a separate polypeptide, as in most bacterial fatty acid synthases (FASs)^3–4, as well as Type I systems in which several domains are encoded on the same polypeptide, as in biosynthetic assembly lines formed by polyketide synthase (PKS)^5–6 and nonribosomal peptide synthetase (NRPS) machinery⁷.

The repeating unit of enzymatic assembly lines is known as the “module”. The domains within a module typically fall into one of four categories: 1) an enzyme that selects and transfers a building block to an ACP, 2) an ACP that covalently shuttles a building block and/or an acyl chain, 3) a processing enzyme that performs chemistry on the acyl chain, or 4) a chain extending/transferring/releasing (ETR) enzyme. Generally, modules extend an acyl chain with a carboxylic acid building block selected by an acyltransferase (AT) or an amino acid building block selected by an adenylation (A) domain. An ACP transporting an α-carboxyacyl group usually collaborates with an upstream ketosynthase (KS) domain to generate a carbon-carbon bond and extend the acyl chain, while an ACP transporting an α-aminoacyl group [a.k.a. peptidyl carrier protein (PCP) or thiolation (T) domain] usually collaborates with an upstream condensation (C) domain to generate a carbon-nitrogen bond and extend the acyl chain.

The biosynthetic community is working to decipher the logic of two large classes of assembly lines that contain PKS machinery - cis-AT assembly lines that primarily rely on embedded ATs^{5, 8} and trans-AT assembly lines that primarily rely on separately-encoded ATs^{6, 9}. From the sequencing of the erythromycin synthase in 1990 until last year, the boundaries of the modules of cis-AT assembly lines were incorrectly defined - comparisons with the domain organization of the mammalian FAS had led to the module being defined with KS at its upstream boundary and ACP at its downstream boundary. The processing enzymes in several cis-AT assembly lines have been shown to evolutionarily co-migrate with the KS downstream of them, leading to the redefinition of cis-AT modules as possessing a downstream KS^{5, 10}. When trans-AT assembly lines started to be discovered fifteen years ago, often from difficult-to-culture, symbiotic bacteria, the use of cis-AT nomenclature resulted in their modules being incorrectly defined as well^11–12. Only a few years later KSs that accept similar substrates in trans-AT assembly lines were oberved to clade together, suggesting that they collaborate most closely with the domains preceding them, not following them¹³.

Bioinformatic analysis of KS domains has helped predict the intermediates and products of trans-AT assembly lines¹³ (analyzing which activities are embedded in the assembly line is insufficient due to the common participation of trans-enzymes). Because the KSs of trans-AT assembly lines often serve as gatekeepers, the clade to which a KS belongs can suggest the chemical moiety most recently added to the acyl chain. However, relying on KS signatures alone has limitations since KS clades do not by themselves always reveal the chemistry of the acyl chain and KSs are often replaced by other ETR enzymes, such as a TE or C domain. To better predict products and decipher how the many module types of trans-AT assembly lines operate, all the domains of the module should be considered. An examination of trans-AT assembly lines in light of the updated module definition, reveals that ACP sequences can help determine the module types present in an assembly line (Figure 1).

Figure 1. Traditional and Updated Module Boundaries for a trans-AT Assembly Line.

The nosperin trans-AT assembly line can be depicted either with traditional modules that start with a KS or C domain and end with an ACP domain or with updated modules that contain a central ACP domain and a downstream extending/transferring/releasing (ETR) enzyme (KS, KS⁰, C, or TE). All but the twelfth module of the nosperin assembly line belong to the module types defined here (a–y). Gatekeeping ETR enzymes that select one intermediate over another and disconnections between assembly line polypeptides are indicated with gate icons and arrowheads, respectively.

That little research on the bioinformatics of the ACP domains of assembly lines has been reported is understandable at least in cis-AT assembly lines where ACPs appear quite homogenous. One study did uncover a signature in the ACPs of both cis-AT assembly lines (curacin and jamaicamide) and trans-AT assembly lines (bacillaene, batumin, mupirocin, pederin, and thiomarinol) that flags their β-ketoacyl substrates for β-branching reactions¹⁴. In each case, the acyl-ACP must associate with a hydroxymethyl-CoA synthase like enzyme (HCS) and two enoyl-CoA hydrolase like enyzmes (ECH1 and ECH2). Presumably, the signature is representative of unique features on the ACP surface that enable it to selectively associate with these enzymes.

