Protein Engineering of Biosynthetic Enzymes Unlocks Libraries of Bioactive Tilimycin Analogs

Abstract

Natural products continue to inspire therapeutic innovation due to their structural complexity and biological potency. Pyrrolobenzodiazepines (PBDs), known for antitumor activity, function by covalently binding to guanine bases in DNA, a mechanism that is inherently species-agnostic. However, their potential as antibiotics remains underexplored, as modest antibiotic activity is commonly seen with this class of compounds. Here, we produce analogs of tilimycin and tilivalline, two PBDs produced by the gut microbe Klebsiella oxytoca. We mutated NpsA, the nonribosomal peptide synthetase (NRPS) pathway protein responsible for initiating biosynthesis through adenylation of 3-hydroxyanthranilic acid on the pathway to form tilimycin and tilivalline, to enhance promiscuity with substrate analogs. Using structure and informatic-guided mutagenesis, we developed a rapid screening method to identify compatible enzyme-building block combinations to generate a panel of tilimycin and tilivalline analogs. We identified compounds that possess the ability to inhibit DNA polymerase and that show growth inhibitor activity with a DNA-repair mutant of E. coli. This work demonstrates the feasibility of NRPS reprogramming to use biocatalytic approaches to access non-natural derivatives with antibiotic potential and highlights tilimycin analogs as candidates for gram-negative antibacterial development.

Graphical Abstract

Introduction

Pyrrolobenzodiazepines (PBDs) are heterotricyclic compounds (Figure 1) that covalently bind to the minor groove of DNA at guanine bases in a sequence selective manner, causing alkylation and double strand breakage within bacterial and tumor cells; as such, they are often referred to as antitumor antibiotics.¹ PBDs are naturally biosynthesized by different bacteria and fungi,^{2, 3} but synthetic analogs have resulted in compounds with higher bioactivity, including derivatives that have proceeded to clinical trials in anticancer therapies.^4–6 PBD monomers have been connected synthetically through a linker to form dimeric molecules that convert DNA alkylators into molecules with the ability to cross-link DNA, which have then been further converted into antibody-drug conjugates and complex antimicrobials.^6–8 While traditional medicinal chemistry with PBDs has produced highly active drug leads, the process of chemically synthesizing PBDs can be laborious due to difficulty in retention of stereochemistry at the C11a position, and stability of the imine bond at the N10-C11 position.^{2, 7, 8}

Figure 1. Pyrrolobenzodiazepine natural products and K. oxytoca PBD synthesis.

A. Examples of PBD natural products. B. Pyrrolobenzodiazepine natural product structure scaffold. C. NRPS pathway to synthesize aldehyde, 2-hydroxy-5-(2-oxo-3-(pyrrolidin-3-ylcarbamoyl)amino)benzamide, involving 1. the adenylation and loading of 3HA substrate onto the PCP ThdA catalyzed by the adenylation domain NpsA, 2. adenylation and PCP loading of L-proline catalyzed by adenylation domain of NpsB, 3. condensation reaction between 3HA and L-proline, and 4. release of the final aldehyde product through an NADPH-dependent reduction reaction performed by NpsB terminal reductase domain. D. NRPS-independent cyclization reaction forms the PBD tilimycin, followed by a spontaneous electrophilic aromatic substitution of indole to form tilivalline.

Klebsiella oxytoca, a gram-negative bacteria from the human gut microbiome that is the causative agent of antibiotic-associated hemorrhagic colitis (AAHC) in Clostridium difficile-negative cases^{9, 10} produces two PBDs called tilimycin and tilivalline. Tilimycin can alkylate host epithelial cell DNA, causing cell damage in the GI tract, leading to apoptosis and colonic inflammation.¹¹ Indole, which is present in K. oxytoca at high concentrations,¹² can react spontaneously with tilimycin to form tilivalline, a second product that binds and stabilizes microtubules, further eliciting epithelial cell damage and contributing to AAHC.¹³ While tilivalline is most significant in a mammalian cell context, tilimycin has been shown to also exhibit antimicrobial activity to compete with surrounding microbiota.^{14, 15}

Tilimycin and tilivalline are produced by a single nonribosomal peptide synthetase (NRPS) biosynthetic pathway.^{10, 11} These modular enzymes act as protein assembly lines for natural products and create diverse peptides using proteogenic and non-proteogenic building blocks.^{16, 17} The NRPSs employ acyl-, aryl-, and peptidyl carrier proteins (often collectively referred to as PCPs) that are modified with a phosphopantetheine moiety derived from coenzyme A to bind the carboxylate substrates and peptide intermediates through a thioester linkage. Generally, a module, composed of the carrier and catalytic domains necessary to incorporate a single amino acid, is present for each building block that is incorporated into the final peptide.

Three NRPS proteins are responsible for the synthesis of tilimycin, NpsA, ThdA, and NpsB.^{10, 11, 18, 19} In this pathway, NpsA, a free-standing adenylation domain, is responsible for introducing 3-hydroxyanthranilic acid (3HA) into the pathway through a two-step reaction between 3HA and ATP that first activates 3HA as an aryl-adenylate, followed by transfer of the substrate to the pantetheine thiol of the carrier domain protein, ThdA (Figure 1C). NpsA facilitates the nucleophilic attack of the pantetheine thiol of ThdA on the adenylated substrate, employing a structural conformational change, covalently loading 3HA onto ThdA, allowing it to be carried to the condensation domain of NpsB.¹¹ The second NRPS module, present on NpsB, adenylates and loads L-proline onto the integrated PCP. The NpsB condensation domain then catalyzes formation of an amide bond between 3HA and the amine of proline, producing a new dipeptide (specifically N-3-hydroxyanthranoylproline) that is bound to the PCP domain of NpsB. This intermediate is delivered to the C-terminal reductase domain of NpsB and released through an NADPH-dependent reduction.^{10, 11} The aldehyde product spontaneously cyclizes to form tilimycin, which can be exported from the cell or, in the presence of indole, react further to form the second PBD product tilivalline (Figure 1D).

