Installation of the Ether Bridge of Lolines by the Iron- and 2-Oxoglutarate-Dependent Oxygenase, LolO: Regio- and Stereochemistry of Sequential Hydroxylation and Oxacyclization Reactions.

The core of the loline family of insecticidal alkaloids is the bicyclic pyrrolizidine unit with an additional strained ether bridge between carbons 2 and 7. Previously reported genetic and in vivo biochemical analyses showed that the presumptive iron- and 2-oxoglutarate-dependent (Fe/2OG) oxygenase, LolO, is required for installation of the ether bridge upon the pathway intermediate, 1- exo-acetamidopyrrolizidine (AcAP). Here we show that LolO is, in fact, solely responsible for this biosynthetic four-electron oxidation. In sequential 2OG- and O2-consuming steps, LolO removes hydrogens from C2 and C7 of AcAP to form both carbon-oxygen bonds in N-acetylnorloline (NANL), the precursor to all other lolines. When supplied with substoichiometric 2OG, LolO only hydroxylates AcAP. At higher 2OG:AcAP ratios, the enzyme further processes the alcohol to the tricyclic NANL. Characterization of the alcohol intermediate by mass spectrometry and nuclear magnetic resonance spectroscopy shows that it is 2- endo-hydroxy-1- exo-acetamidopyrrolizidine (2- endo-OH-AcAP). Kinetic and spectroscopic analyses of reactions with site-specifically deuteriated AcAP substrates confirm that the C2-H bond is cleaved first and that the responsible intermediate is, as expected, an FeIV-oxo (ferryl) complex. Analyses of the loline products from cultures fed with stereospecifically deuteriated AcAP precursors, proline and aspartic acid, establish that LolO removes the endo hydrogens from C2 and C7 and forms both new C-O bonds with retention of configuration. These findings delineate the pathway to an important class of natural insecticides and lay the foundation for mechanistic dissection of the chemically challenging oxacyclization reaction.

Lolines are a class of saturated pyrrolizidines with an unusual ether bridge and a 1-exo-amino group that is variously substituted to differentiate members of the class (Scheme ). In cool-season grasses (Poaceae, subfamily Poöideae), lolines are synthesized by endophytic fungi of the genus Epichloë. The fungi are obligate symbionts with the host plant, affording it protection from herbivores by producing several types of objectionable or toxic alkaloid compounds, including lolines. Lolines exhibit biological activities distinct from those of other well-known toxic alkaloids produced by either plants or their fungal symbionts. For example, the plant necine-type pyrrolizidine and fungal ergot alkaloids are both notoriously toxic to mammals such as grazing livestock,[1] whereas lolines have broad-spectrum antifeedant activity and toxicity against invertebrate herbivores but not mammals.[2,3] It has been proposed that their complete saturation and uniquely compact tricyclic core, which results from the strained ether bridge, render lolines less susceptible to the mammalian metabolism that converts the other compounds into potent hepatotoxins.[4,5] These favorable properties could make the breeding of loline-producing grass cultivars a “green” alternative to the use of chemical insecticides.[4]

Scheme 1

Overview of the Loline Alkaloid Biosynthetic Pathway and Representative Lolines

The color coding tracks the atoms of O-acetyl-l-homoserine and l-proline into the loline alkaloids.

The biosynthetic pathway to the loline alkaloids has recently been mapped in some detail.[3,6−12] A cluster of as many as 11 genes (designated LOL) encodes enzymes that are strictly associated with loline biosynthesis. The first committed step is the condensation of O-acetyl-l-homoserine (derived from l-aspartic acid) and l-proline, probably catalyzed by the γ-type pyridoxal 5’-phosphate(PLP)-dependent enzyme, LolC, to give the diamino diacid, NACPP (Scheme ).[8,11] Ensuing cyclization, decarboxylation, and acetylation steps yield 1-exo-acetamidopyrrolizidine (AcAP).[6] Results of in vivo genetic and biochemical studies established that LolO, predicted from its sequence to be an iron- and 2-oxoglutarate-dependent (Fe/2OG) oxygenase, has an essential role in processing the pyrrolizidine moiety of AcAP into the tricyclic loline core. Inhibition of lolO expression by RNAi resulted in diminished production of lolines and marked accumulation of AcAP.[6] Examination of natural fungal variants that produce AcAP as the end product revealed that they all had mutations in the lolO gene. Introduction of a plasmid harboring the wild-type lolO gene into one such mutant strain restored production of N-acetylnorloline (NANL), the expected product of ether-bridge installation upon AcAP. The somewhat surprising implications of these genetic studies—that AcAP, rather than the simpler aminopyrrolizidine, is the entry point to the tricyclic core and NANL the “gateway” loline—were subsequently confirmed by feeding experiments, in which cultures supplied with AcAP deuterium labeled in both the pyrrolizidine core and the N-acetyl group produced NANL with both labels in place. Further processing of the exocyclic moiety of NANL by fungal N-acetamidase (LolN), methyltransferase (LolM), and cytochrome P450 (LolP) enzymes, with occasional involvement of a plant acetyltransferase, generates the other known lolines.[7,12]