Here, bioinformatic analysis of 526 ACPs from 33 well-characterized trans-AT assembly lines reveals that ACPs from β-branching modules are only one of several groups of ACPs from trans-AT assembly lines that clade together. Indeed, most ACPs from the same module type were found to be related to one another, reflective of their interactions with cognate processing enzymes that have been relatively conserved through evolution. KSs downstream of ACPs from the same module type also clade together, whereas KSs upstream do not. Beyond nomenclature, how the presented analysis impacts our understanding of modules, the evolution of assembly lines, pathway prediction, and assembly line engineering is discussed.

Materials and Methods

Well-characterized, nonredundant trans-AT assembly lines with reasonable biosynthetic models were chosen for analysis (albicidin, Alb, BGC0001088; bacillaene, Bae, BGC0001089; basiliskamides, Bas, BGC0000172; bongkrekic acid, Bon, BGC0000173; bryostatin, Bry, BGC0000174; calyculin, Cal, BGC0000967; chivosazole, Chi, BGC0001069; corallopyronin, Cor, BGC0001091; difficidin, Dif, BGC0000176; disorazole, Dsz, BGC0001093; migrastatin, Mgs, BGC0000177; elansolid, Ela, BGC0000178; enacyloxin, Ena, BGC0001094; etnangien, Etn, BGC0000179; griseoviridin, Sgv, BGC0000459; batumin, Bat, BGC0001099; kirromycin, Kir, BGC0001070; legioliulin, Lgl, BGC0000180; leinamycin, Lnm, BGC0001101; macrolactin, Mln, BGC0000181; mupirocin, Mmp, BGC0000182; myxovirescin, Myx, BGC0001025; nosperin, Nsp, BGC0001071; oocydin, Ooc, BGC0001031; onnamide, Onn, BGC0001105; oxazolomycin, Ozm, BGC0001106; patellazole, Ptz, BGC0001107; psymberin, Psy, BGC0001110; rhizopodins, Riz, BGC0001111; rhizoxins, Rhi, BGC0001112; sorangicin, Sor, BGC0000184; thailandamide, Tai, BGC0000186, and thailanstatin, Tst, BGC0001114). Modules from cis-AT assembly lines were also selected for comparison (amphotericin module 10, Amp10, BGC0000015; concanamycin module 6, Con6, BGC0000040; erythromycin module 4, Ery4, BGC0000055; mycolactone module 4, Myc4, BGC0000103; nystatin module 9, Nys9, BGC0000115; pikromycin module 5, Pik5, BGC0000094; rapamycin module 6, Rap6, BGC0001040; rifamycin module 5, Rif5, BGC0000136; spinosad module 5, Spn5, BGC0000148; and tylosin module 3, Tyl3, BGC0000166)^{6, 9, 15}.

All presented cladograms were similarly generated. First, sequences were aligned with MUSCLE¹⁶. Then RAxML performed rapid bootstrapping and a subsequent maximum-likelihood (ML) search on each alignment¹⁷. To construct trees, the LG amino acid substitution matrix as well as the GAMMA model of rate heterogeneity were utilized and the alpha-parameter was estimated by ML¹⁸. The random seed for parsimony inferences and for rapid bootstraps (500 for ACPs, 100 for KSs) was 22222. ACP and KS trees were rooted by representative domains from cis-AT assembly lines. Finally, cladograms were visualized using Dendroscope 3, with confidence intervals ≥60 displayed¹⁹.

Results and Discussion

Module Types and ACP Families Defined

To color ACPs in the ACP cladogram by the type of module to which they belong a definition for “module type” was developed. Defining modules only by the chemistry they perform does not sufficiently explain the clading observed in the ACP cladogram (Figure S1). Including the classes of enzymes they contain in the definition significantly helps. However, to best explain ACP clading, how ACPs and their cognate enzymes have co-evolved to interact with one another also needs to be considered. Since ACPs that belong to the same major divisions of the cladogram (families) are likely to be evolutionarily related, module types were analyzed in the context of each ACP family. Thus, each of Families I-VIII contain a different set of module types (a–z, with z representing all undefined module types). A cladogram of ACPs that belong to the defined module types was generated and colored by those module types (Figure 2).

Figure 2. Cladogram of ACPs Colored by Module Type.