Given the complexity of many natural products, the use of enzymatic approaches to biosynthesize libraries of compound analogs could facilitate the discovery of natural product-based bioactive molecules that share the original scaffold.²⁰ Combining NRPS domains or subdomains has yielded catalysts that are able to produce altered peptides when heterologously expressed. Pyoverdine, a large siderophore from Pseudomonas aeruginosa, has been engineered through careful assessment of the interdomain boundaries to enable the adenylation domain of the terminal module that incorporates threonine to be replaced with an alternate lysine-specifying domain.²¹ An elegant series of studies used structural biology and phylogenetic analysis to define boundaries within the NRPS condensation or the carrier protein to result in the highly efficient production of peptide analogs.^{22, 23} Alternate fungal cyclodepsipeptides have additionally been produced through domain shuffling and truncation.²⁴ While these studies have examined large NRPS clusters that have been expressed recombinantly allowing product formation in cell culture to be monitored, biochemical reconstitution has also been used with smaller, less complex NRPS pathways. In biochemical reactions, analogs of the aryl-capped antibiotic obafluorin²⁵ and the siderophores fimsbactin²⁶ and bacillibactin²⁷ have been explored. Finally, we note the potential to use cell-free expression systems to assay biosynthetic enzymes in cell lysates,^{28, 29} a strategy we recently employed to make tilimycin.³⁰

Previous studies have shown that tilivalline variants can be produced from biochemical reconstitution with closely related 3HA analogs and wild-type NpsA,^{18, 19} illustrating that NpsA possesses natural promiscuity to incorporate alternate aryl substrates that progress through the pathway. A similar study with the PBD tomaymycin showed that the adenylation domains within TomA and TomB, the homologs of NpsA and NpsB, could load analogs of their respective substrates onto the neighboring carrier domain, as detected by intact protein LC-MS.³¹ The complete biosynthesis of tomaymycin was not examined due to the requirement for downstream tailoring enzymes. Additionally, the tilimycin studies investigated the ability of substituted indole derivatives to react to form additional analogs using either biochemical reconstitution to form analogous products¹⁸ or cellular approaches.¹⁹ While these studies highlighted the NRPS promiscuity, the analogs were not characterized further, with no discussion of the biological activity of any analogs.

Inspired by these promising studies to produce tilimycin and tilivalline analogs, we explored the use of protein engineering to expand the pocket further and create an even larger library of PBDs synthesized using the pathway. Tilimycin and tilivalline have been tested primarily with mammalian and cancer cell lines,^{13, 32} where they are most potent, while analogs are underexplored in terms of bacterial cytotoxicity. Also, while tilimycin has been reported to have minimal activity with gram-positive and gram-negative bacteria found in fecal cultures,¹⁵ there is no reported evidence of using tilimycin analogs to assess potential potency in bacterial cells.

Here, a biocatalytic approach was used to engineer NpsA to produce a panel of PBDs that were tested for gram-negative antibacterial activity. Various anthranilate substrates were initially screened with both wild-type and mutant NpsA enzymes in a full biochemical reconstitution pathway using a NADPH oxidation analysis to quantify throughput. Importantly, by monitoring the final step of the pathway, we ensured that analogs introduced would be compatible with downstream catalytic domains. We additionally explored promiscuity of NpsB at the proline-binding site. Combined, we were able to produce significant amounts (greater than 10% of the native products produced by wild-type enzymes) of 24 of 28 possible tilimycin analogs and 20 of 28 tilivalline analogs, with 14 tilivalline analogs produced with a mutant enzyme at levels higher than the natural product. Compounds that derive from the most promising combinations of substrates and enzymes were then produced and isolated using LC-MS, and subjected to further microbial testing, developing future leads for antibiotic testing and optimization.

Materials and Methods

Cloning and gene annotation

Wild-type K. oxytoca proteins (Table S1) were previously cloned into a pET15b vector containing a N-terminal TEV cleavage site and 5x HisTag.¹¹ NpsA mutations were introduced through PCR site-directed mutagenesis using Phusion ^® Polymerase and following manufacturer’s (NEB) protocol. Oligonucleotides used in mutagenesis reactions are presented in Table S2.

Expression and Protein purification of NpsA and NpsA mutants

Wild-type and mutant NpsA were expressed in BL21 (DE3). A single colony from a transformation was inoculated in a 50 ml LB culture with ampicillin (100 μg/ml). The starter culture grew overnight for ≤ 20 hours at 37°C and 250 RPM. A secondary culture containing one liter LB, ampicillin (100μg/ml) and 1% of the starter culture was grown at 37 °C and 250 RPM until the OD₆₀₀ reached 0.4 to 0.6, at which point the temperature and shaking decreased to 16 °C and 200 RPM respectively. The culture was induced through addition of IPTG (0.5 mM) and incubated for 16-24 hours. Cells were harvested at 6,000 RPM for 10 min at 4°C and stored at −80 °C until needed for purification.

Cell pellets were resuspended in lysis buffer (50 mM HEPES, 150 mM NaCl, 30 mM Imidazole, 0.2 mM TCEP and 10% glycerol pH 7.5) at 10% w/v. The cell mixture was lysed using a Branson SFX250 sonifier with a microtip following a protocol of 30 seconds on, 90 seconds off for 0.1 min per ml of lysate (no more than 50ml of lysate at a time) at 60% amplitude. The lysed mixture was clarified by ultracentrifugation (30 min, 40,000 RPM, 4 °C) to separate soluble and insoluble fractions and the supernatant was filtered using a 0.45 μm syringe filter. The filtered lysate was loaded onto a Bio-Rad NGC FPLC system with a 5 ml His-Trap column connected. The 5 ml column was pre-equilibrated with five column volumes (CV) lysis buffer prior to sample loading. The column was then washed with five CV lysis buffer and then washed with five CV of 5% elution buffer (50 mM HEPES, 250 mM NaCl, 300 mM imidazole, 0.2 mM TCEP and 10% glycerol pH 7.5). The sample was eluted with ten CV of 100% elution buffer. NpsA was assessed using SDS-PAGE and the appropriate fractions were collected and dialyzed twice in ten volumes of dialysis buffer (50 mM HEPES, 250 mM NaCl, 0.2 mM TCEP and 10% glycerol pH 7.5) for 2 h and then overnight with the addition of 1% TEV protease to cleave the His-tag. The following day, imidazole pH 7.5 was added to the dialyzed sample to a final concentration of 20 mM, which was loaded again on the His-Trap column with the same protocol. NpsA was again monitored through SDS-PAGE and dialyzed against two rounds of one liter of dialysis buffer to remove imidazole. NpsA and mutants were concentrated using a 30 kDa centricon concentrator and protein concentration was determined using the Nanodrop^™ One spectrophotometer (Thermo Scientific), using the specific absorbance of NpsA (6.41 for 1% absorbance correction). The protein sample was flash frozen in liquid nitrogen and stored at −80°C.

Expression and Protein purification of ThdA

The NpsA purification expression and purification protocol was used with ThdA with the following changes. Following the incubation overnight with TEV, we added 10 nM Sfp and 10 mM MgCl₂ to the dialyzed ThdA solution to produce holo-ThdA (transfer of the 4′-phosphopantetheine moiety). Coenzyme A (CoA) was added at 2-fold molar excess per ThdA monomer (protein concentration determined by A₂₈₀) and the reaction was incubated at 23 °C for 2 h.³³ holo-ThdA was concentrated using a 3 kDa centricon and protein concentration was confirmed with a 1% absorbance correction of 6.50. The protein sample was flash frozen in liquid nitrogen and stored at −80°C.

