Engineered Biosynthesis of Alkyne-Tagged Polyketides by Type I PKSs.

Polyketides produced by modular polyketide synthases (PKSs) are important small molecules widely used as drugs, pesticides, and biological probes. Tagging these polyketides with a clickable functionality enables the visualization, diversification, and mode of action study through bio-orthogonal chemistry. We report the de novo biosynthesis of alkyne-tagged polyketides by modular type I PKSs through starter unit engineering. Specifically, we use JamABC, a terminal alkyne biosynthetic machinery from the jamaicamide B biosynthetic pathway, in combination with representative modular PKSs. We demonstrate that JamABC works as a trans loading system for engineered type I PKSs to produce alkyne-tagged polyketides. In addition, the production efficiency can be improved by enhancing the interactions between the carrier protein (JamC) and PKSs using docking domains and site-directed mutagenesis of JamC. This work thus provides engineering guidelines and strategies that are applicable to additional modular type I PKSs to produce targeted alkyne-tagged metabolites for chemical and biological applications.

Introduction

Natural products produced by modular polyketide synthases (PKSs) have demonstrated their use as therapeutics, industrial products, pesticides, and biological probes following intense study over the past decades (Hertweck, 2009, Klaus and Grininger, 2018). Some well-known examples of these polyketides include the antibiotic erythromycin and the immunosuppressant rapamycin, both of which were initially isolated from bacterial sources and have been approved for clinical use for decades (Cottens et al., 2019, Hertweck, 2009, Jelic and Antolovic, 2016). The process for discovery, diversification, and mode of action elucidation of polyketides remains challenging and time consuming, although it has been improved in recent years due to many technical advancements. One such technology is to tag polyketides with a clickable functionality, which has been demonstrated to facilitate the study of polyketide biosynthesis, biology, and pharmacology through bio-orthogonal chemistry (DeGuire et al., 2015, Harvey et al., 2012, Hughes et al., 2014, Kalkreuter et al., 2019a, Kalkreuter et al., 2019b, Koryakina et al., 2017, Musiol-Kroll et al., 2017, Riva et al., 2014, Seidel et al., 2019, Zhu and Zhang, 2015). In particular, polyketides can be tagged through semi-synthesis (DeGuire et al., 2015, Seidel et al., 2019), total synthesis (Staub and Sieber, 2008), precursor-directed biosynthesis (Harvey et al., 2012, Koryakina et al., 2017, Musiol-Kroll et al., 2017, Seidel et al., 2017, Yan et al., 2013), or de novo biosynthesis (Zhu et al., 2015a, Zhu and Zhang, 2015). In this work we aim to further develop the strategy of de novo biosynthesis, which offers the unique advantage of not feeding the biorthogonal moiety itself, which could lead to increased background due to the diffusible non-specific nature of feeding starter or extender units. Instead the taggable group is incorporated by enzymatically synthesizing both the complex polyketide scaffolds and the unique clickable functionality allowing in situ bio-orthogonal chemical transformations.

Modular PKSs, often referred to as type I PKSs, have modules with multiple catalytic domains that perform separate enzymatic activities and act as an assembly line to select and incorporate building monomers into polyketide scaffolds (Jenke-Kodama and Dittmann, 2009, Keatinge-Clay, 2012, Khosla et al., 2014, Ladner and Williams, 2016) (Figure 1). The monomers used for extension, typically malonyl- or methylmalonyl-CoA, are recognized by acyltransferase (AT) domains, and the carbon-carbon bond is formed through decarboxylative Claisen condensations catalyzed by the ketosynthase (KS) domains. The megasynthases themselves have been investigated and have undergone extensive engineering efforts due to their modular structures that have captured scientists' imagination with the possibility of producing on-demand, designer molecules (Awakawa et al., 2018, Barajas et al., 2017, Chemler et al., 2015, Harvey et al., 2012, Kalkreuter and Williams, 2018, Klaus and Grininger, 2018, Koryakina et al., 2017, Moss et al., 2013, Ranganathan et al., 1999, Sundermann et al., 2013, Tang et al., 2000, Wlodek et al., 2017, Yonemoto et al., 2012, Yuzawa et al., 2017). Many of these engineering strategies have included efforts geared toward the inclusion of functional chemical handles for subsequent drug discovery or chemical biology studies, albeit often employing fed precursors containing the functionality of interest (Kalkreuter et al., 2019a, Koryakina et al., 2017, Mohammadi-Ostad-Kalayeh et al., 2018).