The observation from these studies that AcAP (but not NANL or another loline) accumulates in the host fungus in the absence of LolO implicated this presumptive Fe/2OG oxygenase in at least initiating, and possibly completing, ether-bridge installation. Published work on other Fe/2OG enzymes would provide precedent for either possibility. In the conserved mechanistic strategy of the enzyme family, a non-heme FeII cofactor serves to couple reductive activation of O2 with oxidative decarboxylation of 2OG.[13,14] In each case, an FeIV–oxo (ferryl) complex is formed and initiates processing of the substrate. Well-studied transformations of aliphatic carbon centers begin with transfer of a hydrogen atom (H•) to the ferryl moiety, producing a carbon-centered radical. In the most common outcome, the radical then couples with the FeIII-coordinated oxygen, resulting in hydroxylation. Other modes of decay of the substrate radical result in halogenation and desaturation outcomes. Several family members catalyze multiple reaction types in sequence. For example, hyoscyamine 6β-hydroxylase (H6H) mediates successive hydroxylation and cyclization steps to install the epoxide moiety of the plant-derived anesthetic drug, scopolamine,[15,16] and clavaminate synthase (CAS) mediates hydroxylation and (after an intervening hydrolytic step catalyzed by another enzyme) oxacyclization and desaturation reactions on the pathway to the antibiotic clavulanic acid.[17−20] Analogously, LolO could effect sequential hydroxylation and oxacyclization steps to install the loline ether bridge. On the other hand, no available data preclude the alternative possibility that LolO merely hydroxylates AcAP and cyclization is then promoted by another enzyme, which, in the aforementioned in vivo studies, would have been present to process the hypothetical LolO product. Indeed, the lolE gene appears also to encode an Fe/2OG oxygenase, and it has been suggested that the LolE protein product could be responsible for a step in ether-bridge installation.[10]

In this work, we have (i) demonstrated in vitro conversion of AcAP to NANL by LolO, (ii) mapped the sequence of events as hydroxylation of C2 followed by ring closure upon C7, and (iii) delineated the stereochemical course of both steps as endo-H abstraction followed by C–O bond formation with retention of configuration. By establishing that LolO is solely responsible for installation of the loline ether bridge, this study provides another example of the switching of regiochemistry and outcome by a member of the versatile Fe/2OG oxygenase class. In addition, the results establish the necessary foundation for dissection of the mechanism of the oxacyclization step, a reaction type that is among the least well understood within the impressive repertoire of the class.

Results and Discussion

LolE Is Dispensible for Ether-Bridge Installation

As a first step toward defining the role of LolO in ether-bridge installation, we explicitly evaluated the published proposal that LolE might share responsibility by catalyzing one of the two required two-electron oxidations.[6] The absence of a functional lolE gene, which we replaced with a selectable marker gene for hygromycin resistance, did not affect the profile of lolines produced by the host fungus.[21] This observation rules out an essential role for LolE in ether-bridge installation and leaves the enzyme without any assigned role in the overall biosynthetic pathway. It is likely that this second presumptive Fe/2OG oxygenase mediates a bifurcating step to other (currently unidentified) metabolites.

