Characterization of TnmH as an O‑Methyltransferase Revealing Insights into Tiancimycin Biosynthesis and Enabling a Biocatalytic Strategy To Prepare Antibody–Tiancimycin Conjugates

Abstract

The enediynes are among the most cytotoxic molecules known, and their use as anticancer drugs has been successfully demonstrated by targeted delivery. Clinical advancement of the anthraquinone-fused enediynes has been hindered by their low titers and lack of functional groups to enable the preparation of antibody–drug conjugates (ADCs). Here we report biochemical and structural characterization of TnmH from the tiancimycin (TNM) biosynthetic pathway, revealing that (i) TnmH catalyzes regiospecific methylation at the C-7 hydroxyl group, (ii) TnmH exhibits broad substrate promiscuity toward hydroxyanthraquinones and S-alkylated SAM analogues and catalyzes efficient installation of reactive alkyl handles, (iii) the X-ray crystal structure of TnmH provides the molecular basis to account for its broad substrate promiscuity, and (iv) TnmH as a biocatalyst enables the development of novel conjugation strategies to prepare antibody–TNM conjugates. These findings should greatly facilitate the construction and evaluation of antibody–TNM conjugates as next-generation ADCs for targeted chemotherapy.

Graphical Abstract

INTRODUCTION

The enediyne antitumor antibiotics are some of the most potent cytotoxic agents found in nature. The biological activity of the enediynes is driven by their shared mechanism of action, electronic rearrangement of the macrocyclic enediyne core to produce a transient benzenoid diradical capable of generating DNA lesions.^1,2 Due to their indiscriminate cytotoxicity, enediynes require a targeting system to be successfully utilized for therapeutic purposes. This has been successfully exemplified by the clinical utilization of neocarzinostatin (NCS, 1) as a poly(styrene-co-maleic acid) conjugate for the treatment of hepatocellular carcinoma and a hydrazide-derivative of calicheamicin (CAL-DMH, 2) as antibody–drug conjugates (ADCs) targeting CD33 (gemtuzumab ozogamicin) and CD22 (inotuzumab ozogamicin) for the treatment of acute myeloid leukemia and acute lymphoblastic leukemia, respectively (Figure 1A).^3–5 The shared mechanism of action of the enediynes and the remarkable clinical success support the wisdom of evaluating additional enediyne scaffolds to identify promising new payloads for targeted chemotherapy.

Figure 1.

Structures of enediynes used clinically for targeted delivery and selected anthraquinone-fused enediynes: (A) neocarzinostatin chromophore (NCS, 1) and a hydrazide-derivative of calicheamicin (CAL-DMH, 2) used in the clinic for targeted delivery; (B) selected anthraquinone-fused enediynes tiancimycin A (3), uncialamycin (4) and amino-UCM (5), yangpumicin A (6), and dynemicin A (7), highlighting the varying substitutions on the A-ring. Active handles introduced by medicinal chemistry or total synthesis to facilitate antibody–drug conjugation, as exemplified by 2 or 5, respectively, are shown in red.

The enediyne natural products are classified based on two structural features: the size of the macrocyclic enediyne cores (9- vs 10-membered) and the pendant peripheral moieties.¹ The 9-membered enediynes, such as NCS, are chromoproteins isolated with their cognate apoproteins stabilizing the labile 9-membered enediyne chromophores. The 10-membered enediynes are discrete small molecules with stable 10-membered enediyne cores and can be further divided into two subfamilies based on the peripheral moieties: the CAL-like and the anthraquinone- fused enediynes. The anthraquinone-fused subfamily consists of tiancimycin A (TNM A, 3),⁶ uncialamycin (UCM, 4),⁷ yangpumicin A (YPM A, 6),^8,9 and dynemicin A (DYN A, 7)¹⁰ (Figure 1B). Currently, clinical use of the anthraquinone-fused enediynes has not been realized.^11,12

Advancement of the anthraquinone-fused enediynes into the clinic has been hindered by their low titers from fermenting the native producers and their lack of functional groups to install appropriate linkers for preparing ADCs. DYN A and YPM A are produced in low titers, at 1.2–1.5 mg/L¹³ and 0.065 mg/L,⁸ respectively, and both are produced by members of the Micromonospora genus that are known to be recalcitrant to common genetic manipulations.^14,15 Both UCM and TNM A are produced by members of the Streptomyces genus, which are known to be genetically amenable.¹⁶ However, submerged fermentation of wild-type Streptomyces uncialis for UCM production is yet to be realized, and UCM titers from fermenting S. uncialis on solid media remain extremely low at ~0.019 mg/L;⁷ wild-type Streptomyces sp. CB03234 produces TNM A at approximately 1–2 mg/L.⁶ Furthermore, advancement of the anthraquinone-fused enediynes as payload candidates is limited by the inability to selectively functionalize the isolated natural products for antibody conjugation. Tremendous efforts have been devoted to overcome some of these challenges, as exemplified by the recent total synthesis of TNM A¹⁷ and several UCM analogues, including the designer amino-UCM (5) with a judiciously installed amino group at C-8 of the A ring (Figure 1B),¹⁸ to enable regiospecific conjugation, thereby evaluating UCM as an ADC payload candidate in preclinical studies.^11,12,18

