Methylations in complex carbapenem biosynthesis are catalyzed by a single cobalamin-dependent radical S-adenosylmethionine enzyme

Abstract

Complex carbapenem β-lactam antibiotics contain diverse C6 alkyl substituents constructed by cobalamin-dependent radical SAM enzymes. TokK installs the C6 isopropyl chain found in asparenomycin. Time-course analyses of TokK and its ortholog ThnK, which forms the C6 ethyl chain of thienamycin, indicate that catalysis occurs through a sequence of discrete, non-processive methyl transfers.

Cobalamin (Cbl)-dependent radical S-adenosyl methionine (rSAM) enzymes catalyze methyl transfer reactions on an array of sub-strates important in the biosynthesis of many classes of natural products.^1–3 ThnK, which is produced by Streptomyces cattleya (NRRL 8057), installs the two carbons that form the ethyl chain at C6 of thienamycin (1), the paradigm carbapenem antibiotic.⁴ There are approximately 50 naturally occurring carbapenems produced by various Streptomycetes, which possess structurally diverse C6 side chains.⁵ As shown in Fig. 1, northienamycin (2) contains only one carbon at C6,⁶ and the asparenomycins⁷ (3) and carpetimycins⁸ (4) contain three carbons extended from C6, diverging from the canonical ethyl chain found in thienamycin and MM 4550 (5).⁹ Known carbapenem biosynthetic gene clusters (BGCs) contain three putative Cbl-dependent rSAM enzymes annotated as methyl transferases, all three of which are essential for the production of the carbapenem MM 4550.^9,10 In early biosynthetic investigations of thienamycin it was shown that the only two methionine-derived carbons in the antibiotic make up the C6 ethyl chain.^11,12 However, the origin of the third carbon in carbapenems with a three-carbon group extended from C6, such as asparenomycin A, has not been addressed. In this paper, we demonstrate that TokK,^§ the ThnK ortholog from Streptomyces tokunonensis (ATCC 31569), is capable of three methylations to form an isopropyl group appended to C6 of a pantetheine-containing carbapenam substrate. These results verify that a single rSAM methyltransferase is sufficient for the biosynthesis of complex carbapenems, and the remaining two rSAM enzymes present in each BGC play other as yet undeter-mined biosynthetic roles.

Fig. 1.

(A) Representative carbapenem natural products with diverse C6 side chains. (B) Proposed mechanism for Cbl-dependent rSAM methyl transfer by TokK.

The tokK and thnK genes were inserted into pET29b to express proteins with a C-terminal His₆-tag and an intermediate TEV protease cleavage site to aid purification. To ensure proper installation of the [4Fe–4S] cluster and the Cbl cofactor, the plasmid for each ortholog was maintained in Escherichia coli BL21 (DE3) along with pDB1282, which codes for the isc operon,¹³ as well as the Cbl-uptake system encoded on pBAD42-BtuCEDFB.¹⁴ Growing each three-plasmid strain in M9-ethanolamine medium yielded soluble, active protein. Although ThnK can be expressed in modest but experimentally usable amounts without the co-expression of pBAD42-BtuCEDFB,⁴ the yield of TokK without this plasmid was very low. Therefore, the marked increase in soluble TokK upon improved Cbl uptake was essential to compare these two enzymes (Fig. S1, ESI^†).

As illustrated in Fig. 1B, Cbl-dependent rSAM methylases such as ThnK and TokK use SAM in two distinct ways.² The first equivalent is used as a methyl donor to achieve the active MeCbl cofactor, and the second is reductively cleaved by the [4Fe–4S] cluster to generate a 5′-deoxyadenosyl radical (5′-dA^•). This radical, recently observed for the first time,¹⁵ abstracts a hydrogen atom from the substrate. The resulting substrate radical then accepts a methyl radical from MeCbl, completing one methylation cycle. The fate of each molecule of SAM manifests in two different coproducts: S-adenosylhomocysteine (SAH), and 5′-deoxyadenosine (5′-dA). When TokK is incubated with substrate 6a in a 30 minute fixed time assay (using conditions under which only one methylation event is observed), SAH, 5′-dA, and methylated product 7 are produced in a 1.0:1.1:0.9 ratio (Fig. S2, ESI^†), consistent with the general mechanism described above.