Recombinant LolO Catalyzes the Conversion of AcAP to NANL in Vitro

To test the alternative hypothesis that LolO alone is fully competent to convert AcAP to NANL, we sought to reconstitute the reaction in vitro with purified enzyme (expressed heterologously in Escherichia coli) and small-molecule components. When supplied with FeII, (±)-AcAP, and excess 2OG under an air atmosphere, purified LolO converted AcAP to NANL, with an associated change in the mass-to-charge ratio [Δ(m/z)] of +14 (Figure A, green and blue traces). This result establishes that LolO alone catalyzes oxidation of both C2 and C7 of AcAP to introduce the strained ether bridge. The enzyme can thus promote the final transformation of a five-step chemoenzymatic synthesis of a loline alkaloid from proline (the four chemical steps to AcAP are outlined below); this route is markedly more concise than any previously reported, purely chemical, synthetic procedure.[22−28]

Figure 1

In vitro conversion by LolO of AcAP into NANL via a hydroxylated intermediate. (A) Liquid chromatography–mass spectrometry (LC–MS) analysis of products from incubation of purified LolO with FeII, (±)-AcAP, and varying quantities of the co-substrate, 2OG, under an air atmosphere at room temperature. (B) Variation of the intensities of LC–MS peaks for AcAP (green), the OH-AcAP intermediate (red), and NANL (blue) in analysis of LolO reactions with varying 2OG levels. Anoxic reaction solutions contained 0.13 mM LolO, 0.10 mM FeII, 0.40 mM (±)-AcAP, and 0–0.80 mM 2OG (0–8 equiv of LolO·FeII reactant) in 50 mM sodium HEPES buffer (pH 7.5). Following injection of an equal volume of air-saturated buffer (∼0.19 mM O2 after mixing), reactions were allowed to proceed at room temperature for 10 min prior to LC–MS analysis.

Detection of a Hydroxylated Intermediate Produced from AcAP by LolO

With a few documented exceptions,[29,30] Fe/2OG oxygenases balance the four-electron reduction of O2 with two coupled two-electron oxidations, the first fragmenting 2OG to CO2 and succinate and the second targeting a carbon center in the primary substrate. The conversion of AcAP to NANL is a four-electron oxidation and requires removal of hydrogens from two unactivated carbon centers. This conversion is, therefore, expected to require two sequential reactions, each consuming 1 equiv of 2OG. By limiting 2OG in reaction mixtures containing excess O2 and (±)-AcAP, we resolved the two steps. Liquid chromatography–mass spectrometry (LC–MS) chromatograms from reactions with 2OG:LolO·FeII ratios of ≤1 exhibited a new peak with a Δ(m/z) of +16 relative to the substrate (Figure A, red traces) but no detectable peak at a Δ(m/z) = +14 for NANL. When we substituted 16O2 with 18O2 in the same experiment, the Δ(m/z) of the new peak increased to +18 relative to the substrate (Figure S1). The observed shift confirmed that the associated compound resulted from hydroxylation of AcAP.

To demonstrate that the hydroxylated compound is an authentic intermediate on the pathway to the NANL product, we analyzed the products of LolO reactions performed with varying quantities of 2OG (Figure B). At 2OG:LolO·FeII ratios of <1, the intensity of the Δ(m/z) = +16 peak increased with increasing [2OG], consistent with the strict dependence of the hydroxylation reaction on the co-substrate. By contrast, once the 2OG:LolO·FeII ratio was increased further in excess of 1, the intensity of the Δ(m/z) = +16 peak decreased with increasing [2OG], becoming barely detectable by a 2OG:LolO·FeII ratio of 4. The diminished yield at high 2OG:LolO·FeII ratios is indicative of further processing of the hydroxylated compound in a second 2OG-dependent step. Importantly, in this phase of the titration, the Δ(m/z) = +14 peak, corresponding to NANL, increased with increasing [2OG], showing that the hydroxylated compound was converted to NANL. The overall conversion catalyzed by LolO can thus be depicted as in Scheme . As noted above, similar hydroxylation–oxacyclization sequences have been reported for the related Fe/2OG oxygenases, H6H and CAS.