Expression and Protein purification of NpsB

NpsB was expressed similarly to NpsA and ThdA except for using Terrific Broth for secondary culture growth (24 g/L yeast extract, 20 g/L tryptone, 4 mL/L glycerol, 0.017 M KH₂PO₄, 0.072 M K₂HPO₄). This changed the induction OD₆₀₀ to 0.6 – 0.8 and only 0.25 mM IPTG was added for a 16–24-hour induction (same temperature and conditions as previously stated).

Because of lower expression levels and stability, NpsB was used without cleaving the His-tag. The NpsB purification followed the same protocol as NpsA and ThdA, but following elution, tagged NpsB was dialyzed three times in 10 volumes of dialysis buffer to reduce imidazole content and concentrated using a 100 kDa centricon to ≤ 5 ml. Following the concentration steps, NpsB was loaded onto a Sephacryl S-300 HR (Cytiva) with a running buffer of 25 mM HEPES (pH 6.8), 125 mM NaCl, 0.1 mM TCEP, and 5% glycerol, to isolate appropriate monomeric peak and elution. The elution fractions were concentrated using a 100 kDa centricon and flash frozen in liquid nitrogen and stored at −80 °C. The lysis and elution buffers for NpsB follow the same buffer recipes as previously stated with pH adjusted to 6.8.

NADPH Oxidation Assay

Reactions were performed in 96-well plates in triplicate with either wild-type or mutant NpsA. A reconstitution master mix was created with reagent concentrations at 1.25× the final working concentrations. After addition of master mix to substrates, the final reaction concentrations were 50 mM Tris pH 8.0, 150 mM NaCl, 10 mM MgCl_2, 5 mM ATP, 1 mM NADPH, 0.5 μM NpsA, 10 μM ThdA (based on kinetic studies¹¹), and 0.5 μM NpsB. The master mix was incubated at 37 °C for 5 minutes. The substrates, 10 μl of 3HA or analog (dissolved in 20% v/v 1 N HCL, 1 M Tris and 1mM TCEP pH 8.0) and 10 μl of Proline or analog (dissolved in 1 M Tris pH 8.0), were added to opposite sides of a well and the reaction was initiated with 80 μl of pre-warmed master mix. The plate was initially shaken for 30 s and readings at 340 nm were monitored every 20-23 s for 7 min (with shaking every interval for 10 s) using a Biotek Cytation One Multi-Mode Imaging Plate Reader. The rate of NADPH oxidation was calculated after subtracting a blank spectrophotometric value of a no NpsA control reaction. Using the molar extinction coefficient of NADPH of 6.22 L mmol⁻¹ cm⁻¹ (ref³⁴) to convert to NADPH μM consumed per min.

Reconstitution Assay

Product reconstruction for both LC-MS analysis as well as E. coli growth inhibition used final concentrations of 20 mM Tris pH 8.0, 100 mM NaCl, 5 mM MgCl₂, 3 mM ATP, 3 mM NADPH, 20 μM ThdA, 2 μM NpsB, 2 μM NpsA (wild-type or mutant), 2 mM proline or analog, and 2 mM 3HA or analog. When needed, indole was added at 2 mM to enable tilivalline (or analog) formation. Biochemical reactions were set up in either 50 μl volume for reconstitution analysis and LC-MS, or 150 μl for product isolation or E. coli screening. Reactions were incubated at 200 RPM at 30 °C for 2 hr, with 10 μl was used for LC-MS analysis.

In the biochemical reconstitution reactions used to screen for the incorporation of proline analogs in the full pathway for tilimycin and tilivalline, as well as to produce tilimycin analogs that were used in bioactivity screening, reactions that used 3-hydroxyanthranilic acid and 3-fluoroanthranilic acid were performed with wild-type NpsA. Reactions used to produce analogs derived from anthranilic acid, 3-methylanthranilic acid, and 3-methoxyanthranilic acid were performed with the N207A mutant of NpsA. Finally, reactions that generated analogs from 4-hydroxyanthranilic acid, 5-hydroxyanthranilic acid, and 5-fluoroanthranilic acid were performed with the N207T NpsA mutant.

LC-MS Protocol

Products were analyzed using the Agilent 1200 Infinity II with attached single quad mass spectrometer. The entire resuspension was loaded onto a Poroshell 120 EC-C18 4 μm HPLC column in-line with an Agilent Mass Spectrometer with ESI. The protocol for the LC portion of this method varied the percentage of acetonitrile, 0-50 % 0-13 minutes, 50-95 % over 13-15 minutes, 95 %-5 % 15-16 minutes. A 10-minute equilibration with 5 % acetonitrile was set up between consecutive runs. UV, total ion chromatograms (TIC) and extracted ion chromatograms (EIC) [M+H] were analyzed for peak height intensities.

Tilimycin Analog Isolation

Isolation of tilimycin analogs involved 150 μl reaction setup described in the reconstitution assay method in triplicate. Following the 2-hour incubation, 500 μl of anhydrous acetonitrile was added to extract tilimycin products. This organic portion was evaporated at 60 °C until dry and resuspended in 30 μl acetonitrile. This suspension was loaded onto the LC-MS following the above protocol, collecting the appropriate peak elution (found by reconstitution assay). 500 μl of acetonitrile and 500 μl 5 M NaCl were used to extract and salt out product. The rganic phase was collected and dried to 150 μl. Estimated quantity and purity samples were loaded on LC-MS using the same protocol and monitored by UV, TIC, and EIC. The estimated quantity of sample was compared to 660 μM of the respective anthranilate substrate as an internal standard.

PCR Inhibition Assay

The pET28b_NpsB plasmid and primers (Table S3) were used for PCR inhibition testing with isolated tilimycin products. The plasmid was diluted to 20 ng/μl and added to a 20 μl reaction with 10x DNA buffer (final concentration 10 mM Tris, 50 mM NaCl), the potential alkylator, and water. Before being added to the PCR reactions, the tilimycin analog products of reconstitution reactions were evaporated to remove acetonitrile and resuspended in 100% water. DNA and either MMS or tilimycin analogs were incubated at 37 °C for 30 minutes. PCR reactions using Phusion Polymerase and following NEB protocol, using 1 μl of the DNA reaction mixture as template in 25 μl. These PCR reactions were analyzed on 1% agarose gels at a molecular weight of 1.4 kB. The PCR reactions were performed twice, with only the highest concentration used with 3HA-tilimcyin, 4HA-tilimycin and 3MA-tilimycin, where isolated product was limiting.