Significant advancements have been made to facilitate the reliable supply of anthraquinone-fused enediynes by microbial fermentation. Strain improvement of wild-type S. sp. CB03234 has afforded engineered strains, upon subsequent medium optimization, with improved TNM titers exceeding 20 mg/L.^19–21 Comparative analysis of DYN, TNM, UCM, and YPM biosynthesis has enabled formulation of a unified biosynthetic pathway for the anthraquinone-fused enediynes, setting the stage to develop S. sp. CB03234 as a biotechnology platform for production of the natural enediynes and engineering of designer analogues (Figure 2).²² Specifically, the ΔtnmL mutant strain S. sp. SB20020 accumulated TNM B (8) and TNM E (9), whose lack of hydroxyl groups at both C-6 and C-7 would support the predicted function of TnmL as a P450 hydroxylase.^6,22 The ΔtnmH mutant strain S. sp. SB20002 afforded TNM F (10) and TNM C (11), both of which are characterized with a free hydroxyl group at C-7, a structural feature that would be consistent with the predicted function of TnmH as an S-adenosyl-L-methionine (SAM)-dependent O-methyltransferase.^6,22 Together with the isolation of TNM D (13), as well as shunt metabolites supporting the intermediacy of TNM G (12) in TNM biosynthesis, from the wild-type S. sp. CB03234 strain, these findings suggested substrate promiscuity of TnmH, TnmL, or both but fall short of defining their extact timing in the post-enediyne polyketide synthase (PKS) tailoring steps for TNM A biosynthesis as depicted in Figure 2 (path I vs path II).²²

Figure 2.

Unified biosynthetic pathway for the anthraquinone-fused enediynes and the post-enediyne polyketide synthase tailoring steps for TNM A (3) biosynthesis as revealed by the isolation of TNM B (8) and TNM E (9), from the ΔtnmL mutant strain S. sp. SB20020, TNM F (10) and TNM C (11), from the ΔtnmH mutant strain S. sp. SB20002, and TNM D (13), as well as shunt metabolites supporting the intermediacy of TNM G (12), from wild-type S. sp. CB03234. Paths I and II highlight different timing of TnmL in installing the hydroxyl groups at C-6 and C-7 and of TnmH in installing the methyl group at the hydroxyl group of C-7, respectively, and in vitro characterization of TnmH in the current study supports path I.

TnmH as a SAM-dependent methyltransferase presents a unique opportunity to functionalize the A-ring of TNMs (Figures S1 and S2). Inspired by literature precedent that methyltransferases could accept S-alkylated SAM analogues containing reactive alkyl handles,²³ we sought to characterize TnmH in vitro to determine its timing in the post-enediyne PKS tailoring steps for TNM A biosynthesis, investigate its substrate specificity and promiscuity, and explore its utility as a biocatalyst to functionalize the A-ring of the TNM scaffold for andibody–TNM conjugation. Here we report the biochemical and structural characterization of TnmH, revealing that (i) TnmH catalyzes regiospecific methylation at the C-7 hydroxyl group, with 11 as the preferred substrate to 10, (ii) TnmH exhibits a broad substrate promiscuity toward both hydroxyanthraquinones and S-alkylated SAM analogues and catalyzes efficient installation of reactive alkyl handles regiospecificially at C-7 of the TNM scaffold, (iii) the X-ray crystal structure of TnmH, which is a homodimer in solution, provides the molecular basis to account for its broad substrate promiscuity, and (iv) TnmH as a biocatalyst enables the development of novel conjugation strategies to prepare antibody–TNM conjugates. These findings should greatly facilitate the construction and evaluation of antibody–TNM conjugates as next-generation ADCs for targeted cancer chemotherapy.

RESULTS AND DISCUSSION

Construction of the Second Generation ΔtnmH Mutant Strain S. sp. SB20024 Enabling a Reliable Supply of TNM C (11).

The original ΔtnmH mutant strain S. sp. SB20002, from which 11 was first isolated (0.7 mg from 6 L) and characterized, was constructed in the wild-type S. sp. CB03234 strain.⁶ Subsequent medium optimization improved TNM production, leading to the isolation of 3 (7.2 mg from 14 L), together with 13 (2.3 mg from 14 L) and shunt metabolites supporting the intermediacy of 12 in TNM A biosynthesis, from S. sp. CB03234²² and 11 (4.1 mg from 14 L), together with 10 (1.2 mg from 14 L), from S. sp. CB20002,²² respectively. While isolation of 10, 11, 13, together with the proposed intermediacy of 12, supported the predicted function of TnmH as a SAM-dependent O-methyltransferase (Figures S1 and S2), it also raised the question for the timing of TnmH to catalyze methylation of the C-7 hydroxyl group in TNM A biosynthesis (i.e., path I vs path II, Figure 2).²² It was also noticed during medium optimization that both S. sp. CB03234 and S. sp. SB20002 co-produced a family of metabolites in high titers that interfered with the isolation of TNMs. We have since characterized this family of metabolites as tiancilactones and cloned and characterized its biosynthetic gene cluster (Figure S3).²⁴ Progress has also been made in improving TNM production by subjecting the wild-type S. sp. CB03234 to multiple rounds of strain improvement. Most significantly, two variants, S. sp. CB03234-R, with titer of 3 exceeding 22 mg/L,¹⁹ and S. sp. CB03234-S, with a combinted titer of 3 and 13 exceeding 33 mg/L,²¹ have been isolated. Fortuitously the S. sp. CB03234-S strain lost a large portion of its genome resulting in lose of its ability to produce the tiancilactone family of metabolites as well as other unknown metabolites resulting in a sigfinicantly cleaner background (Figure 3).

Figure 3.

Second generation ΔtnmH mutant strain S. sp. SB20024 provides a reliable supply of TNM C (11). HPLC analysis of (A) the original ΔtnmH mutant strain S. sp. SB20002 producing both TNM F (10) and TNM C (11) (at a combined isolated yield of ~0.4 mg/L) co-produced with the tiancilactones and (B) the second generation ΔtnmH mutant strain S. sp. SB20024 producing TNM C (11) as the sole metabolite (at a titer of ~4 mg/L) free from the tiancilactones. While UV–vis at 254 nm detects both the TNMs and tiancilactones, UV–vis at 539 nm detects the TNMs only. TNM F (10) (●); TNM C (11) (◆); tiancilactone A (▽); tiancilactone B (■).