In addition to mechanistically resembling ThnK, TokK also prefers the same substrate. In 90 minute fixed-time assays monitored by ultraperformance liquid chromatography-high resolution mass spectrometry (UPLC-HRMS), a single methylation is observed with both substrate 6a and its (2S)-diastereomer 6b. However, di- and tri-methylated products were only observed in TokK reactions with the (2R)-diastereomer 6a (Fig. 2A), which therefore was used as the substrate in all subsequent experiments. As shown in Fig. 2B, extracted-ion chromatograms for the product of each additional methylation show a corresponding increase in retention time as the products become more hydrophobic. The first two methylations catalyzed by both TokK and ThnK result in identical products 7 and 8, but only TokK is capable of a third methylation to produce 9, which is on pathway to asparenomycin. Mass spectrometric fragmentation of carbapenams and carbapenems is known to occur through a formal retro-[2+2] cleavage,⁴ which leads to formation of a ketene and an endocyclic imine. Any modifications at C6 remain with the ketene fragment, while the imine fragment retains the charge detectable by positive-ion mass spectrometry. Despite increase in the mass of the parent ion for each methyl added to the carbapenam, the imine fragments are identical for all with the increased mass owing to modification specifically at C6 (Fig. S3, ESI^†). To confirm that the third methyl transfer by TokK leads to the branched isopropyl chain found in asparenomycin, an authentic standard was prepared and was found to match the chromatographic and mass spectrometric properties of the tri-methylated TokK product (Fig. 2B inset, UPLC-MS and MS/MS data in both positive and negative modes are displayed in Fig. S4–S9, ESI^†). As shown in Fig. 2C, applying an established strategy¹⁶ the cis-disubstituted azetidinone 10 was first formed in a stereocontrolled manner using the method of Lectka.¹⁷ The acid 13 was homologated to 14 and extended to the diazoketoester 16.¹⁶ Rhodium(II) catalyzed carbene insertion gave the 2-oxo-carbapenem, which was readily reduced and dehydrated to 17.¹⁸ β-Addition of pantetheine⁴ gave a mixture of C2 diastereomers, which could be separated by preparative HPLC whereupon an unambiguous specimen of 9 was obtained by hydrogenolysis.

Fig. 2.

Activity assays of TokK and ThnK. (A) Fixed time assays of TokK incubated with both (2R)- and (2S)-pantetheinylated carbapenams (6a and 6b) show that monomethylation is possible with both stereoisomers, but that additional methylation events only occur with the (2R)-diastereomer 6a. (B) Representative extracted-ion chromatograms (EICs) of sequentially methylated carbapenams detected in TokK (top) and ThnK (bottom) assays containing equimolar enzyme and substrate 6a. Dashed traces are EICs from the same assay without SAM. From left to right, EICs are for m/z = 432.18 ± 0.01, 446.20 ± 0.01, 460.21 ± 0.01, and 474.23 ± 0.01, respectively. (C) Synthesis of TokK product standard 9.

Cbl-dependent rSAM enzymes are notoriously difficult to work with, and there are relatively few characterized examples despite their ubiquity in nature (7000 predicted members of the Cbl-dependent subfamily²). Although significant progress has been made in expressing and purifying these enzymes, rigorous kinetic analysis of Cbl-dependent rSAM enzymes remains elusive owing to the batch-to-batch variation in enzyme activity. However, to further probe the functional differences between TokK and ThnK (79% identical, Fig. S10, ESI^†), 72 h time-course assays containing equimolar enzyme and substrate 6a (100 μM each) were conducted and the product profiles analyzed. In the dilute, buffered conditions of the assay, the carbapenam substrate 6a and alkylated 7, 8 and 9 were found to be stable to detectable hydrolysis of the β-lactam ring for up to 72 h. The successive addition of CH₂ units (hydrocarbon) was reasonably taken not to affect ion formation by electrospray and, therefore, peak areas normalized to the sum of all carbapenam species present were proportional to concentration. Throughout the time course, the sum of ion counts of all β-lactam species remained approximately constant, providing an internal check of stability. As shown in Fig. 3C, ThnK produces an early rise in the concentration of mono-methylated product 7, with full conversion to the di-methylated product 8 by approximately 48 h. The tri-methylated product does not accumulate, which is consistent with the principal observation of a C6 ethyl side chain in thienamycin. The TokK time course (Fig. 3D) gave very different results, with rapid conversion of the available substrate to the methyl product 7, then the slightly delayed ethyl product 8 (Fig. 3E) followed by a slower third methylation to yield the C6 isopropyl carbapenam 9.