Figure 1

Overview of the JamABC Cassette and Type I PKSs in This Work

(A) JamABC works together to produce 5-hexynoyl-JamC as a starter unit for the downstream PKS/nonribosomal peptide synthetase assembly line in jamaicamide B biosynthesis.

(B) Native LipPKS1 and DEBSM6 domain organization and associated polyketide products.

(C) Engineered LipPKS1∗ and DEBSM6∗ used as representative modular PKSs in the current study.

The terminal alkyne is a canonical bio-orthogonal functional group as it is small, stable, and can be selectively reacted via copper-catalyzed azide-alkyne cycloaddition, where an azide containing a fluorophore, mass tag, or other chemical moiety is attached (Prescher and Bertozzi, 2005, Zhu and Zhang, 2015). The bio-orthogonality of alkynes is due to its chemical stability in biological environments and its rarity in biology where only a small number of terminal alkyne-bearing secondary metabolites have been discovered and even fewer biosynthetic pathways have been elucidated (Edwards et al., 2004, Fritsche et al., 2014, Haritos et al., 2012, Lee et al., 1998, Marchand et al., 2019, McPhail et al., 2007, Minto and Blacklock, 2008, Moss et al., 2019, Ross et al., 2014). We recently identified and characterized an acyl carrier protein (ACP)-dependent, three-protein pathway to generate the terminal alkyne functionality in E. coli (Zhu et al., 2015a, Zhu et al., 2015b, Zhu et al., 2016). For example, in the biosynthesis of the cyanobacterial jamaicamide B, JamA, an acyl-ACP synthetase, activates and loads 5-hexanoic acid onto JamC, a dedicated ACP. The resulting 5-hexanoyl-JamC is modified by JamB, a membrane-bound desaturase/acetylenase, to yield 5-hexynoyl-JamC as a starter unit for the downstream PKS/nonribosomal peptide synthetase assembly line (Figure 1A) (Edwards et al., 2004, Zhu et al., 2015a). JamABC thus represents a portable tri-gene cassette that may be useful for in situ generation and incorporation of terminal alkynes into various molecular scaffolds on demand. Toward this end, we demonstrated that PKS starter unit engineering is a feasible strategy to install the fatty alkynyl starter unit generated by JamABC onto polyketide scaffolds, such as those generated by promiscuous type III PKSs, which recognize both the acyl group and the acyl carrier (JamC) (Zhu et al., 2015a, Zhu et al., 2015b). However, to generalize this strategy to other polyketide scaffolds, in particular those synthesized by modular type I PKSs, additional model systems and protein engineering methods need to be explored.

Here we employ two well-studied type I PKSs, LipPKS1 and DEBSM6, to explore engineering strategies to make alkyne-tagged polyketides. LipPKS1 is the first module in lipomycin biosynthesis that natively utilizes an isobutyl starter unit presented by a loading ACP (Figure 1B) (Bihlmaier et al., 2006). DEBSM6 is the last PKS module from the erythromycin biosynthetic pathway (Figure 1B) (Rawlings, 2001). In addition, engineered LipPKS1 and DEBSM6 have been obtained to utilize malonyl-CoA instead of methylmalonyl-CoA as the extender unit with the promiscuous DEBS thioesterase to promote the acid product release as demonstrated from both in vitro biochemical studies and in E. coli (Yuzawa et al., 2017). These two engineered modules are thus simple and convenient systems for in-depth assessment of the interaction between representative module PKSs and JamABC for alkyne-tagged polyketide biosynthesis. Considering the known critical role of the cognate ACP (JamC) in the alkyne biosynthetic machinery (Su et al., 2018, Zhu et al., 2015a), the recognition of JamC by PKSs is expected to play a key role in alkyne-tagged polyketide synthesis and therefore is the focus of the present study.