The improved overall titer of TNMs, together with its simplified metabolite profile, motivated us to construct the second generation ΔtnmH mutant in the S. sp. CB03234-S strain to facilitate the production and isolation of 10 and 11. We inactivated the tnmH gene in S. sp. CB03234-S by replacing it with a kanamycin resistance cassette, and gene replacement was carried out by following previously described methods (Supporting Information Materials and Methods).⁶ The resultant second generation ΔtnmH mutant strain S. sp. SB20024 genotype was confirmed by PCR and Southern blot analysis (Figure S4). Under the optimized conditions for TMN production,^21,22 S. sp. SB20024 produces 11 at a titer of ~4 mg/L, free from the inference of the tiancilactones, enabling a reliable supply of 11. Intriguingly, in contrast to S. sp. SB20002, which produces both 10 and 11 in a 1:3.4 ratio,²² S. sp. SB20024 produces 11, with little or undetectable amount of 10 (Figure S3C). This further facilitates the isolation and purification of 11 for in vitro characterization of TnmH (Figures 4, 5 and Table 1) and the utilization of TnmH as a biocatalyst enabling the preparation of antibody–TNM conjugates (Figure 6).

Figure 4.

TnmH as a SAM-dependent O-methyltransferase catalyzes the regiospecific methylation of the hydroxyl group at C-7 of TNM F (10) and TNM C (11) in vitro. (A) HPLC analyses of TnmH-catalyzed reaction in the presence of SAM with TNM F (10) as a substrate: (I) an authentic standard of 10; (II) complete reaction with boiled TnmH; (III) complete reaction at 28 °C for 20 min; (IV) an authentic standard of 12. (B) HPLC analyses of TnmH-catalyzed reaction in the presence of SAM with TNM C (11) as a substrate: (I) an authentic standard of 11; (II) complete reaction with boiled TnmH; (III) complete reaction at 28 °C for 5 min; (IV) an authentic standard of 13. TNM F (10) (●); TNM G (12) (○); TNM C (11) (◆); TNM D (13) (◇).

Figure 5.

Crystal structures of TnmH (PDB entries 6CLW, 6CLX, and 6BBX). (A) Overall structure of TnmH homodimer in complex with SAM (PDB entry 6CLX). (B) Model of TNM C (11) docked into the X-ray structure of holo-TnmH in complex with SAM (PDB entry 6BBX), revealing a large active site pocket. SAM and the surface of 11, shown in purple, and the active site residues His246 and Asp247 are labeled. (C) Close-up view of the S-methyl group of SAM seen in the holo-TnmH structure, with proximal residues labeled and colored by b-factor according to the color key shown.

Table 1.

(A) TnmH-Catalyzed O-Methylation Exhibiting Broad Substrate Promiscuity toward Hydroxyanthraquinones^a and (B) TnmH-Catalyzed O-Alkylation of Hydroxyanthraquinone, As Exemplified by 11, Exhibiting Broad Substrate Promiscuity toward S-Alkylated SAM Analogues

hydroxyanthraquinone	cofactor	product	relative activity (%)
11	SAM	13b	100
14	SAM	19b	5.9
15	SAM	20b	64
16	SAM	21b	49
17	SAM	22b	3.4
18	SAM	23b	0.4
11	SAM	13c	100
11	24	27c	11
11	25	28c	6.5
11	26	29c	26

^aThe A-ring of 14–18 following the same numbering as 11.

^bConditions: 5 μM TnmH, 100 μM hydroxyanthraquinone, 1.0 mM cofactor, pH 7.5, 15 min.

^cConditions: 1 μM TnmH, 250 μM hydroxyanthraquinone, 1.0 mM cofactor, pH 7.5, 30 min.

Figure 6.

Leveraging the TnmH-installed reactive alkyl handles at C-7 of TNMs to develop novel conjugation strategies to prepare antibody–TNM conjugates as exemplified by the synthesis of 27, 28, 29 from 11 and subsequent preparation of 31 from 27.

In Vitro Characterization of TnmH as a SAM-Dependent O-Methyltransferase Catalyzing Regiospecific Methylation at the C-7 Hydroxyl Group, with TNM C (11) as the Preferred Substrate to TNM F (10).

To characterize TnmH in vitro, we cloned the tnmH gene from S. sp. CB03234 and expressed it heterologously in E. coli (Supporting Information Materials and Methods). The overproduced TnmH protein was purified to homogeneity (Figure S5) and found to homodimerize in solution on the basis of size exclusion chromatography (Figure S6). In the presence of SAM, TnmH catalyzed time-dependent conversion of 10 or 11 into a new product, respectively, and these products were absent in the negative controls using boiled TnmH (Figure 4). To establish their identity, the new products were subjected to high-resolution electrospray mass spectrometry (HR-ESI-MS) analysis, revealing their molecular weights to be 14 Da higher than the corresponding substrates, consistent with that of methylated products (Figure S7). Since regiospecifc methylation at the C-7 hydroxyl group of 11 would afford 13, which has been isolated and structurally characterized from the wild-type S. sp. CB03234 strain previously,²² TnmH-catalyzed O-methylation of the C-7 hydroxyl group of 11 was established by HPLC and HR-ESI-MS analysis by co-injection with an authentic standard of 13. Regiospecific methylation of the C-7 hydroxyl group of 10 would afford 12. Although the intermediacy of 12 in TNM biosynthesis has been proposed based on the shunt metabolites isolated from the wild-type S. sp. CB03234 strain,²² 12 has neither been isolated nor structurally characterized to date. Thus, we scaled up the reaction of TnmH-catalyzed methylation of 10, in the presence of SAM, and prepared and isolated the resultant product for full structural characterization (Experimental Section). The identify of the new product 12 was fully established by 1D and 2D NMR spectroscopic analysis (Table S4 and Figures S21–S27). On the basis of the NMR data, the only difference between 10 and 12 was the presence of a CH₃O group (δ_C 56.5, CH₃ and δ_H 4.03, s) at C-7 (δ_C 165.9, C). Specifically, the HMBC correlation between CH₃O at C-7 (δ_H 4.03) and C-7 (δ_C 165.9) and the ROESY correlation between CH₃O at C-7 (δ_H 4.03) and H-8 (δ_H 7.39) supported the assignment of a CH₃O group at C-7 of 12, unambiguously establishing TnmH-catalyzed regiospecific O-methylation of the C-7 hydroxyl group of 10 to yield 12 (Figure 4).