Scheme 2

Overall Transformation Catalyzed by LolO

The observation that far more than half of the synthetic, racemic AcAP substrate is consumed at the highest 2OG:LolO·FeII ratios interrogated in the experiments depicted in Figure is surprising, given that most enzymes process only one enantiomer of a chiral substrate. The ability of LolO to process the unnatural AcAP enantiomer (albeit with diminished efficiency) hints at an unusual plasticity of its active site. However, this enantiomer is unlikely to be converted to the enantiomer of NANL, as the intensity of the +14 peak did not increase concomitantly with consumption of the second half of the substrate.

Identification of the Hydroxylated Intermediate as 2-endo-Hydroxy-1-exo-acetamidopyrrolizidine (2-endo-OH-AcAP)

To define the structure of the hydroxylated intermediate, we performed the LolO-mediated oxidation of AcAP on a large scale under conditions maximizing the intensity of the Δ(m/z) = +16 peak, and we purified the associated compound by high-performance liquid chromatography (HPLC). Upon isolation, the compound contained ammonium acetate (from the mobile phase) and at least one unidentified minor contaminant, but it was sufficiently pure to permit identification from its 1H nuclear magnetic resonance (NMR), 1H–1H COSY, and nuclear Overhauser effect (NOE) spectra (Figures S2–S5). The 1H NMR spectrum exhibited two doublets of doublets at δ 3.25 (J = 12.6 and 6.6 Hz) and 3.90 (J = 12.7 and 5.7 Hz), which can be attributed to the geminal pair of H atoms at C3 (Figure ); in the COSY spectrum, these two resonances correlated only to one another and to the most downfield resonance, a quartet at δ 4.43, assigned to the H atom attached to C2. In the COSY spectrum, the quartet at δ 4.43 also correlated to a triplet at δ 4.16, assigned to the H atom at C1. These data are consistent with the presence of the hydroxyl group on C2 but not with its presence on C7. Furthermore, irradiation of the resonance attributed to the H atom attached to C2 (quartet at δ 4.43) caused strong NOE enhancements of two resonances, one assigned to the hydrogen attached to C8 (quartet at δ 4.08) and one attributed to one of the two hydrogens attached to C3 (dd at δ 3.90), but only a very small NOE enhancement of the resonance attributed to the hydrogen at C1 (triplet at δ 4.16). These data are uniquely consistent with installation of the C2 hydroxyl group with endo stereochemistry, thus identifying the intermediate as 2-endo-OH-AcAP.

Figure 2

Selected 1H–1H COSY and NOE correlations and assignment of the structure of 2-endo-OH-AcAP.

Order of C–H Cleavage Steps Determined by Transient Kinetic/Spectroscopic Experiments with Deuterium-Labeled Substrates

Previous studies of Fe/2OG hydroxylases, halogenases, and the epimerase/desaturase, CarC, demonstrated accumulation of the C–H bond-cleaving ferryl complexes and substrate deuterium kinetic isotope effects (2H KIE) of 20–60 on their decay by H• abstraction.[14,31,32] To verify that C2, rather than C7, is hydroxylated in the first step of ether-bridge installation by LolO, we synthesized site-specifically deuterium-labeled AcAPs and tested the prediction that the reaction with the C2-labeled substrate would exhibit a large 2H-KIE upon decay of the ferryl intermediate, but the reaction of the C7-labeled substrate would not.

Synthesis of (±)-7,7-[2H2]AcAP

Installation of deuterium atoms at C7 of AcAP required 3,3-[2H2]Pro. The literature route to it was unsatisfactory,[36] and so we prepared it in racemic form by reducing 3-keto proline derivative 1(37) with NaBD4, tosylating the resulting alcohol, reducing the resulting tosylate with NaBD4, and deprotecting with 6 M HCl (Scheme ). Using procedures that we previously reported,[9] we then esterified (±)-3,3-[2H2]Pro and allowed the resulting ester to undergo conjugate addition to ethyl acrylate to give (±)-3,3-[2H2]2. We then subjected this diester to Dieckmann condensation, hydrolysis, decarboxylation, and NH2OH·HCl condensation to give oxime (±)-7,7-[2H2]3. Previously, we had reduced (±)-3,3-[2H2]3 with Raney nickel and then trideuteroacetylated the nascent amine in an attempt to make (±)-2′,2′,2′,3,3-[2H5]AcAP.[7] This step gave a modest yield and also caused a considerable loss of deuterium, resulting in a product with either one (85%) or no (13%) deuterium atoms. Loss of deuterium was even more pronounced in the attempted reduction of (±)-7,7-[2H2]3. After much experimentation with alternative reducing agents, which either gave unacceptably poor yields or favored the undesired diastereomer, we followed a precedent from ref (38) by treating (±)-7,7-[2H2]3 with nickel boride generated in situ from NiCl2 and NaBH4. Subsequent acetylation gave (±)-7,7-[2H2]AcAP in a yield (13%) that was better than that of the Raney Ni reaction and without a detectable loss of deuterium [97% d2 by high-resolution mass spectrometry (HRMS)].