An analysis of the ACPs from 33 trans-AT assembly lines and 10 ACPs from cis-AT assembly lines revealed that ACPs from the same types of modules clade together and helped to define the most common module types (Figure S1). Shown here is the cladogram of the ACPs from twenty-five module types (a–y), characterized by the moiety they add as well as the classes of enzyme they contain (KSs and KRs are divided into four clades as in Figures S2 & S3). ACP families (I–VIII) that contain related module types were also defined. Suspected gatekeeping enzymes are conservatively indicated with gate icons.

Within trans-AT assembly lines, ACPs most commonly collaborate with KSs and ketoreductases (KRs) that themselves are divided into clades. From cladograms of KSs in the analyzed assembly lines, four clades were defined similar to a previous analysis (Figure S2)²⁰. The KS_I clade contains KSs that gatekeep for C_α/C_β-double bonds. The KS_II clade contains KSs that gatekeep for C_α/C_β-single bonds. With the exception of KSs in the KS_III clade associated with β-branching, members of the KS_III and KS_IV clades generally do not appear to gatekeep. The structural biology of KSs supports these clade divisions (KS_I, MgsKS6, PDB 4TKT; KS_II, PksKS2, PDB 4NA1; KS_IV, OzmKS9, PDB 4OPF)^20–21. From a cladogram of KRs from the defined module types, four main clades were observed (Figure S3). The KR_I clade contains KRs that are from methyltransferase (MT)-less modules and generally A-type (A-type KRs generate L-β-hydroxy groups), the KR_II contains KRs that are from MT-less modules and generally B-type (B-type KRs generate D-β-hydroxy groups), the KR_III clade contains KRs that are from MT-containing modules and can be A-type or B-type, and the KR_IV clade contains KRs that are from the first module of dehydrating bimodules as well as some α-oxidizing modules and are A-type. The structural biology of KRs supports these clade divisions (KR_II, MlnKR7, PDB 5D2E; KR_IV, PksKR4, PDB 5KTK)^22–24.

ACPs Evolutionarily Co-migrate with Downstream KSs, not Upstream KSs

Cladograms were generated for the KSs downstream and upstream of ACPs that were assigned a module type (Figure 3). In the cladogram of downstream KSs, KSs were colored by the module type of the ACP they follow, and in the cladogram of upstream KSs, KSs were colored by the module type of the ACP following them. From this analysis it is apparent that KSs downstream of ACPs from the same module type clade with one another and that KSs upstream of ACPs from the same module type do not, with the expected exception of those in the first module of a bimodule¹³. This analysis supports the redefinition of trans-AT PKS modules such that ACPs are grouped with the KSs downstream of them^{5, 10}.

Figure 3. Cladograms of KSs Upstream and Downstream of ACPs Support the Updated Module Boundaries.

A) The lack of clustering for KSs upstream of ACPs that clade together, with the anticipated exception of KSs from the first module of bimodules, indicates ACPs do not usually evolutionarily co-migrate with the KS upstream of them. B) Clustering is observed for KSs downstream of ACPs that clade together. This observation suggests that ACPs evolutionarily generally co-migrate with the KS downstream of them, reaffirming the updated module boundaries. Colored dots indicate condensation-incompetent KS⁰s.

The collaboration of processing enzymes with a downstream KS (or other ETR enzyme) is often required to ensure that a single acyl intermediate is passed to the next module in the assembly line. While processing reactions driven by the high NADPH/NADP⁺ and SAM/SAH ratios in the cell [e.g., those mediated by KR, enoylreductase (ER), and MT], essentially go to completion given a chance, others are in an equilibrium such that an ACP presents more than one acyl chain to the KS downstream of it. Thus, KS_Is play a gatekeeping role in modules that generate double bonds, accepting an α/β-unsaturated intermediate but not the more abundant β-hydroxy intermediate [interconverted by a dehydratase (DH) in these modules]. KS_IIs play a gatekeeping role in modules that generate β/γ-unsaturated intermediates from α/β-unsaturated intermediates [through a DH^*; KSs from a separate clade similarly gatekeep in the second module of enoyl-isomerase (EI) bimodules]²⁵. Some KS_IIIs in β-branching modules apparently gatekeep to select an intermediate with a β-exomethylene group over one with an α/β-olefin. In this work, gate icons have been conservatively placed next to ETR enzymes suspected of gatekeeping.