To establish the conditions for kinetic analysis of the TnmH-catalyzed O-methylation of 10 and 11, we used alizarin (14) as a substrate surrogate to determine optimal pH and metal ion dependence (Figure S8) and, following reaction optimization, carried out a time course to determine the linear range of conversion rate for TnmH-catalyzed O-methylation of 11 to 13 (Figure S9). The kinetic parameters for the TnmH-catalyzed O-methylation of 10 or 11 were first determined with saturating SAM (at 1 mM) and varying concentrations of 10 or 11 (from 7.8 to 500 μM) (Supporting Information Materials and Methods). TnmH-catalyzed conversion of 10 and 11 to 12 and 13 followed Michaelis–Menten kinetics with a K_m value of 11.7 ± 1.2 μM and a k_cat value of 0.34 ± 0.076 min⁻¹ for 10 and a K_m value of 31.1 ± 4.1 μM and a k_cat value of 2.69 ± 0.09 min⁻¹ for 11, respectively (Figure S10). These experiments were repeated with saturating 11 (at 500 μM) and varying concentration of SAM (from 7.8 μM to 1.0 mM), and the results displayed similar Michaelis–Menten kinetics with a K_m value for SAM of 62 ± 4.8 μM and a k_cat value of 2.46 ± 0.1 min⁻¹ (Figure S10). The k_cat value determined with saturating SAM is in good agreement with that obtained with saturating 11. Steady-state kinetics of TnmH revealed 3-fold higher catalytic efficiency (k_cat/K_m) for 11 over 10, revealing 11 as the preferred substrate of TnmH. These results are consistent with 11 as the major metabolite accumulated in the ΔtnmH mutants (Figure 3), supporting that sequential hydroxylation of 9 and 10 by TnmL to afford 11 occurs prior to TnmH-catalyzed methylation of 11 to afford 13 in TNM biosynthesis (path I, Figure 2).

TnmH Exhibits a Broad Substrate Promiscuity toward Both Hydroxyanthraquinones and S-Alkylated SAM Analogues.

Substrate promiscuity, a hallmark feature of natural product biosynthetic pathways, has been extensively exploited to generate natural product structural diversity by combinatorial biosynthetic strategies.²⁵ Motivated by the fact that TnmH can catalyze efficient O-methylation of both 10 and 11, we set out to investigate if TnmH could also accept other hydroxyanthraquinones as potential substrates (Supporting Information Materials and Methods, Figure S11 and S12). A panel of five additional hydroxyanthraquinones (14–18) were selected (Table 1A). Compounds 15 and 16, featuring the same A-ring hydroxyl substitutions as 10 and 11, respectively, are cycloaromatized products of 10 and 11 isolated previously from S. sp. SB20002.²² Compounds 14, 17, and 18 are alizarin analogues mimicking the anthraquinone moiety of the anthraquinone-fused subfamily of enediynes with varying hydroxyl substitutions on the A-ring as 10 and 11, as well as 6 and 7 (Figure 1). Under the identical condition used to assay 10 and 11 in the presence of SAM, TnmH recognized all five hydroxyanthraquinones as substrates and catalyzed efficient formation of the O-methylated products (Table 1A), the identities of which were confirmed by HPLC, co-injected with authentic standards, HR-ESI-MS analysis, or both (Figure S13). Similar to 10 and 11, TnmH regiospecifically methylated the C-7 hydroxyl group of 15, exhibiting a higher relative activity for 15 compared to 16. In contrast, while the same regiospecific methylation of the C-7 hydroxyl group of 14 was observed, TnmH displayed similar activities toward 14 and 17. In fact, 18, which features a C-6 hydroxy group, was also O-methylated with a lower relative activity in comparison to 14 and 17 (Table 1A). These findings unveil the intrinsic substrate promiscuity of TnmH toward the hydroxyanthraquinone scaffold and will surely inspire future efforts to engineer new members of the anthraquinone-fused subfamily of enediynes by manipulating the post-enediyne PKS steps in their biosynthesis (Figure 2).