Scheme 3

Synthesis of (±)-7,7-[2H2]AcAP

Synthesis of (±)-2,2,8-[2H3]AcAP

To install deuterium atoms at C2, we subjected diester 2(11) to Dieckmann condensation and then used DCl in D2O to effect hydrolysis, decarboxylation, and deuterium incorporation (Scheme ). Formation of oxime 2,2,8-[2H3]3 and nickel boride reduction proceeded without any loss of deuterium, and subsequent acetylation gave (±)-2,2,8-[2H3]AcAP (88% d3 by HRMS). We did not view the incorporation of the additional deuterium at C8 as a detriment, because our prior demonstration of the retention of this hydron in the biosynthetic conversion of AcAP to NANL[6] implied that it would not impact kinetic analysis of the LolO reaction. Moreover, the C8–H/D exchange demonstrated here validates our suppositions in prior studies that C2 of the proline starting material was racemized and the AcAP derived therefrom was thus racemic.

Scheme 4

Synthesis of (±)-2,2,8-[2H3]AcAP

Kinetic and Spectroscopic Evidence for C2 Hydroxylation by a Ferryl Intermediate

The conserved experimental hallmarks of the ferryl complexes in Fe/2OG oxygenases are their transient ultraviolet absorption at ∼320 nm and Mössbauer quadrupole doublet spectra with isomer shift parameters, δ, of 0.2–0.3 mm/s and quadrupole splitting parameters, ΔEQ, of approximately −1 mm/s.[14,32−35] With the required deuterium-labeled substrates in hand, we tested for both accumulation of the ferryl complex and a 2H-KIE on its decay by stopped-flow absorption and freeze-quench Mössbauer spectroscopies, as we have previously done for other Fe/2OG oxygenases.[29,34,35,39,40] Mixing an O2-free solution containing LolO, FeII, 2OG, and (±)-AcAP with air-saturated buffer at 5 °C resulted in transient absorption at 320 nm (Figure A, red trace), and the presence of the characteristic quadrupole doublet features in the 4.2 K, 53 mT Mössbauer spectra of samples freeze-quenched at short reaction times (Figure B, spectra i–iii) confirmed that the transient ultraviolet absorption arose from the ferryl complex. The kinetics of the intermediate were hardly altered when we used 7,7-[2H2]AcAP in place of the unlabeled substrate (in Figure A, compare red and blue traces), but the use of 2,2,8-[2H3]AcAP resulted in enhanced accumulation and slower decay of the transient ultraviolet (UV) absorption (Figure A, green trace) and Mössbauer quadrupole doublet features associated with the ferryl complex (Figure B, spectra iv and v). The marked stabilization of the ferryl complex by deuterium at C2, and the absence of a significant effect of deuterium at C7, confirm that the first step in ether-bridge installation by LolO is transfer of H• from C2 of AcAP to initiate hydroxylation of that position.

Figure 3

Kinetic and spectroscopic evidence for abstraction of hydrogen from C2 of AcAP by a ferryl intermediate. (A) Change in absorbance at 320 nm following mixing at 5 °C of an anoxic solution containing 1.1 mM LolO, 0.80 mM FeII, 5.0 mM 2OG, and the indicated (±)-AcAP isotopolog (4.0 mM) in 50 mM sodium HEPES (pH 7.5) with an equal volume of air-saturated buffer (∼0.19 mM O2 after mixing). Kinetic constants are listed in Table . (B) The 4.2 K, 53 mT Mössbauer spectra of samples freeze-quenched at the indicated reaction times after mixing of an anoxic solution containing 2.3 mM LolO, 1.8 mM 57FeII, 12 mM 2OG, and the indicated (±)-AcAP isotopolog (3.6 mM) in 50 mM sodium HEPES (pH 7.5) with an equal volume of O2-saturated buffer (∼0.90 mM O2 after mixing). Black vertical bars represent experimental data. The red lines are simulations of the quadrupole doublet of ferryl species with the parameters of isomer shift (δ = 0.28 mm/s) and quadrupole splitting parameter (|ΔEQ| = 0.89 mm/s) accounting for 16, 3, 28, and 15% of total iron in spectra ii–v, respectively.