The abundance of condensation-incompetent KSs, termed KS⁰s (indicated by colored dots in Figure 3; usually these possess deviations in the HGTGT motif that in condensation-competent KSs contains the first histidine of the Cys/His/His triad), in trans-AT assembly lines illustrates that KSs have several important roles besides forming carbon-carbon bonds. Viewed from the perspective of cis-AT assembly lines that are comprised of very few KS⁰s, these domains seemingly perform unnecessary reactions, transferring the acyl chain from one ACP to another through an acyl-KS⁰ intermediate. However, KS⁰s can play a gatekeeping role and/or transfer acyl chains between different types of ACP such that the chain can be operated on by other processing or ETR enzymes. Catalytically-competent KSs can also perform these functions but mandate a chain extension; thus, KS⁰s also provide chemical flexibility to trans-AT assembly lines.

ACP Sequence Motifs and Structure

To better understand how ACPs flag their cognate enzymes, sequence alignments and sequence logos were generated for ACPs of each of the common module types (Figures 4 and S4)²⁶. As noticed with ACPs of β-branching modules (Family V)¹⁴, sequence motifs for many of the module types are evident. To obtain a better physical understanding of the motifs, the structural biology of assembly line ACPs was investigated.

Figure 4. ACP Signatures.

A stereodiagram of an ACP (Mmp8a from the mupirocin trans-AT assembly line, PDB 2L22) shows which positions in assembly line ACPs most frequently contact cognate enzymes in the six available complex structures (red indicates that in 6/6 structures a residue in that position is within 4 Å of the cognate enzyme). The four helices (I–IV), the phosphopantetheinylated serine (*), and the boundaries of the ACP domain (N and C) are marked. Sequence logos show the ACP signatures for each of the module types, which are indicated by the neighboring cartoon. Triangles indicate relative gaps or insertions: in Family I the loop between helices II and III is generally one residue shorter, in Family IV the loop between helices I and II is generally two residues shorter, in cis-AT ACPs the loop between helices I and II is generally one residue longer and contains a conserved histidine. The primary and secondary structure of Mmp8a at the bottom helps show where residues in the ACP signatures are structurally positioned and how likely they are to make contact with cognate enzymes.

Many examples of assembly line ACPs docking with cognate enzymes are known, even though all are from NRPS modules. Representative interactions between ACPs and embedded assembly line enzymes have been observed for the C domain (donor site: PDB 5EJD²⁷, acceptor site: PDB 4ZXH²⁸), the A domain (PDBs 4DG9²⁹), the epimerization (E) domain (PDB 5ISX³⁰), and the thioesterase (TE) domain (PDB 3TEJ³¹). Representative interactions between an ACP and a non-embedded P450 monooxygenase have also been observed (PDB 4PXH³²). With these six complexes an interaction map was constructed (Figure 4). Residues on a structurally-elucidated ACP from the mupirocin trans-AT assembly line (MmpACP8a, PDB 2L22)¹⁴ were colored from blue to red (with red representing that in 6 out of 6 structures the equivalent residue is within 4 Å of the cognate enzyme). While each interaction with a cognate enzyme is unique, the majority of contacts are made with ACP helices II and III. No structures of assembly line ACPs interacting with the enzymes of PKS modules are available; however, the VinK/VinL complex (AT_L and ACP_L from the vicenistatin cis-AT assembly line) may be representative of how ACPs and ATs associate (PDB 5CZD)³³.

While significant differences between module types are apparent, the largest distinctions occur between families (Figure 4). Compared to the average ACP, most ACPs in Family I contain one fewer residue in the loop connecting helices II and III, and most ACPs in Family IV contain two fewer residues in the loop connecting helices I and II. The loop connecting helices I and II is one residue longer in ACPs from cis-AT modules and contains a highly conserved histidine. On the surface, b modules from Family I and Family III appear similar, containing KR+MT+ACP+KS architectures and generating α-methyl, β-hydroxy moieties; however, their ACPs possess different signatures that likely reflect the distinct interfaces that each ACP forms with its cognate KR and MT.

Multiple ACPs in the same module, called tandem ACPs, are thought to increase the flux through an otherwise rate-limiting portion of the assembly line³⁴. This is evident in β-branching modules that typically contain two to three ACPs - for each of the individual ACPs to interact with HCS, ECH1, and ECH2 as well as AT, an upstream KS, and an ETR enzyme, they possess very similar sequences. In other types of modules, tandem ACPs are also highly homologous, appearing adjacent to one another in the ACP cladogram (Figure 2). Even ACPs separated by an enzymatic domain, like the two ACPs in the seventh module of the leinamycin assembly line located before and after an MT, are highly homologous. Divergent sequences of ACPs within the same module indicate they play different roles. The two ACPs in the termination module of the leinamycin assembly line apparently help with different steps in the construction of a 1,3-dioxo-1,2-dithiolane moiety³⁵. The two ACPs that appear to belong to the ninth module of the oxazolomycin assembly line actually interact with different downstream KSs, as discussed in the next section.