Inspired by the substrate promiscuity toward hydroxyanthraquinones, we decided to investigate if TnmH could use S-alkylated SAM analogues to regiospecifically alkylate the C-7 hydroxyl group, thereby installing reactive alkyl handles to the TNM scaffold. Methyltransferase-mediated exploitation of S-alkylated SAM analogues to install reactive alkyl groups for targeted natural product functionalization is well precedented in the literature.^23,26,27 Three S-alkylated SAM analogues, i.e., S-propargyl (24), S-2-oxopropyl (25), and S-allyl (26), were selected, the choice of which was motivated by the potential of further functionalization of the resultant alkylated TNM products to develop novel conjugation chemistry^28–31 (see Figures 6 and S20). The three S-alkylated SAM analogues were synthesized according to literature procedures (Supporting Information Materials and Methods, Figure S14).³² Under the identical conditions used to assay 11, but substituting SAM with the diasteromeric preparations of 24, 25, or 26, TnmH utilized all three S-alkylated SAM analogues and catalyzed regiospecific alkylation of 11 to afford the corresponding C-7 O-alkylated product 27, 28, or 29, respectively (Table 1B), the identities of which were confirmed by HPLC, HR-ESI-MS, or 1D and 2D NMR spectroscopic analysis (Experimental Section, Figures S15 and S28–S42). As is well-known in the literature for other methyltransferases,^23,33,34 TnmH also suffered from reduced activities toward the three S-alkyl-SAM analogues relative to SAM. Nonetheless, these findings validated our initial hypothesis and warrant further optimization of TnmH as a biocatalyst to enable the development of novel conjugation strategies toward antibody–TNM conjugates.

Crystal Structures of TnmH Reveal a Large Active Site Pocket and Provide a Molecular Basis for the Observed Substrate Promiscuity.

Given the observed unusually broad substrate promiscuity, we set out to characterize the TnmH structure for a deeper understanding of the molecular determinants of substrate promiscuity toward both hydroxyanthraquinones and S-alkylated SAM analogues. TnmH crystals were obtained by using the hanging drop vapor-diffusion method.³⁵ The X-ray crystal structure of TnmH (PDB entry 6CLW) was determined by the molecular replacement method,³⁶ using the structure of the NcsB1 O-methyltransferase (PDB entry 3I53)³⁷ as a search model, and solved at a resolution of 2.74 Å. Crystals of TnmH in complex with SAM were obtained by soaking TnmH crystals with SAM. The holo-TnmH X-ray crystal structure in complex with SAM (PDB entry 6CLX) was determined using the X-ray crystal structure of TnmH as a search model and solved at a resolution of 2.73 Å. A summary of the crystallographic data and refinement statistics is given in Table S5. To generate the ternary structure model of TnmH in complex with both SAM and 11 (Figure 5), the 3D model of 11 was extracted from the TnmS3 and 11 complex structure (PDB code 6BBX) we solved previously,³⁸ docked into the holo-TnmH structure in complex with SAM, and energetically minimized (Supporting Information Materials and Methods).

Consistent with the oligomeric state in solution, as determined by gel filtration chromatography (Figure S6), X-ray crystallography revealed that TnmH is homodimeric and displays an overall architecture similar to the large family of SAM-dependent methyltransferases (Figure 5A).³⁹ Each monomeric subunit is made up of two domains totaling 20 α-helices and nine β-sheets. The dimeric interface primary consists of hydrophobic residues, and the C-terminal domain exhibits a Rossman-like fold conserved in structural homologues,³⁹ which comprises most of the SAM-binding site. While the methyl group of SAM is relatively solvent accessible, the nearest residues impacting SAM binding include Asp247 (3.5 Å), Arg153 (3.4 Å), and Met151 (3.1 Å) (Figure 5B,C). Representation of residues proximal to SAM binding by their flexibility (b-factor) revealed flexibility among residues in the active site, most notably Arg153 (Figure 5C), which may contribute to the observed substrtate promiscuity toward the S-alkylated SAM analogues.

The active site of TnmH reveals a large cavity capable of accommodating diverse substrates. A docking model of the ternary complex generated using the TnmH-SAM holo structure revealed the distance separating the donor methyl group of SAM and the C-7 and C-6 hydroxyl groups of 11 to be 2.7 and 3.8 Å, respectively, accounting for the observed regiospecificity (Figure 5B). Moreover, the distal side of the molecule (consisting of the enediyne core and additional side chain at C-26) is bound in a large, solvent- accessible cavity, providing a rationale for the observed methylation of simple hydroxyanthraquinones and cycloaromatized enediyne congeners that differ substantially in their structural features. This large cavity, with numerous flexible residues at the interface of 11 and SAM, may enable rotations for productive methylation at the C-6 hydroxyl group when simplified substrates that lack the C-7 hydroxyl group, such as 18, are presented. Catalytic roles for residues in the active site may include activation of the phenol for nucleophilic attack via general acid–base catalysis, as implicated previously for phenolic O-methyltransferases.^37,40 In TnmH, two residues in the proximity of the donor methyl group of SAM may play this role; His246 and Asp247 may function to activate the C-7 hydroxyl group of 11, as well as other hydroxyanthraquinones, for nucleophilic attack and/or participate in proton shuttling, as shown by null mutagenesis of each residue independently demonstrating their contribution to catalysis, and abolished activity seen in the double mutant (Figure S16).

TnmH as a Biocatalyst Enables the Development of Novel Conjugation Strategies To Prepare Antibody–TNM Conjugates.

The broad substrate promiscuity observed for TnmH led us to explore the utility of TnmH as a biocatalyst to install reactive alkyl handles to the C-7 hydroxyl group of 11 and develop novel strategies to prepare antibody–TNM conjugates. In an attempt to better accommodate the S-alkylated SAM analogues, four TnmH mutants, Asn243Ala, Asn243Gly, Val244Ala, and Arg153Ala, were constructed to expand the SAM binding site (Supporting Information Materials and Methods, Figure 5C). The β-carbons of Asn243 and Val244 and the η²-nitrogen of Arg153 were found at distances of 4.6, 5.4, and 3.4 Å, respectively, from the donor methyl group of SAM and are hypothesized to sterically clash with or disrupt binding of S-alkylated SAM analogue 24 or 25. The four TnmH mutants were similarly overproduced in E. coli, purified to homogeneity, and subjected to an extra step of activated charcoal chromatography to remove any endogeneously bound SAM (Figure S5D). When assayed under the identical conditions as wild-type TnmH, using 11 and 24 or 25 as substrates, the four TnmH mutants were all catalytically competent, none of which, however, gained any apparent enhancement of 28 or 29 formation relative to wild-type TnmH (Figure S17). Moreover, the ratio of turnover of 24 verse 25 remained relatively constant for all assayed mutants, suggesting there is no preference for either cosubstrates with the engineered mutants. Wild-type TnmH was therefore used as the preferred biocatalyst to further optimize the enzymatic reaction to obtain both 28 and 29.