Table 1

Observed Kinetic Constants for Formation and Decay of the Ferryl Intermediate during Hydroxylation of C2 of the (±)-AcAP Isotopologs by LolO

substrate	kformation (105 M–1 s–1)	kdecay (s–1)	kH/kD
AcAP	1.3 ± 0.2	48 ± 4	–
7,7-[2H2]AcAP	1.3 ± 0.2	47 ± 5	1.0 ± 0.2
2,2,8-[2H3]AcAP	1.3 ± 0.2	2.0 ± 0.1	23 ± 5

In the freeze-quench Mössbauer experiment with 2,2,8-[2H3]AcAP, the ferryl complex was observed to accumulate to a maximum of only ∼29% of the total iron in the sample, less than half the quantity predicted on the basis of its relatively fast formation and much slower decay (Table ). This sub-stoichiometric accumulation arose from the use of slightly less than saturating substrate in that experiment [1 equiv of the correct enantiomer, giving ∼92% saturation (see Figure S6)] and, more importantly, from the fact that only 0.4 ± 0.1 of the LolO·FeII·2OG·AcAP complex is competent to react rapidly with O2 to generate the intermediate. Such fractional reactivity has been encountered for every other Fe/2OG enzyme that has been subjected to this level of transient-state kinetic analysis.[29,31,34,41]

By global simulation of the kinetic data for the reactions with the three substrate isotopologs under varying experimental conditions (e.g., limiting O2, limiting 2OG, etc.), we obtained values for the rate constants for formation and decay of the ferryl complex and observed 2H-KIE on the C2–H cleavage step (Table ). The intrinsic 2H-KIE for this step may be greater than the observed value of 23 ± 5, if pathways not involving C2–H cleavage contribute significantly to ferryl decay when the intermediate is challenged by the heavy hydron at C2. With the parameters listed in Table , both the quantities of the ferryl complex determined in the freeze-quench Mössbauer experiments and the kinetic traces from stopped-flow experiments performed under identical reaction conditions could be accounted for (Figure S7), assuming a reactive fraction of 0.4 and a molar absorptivity of the ferryl complex of 2000 M–1 cm–1. The latter value is similar to (albeit somewhat larger than) those determined previously for ferryl complexes in other Fe/2OG oxygenases.

Determination of the Stereochemical Courses of the Steps in Ether-Bridge Installation by LolO

In formation of the ether bridge, C–H bonds to C2 and C7 of AcAP are cleaved and a new C–O bond to each carbon is formed. The structure of the NANL product implies that the ether oxygen is added to AcAP at the positions initially occupied by the endo hydrogens, but the free-radical nature of this chemistry makes either of two cases possible for each step: the endo hydrogen could be abstracted and the C–O bond formed with retention, or the exo hydrogen could be removed and the C–O bond formed with inversion. Although the oxygen-rebound mechanism for hydroxylation makes retention of configuration at C2 far more likely than inversion, the situation is less clear for C7, because the mechanism of ring closure is less well understood. The simplest possibility would involve coordination of the newly installed C2 oxygen to the FeII cofactor in the reactant complex for the oxacyclization step, followed later by radicaloid coupling of the ferryl-generated substrate radical with the alkoxo, rather than the hydroxo, ligand. This mechanism would be analogous to that proposed for thiazolidine ring closure in the reaction of isopenicillin N-synthase (IPNS).[42,43] However, it is unclear for the case of LolO how the cofactor, which is expected to have one carboxylate and two histidine ligands from the protein as well as a carboxylate and carbonyl oxygen from 2OG (Figure S8), could coordinate the C2–O bond and still activate O2. A second possibility would involve formation of a carbocation by the transfer of an electron from the carbon radical to the Fe(III)–OH cofactor state and subsequent polar ring closure by attack of the substrate oxygen. Indeed, this mechanism was recently proposed for closure of the epoxide ring in the antibiotic fosfomycin by (S)-2-hydroxypropylphosphonate epoxidase (HppE).[44,45] Among the evidence cited for this mechanism was the inversion of configuration at C1 during ring closure. This example shows that the stereochemistry of H• abstraction and C–O coupling is a crucial mechanistic issue in the oxacyclization reactions. To determine the stereochemical courses of both steps in installation of the ether bridge by LolO, we fed stereoselectively deuterium-labeled precursors, Pro and Asp, to loline-producing cultures and tested for the presence of deuterium in the products.