Results and Discussion

Alkyne-Tagged Polyketide Synthesis In Vitro

To probe the possible recognition of the 5-hexynoyl-JamC by PKSs, in vitro assays were initially performed using the engineered LipPKS1 and DEBSM6 modules. The reported engineered LipPKS1 was further modified by removing the AT and ACP loading domains to create a truncated version to facilitate the alternative starter unit incorporation. We hypothesized that these engineered PKSs (termed LipPKS1∗ and DEBSM6∗, Figure 1C) without the loading domains would result in JamC to act in trans to selectively load and extend JamC-linked acyl chains. For in vitro assessment we purified JamA, holo-JamC, and LipPKS1∗/DEBSM6∗ from E. coli after overexpression (Figure S1), or an E. coli BAP1 strain that contains a chromosomal copy of the phosphopantetheinyl transferase Sfp that was used to ensure the post-translational modification of carrier proteins to the pantetheinylated forms (Pfeifer et al., 2001). Purified enzymes were incubated with 5-hexynoic acid, ATP, malonyl-CoA, and NADPH for alkyne-tagged polyketide biosynthesis in vitro (Figure 2A). JamB activity for alkyne biosynthesis was not assessed in vitro due to the difficulty of obtaining active and purified membrane proteins and was assessed later in vivo. The expected product, 3-hydroxy-7-octynoic acid (1), was successfully produced by both engineered PKSs as confirmed by comparing with the synthetic chemical standard (Figures 2B, 2C, and S2–S4, Scheme S1). Interestingly, replacement of 5-hexynoyl-JamC by 5-hexynoyl-CoA, which was generated in situ using a promiscuous acyl-CoA ligase Orf35 (Zhang et al., 2010), dropped the formation of 1 to trace amounts, demonstrating a preference of these two PKSs toward JamC over CoA as the acyl carrier (Figures 2B, 2C, and S2).

Figure 2

In Vitro Assessment of Alkyne-Tagged Polyketide Biosynthesis Using LipPKS1∗ and DEBSM6∗

(A) Overview scheme of in vitro reactions between JamA/JamC and engineered PKSs to produce 3-hydroxy-7-octynoic acid (1).

(B) Formation of 1 by LipPKS1∗ under various reaction conditions and engineering strategies.

(C) Formation of 1 by DEBSM6∗ under various reaction conditions and engineering strategies. Engineered PKS cartoon is truncated for clarity. All graphs are shown as relative product formed compared with the JamC/PKS with no modifications calculated from integration of the extracted ion chromatogram (EIC) for compound 1 (set as 1). Error bars indicate SEM for n ≥ 2 independent experiments.

Evaluation of Docking Domain Strategy to Improve JamC-PKS Interactions

As protein-protein interactions are known to dominate the turnover of chimeric PKS assembly lines (Klaus et al., 2016), we proposed that improved communication between the upstream JamC and the downstream KS could lead to a more efficient alkyne-tagged polyketide biosynthesis. Docking domains, often found on the C terminus of ACPs (ddACP) and the N terminus of KSs (ddKS), have been shown to be important for protein-protein interactions in PKSs (Gokhale et al., 1999, Tsuji et al., 2001, Zeng et al., 2016). We then set out to evaluate the strategy of fusing known docking domains to the C terminus of JamC and the N terminus of the LipPKS1∗/DEBSM6∗ KS domains to improve protein recognition. In particular, we chose to utilize the class 2 docking domains from the cyanobacterial curacin pathway as the pair CddCurK (ddACP) and NddCurL(ddKS) was shown to be modular and portable (Whicher et al., 2013). We also chose the related docking domain pair CddJamK (ddACP) and NddJamL(ddKS) from the jamaicamide pathway as moving docking domains within pathways was shown to be more successful than inter-pathway swapping (Klaus and Grininger, 2018, Klaus et al., 2016, Whicher et al., 2013). The fusion of these docking domains to JamC and PKSs did not significantly impact the expression and folding of these proteins (Figures S1 and S5). In vitro product formation assays using purified proteins demonstrated the success of this strategy in generating product 1 (Figures 2 and S6). The adoption of the pair of CddJamK and NddJamL had minimal effect on the production of 1, whereas the pair of CddCurK and NddCurL led to significantly more amount of 1 in both PKS systems (∼3-fold for LipPKS1∗ and ∼40-fold for DEBSM6∗) (Figures 2B and 2C). Control experiments using only one of the docking domains produced less products than using the pair for CddCurK and NddCurL. In addition, the poor production of 1 with the docking domain fused to JamC excluded the possibility of improved recognition of modified JamC by JamA (Figures 2B and 2C), indicating that the improved communication between the engineered JamC and KS due to the docking domains is the main contributor for higher production of 1 in vitro.