Pathway Prediction

The generation of biosynthetic models for trans-AT assembly lines, such as the basiliskamide, myxovirescin, and oxazolomycin is greatly facilitated by determining the module type to which its ACPs belong (Figure 5).

Figure 5. Pathway Prediction.

Three examples show how the new understanding of trans-AT modules can be employed to generate better biosynthetic models. A) The order and function of the basiliskamide polypeptides had been mysterious. Assembling them to generate module types defined here enabled the unambiguous ordering BasG, BasF, BasE, BasD, BasC. An unusual feature of the assembly line is the final C domain that apparently gatekeeps for an α/β-cis, γ/δ-trans diene, in contrast to KSs in its position that usually gatekeep for the α/β-trans, γ/δ-cis diene, but like the similarly-positioned macrolactin TE. B) The order and function of the first polypeptides (TaI, TaL, and Ta-1) of the myxovirescin pathway had also been cryptic. Assembling the polypeptides to form module types defined here, the order could be confidently determined. The first v module generates a propionyl group, possibly through the methylation and decarboxylation of a malonyl group. The second d module reduces the β-carbon and oxidizes the α-carbon. The third module is an uncategorized module type (z) that likely generates an α-hydroxy group through a DH-mediated dehydration to an α-keto group followed by KR-mediated α-ketoreduction similar to PksKR3 (Ref. 22). C) The unusual order of the domains of 9^th-11^th modules of the oxazolomycin pathway made deciphering that part of the pathway difficult. From their signatures, each of the domains in these modules reveal they are mixed with one another. Even so, the ACPs specifically collaborate with the domains of the module to which they belong. Other unusual features are present in the thirteenth module: a DH+ER+KR tridomain that possesses cis-AT architecture, an ACP that clades with cis-AT ACPs, and a docking interaction mediated by catalytically inactive O-methyltransferase (OMT⁰) and ACP⁰ domains. Icons and arrowheads are explained in Figure 1.

The current biosynthetic model for the basiliskamide pathway contains several problems (Figure 5a)^{6, 36}. Through elucidating the module types of this assembly line, a correct ordering of its polypeptides can be established (BasG, BasF, BasE, BasD, and BasC). The ACP in BasG is from a v module [GCN5-related acyltransferase (GNAT)-containing] and is thus the first in the assembly line. The ACP in BasF is from a j module, and the DH+KR_III missing from BasF is located at the C-terminal end of BasG. The first ACP of BasE is from a b module, and the KR_III missing from BasE is located at the C-terminal end of BasF. The tandem ACPs in BasD belong to the second module of a type B dehydrating bimodule of which the first module is located at the C-terminal end of BasE. The C domain encoded by BasC apparently substitutes for the second KS of type B dehydrating bimodules. Interestingly, this C domain and a TE that makes an equivalent substitution in the macrolactin assembly line both gatekeep for a α/β-cis, γ/δ-trans diene in contrast to the α/β-trans, γ/δ-cis diene normally selected by type B dehydrating bimodules (the dienes may be interconverted by DH)²⁴.

As the first part of the myxovirescin assembly line generates an unusual intermediate, the order and function of its first polypeptides (TaI, TaL, and Ta-1) had been difficult to assign (Figure 5b)^{9, 37}. However, by determining the module type to which its ACPs belong (the last ACP of TaI and the first KS of TaL collaborate in a d module and that the last ACP of TaL and the first C domain of Ta-1 collaborate in a u module) the polypeptides of the assembly line can be ordered TaI, TaL, Ta-1. The first module, which contains an MT, a GNAT, an ACP, and a KS, likely generates a propionyl group similar to the saxitoxin, enacyloxin, and YM-47522 (related to basiliskamide) pathways³⁸. The second module likely reduces the β-ketone with a B-type KR to generate a D-β-hydroxy group and oxidizes the α-methylene with a trans-Ox to generate a D-α-hydroxy group (d modules are only known to generate D-α, D-β-dihydroxy and L-α, L-β-dihydroxy moieties). The third module may operate similar to Cal13 to generate an α-keto group with its DH³⁹ before its KR performs an α-ketoreduction similar to PksKR3 to install an L-α-hydroxy group²². The KS⁰ of the third module serves as an adapter, passing the acyl chain to an ACP that functions with its cognate C domain in the fourth module. Another correction to the myxovirescin biosynthetic model is in the seventh module, which possesses the signatures of a d module and generates a D-α, D-β-dihydroxy moiety.