Stereochemistry of C7–H Cleavage

Previous work showed that C7 of N-formylloline (NFL) is derived from C3 of Pro (Scheme , left side).[8] With the reasonable assumption that the C3 configuration of Pro should remain unchanged until the C–H abstraction event, we expected that feeding trans-3-[2H]Pro would lead to the intermediate 7-endo-[2H]AcAP, and cis-3-[2H]Pro to 7-exo-[2H]AcAP. Therefore, if LolO removed the endo hydrogen from C7 of AcAP (blue highlight), then trans-3-[2H]Pro would give NFL lacking deuterium at C7, and cis-3-[2H]Pro would give 7-[2H]NFL (blue arrows). By contrast, if LolO abstracted the exo H atom at C7 of AcAP (green highlight), then trans-3-[2H]Pro would give 7-[2H]NFL, and cis-3-[2H]Pro would give NFL lacking deuterium (green dashed arrows).

Scheme 5

Predicted Hydrogen Isotopic Composition of NFL Obtained from Cultures Fed Stereoselectively Deuterium-Labeled Precursors, Pro and Asp

Green and blue spheres depict the relevant hydrons and the positions that they come to occupy in AcAP. The solid blue arrows depict the results of abstraction of the endo hydrons from C2 and C7 of AcAP by LolO, and the dashed green arrows depict the results of removal of the exo hydrons.

Syntheses of cis- and trans-3-[2H]Pro

We began the syntheses of regio- and stereoselectively deuteriated prolines[36] by protecting commercially available trans-3-hydroxy-l-proline with Boc and tert-butyl groups to give 4 (Scheme ).[11] We divided 4 into two portions, allowing one portion to undergo a Mitsunobu reaction to give 5. We then subjected both 5 and the remaining 4 to tosylation, reduction with NaBD4, and deprotection with 6 N HCl to give cis- and trans-3-[2H]Pro·HCl, respectively.

Scheme 6

Syntheses of cis- and trans-3-[2H]Pro·HCl

Culture Feeding and Analysis of Products

We fed loline-producing cultures of Epichloë uncinata e167 with the two Pro isotopologs and measured the deuterium content of the resultant NFL by GC–MS (Figure ). The spectrum of NFL lacking deuterium had a prominent m/z = 183 peak, previously assigned to the [M + H] ion,[3,8,46] and a small m/z = 184 peak from species with one atom of a heavier isotope of C, N, or O (Figure A). The spectrum of NFL from the culture with cis-3-[2H]Pro displayed a significant enhancement of the 184 (+1) peak (30.5 ± 2.3%) (Figure B), consistent with the incorporation of 2H into much of the NFL, whereas the spectrum of NFL from the culture with trans-3-[2H]Pro had a barely detectable peak (2.9 ± 0.3%) at this m/z value (Figure C). The small but significant enrichment of 2H from the feeding of trans-3-[2H]Pro likely arose from contamination of the primary isomer with a small amount of cis-3-[2H]Pro, as shown in the 1H NMR spectrum of the synthetic material (Figure S9). The incorporation of one deuterium atom into NFL produced from cis-3-[2H]Pro implies that LolO abstracts the endo hydrogen from C7 of AcAP (Scheme ).

Figure 4

Mass spectra of NFL from loline-producing cultures fed with (A) natural-abundance Pro, (B) cis-3-[2H]Pro, and (C) trans-3-[2H]Pro.