Fig. 3.

Time course analysis of TokK and ThnK. (A) Reaction scheme for sequential methylations, which serves as the basis for the COPASI mass action kinetic model. (B) Table of first-order rate constants (h⁻¹) from COPASI parameter estimation tool fitted to the respective time course data. (C) Time course of ThnK with substrate 6a. (D) Time course of TokK with substrate 6a. (E) Expansion of the first 12 h of the TokK time course. For (C–E) symbols represent experimental data and curves correlate to Vcell simulations of flux through the pathway with the given estimated rate constants.

The C6 alkyl chains formed by ThnK and TokK are the result of consecutive methyl transfers that each require the coordinated organization of two molecules of SAM, internal transfers, release of SAH and 5′-dA, protein conformational changes and substrate binding for successful turnover. Such a complex reaction cycle can be visualized acting at two possible extremes. The first would be a purely processive model in which the substrate 6a would bind, three methyl transfers take place with no or minimal release of partially alkylated intermediates and, finally, the isopropyl product 9 would be released. Alternatively, the second mechanism would posit free dissociation/reassociation of initial substrate and all intermediates and 9 in a sequential model (Fig. 3A). The fact that a mixture of successively methylated intermediates was detected in the reactions of ThnK and TokK gave qualitative support to the latter mechanism. In keeping with this observation, S. cattleya produces a small amount of northienamycin (2, 5% of the total carbapenems⁶), which is singly methylated at C6, as well as thienamycin (1), indicating that the mono-methylated product dissociates from ThnK and can proceed through the rest of the biosynthetic pathway. The only other Cbl-dependent rSAM methylase that has been experimentally demonstrated to perform consecutive methylations on its substrate is CysS, which methylates a cystobactamid precursor three times to install t-butyl and s-butyl groups.¹⁹ The library of 25 cystobactamids containing various extents of methylation by the Myxobacterium that encodes CysS also lends support to the sequential hypothesis.

To explore the mechanism more precisely, a mass-action kinetic model of the overall reaction was built using COPASI.²⁰ This model treats each methylation event as a discrete, irreversible step with its own first-order rate constant, as summarized in Fig. 3A. As noted above, the complex set of catalytic events as well as the actual C₁-transfer to substrate are subsumed into k₁–k₃. The parameter estimation tool was then used to fit the model to the experimental data for both ThnK and TokK and determine rate constant values that would give rise to the observed flux of increasingly methylated products. VCell²¹ was then applied to simulate curves using the values from the COPASI parameter estimation. These curves overlay closely with the experimental data, supporting the restrictive model that each methylation event for both enzymes is discrete, not processive. With both ThnK and TokK, the product distribution at 72 h changes very little compared to the 48 h time point. Considering the oxygen sensitivity and instability inherent to the rSAM superfamily, it is remarkable that ThnK and TokK activity is sustained for 48 h.

Because of the variability in enzyme activity between batches, the absolute values of the rate constants are not meaningfully obtained. However, the relative rates of each sequential methylation are informative. The two methylations catalyzed by ThnK take place at similar rates, with the second methylation occurring approximately three-fold slower than the first. The first methylation event catalyzed by TokK occurs at a similar rate to that of ThnK. However, the second methylation by TokK happens more than twice as fast as the first TokK methylation, and approximately 10-fold faster than the second ThnK methylation. The third methylation by TokK is approximately 60 times slower than the second, but is fast enough for the isopropyl carbapenam to be the major product by the end of 48 h. Evolutionary selection for a faster second methylation potentially allowed for the slower third methylation to become product determining on the biosynthetic time scale. Installation of a third methyl group is more sterically demanding than the second, requiring the second methyl group to be positioned in the active site in a manner that accommodates hydrogen atom abstraction and methyl transfer. To date, no Cbl-dependent rSAM methyltransferases have been structurally characterized, and without that information it is not possible to discern which residues enable TokK to overcome this steric challenge and kinetically out-strip ThnK to perform a third methylation.

In conclusion, Cbl-dependent rSAM enzymes orthologous to ThnK are responsible for generating C6 alkyl diversity in complex carbapenems by stepwise methyl transfers taking place principally, if not exclusively, by a sequential mechanism. The other two annotated enzymes of this class in carbapenem biosynthetic gene clusters play essential and likely unprecedented roles in the construction of these important β-lactam antibiotics.