E. coli Screening

Tilimycin reconstitution reactions were set up and incubated for an initial 2-hour period at 30 °C. Cells were grown to saturation overnight in Mueller Hinton broth, then diluted and allowed to reach mid-log phase, at OD₆₀₀ of 0.4-0.5. This log-phase culture was diluted to OD₆₀₀ of 0.1 in the same media, and 50 μl of cells were combined with 50 μl of the TM reaction at 37 °C and monitored for 8 hours. The biochemical reaction was not quenched, allowing for continued formation of product while mixing with the growing cells. The OD₆₀₀ was measured and plotted over time. To assess viability, the area under the growth curve was calculated and compared to untreated cell growth.^{15, 35} Two biological triplicate (six total) experiments were performed. Where indicated, 34 μg/ml chloramphenicol was used as a positive control antibiotic for cell growth assays.

Results

Structure-Guided Design of NpsA Mutants

In a previous study that examined the biosynthetic pathway and potential NpsA inhibitors, we determined the structure of full length NpsA bound to 3HA-AMS, a mimic of the adenylate intermediate that employs an adenosylsulfonamide (AMS) group to replace the reactive adenylate, and as a complex with the ThdA carrier protein through formation of a fusion protein and the use of a mechanism-based vinylsulfonamide inhibitor.¹¹ The structure illustrated the conventional adenylation domain architecture with a larger A_core domain connected to the smaller C-terminal A_sub domain, which adopts different conformations to catalyze the two-step reaction. While the structures of the full length and ThdA complex were solved at modest resolution (2.9-3.0 Å), we additionally determined the structure of the A_core domain bound to the AMS derivatives of 3HA (Figure 2A), 3-hydroxybenzoic acid, and anthranilic acid at 1.8 – 2.1 Å resolution, allowing us to accurately model the ligand.

Figure 2. NpsA mutagenesis and biological activity.

A. The active site of NpsA with 3HA-AMS bound as a non-hydrolyzable inhibitor (PDB: 6VHX).¹⁸ Key residues specifically interacting with 3HA moiety are labeled, with key hydrogen bond interactions between the primary amine and Ser271, and the 3-hydroxyl substituent and Ser271 and Asn207. B. Sequence alignment with NpsA substrate binding residues and homologs of NpsA that also form homologous PBD products. Highlighted NpsA residues include Ala203, Ala204, Asn207, Ser271, Gly296, and Ala302. C. List of anthranilic acid (AA) analogs with substituents on the defined R-positions.

The 3HA binding pocket is formed by Ala203, Ala204, Asn207, Ser271, Gly296, and Ala302.¹¹ The three alanine residues form one wall of the pocket that interacts with the aromatic ring of 3HA, while Ser271 hydrogen bonds with both the C2 amine and the C3 hydroxyl of 3HA; this residue is conserved in characterized PBD anthranilate adenylation domains (Figure 2B).¹¹ Considering the amine and hydroxyl constituents, the aryl ring bound to NpsA is flipped relative to the position observed in siderophore biosynthetic aryl-adenylating enzymes such as EntE, BasE, and FbsH that activate salicylic acid or 2,3-dihydroxybenzoic acid.^{26, 36, 37} Asn207 adds a second hydrogen bond interaction with the hydroxyl group on 3HA.¹¹ Binding studies with the three AMS inhibitors illustrate that the removal of the hydroxyl led to a 20-fold decrease in affinity, while removal of the 2-amino group had a smaller 4-fold reduction, suggesting that the interactions between the C3 hydroxyl and Ser271 and Asn207 contribute significantly to substrate binding.

We identified potential residues that might hinder the ability of the wild-type enzyme to bind larger, non-native substrates. Modeling analogous substrates into the pocket based on the solved N-terminal structure of NpsA identified clashes of new substituents with several residues, particularly between the C4 position and the side chain of Asn207 (Figure 2A). The NpsA homologs Lim1 (involved in the production of limazepine), TomA (tomaymycin), SibE (sibiromycin), and AntA (anthramycin) possess substitutions at the Asn207 position (Figure 2B).^{38, 39} The PBD sibiromycin has a similar scaffold to tilimycin, with an extra methyl group on the C4 position of the aromatic moiety. Using AlphaFold Google Colab,⁴⁰ we examined a model for the adenylating enzyme SibE created (Figure S1) bound to 3-hydroxy-4-methylanthranilic acid (3H4MA). SibE contains both the adenylation domain as well as the carrier protein, in effect a fusion of NpsA and ThdA from the tilimycin cluster. The overall pLDDT score for the full length protein is 83.4, reflecting high overall confidence in the structure. The dynamic A_sub domain and the linker joining the carrier protein domain reduce the overall pLDDT, as the A_core domain, residues 1-394 possess a pLDDT of 90.9. As the entirety of the substrate binding pocket resides within the A_core domain, this offered high confidence in the SibE model. The position of the 3H4MA ligand was manually docked by superposition with the NpsA structure 6VHX. Asn207 in NpsA is replaced by Thr208 in SibE, suggesting that a shorter alkyl-chain side group could open the pocket for new substituents on the C3 position as well as C4 position of the aromatic ring. Modeling also suggested that a potential second site for inclusion of analogs with C4 or C5 position substituents might be influenced by Ala302, which also shows sequence variance with NpsA homologs (Figure 2B). We therefore created five NpsA mutations for further activity experiments, serine, threonine, and alanine substitutions in the Asn207 position, and serine and threonine substitutions for Ala302.

We identified a panel of aromatic acid analogs (Figure 2C) to be tested with NpsA that contained a six-member aromatic ring and the amine at the C2 position to maintain the potential for spontaneous cyclization of the released aldehyde product. Analogs with substituents added to the aromatic ring such as hydroxyl, methoxy, methyl, and fluoro groups on the C3, C4, C5, or C6 position. We also identified a series of heterocyclic aromatic acids that were also tested.

NADPH Consumption Assay Shows 3HA Analog Incorporation in wild-type and Mutant NpsA

Prior studies to explore promiscuity of wild-type and mutant adenylation domains within NRPS pathways have directly monitored adenylation domain activity as an initial screening tool. These initial biochemical assays allow identification of active mutant/substrate combinations before proceeding to the more laborious pathway reconstruction assays. Previous work from our lab^{26, 37} and other labs^41–44 have used a variety of different adenylate forming assays to analyze engineered enzymes. While this approach enables characterization of the kinetic activity of the adenylation domain, a potential drawback is that substrate analogs identified as promising building blocks for the adenylation domain may fail to be accepted by downstream catalytic domains. This raises the possibility that downstream library diversity may be more limited than expected on the basis of adenylating activity.