Exploitation of methyltransferases as biocatalysts is often complicated by S-adenosylhomocysteine (SAH), the co-product of SAM-dependent methyltransferase that is known to exhibit product inhibition, and this is particularly problematic when using S-alkylated SAM analogues as substrates due to their decreased activity relative to SAM.⁴¹ To overcome this problem, we cloned the mtnN gene from E. coli, which encodes a SAH nucleosidase,⁴² and expressed it in E. coli; the overproduced SAH nucleosidase was purified to homogeneity (Supporting Information Materials and Methods, Figure S5B). SAH nucleosidase is known to alleviate the inhibitory effect of SAH on methyltransferases by degrading SAH into S-ribosylhomocysteine and adenine.^41,43 Under the optimized conditions and in the presence of the recombinant SAH nucleosidase, TnmH-catalyzed reactions of 11 with 24 or 25, on 10 mg scale, afforded 27 or 28, respectively, with 50–80% yields (Experimental Section, Figure 6).

Next, we subjected 27 and 28, together with 12 and 13, in comparison with 10 and 11, as well as 3, to determine the effect of the varying alkylations of the C-7 hydroxyl group on their cytotoxicity (Experimental Section, Figure S18). As noted previously,²² methylation at the C-7 hydroxyl group enhanced cytotoxic activity across all cell lines examined. This was seen with the increased potency of 12 and 13 relative to 10 and 11, and 27 and 28 exhibited comparable or enhanced potency relative to 13, with 28 being as potent as 3 (Table 2). The observed relative cytotoxicity for each of these TNM analogues correlated well with their DNA cleavage activity as seen with the plasmid relaxation assays (Experimental Section, Figure S19). Taken together, these findings support the wisdom of exploiting the reactive alkyl handles installed at C-7 by TnmH to develop novel conjugation strategies to prepare antibody–TNM conjugates.

Table 2.

Cytotoxicity (EC₅₀ in nM) of 27, 28, and Other TNM Analogues with Varying C-7 Substitutions (10, 11, 12, 13), in Comparison with TNM A (3), against four Selected Human Cancer Cell Lines

cell line	SF-295	SKMEL-5	MDA-MB-231	NCI-H226
3b	0.45 ± 0.14	0.74 ± 0.2	1.1 ± 0.3	4.2 ± 1.0
10a	51 ± 3.0	36 ± 16	37 ± 3.0	75 ± 9.0
11a	45 ± 12	32 ± 13	34 ± 6.0	66 ± 14
12b	21 ± 14	19 ± 6.5	14 ± 1.0	68 ± 10
13a	4.0 ± 1.6	2.8 ± 0.6	5.0 ± 0.3	32 ± 6.0
27b	2.8 ± 0.7	2.8 ± 0.7	3.7 ± 2.1	26 ± 4.0
28b	1.1 ± 0.4	0.8 ± 0.3	0.9 ± 0.5	3.7 ± 1.0

^aN = 2.

^bN = 3.

Finally, as a proof-of-concept, we assessed the reactivity of 27 under conditions for copper-catalyzed azide–alkyne cycloaddition (CuAAC) to demonstrate the functional utility of the installed propargyl handle in preparing antibody–TNM conjugates. Terminal alkynes are common functional handles for chemoselective bioconjugation due to their reactivity with organic azides in the presence of a Cu(I) catalyst,⁴⁴ but the compatibility of this reaction with enediynes (sensitive to reducing and acidic conditions) was previously unknown. We subjected 27 to conventional CuAAC conditions with a heterobifunctional linker 30 we have reported previously that contained an azide and β-lactam functionality used for site-specific conjugation to uniquely reactive lysine residues of engineered antibodies (Supporting Information Materials and Methods).^45,46 The cycloaddition product 31 was readily observed by HPLC analysis, and upon reaction optimization, we reliably prepared 30 with a >80% yield based on recovered substrate 27 (Experimental Section, Figure 6). The identity of 31 was fully characterized by HR-ESI-MS and 1D and 2D NMR spectroscopic analysis (Figures S43–S51). The efficient synthesis of 31 demonstrates the utility of the reactive alkyl handles installed at C-7 by TnmH in enabling downstream bioconjugation. Similarly, the 2-oxopropyl handle of 28 could be functionalized with a hydrazine as has been exemplified by the CAL-based ADCs.⁴ Allyl-group-enabled bioconjugation, via tetrazine ligation³⁰ or photoclick chemistry,³¹ has also been demonstrated and should be applicable to 29 (Figure S20).