The oxazolomycin assembly line contains several unusual features that had made proposing its biosynthetic model challenging (Figure 5c)^{6, 40}. In OzmH, the 9^th–11^th modules mix with one another. All of the domains of a type A dehydrating bimodule that comprises the 9^th and 10^th modules (k & m modules: KR_IV+ACP+KS⁰_IV+DH+ACP+KS_I,) and all of the domains of the 11^th module (b module: KR_III+MT+ACP+KS_III) possess their expected signatures. An unexpected activity is catalyzed by DH, which apparently helps to generate an α/β-trans double bond instead of an anticipated α/β-cis double bond (a leucine substitutes for the conserved active site aspartate). Even with the domains of three modules mixed together, the ACPs only collaborate with their cognate enzymes. The thirteenth module, which contains the DH+ER+KR architecture of cis-AT modules and is split between polypeptides, apparently uses inactive OMT⁰ and ACP⁰ domains to noncovalently assemble.

Predicting the stereochemistry of the moiety a module adds is not always possible through a determination of its module type alone. Some module types (e.g., b, d, e, f) include both modules containing A-type KRs that generate L-β-hydroxyl groups and modules containing B-type KRs that generate D-β-hydroxyl groups. In modules with an active DH (e & f) this results in the formation of α/β-trans double bonds from D-β-hydroxyacyl substrates and α/β-cis double bonds from L-β-hydroxyacyl substrates (e.g., Dsz5, Dsz6, Dsz8, Mgs2, Ozm4, and Sor24, which represent a significant fraction of the identified e modules, generate cis-double bonds). KSs of b and d modules are apparently quite tolerant of the stereochemical orientations of α- and β-substituents in the acyl chains they accept, while the KS_Is of e and f modules apparently gatekeep for planar geometry at the α- and β-carbons and are able to select for both α/β-trans double bonds and α/β-cis double bonds. The absence or presence of a signature aspartate in the KR usually indicates whether it is A-type or B-type, respectively, although the A-type KR_IVs of d modules are an exception to this^{22, 41}.

Insights into Module Function

ACPs that belong to different module types within the same family often possess similar signatures (Figure 4). In Family III, the ACPs from the j module type collaborate with a DH, KR, MT, and an ER. A significant interface is likely formed by ACP with each of these processing enzymes through helices II and III. Even though the ACPs of the b and x module types that are also in Family III do not interact with all of these processing enzymes, their signatures are very similar to those of the j module type. Thus, b and x modules, which likely evolved from j modules, still rely on interactions between their processing enzymes and ACP surfaces generated by the signature residues of Family III.

While ACPs generally co-evolve with the KS of their module type, they can function with other ETR enzymes that have naturally replaced the KS. If the KS that was substituted was not a gatekeeper the replacement does not impact the structure of the intermediate passed to the next module. An example of this innocuous substitution is provided by the bacillaene assembly lines of Bacillus subtilis and Bacillus amyloliquefaciens. After the a-type 11^th module of the B. subtilis assembly line has added an L-β-hydroxy unit to the chain, its KS⁰ transfers the acyl chain to the ACP of the u-type 12^th module, which interfaces with a C domain. In the B. amyloliquefaciens assembly line, the 11^th module contains a C domain equivalent to that of the 12^th module of the B. subtilis assembly line. Presumably, both the KS⁰ of an ancestral 11^th module as well as the ACP of an ancestral 12^th module were deleted, and the ACP of the a module was able to collaborate with the C domain of the u module. When gatekeeping KSs are replaced, the new ETR enzyme must select between interconverting acyl chains for the hybrid module to generate a single intermediate. Only four examples of this are present in the 33 analyzed assembly lines (the TE in Dif15 may select for an β-exomethylene, the C domain in Sgv3 may select for an α/β-olefin, the TE in Mln12 may select for an α/β-cis, γ/δ-trans diene, and the C domain of Bas6 may select for an α/β-cis, γ/δ-trans diene).