We realized that an assay that rapidly confirms pathway throughput would be advantageous. The unique termination feature of the tilimycin/tilivalline biosynthetic pathway enabled a facile initial screening of the complete pathway, enabling the initial screening of enzyme/substrate combinations. The measurement of NADPH oxidation has been used to determine apparent kinetic constants for NpsB in the presence of saturating concentrations of ATP and proline, with stoichiometric concentrations of loaded ThdA.¹¹ We used this assay as a tool to assess pathway throughput with varying NpsA enzymes and substrates by monitoring NADPH consumption spectrophotometrically. Given an apparent K_M value of 63 μM¹¹ for 3HA and potentially limiting substrate solubility, we used 300 μM aromatic analogs to measure the relative rate of the aldehyde release following incorporation and condensation of both substrates in the NRPS pathway.

Wild-type NpsA was found to recognize multiple anthranilate substrate analogs (Figure 3A, Figure S2), as previously observed.¹⁸ In our studies, the wild-type enzyme showed activity with five of nine analogs tested, with activity reduced by more than 10-fold with anthranilic acid (AA), 3-methylanthranilic acid (3MA), 5-hydroxyanthranilic acid (5HA), 3-fluoroanthranilic acid (3FA), and 5-fluoroanthranilic acid (5FA). We tested phenylglycine as a more diverse aromatic analog that could possibly have been used as a potential substrate analog, but no NADPH reduction was observed. Most likely, the introduction of the extra carbon atom that separates the phenyl and amine groups from the carboxylate compared to the other homologs decreased relative activity of either the NpsA adenylation domain, or the NpsB condensation or reductase domain (Figure 3A). Importantly, this compound serves as a control that NpsB is unable to reduce loaded proline to any appreciable level, suggesting that the remaining active compounds resulted from the release of the N-arylproline analog aldehydes.

Figure 3. Biochemical activity assays to evaluate analog inclusion into the TM pathway.

A. NADPH oxidation values for each substrate with either wild-type, N207A, N207T, or N207S mutant NpsA. Values are plotted in triplicate and depicted as mean ± standard deviation. Oxidation activity was monitored by measuring the negative slope at 340 nm, indicating NADPH consumption. B. Spontaneous conversion of tilimycin into tilivalline with the introduction of 2mM indole in the biochemical reaction. C. Superimposed extracted ion chromatograms (EIC) traces of wild-type NpsA favoring substrates, 3HA and 3FA, showing the reconstituted tilimycin (black) and tilivalline (pink) analog products, obtained in reactions in the absence and presence of indole, respectively. D. Superimposed EIC traces of N207A NpsA favoring substrates, 3Methxy, 3MA, and AA, showing the reconstituted tilimycin (black) and tilivalline (pink) analog products obtained in reactions in the absence and presence of indole, respectively. E. Superimposed EIC traces of N207T NpsA favoring substrates, 4HA, 5HA, and 5FA, showing the reconstituted tilimycin (black) and tilivalline (pink) products obtained in reactions in the absence and presence of indole, respectively.

The same set of substrates were tested with the three Asn207 NpsA mutants (Figure 3A). The Asn207 mutants were found to have ~5-fold lower activity than the wild-type enzyme with 3HA, likely due to the loss of a stabilizing hydrogen bond between the Asn207 residue and hydroxyl moiety on the aromatic ring. We were encouraged to see, however, that the mutant activity with analog substrates had higher NADPH oxidation turnover with analogs, except for 6-methylanthranilic acid (6MA) which had no activity with any NpsA constructs and 3-fluoroanthranilic acid (3FA) which maintained a higher activity with the wild-type enzyme (Figure 3A). None of the reactions with mutant NpsA and analogs restored activity to the NADPH consumption rate comparable to the wild-type enzyme with 3HA. While this may reflect reduced adenylating activity by the mutant NpsA, it is also very likely some level of discrimination by the NpsB condensation or reductase domains is still observed with the 3HA analogs during the peptide bond forming or reductive cleavage steps.

Using the NADPH consumption assay to compare the different enzymes, the N207A mutant had better activity than wild-type with several analogs, including AA, 3MA, and 3-methoxyanthranilic acid (3Methoxy). An alanine at the 207 position would allow for not only higher active site volume for a substituent in the C3 position, such as 3Methoxy, but also a hydrophobic interaction with a smaller hydrophobic residue such as a methyl group of 3MA. AA has no substituents on the benzyl ring but favors the N207A mutant slightly more than the N207S mutant versus the activities of wild-type and N207T. This may imply a non-polar residue within the pocket may alleviate repulsion that can be seen with a polar group.

N207T had the highest activity with analogs 4-hydroxyanthranilic acid (4HA), 5HA, and 5FA. 4HA is similar to the native substrate of SibE, 4-hydroxyl-3-methylanthranilic acid (4H3MA), without the presence of a 3-methyl group, and was inactive with the wild-type enzyme. Excitingly, 4HA could be recognized as a substrate by all Asn207 mutants, while the wild-type enzyme showed no activity with this analog. This implies a significant hindrance to incorporating a hydroxyl in the C4 position relates to the distance between substituent and residue, rather than polarity of the active site residue. Considering this, the mutation to threonine in the 207 position seems to better incorporate polar substituents in the C4 and C5 positions of the aromatic ring, while alanine in the same position better incorporates a methyl, methoxy, or no substituent in the C3 position.

Ala302 Mutant and Heterocyclic Compounds are Not Compatible in Tilimycin Pathway

TomA, the NpsA homolog responsible for the biosynthesis of tomaymycin contains a serine residue in place of Ala302 of NpsA (Figure 2B). The TomA substrate contains a hydroxyl in the C4 position and a methoxy in the C5 position of the aromatic ring (Figure 1A). Although the side chain of Ala302 is not oriented toward the aromatic substrate,¹¹ we designed Ala302 mutants with serine and threonine in NpsA to potentially mimic the interaction of TomA with the hydroxyl group of 4HA and 5HA. When A302S and A302T mutants of NpsA were tested in the NADPH oxidation reconstitution assay with 4HA and 5HA, no activity was observed (Figure S3). The different orientation of Ala302 in the active site of NpsA (Figure S3) suggests that additional changes to this loop in TomA might be responsible for accommodating the additional substitutions in the aromatic substrate. Comparing the active site volume of a TomA model with NpsA displays volumetric differences in the substrate binding pocket, further proving that a single residue mutation in this position would not increase reactivity with C4 and C5 position substituents in the context of NpsA.