Evaluation of Site-Directed Mutagenesis of JamC to Improve JamC-PKS Interactions

In addition to docking domains, we also wanted to identify a less-intensive engineering strategy to improve JamC-PKS communication. Mutating JamC without perturbation to the large megasynthase would make this strategy more easily adaptable to different systems. From the well-studied DEBS system, it has been shown that direct ACP-KS protein-protein interactions during translocation are selective, and key residues within helix I of ACP have been identified that contribute to chain translocation specificity (Kapur et al., 2012, Klaus et al., 2016). Inspired by the previous successful studies, we identified the corresponding residue in JamC (E32) that may play an important role in ACP-KS interactions through sequence alignments and structural modeling (Figure S7). To mimic the native upstream ACP, we chose the mutations E32T for LipPKS1∗ and E32H for DEBSM6∗ based on alignments to the respective ACPs found upstream in the native systems (Figure S7). These two JamC mutants were cloned, overexpressed, and purified from BAP1 with a similar yield to the wild-type protein (Figure S1). In vitro product formation assays showed that the formation of 1 increased approximately 7-fold with LipPKS1∗ (Figure 2B) and 2-fold with DEBSM6∗ (Figure 2C). These fold increases demonstrated the effectiveness of this strategy in improving the production of alkyne-tagged polyketides in vitro, most likely due to an improved JamC communication with modular PKSs.

Perturbation of JamB Activity by JamC Engineering

In vitro biochemical assays demonstrated the success of protein engineering in improving the recognition of JamC by PKSs to promote the translocation of the alkynyl starter unit. However, the potential impact of JamC modification on the activity of JamB, the desaturase/acetylenase that functions on a JamC-tethered substrate to form a terminal alkyne, is unclear. As it is difficult to reconstitute and quantify the activity of the membrane-bound JamB in vitro, we then tried to implement the biosynthetic machinery of alkyne-tagged polyketides in E. coli to assess the possible impact. In addition, the titers of relevant products were also quantified in E. coli to probe the effectiveness of two engineering strategies to improve JamC-PKS interactions in vivo. Combinations of JamA, B, C, PKSs, and their variants were expressed in an E. coli BAP1 strain under a T7 promoter to obtain various engineered strains. A single mutation in JamB (M5T) identified in previous work, presumably with an improved interaction with the electron donor, was used in all strains to increase the alkyne titer in E. coli (Zhu et al., 2016). All strains were grown with 5-hexenoic acid feeding, followed by extraction and quantification of 3-hydroxy-7-octenoic acid (2) and 3-hydroxy-7-octynoic acid production (1), by fitting to a standard curve of synthesized standards generated through liquid chromatography-high-resolution mass spectrometric analysis (Scheme S1, Figures S3, S4, S8, and S9). The product 2 was expected to be a side product due to the activities of JamA, C, and PKS without the action of JamB (Figure 3A). Other possible products were also analyzed, as it is conceivable that the PKSs accept different fatty acyl starter units in vivo via JamC or other acyl carriers (Figure S10).

Figure 3

In Vivo Assessment of Alkyne-Tagged Polyketide Biosynthesis Using LipPKS1∗

(A) Overview scheme of in vivo reactions between JamABC and LipPKS1∗ to produce 3-hydroxy-7-octynoic acid (1) and 3-hydroxy-7-octenoic acid (2).

(B) Quantification of alkyne product titers resulting from the engineered JamC and LipPKS1∗. Alkyne 1 product titers are shown in blue (left y axis), and the relative JamB activities are shown in green (right y axis).

(C) Quantification of total product titers resulting from the engineered JamC and LipPKS1∗. LipPKS1∗ cartoon is truncated for clarity. All titers shown have subtracted background from a control strain lacking JamC to better reflect the interaction between JamC and LipPKS1∗. All error bars represent SEM for n ≥ 3 biological replicates.

An initial investigation of the titer of compound 1 produced by the co-expression of JamA, B, C, and LipPKS1∗/DEBSM6∗ demonstrated that DEBSM6∗ produced compound 1 (0.014 mg/L) significantly less than LipPKS1∗ (0.071 mg/L). Much higher amounts of products other than 1 and 2 with a longer acyl chain were generated by DEBSM6∗ in vivo (Figure S10), consistent with the native acyl chain length accepted by LipPKS1∗/DEBSM6∗ (C4 versus C13). We concluded that DEBSM6∗ would not be an effective in vivo model system to probe the activity of JamB due to complicated product profiles and thus limited the in vivo study to LipPKS1∗.