CONCLUSION

The anthraquinone-fused subfamily of enediynes are promising anticancer drug candidates for targeted delivery. Advancement of the anthraquinone-fused enediynes into the clinic, however, has been hindered by their low titers from fermenting the native producers and lack of functional groups that enable ADC preparation. We have now constructed the second generation ΔtnmH mutant strain S. sp. SB20024 in the TNM overproducing S. sp. CB03234-S strain, which reliably produces 11 with a titer of ~4 mg/L (Figure 3). In vitro biochemical characterization of TnmH as a SAM-dependent methyltransferase not only confirmed its precise role in TNM biosynthesis, catalyzing regiospecific O-methylation at the C-7 hydroxyl group of the TNM scaffold with 11 as the preferred substrate (Figures 2, 4), but also unveiled its broad substrate promiscuity toward both other hydroxyanthraquinones and S-alkylated SAM analogues (Table 1). Structural characterization of TnmH provided a molecular basis to account for the observed broad substrate promiscuity (Figure 5). Significantly, we demonstrated the feasibility of installing active alkyl handles to the C-7 hydroxyl group of the TNM scaffold by exploiting TnmH’s ability to utilize selected S-alkylated SAM analogues. The resultant TNM analogues retain full cytotoxicity, supporting the C-7 hydroxyl group as an excellent site for functionalization and antibody–TNM conjugation (Table 2). Finally, we showcased TnmH as a biocatalyst enabling the development of novel conjugation strategies by attaching a β-lactam linker using CuAAC chemistry to the C-7 O-propargylated TNM (Figure 6). Taken together, these findings set the stage to prepare antibody–TNM conjugates and should greatly facilitate the advancement of the anthraquinone-fused subfamily of enediynes as next-generation ADCs for targeted chemotherapy.

While the current study highlighted TnmH-catalyzed regiospecific installation of reactive alkyl groups, thereby functionalizing the TNM scaffold to enable antibody conjugation, this strategy should be readily applicable to the other members of the anthraquinone-fused subfamily of enediynes on the basis of their unified biosynthetic pathway and the demonstrated feasibility to engineer designer analogues in S. sp. CB03234 and its engineered overproducers as platform strains by combinatorial biosynthesis strategies (Figures 1 and 2). It should also inspire exploitation of other enzymes of the TNM biosynthetic pathway or enzymes of any other natural product biosynthetic pathways in general as biocatalysts to overcome difficulties in functionalizing natural products, further enriching the toolbox⁴⁷ for antibody–drug conjugation.

EXPERIMENTAL SECTION

General Experimental Procedures.

All reactions containing water- or air-sensitive reagents were performed using flame-dried glassware under nitrogen or argon. Tetrahydrofuran and dichloro-methane were passed through two columns of neutral alumina. Compounds 3, 8–11, 15, and 16 were isolated following known protocols.^21,22 Compounds 19, 22–26, and 30 were prepared according to literature procedures.^32,46,48,49 All other chemicals were purchased from commercial sources and used without further purification. All ¹H, ¹³C, and 2D NMR (¹H–¹H COSY, ¹H–¹³C HSQC, ¹H–¹³C HMBC, ¹H–¹H ROESY) experiments were run on a Bruker Avance III Ultrashield 700 with QCI cryoprobe at 700 MHz for ¹H and 175 MHz for ¹³C nuclei, a Bruker Avance NEO 600 with TCI cryoprobe at 600 MHz for ¹H and 150 MHz for ¹³C nuclei, or a Bruker Avance 400 at 400 MHz for ¹H nuclei and 100 MHz for ¹³C nuclei. MPLC separation was conducted on a Biotage Isolera One equipped with a Biotage SNAP cartridge KP-C18-HS column (30 g). Semipreparative HPLC was performed on a Varian liquid chromatography system equipped with a YMC-pack ODS-A column (250 mm × 10 mm, 5 μm), unless otherwise specified. LC–MS was performed on an Agilent 1260 Infinity LC coupled to a 6230 TOF (HRESI) equipped with an Agilent Poroshell 120 EC-C18 column (50 mm × 4.6 mm, 2.7 μm). Optical rotations were obtained using an AUTOPOL IV automatic polarimeter (Rudolph Research Analytical). UV was measured with a NanoDrop 2000C spectrophotometer (Thermo Scientific). All compounds tested were >95% pure.

Enzymatic Synthesis and Structural Elucidation of 12, 27, and 28.

For enzymatic synthesis of 12, 27, or 28, a typical 20 mL batch of the reaction mixture contained 100 μM TnmH, 200 μM SAH nucleosidase, 2.5 mM SAM or SAM analog (24 or 25), and 500 μM 10 or 11, respectively, in 100 mM Na₂HPO₄ buffer (pH 7.5) with a final DMSO concentration of 5%. The reactions were incubated at 28 °C overnight with occasional shaking and terminated by the addition of methanol to a final concentration of 50%. The resulting mixture was centrifuged at 3750g at 4 °C for 10 min. The supernatant was extracted with 5 mL of methanol twice and concentrated to afford a crude product that was subsequently subjected to HPLC purification to afford pure products of 12, 27, and 28. Preparative HPLC was performed on an Agilent 1260 Prep Infinity LC with a MWD detector equipped with an Agilent Eclipse Plus phenyl-hexyl column (250 mm × 9.4 mm, 5 μm). Reaction products using 10 and SAM were purified using CH₃OH/H₂O (65/35) as the mobile phase to afford 12 (2.0 mg). Reaction products containing using 11 and 24 or 25 were purified using CH₃OH/H₂O (65/35) to afford 27 (7.2 mg) and 28 (11.0 mg), respectively.

TNM G (12).

Isolated as a purple film; [α]_D²⁵ +1587 (c = 0.00001, CH₃OH); UV (CH₃OH) λ_max (log ε) 209 (0.69), 215 (0.93), 252 (0.73), 270 (0.82), 285 (0.79), 300 (1.1), 374 (0.34), 539 (1.4) nm; ¹H and ¹³C NMR data, see Table S4; HR-ESI-MS (positive ion) affording the [M + H]⁺ ion at m/z 558.1369 (calcd [M + H]⁺ ion for molecular formula C₃₀H₂₃NO₁₀ at m/z 558.1395).

Propargyl-TNM C (27).