We also tested a series of heterocyclic compounds in the NADPH oxidation assay, as pyridine carboxylic acids contain an aromatic ring, although with a nitrogen that might contribute to active site interactions. None of these compounds displayed any appreciable activity with wild-type or any of the NpsA Asn207 or Ala302 mutants (Figure S4).

Tilimycin and Tilivalline Analogs synthesized using NpsA Asn207 Mutants

Encouraged by the NADPH screening results, the production of tilimycin and tilivalline analogs through biochemical reconstitution was next evaluated by LC-MS with the enzyme and substrate combinations that illustrated the highest activity from the NADPH oxidation assay (Figure 3C–E). (We note that analogs of tilimycin (TM) are described by the 3HA analog used in the reaction. For example, the TM analog derived from 3-methylanthranilate will be referred to as 3MA-tilimycin). Reactions were analyzed by LC-MS to determine whether product peaks were observed on extracted-ion chromatograms (EIC) with expected (M+H) m/z values (Figure 3C–E). 6MA was excluded from LC-MS testing due to poor activity in the NADPH oxidation assay (Figure 3A).

All expected products were observed except for the tilimycin analog derived from 5HA, which was assayed with NpsA N207T (Figure 3C–E). A doublet peak is present in some LC-MS samples that we attribute to culdesacin, an isomer of the TM analog derived from the spontaneous alternate ring closure mechanism (Figure S5), as observed previously.⁴⁵ In a separate reaction, 2 mM indole was added to produce tilivalline (TV) analogs (Figure 3C–E). TV analogs were observed in all reactions except for the 5HA-tilimycin reaction, which was expected due to its failure to show conversion to 5HA-tilimycin.

In the NADPH screening and biochemical reconstitution experiments, some analogs, such as those derived from 3MA and 5FA, were produced at higher levels with a mutant adenylation domain than with the wild-type NpsA (Figure 4A). Additionally, wild-type NpsA displayed little to no activity using 4HA as a substrate and N207T was able to incorporate 4HA in the pathway with the highest activity. We tested product formation of 4HA-tilimcyin over multiple time points using the wild-type and N207T mutant, to compare the rate of 4HA-tilimycin formation. It was observed that not only was there a higher amount of 4HA-tilimycin produced with N207T, but also a higher rate of product formation was observed (Figure 4B). This demonstrates the engineered mutants show an enzymatically faster process synthesizing the TM analogs versus wild-type NpsA. Because the reactions were performed with the same, wild-type NpsB enzyme, we attribute the increased activity to the mutation in NpsA.

Figure 4. Wild-type vs. mutant TM analog production.

A. LC-MS EIC peaks after 1 hour for 3MA, 5FA, and 4HA tilimycin using wild-type or respective high activity mutant NpsA enzymes. B. 4HA product normalized with 1mM phenylalanine internal standard for time points 0.5, 1, 2.5, 4, and 15 h using wild-type or N207T NpsA.

NpsB also exhibits substrate promiscuity and expands library of PBD analogs

While NpsA has been the primary focus in prior studies for substrate variants,¹⁸ exploration into NpsB catalytic versatility is limited to a single study study using 2,3-dehydroproline, which produce small amounts of a tilivalline derivative.¹⁹ NpsB is produced at lower levels and is less stable^{11, 30} (Figure S6), which may have limited its investigation. We leveraged the NADPH oxidation assay to test NpsB promiscuity with various proline-like analogs with varying substituents around the pyrrolidine ring (Figure 5A). We also tested several analogs with changes to the ring, including a four-membered ring, a dehydrogenated five-membered ring, a six-membered ring, and a thiazolidine analog containing a sulfur atom (Figure 5A). These alternative substrates have been described as bioactive proline analogs making them interesting, while also providing insight on NpsB adenylation domain active site binding.^46–49

Figure 5. Substrate scope of NpsB and resulting PBD analog library.

A. Proline analogs with noted substituents and ring modifications tested in reconstitution reactions. B. NADPH oxidation activity of reconstitution with 3HA natural substrate and proline analogs. Values are plotted in triplicate and values are depicted as mean ± standard deviation. Oxidation activity was monitored by measuring the negative slope at 340nm. C. Heat map representation of 3HA and proline tilimycin variants. The heatmap scale reflects production of analogs, which are quantified based on EIC heights of the analogous product compared to the natural tilimycin product produced with wild-type NpsA. D. Heat map representation of 3HA and proline tilivalline variants in biochemical reactions containing indole. Ratios are quantified based on EIC heights of the analogous product over the natural tilivalline product produced with wild-type NpsA.

We screened the pathway with the NADPH reduction assay, highlighting that NpsB is able to incorporate several analogs that progress to the condensation and reductase domains, including the hydroxylated cis-L-4-hydroxyproline [(2S,4S)-4-hydroxypyrrolidine-2-carboxylic acid, cis-L-4-HP], the oxidized 3,4-dehydroproline (3,4-DH-LP), and an analog with a smaller ring, 2-aziditidine carboxylic acid (2-AzCA) into the natural NRPS pathway at an activity comparable to L-proline (Figure 5B).

Having demonstrated the ability of both NpsA to accept alternate substrates, with mutant enzymes possessing higher activity, and of NpsB to use proline analogs and retain catalytic competence with a variety of analogs of both substrates, we proceeded to test the production of analogs from mixtures of 3HA and proline analogs. Combinations of substrates were then provided to the TM and TV reconstitution assay to form secondary analogous products using the NpsA mutant that possessed the highest activity with each anthranilate substrates.

Consistently, tilivalline is produced at lower levels that tilimycin in our reconstitution assay, as detected by LC-MS. However, we were encouraged to see that analogs of tilimycin were observed at levels greater than 10% of the amount of tilimycin produced with wild-type NpsA with all but four combinations, AA or 3FA with cis-L-4-HP, 4HA with 3,4 DH-LP, and 5FA with 2AzCA (Figure 5C). The addition of indole also led to formation of tilivalline product analogs, displaying a novel library of PBD analogs (Figure 5D, Figure S7). Although these proline analogs could be incorporated via biochemical reconstitution, they did not produce a high enough quantity to be considered for subsequent microbial experiments

Examination of the bioactivity of tilimycin analogs

We next wanted to test the bioactivity of the tilimycin analogs that could now be produced. We initially tested the alkylation ability of analogs in a PCR inhibition assay, reasoning that alkylation of a plasmid template would hinder DNA amplification by Phusion polymerase. We used the alkylating agent methyl methanesulfonate (MMS), which forms guanine adducts, as a positive control at concentrations ranging from 0 to 500 mM based on previous alkylation studies done in yeast, with variable results at intermediate concentrations.⁵⁰ PBDs have a lower concentration requirement than MMS for alkylation due to instability of the smaller alkylator and PBDs being able to form stable covalent adducts.^{1, 51} We used three 150 μl biochemical reconstitution reactions to produce analogs from multiple substrate building blocks and the preferred mutant enzymes, and isolated the molecules through HPLC (Figures S8–S12).