Evolutionary Relationship between trans-AT and cis-AT Assembly Lines

trans-AT and cis-AT assembly lines have been proposed to have arisen through convergent evolution^{9, 42}; however, the compatibility of trans-AT and cis-AT machinery is suggestive of divergent evolution. Several hybrid assembly lines composed of both trans-AT and cis-AT modules are now known. The chivosazole, enacyloxin, kirromycin, and nosperin assembly lines are primarily comprised of trans-AT modules, the ambruticin, DKxanthene, and pyoluteorin assembly lines are primarily comprised of cis-AT modules, and the neocarzilin assembly line contains an equal mix of trans-AT and cis-AT modules⁴³.

Intriguingly, some modules contain both trans-AT and cis-AT components. The fourteenth module of the kirromycin assembly line appears to be a trans-AT module until closer inspection. The polypeptides KirV (containing the DH+KR+ACP of the fourteenth module) and KirVI (containing the KS of the fourteenth module) noncovalently associate through the same type of docking domains utilized in cis-AT assembly lines⁴⁴. Its ACP and KS also clade with the ACPs and KSs of cis-AT assembly lines (Figures 2 & 3). The thirteenth module of the oxazolomycin trans-AT assembly line has characteristics of a cis-AT module: its DH, KR, and ER domains are organized as in cis-AT modules, and its ACP contains the “DSLTAVELRN” sequence motif characteristic of ACPs from cis-AT modules (Figure 4). The second and seventh modules of the enacyloxin trans-AT assembly line appear to be cis-AT modules except for their docking domains, which are more like the four-helix bundle docking domains utilized by trans-AT assembly lines²³.

cis-AT modules may have evolved through an ancestral trans-AT module. In comparison to the ACPs and KSs from trans-AT modules, the ACPs and KSs of cis-AT modules are much less diverse and cluster tightly with one another in the ACP and KS cladograms (Figures 2 & 3). As the ACPs of cis-AT assembly lines have a “DSLTAVELRN” sequence motif, which is most similar to the signature motif of ACPs from j modules (DH+KR_III+MT+ACP+ER+KS_III), perhaps cis-AT modules arose from such a module (the thirteenth module of the oxazolomycin assembly line, in which the ER domain is positioned between the structural and catalytic subdomains of KR, as in most cis-AT modules, is a good candidate) (Figure 5)⁴⁵. The “DSLTAVELRN” sequence motif of cis-AT ACPs seems to be important for ACPs that cooperate with an embedded DH, KR, and ER since it is highly conserved in modules that possess these three enzymes and poorly conserved in modules lacking them. That the ACP of the ninth module of the FD-891 cis-AT assembly line possesses the “DSLTAVELRN” motif even though ER is positioned within DH instead of KR indicates that the location of ER within the module does not significantly affect how it binds with ACP⁴⁶.

Nomenclature

Since transitioning from the traditional module boundaries to the updated module boundaries will likely be a slow process, authors should make the numbering system they are employing evident. The new module boundaries do not change the numbering of KS or C domains (Figure 1). The first module of an assembly line includes the KS or C domain that performs the first chain extension and all the embedded domains upstream of it. All of the ensuing KS, KS⁰, and C domains are consecutively numbered. The chain releasing enzyme also receives a module number. The numbering of all embedded processing enzymes and ACPs increases by one relative to the traditional numbering. The analysis presented here was aided by assigning letters to module types and numbers to ACP families as well as KS and KR clades; these labels are a temporary convenience and will change as more is learned about the modules of trans-AT assembly lines.

The updated PKS module boundaries affect the boundaries of NRPS modules within trans-AT assembly lines. The NRPS modules of the albicidin assembly line, traditionally defined as C+A+ACP, become A+ACP+C (Ref. 6). As many C domains are known to act as gatekeepers that collaborate with upstream processing enzymes (e.g., E domains), these boundaries make sense⁷; however, the NRPS community will need to judge whether the NRPS module should be redefined.

Conclusion

Many attempts to Lego-ize enzymatic assembly lines have been made through the years^47–48. However, re-engineered assembly lines have traditionally suffered from inactivity or low activity. The newly defined module boundaries afford exciting new opportunities. By keeping together the processing enzymes, ACP, and KS that have co-evolved, it may be possible to engineer assembly lines with activities comparable to those naturally observed. Such evolution-guided engineering could ultimately lead to the programmed synthesis of designer molecules and medicines.