Isolated as a purple film; [α]_D²⁵ +1977 (c = 0.00001, CH₃OH); UV (CH₃OH) λ_max (log ε) 209 (1.2), 219 (1.2), 242 (1.6), 260 (1.5), 417 (0.21), 544 (0.79), 576 (0.68) nm; ¹H and ¹³C NMR data, see Table S4; HR-ESI-MS (positive ion) affording the [M + H]⁺ ion at m/z 598.1338 (calcd [M + H]⁺ ion for molecular formula C₃₂H₂₃NO₁₁ at m/z 598.1344).

Keto-TNM C (28).

Isolated as a purple film; [α]_D²⁵ +3,333 (c = 0.00001, CH₃OH); UV (CH₃OH) λ_max (log ε) 209 (1.3), 215 (1.2), 245 (2.0), 260 (1.9), 408 (0.20),544 (1.0), 578 (0.83) nm; ¹H and ¹³C NMR data, see Table S4; HR-ESI-MS (positive ion) affording the [M + H]⁺ ion at m/z 616.1461 (calcd [M + H]⁺ ion for molecular formula C₃₂H₂₅NO₁₂ at m/z 616.1450).

Functionalization of Propargyl-TNM C (27).

As a proof-of-concept for further functionalization and downstream bioconjugation, propargyl-TNM C (27) was subjected to copper-catalyzed azide–alkyne cycloaddition (CuAAC)⁴⁴ to attach a β-lactam functional group (30) that can serve as an attachment handle for antibodies.

β-Lactam TNM C (31).

To a solution of 30 (4.3 mg, 8.0 μmol) and 27 (4.0 mg, 6.7 μmol) in THF (250 μL) were added tris(3-hydroxypropyltriazolylmethyl)amine (THPTA, 0.73 μg, 1.7 μmol), copper sulfate (0.42 μg,1.7 μmol), and sodium ascorbate (0.66 μg, 3.4 μmol). The reaction was stirred at room temperature for 16 h and purified by reverse phase HPLC (10–80% CH₃CN/H₂O) resulting in a purple film (3.6 mg, 83% BRSM). [α]_D²⁵ +348.0 (c = 0.00008, CH₃CN); UV (CH₃CN) λ_max (log ε) 205 (0.36), 215 (0.10), 225 (0.27), 243 (0.23), 277 (0.23), 547 (0.41), 585 (0.36) nm; ¹H and ¹³C NMR data, see Table S6; HR-ESI-MS (positive ion) affording the [M + H]⁺ ion at m/z 1132.3778 (calcd [M + H]⁺ ion for molecular formula C₃₂H₂₃NO₁₁ at m/z 1132.3782).

Cell Viability Assay.

Compounds 3, 10, 11, 12, 13, 12, 27, and 28 were tested for cell viability against four human cancer cell lines including melanoma (SK-MEL-5), breast (MDA-MB-231), central nervous system (SF-295), and non-small-cell lung cancer (NCI-H226) (Figure S18). Suspended cultures of cells were diluted to a concentration of 5 × 10⁴ cells per mL in either RPMI 1640 or DMEM medium supplemented with 10% fetal bovine serum, 100 μg per mL of streptomycin, and 100 unit per mL of penicillin. The suspended cultures were dispensed into 96-well plates (100 μL per well), and the plates were incubated for 24 h at 37 °C in an atmosphere of 5% CO₂, 95% air, and 100% humidity. The original medium was then removed, and 100 μL of fresh medium was added, followed by addition of serial dilutions of drugs (1 μL in DMSO with final concentrations ranging from 0 to 1000 nM). Plates were incubated under the above conditions for 72 h. Finally, 20 μL of CellTiter 96 AQueous One Solution Reagent (Promega) was added to the plates and incubation continued at 37 °C in a humidified, 5% CO₂ atmosphere for 30–60 min. The absorbance at 490 nm was recorded using a SpectraMax M5 plate reader. Each point represents the mean ± SD of three replicates, and the EC₅₀ was determined by computerized curve fitting using GraphPad Prism. The same software was used to determine statistical significance using an extra sum-of-squares F-test. All new compounds in this study (12, 27, and 28) were assayed three times independently against all cell lines examined.

Plasmid Relaxation Assay.

Compounds 3, 10–13, 27, 28 were assayed for their efficacy to relax supercoiled plasmid DNA by generating single- and double-strand DNA breaks, as previously described.⁵⁰ Briefly, assays were performed in 20 μL (total volume) of Dulbecco’s phosphate buffered saline (pH 7.4) with 5% DMSO, containing 0.8 μg (25 nM plasmid, 120 μM base pair) of pUC19 (New England Biolabs), and varying concentrations of compounds. The assays were performed in triplicate in the presence and absence of 1 mM reduced glutathione. The reactions were incubated at 37 °C for 24 h, after which 5 μL of loading dye was added and samples were analyzed by electrophoresis on a 1% agarose gel containing SYBR Green I nucleic acid gel stain (Thermo Fisher). Gel electrophoresis was carried out in TAE buffer at 100 V for 1 h. The relative activity was determined by quantifying the intensity of bands corresponding to supercoiled plasmid (form I, Figure S19B) normalized to DMSO-treated controls. Band intensity was quantified using ImageJ software.⁵¹ The quantified band intensities were plotted using GraphPad Prism (Figure S19C).

ABBREVIATIONS USED

ADC

antibody–drug conjugate

BGC

biosynthetic gene cluster

CAL

calicheamicin

CuAAC

copper-catalyzed azide–alkyne cycloaddition

DYN

dynemicin

HR-ESI-MS

high-resolution electrospray ionization mass spectrometry

NCS

neocarzinostatin

PKS

polyketide synthase

SAH

S-adenosylhomocysteine

SAM

S-adenosyl-l-methionine

TNM

tiancimycin

UCM

uncialamycin

YPM

yangpumicin