Doses of 1 μM, 30 μM, and 100 μM for the isolated PBD analogs were tested in polymerase, consistent with previously reported concentrations in related assays.^{2, 13} While the natural product, 3HA-tilimycin, and 3MA-tilimycin have no effect on amplicon replication at the concentrations tested, the analogs AA-tilimycin, 3FA-tilimycin, and 5FA-tilimycin reproducibly decreased amplification almost completely at 100μM (Figure 6A). Due to low yields on product isolation, 3HA-tilimycin and analogs 3MA-tilimycin and 4HA-tilimycin were only tested at higher concentrations. However, we note that in a single replicate, 4HA was also able to inhibit PCR amplification at 100 μM (Figures S13, S14). 3-Methoxytilimycin was not tested due to low product yield. We note that PCR amplification often exhibits a degree of variability resulting in some replicates where no DNA was amplified even in the absence of an alkylator (Figure S14). However, taken as a whole, the analysis with tilimycin analogs shows no inhibition of DNA polymerization for natural tilimycin in all experiments tested, but a clear trend for inhibition of DNA polymerization at the higher concentrations of several analogs.

Figure 6. TM analog biochemical and microbiological alkylation screening.

A. PCR inhibition gels of positive control MMS at 0mM, 5mM and 500mM and TM analogs at 1μM, 30 μM, and 100 μM. Full gels and replicates are shown in Figure S13 and summarized in Figure S14. B. Growth curves of Δ3 cells with 3HA-TM and lower viability analogs. No substrate represents the growth control and MMS at 0.25mM is shown as a positive control. C. Apparent viability of ΔalkA, Δtag, and ΔdinB E. coli cells incubated with TM-analog reactions. MMS is shown for reference as positive control. D. Growth curves of wild-type cells with 3HA-TM and lower viability analogs. No substrate represents growth control and MMS at 0.25mM is shown to have a reproducible increase in growth. E. Apparent viability of wild-type E. coli cells incubated with TM-analog reactions. Statistical significance compared to no substrate control is represented as ns (p-value>0.05), * (p-value<0.05), ** (p-value<0.01), *** (p-value <0.001), **** (p-value<0.0001).

To evaluate the ability of TM analogs to act as antibiotics, a screening method was developed to test the impact of the biochemical reaction for tilimycin analogs on growth of both a wild-type and DNA repair mutant strain of E. coli MG1655. A mutant cell line with alkA, tag, and dinB gene knockouts is sensitive to DNA alkylation, as these genes are major components of the base excision repair (BER) and SOS DNA repair pathway.⁵² We used MMS, which has been observed to decrease growth in these cells, as a positive control.⁵² In contrast to the DNA repair defective mutant cell line, wild-type cells are able to efficiently remove the adducts from the DNA (Figure S15) and growth in the presence of MMS is not affected.

Our strategy was to perform the biochemical reconstitution reactions to produce the tilimycin analog and then combine the reaction directly with cells in minimal media, allowing continued biosynthesis of product during cell growth. Because the biochemical reconstitution reactions were directly used to treat cells, we first tested to see if the 3HA analogs were toxic, to rule out that any observed impact on growth was caused by residual substrate. Of the anthranilic acid analogs tested at 1 mM, only 4HA affected the alkylation sensitive E. coli (Figure S16). We also performed a reaction in the absence of NpsB, to test the possibility that aryl-adenylate compounds might be formed and leak off the NpsA adenylating domain. Reactions performed in the absence of NpsB indeed had an apparent growth effect on the knockout cells (Figure S17). We attributed the impact on cell growth with analog reaction reconstituted with no NpsB to the presence of the aryladenylates. We therefore analyzed all full reconstitution reactions (including NpsB) to monitor the presence of the adenylate that might impact cell growth. We observed that in the full reconstitution reactions the aryl-adenylate compounds were present at very low concentration, with the exception of the 3HA-tilimycin reaction (Figure S17F). However, the 3HA-tilimycin reaction resulted in no effect on growth. This suggests, the concentration of aryl-adenylate formed in the natural reaction does not impede the growth of the alkylation sensitive E. coli strain and that the impact on growth of the DNA-repair defective mutant was due to the presence of the tilimycin analogs.

The biochemical reconstitution reactions that produced the tested TM-analogs were observed to have a small but reproducible decrease in growth with the triple knockout strain, except for the natural 3HA-tilimycin, which had no effect on the growth (Figure 6B and C). We also tested the effect of the tilimycin analogs on wild-type E. coli. No effect was observed on wild-type cells, implying the analogs were acting as alkylating agents (Figure 6D and E). Surprisingly, a small decrease in growth was observed for the analogs 5FA-TM with wild-type cells, suggesting either a potent alkylation higher than other analogs or a possible off target killing effect.

Discussion

Our investigation of the promiscuity of the tilimycin and tilivalline biosynthetic pathway has expanded the capability of NpsA to accept variant substrates and also demonstrated promiscuity within the catalytic domains of NpsB to generate 44 new PBDs. Informed by prior structures of NpsA and further supplemented by multiple sequence alignments with other PBD anthranilate adenylation domains, we employed directed mutagenesis within the active site to change the volume and endogenous interactions between NpsA and the analogous substrates. Using the NADPH oxidation assay to screen for preferable analog and mutant pairs, we were able to conclude not only that NpsA could incorporate these analogs for the adenylation reaction, but also that the incorporated substrate analog progressed through the entire pathway, being recognized by both the NpsB condensation and reductase domains. While prior studies have focused on adenylation screens to assess initial substrate inclusion, these assays provide no insight into the ability of a substrate to be incorporated in subsequent NRPS steps, often limiting throughput caused by roadblocks in downstream domains.^{26, 53} This assay therefore provides a more comprehensive evaluation by enabling full pathway reconstitution. Having produced multiple tilimycin analogs, we further tested several for biochemical and microbial analysis showing analogs have higher alkylation activity and appear to be more potent than the native tilimycin.