The products 1 and 2 were produced by LipPKS1∗ in an approximately 1:5 ratio, and this efficiency was set to be a relative JamB activity of 100% (Figure 3B). This product ratio was dropped ∼4-fold when either docking pair was used, suggesting that the fusion of a docking domain to JamC affected its recognition by JamB (Figure 3B). In contrast, the E32T point mutation of JamC had minimal effect on the product ratio while increasing the titer of 1 ∼6-fold to 0.42 mg/L, consistent with previous observations that the helix I of ACP did not play an important role in interacting with JamB (Su et al., 2018, Zhu et al., 2016). We next probed the combined product titer of alkyne 1 and alkene 2 to assess the effectiveness of the two engineering strategies in their ability to improve JamC-LipPKS1∗ interactions in vivo (Figure 3C). Consistent with the trends observed in vitro, the combined titer improved more than 10- and 20-fold using docking domains CddCurK/NddCurL and CddJamK/NddJamL, respectively, and ∼10-fold using JamC (E32T), demonstrating the success of either strategy in improving JamC-PKS interactions in vivo (Figure 3C).

Finally, we probed the possible synergistic effects of the two engineering strategies in improving the alkyne-tagged polyketide biosynthesis in vivo. We observed additive effects when using docking domains and the JamC point mutation in improving JamC- LipPKS1∗ interactions. The combined titer of 1 and 2 roughly equaled the sum of that when either engineering strategy was used. The maximum amount of product obtained was ∼16 mg/L from JamC(E32T)-CddJamK/NddJamL-LipPKS1∗, an approximately 39-fold increase from unmodified JamC/LipPKS1∗ (Figure 3C). However, due to the expected disruption of JamB activity when the docking domain is fused to JamC, the absolute titer of the alkyne product 1 was not increased when using both engineering strategies compared with the JamC mutagenesis alone (Figure 3B). These results further highlight the importance of JamB efficiency in de novo alkyne synthesis, which remains to be a limiting step in the production of alkyne-tagged polyketides.

Limitations of the Study

Although the current results demonstrate a great potential of de novo biosynthesizing alkyne-tagged polyketides by engineering both the alkyne biosynthetic machinery and modular type I PKSs, the strategy is limited to incorporate an alkynyl starter unit, which needs to be tolerated by PKSs. It is expected to work well with PKSs with a native starter unit resembling the alkyne-containing acyl group presented by the alkyne biosynthetic machinery, such as in the case of LipPKS1, but may not work with PKSs recognizing very different starter units, such as in the case of DEBSM6. This is particularly exemplified by the in vivo results of DEBSM6, in which a complex metabolic background significantly decreased the efficiency of alkyne-tagged polyketide biosynthesis by these PKSs.

Conclusion

We have successfully demonstrated that carrier protein-dependent alkyne biosynthetic machinery can work as a trans loading system for truncated Type I PKSs to produce alkyne-tagged polyketides both in vitro and in vivo. Two protein engineering strategies were explored to improve the interaction between the carrier protein within the alkyne biosynthetic machinery (JamC) and modular PKSs. This included the employment of PKS docking domains and site-directed mutagenesis of JamC to increase acyl chain translocation specificity. Both strategies were shown to be successful, leading to enhanced recognition of JamC by modular PKSs and thus improved alkyne-tagged polyketide production. In addition, the effects of both engineering strategies to improve protein-protein interactions were additive, leading to an ∼39-fold increase in the polyketide production by an engineered LipPKS1 in E. coli. It is also notable that the installation of a docking domain on JamC, but not the site-directed mutagenesis, disrupted its recognition by JamB in alkyne-tagged polyketide production. Furthermore, the native acyl group specificity of modular PKSs was suggested to be important for alkyne-tagged polyketide production, in particular in vivo where competing acyl groups were present. In summary, this work has shown the first examples of de novo biosynthesis of alkyne-tagged polyketides by modular type I PKSs through starter unit engineering and further provided engineering guidelines and strategies that are expected to be applicable to other modular PKSs to produce targeted alkyne-tagged metabolites for drug discovery and chemical biology studies.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.