The heterocylic pyridine compounds were examined and did not result in any detectable NADPH oxidation, suggesting that they were either unable to be adenylated and loaded onto ThdA by NpsA or failed to react further in the NpsB condensation or reductase domains. The carboxylic acid reductase (CAR) proteins are three-domain NRPS-like proteins that contain an adenylate, carrier protein, reductase architecture.⁵⁴ Five CAR homologs that share 25-30 % identity with the NpsA adenylation domain the NpsB reductase domain were examined for the ability to use pyridine-2-carboxylic acid as a substrate. The activity of the adenylation and reduction steps were not monitored separately. Three CAR homologs failed to react with any of the pyridine derivative while the remaining two CAR homologs showed an apparent catalytic efficiency that was reduced by several orders of magnitude compared to the best substrate.⁵⁴ The failure of heterocycles to be incorporated into the tilimycin pathway similarly could result from improper or lack of adenylation of these compounds using NpsA, or the substrates are able to continue through the pathway but are unable to be released as the final aldehyde product.

The inability of any of the mutant enzymes to incorporate 6-methylanthranilic acid into a tilimycin derivative further suggests that mutations to the hydrophobic side of the NpsA substrate binding pocket may still be required to further expand the potential to generate 6-substituted analogs. Additionally, the possibility remains that 6MA was adenylated by NpsA but failed to progress further into the pathway, with a potential roadblock at the condensation or reductase domains.

The only tilimycin analog that appeared promising in the initial screening using NADPH oxidation that failed LC-MS confirmation is 5HA-tilimycin. This analog presented moderate activity with wild-type and mutant NpsA proteins but yielded no peak for the appropriate EIC measurement. We speculate that the released aldehyde might fail to cyclize to form the tilimycin analog or could further react to an unidentified secondary product. Several peaks were found in the total ion chromatogram but were not explored further.

Two previous studies published in 2018 used precursor directed biosynthesis of the tilimycin/tilivalline analogs in bacterial cells expressing the biosynthetic pathway. In one, the three genes from K. oxytoca were heterologously expressed E. coli and cells were simultaneously induced for protein expression and treated with alternate substrate building blocks, including proline, and analogs of 3HA and indole.¹⁸ A second study explored the homologous pathway from Xenorhabdus indica. Precursor directed biosynthesis was used with both wild-type X. indica cells and in an E. coli strain that heterologously overexpressed the tilivalline pathway.¹⁹ In both studies, multiple tilivalline analogs were identified in culture media that resulted from the building blocks provided in the media. Both studies used wild-type enzymes and neither tried to quantify the tilimycin analogs lacking the indole moiety, which had only been identified as an alternate product^{45, 55} shortly before publication of these two studies. A sibiromycin analog, lacking the hydroxyl in the C4 position of the substrate, has also been produced using a mutasynthesis method similar to the Bode group¹⁹, showing the potential for producing other PBDs purely through biological means.⁵⁶

Including phenylglycine and the three pyridines, we screened a total of 12 aromatic acids and nine proline analogs as potential substrates for the biochemical reaction. The initial screening with the NADPH oxidation assay allowed us to eliminate five aromatic molecules and six proline analogs. From the most promising 7 anthranilic acid and 4 proline analogs, including the natural substrates 3HA and proline, we performed a high-throughput assay to test the production of 28 tilimycin and 28 tilivalline analogs. Of the 56 total compounds, we were able to generate a library of 44 tilimycin/tilivalline analogs using biochemical reconstitution with purified enzymes and isolate several of the highest producing products. The tilimycin/tilivalline pathway appears robust, with quantities of alternate products similar to or, for several tilivalline analogs, exceeding the amount of natural product produced in biochemical experiments. Based on the NADPH preliminary screen and, in limited exploration with the LC-MS comparison of wild-type and mutant enzymes, the use of mutant NpsA adenylation domains appears to improve product yield and potentially enable the production of analogs that could otherwise not be studied.

Because of limitations in NpsB solubility and stability, we have previously described a cell-free lysate method that may allow easier scale-up of the production of tilimycin compared to biochemical reconstitution.³⁰ These results highlight the possibility of producing a library of novel PBDs at a high concentration using a greener method to generate these compounds for microbial testing.

The biosynthetic approach highlighted here does contain limitations. Most significantly, the amount of tilimycin analogs produced was limited to ~1 mg of material. This prevents rigorous chemical analysis at this stage to confirm purity and chemical structure, and also challenges more complete bioactivity screening. However, the system does allow us to produce enough material quickly, cheaply, and cleanly for initial testing in the DNA polymerization and growth inhibition assays. The broad goal of our approach is to identify analogs that could be prioritized for synthesis for more complete chemical and biological analysis.

Finally, tilimycin has not been found to have prominent antibiotic activity, but has been studied for its cytotoxicity in mammalian cells.¹³ We were intrigued by our findings through the E.coli screening that the alkylation sensitive strain was significantly affected by the presence of tilimycin analogs that, during reconstitution, produced five-times less product (3FA-tilimycin). Complemented with the PCR assay, analogs such as AA-tilimycin, 3FA-tilimycin, and 5FA-tilimycin could be potentially high affinity alkylators that could be carried over into other cytotoxicity studies, such as antifungal or anticancer cell testing. These analogs also benefit from a lack of hydroxy group in the C3 position, which is associated with cardiotoxicity with related PBDs.¹ While the bioactivity of pyrrole derivatives of tilimycin were not further explored, mutagenesis studies involving NpsB may also be beneficial for increasing production of these proline derivatives or expanding the scope of proline analogs that can be incorporated into novel natural products.

Combined, this study explored biochemical techniques to produce a large library of PBD compounds, independent from traditional organic chemical methods. We expanded upon a naturally promiscuous NRPS system by structure-guided mutagenesis that not only demonstrated the possibility to engineer catalysts for new chemical analogs with potential for antimicrobial use, but also the flexibility and feasibility of synthesizing said compounds in a biochemical fashion. We note that we were able to progress to the bioactivity screening of 28 compounds in a single day, saving time over traditional chemical synthesis and isolation. Our approach can establish that biocatalytic processes can result in sufficient quantities of compounds for initial bioactivity screening, identifying the most interesting compounds that can progress to chemical synthesis, focusing synthetic efforts on the most promising analogs. Additionally, continued investigation of the engineering of NRPS enzymes with smaller and relatively simpler systems such as the two module PBD biosynthetic pathways described here can establish the rules necessary for generation of biocatalysts that can produce expanded libraries of novel natural products for more complex biosynthetic pathways for bioactivity screening.

Supplementary Material

Supporting Information is available free of charge at

• Modeling of NpsA homolog SibE

• Biochemical assays including controls, A302 mutant NpsA, heterocyclic substrates, and LC-MS traces for products with proline analogs.

• Analysis of isolated tilimycin analogs

• DNA polymerization assay full gels and replicates

• Cell growth curves, including controls with substrates and adenylate intermediates

• Protein, primer, and polymerization template sequences