Stereochemistry of C2–H Cleavage

Previous work showed that C2 of NFL is derived from C3 of Asp (Scheme , right side) via homoserine.[8] As analyzed in detail in the Supporting Information, the sequence of biosynthetic steps converting Asp to AcAP (in particular the step attributed to the PLP-dependent enzyme, LolC) is expected to exchange the C2 hydron and one of two diastereotopic C3 hydrons with solvent (Scheme S1) but not to exchange the other C3 hydron nor to invert or racemize C3. We thus anticipated that (1) in principle, (3S)-3-[2H]Asp would yield 2-endo-[2H]AcAP and (3R)-3-[2H]Asp 2-exo-[2H]AcAP but (2) in actuality, one of the two diastereomers would yield AcAP lacking deuterium (Scheme S2). Therefore, if LolO were to remove the endo hydron from C2 of AcAP (blue highlight), then feeding cultures with (3S)-3-[2H]Asp would necessarily yield NFL lacking deuterium, whereas feeding with (3R)-3-[2H]Asp could yield 2-[2H]NFL (blue arrows), provided that the other biosynthetic enzymes did not cause exchange of the deuteron with solvent during production of AcAP. Conversely, were LolO to abstract the exo hydron (green highlight), only for (3S)-3-[2H]Asp could deuterium possibly be retained in the NFL product of the LolO reaction (green dashed arrows).

We prepared (3S)-2,3-[2H2]Asp (89% d2) and (3R)-3-[2H]Asp (95% d1) chemoenzymatically by literature procedures, except that we used aspartase B[47] instead of aspartase A[48,49] or 3-methylaspartase[50] (Scheme ). As noted in the analysis in the Supporting Information, we expected that the C2 hydron of the former compound would be lost to solvent in conversion to AcAP, thus obviating complications from the extraneous deuteron. The two Asp isotopologs were applied to loline-producing cultures, and the isotopic composition of the resultant NFL was determined by GC–MS (Table ). In the feeding experiments with (3R)-3-[2H]Asp, the NFL product was significantly enriched in a species with m/z +1 relative to the product from the natural-abundance control, implying the retention of a single deuteron in a significant fraction of the product. By contrast, experiments with (3S)-2,3-[2H2]Asp gave no more [2H]NFL (nor [2H2]NFL) than did the control. These results are most consistent with abstraction of the endo hydrogen from C2 of AcAP (Table S1). In addition, the absence of any residual deuterium in the NFL formed from the (3S)-2,3-[2H2]Asp precursor confirms the expected exchange of the C2 hydron during the steps leading to AcAP (Scheme S1). Hence, as for the oxidation of C7, LolO also abstracts the endo hydrogen from C2 and forms the C2–O bond with retention.

Scheme 7

Syntheses of (3R)-3-[2H]- and (3S)-2,3-[2H2]Asp

Table 2

Analysis of Isotopic Compositions of the NFL Products from Cultures Fed with Stereospecifically Deuterium-Labeled Aspartic Acids

Asp isotopolog	amu shift	% enrichment of NFL (mean ± SD)	p value vs control
(3R)-3-[2H]	+1	4.1 ± 0.9	0.0029
(3S)-2,3-[2H2]	+1	1.1 ± 0.4	0.9137
(3S)-2,3-[2H2]	+2	0.2 ± 0.3	0.8139

Conclusions

In summary, we have shown that LolO, an Fe/2OG oxygenase, is solely responsible for introducing both C–O bonds of the ether bridge of the loline alkaloids. LolO employs a ferryl complex to abstract the endo H atom from C2 of AcAP, adds HO• to the nascent radical with retention to give 2-endo-OH-AcAP, abstracts the endo H atom from C7 (presumably with a second ferryl complex), and then couples the C2–O atom to the C7 radical with retention, giving NANL. These findings imply that the bicyclic AcAP substrate binds within LolO so that it effectively cradles the iron cofactor in its concave face, directing both endo hydrogens for abstraction from sequentially formed ferryl intermediates. Whether the geometries of the ferryl intermediates are identical in the two steps and precisely how the fate of the substrate radical is switched from coupling with the FeIII–hydroxo ligand in the first step to coupling with the C2 alkoxyl group in the second step remain to be established by future structural and mechanistic studies.

Materials and Methods

Overexpression of LolO in E. coli

A gene encoding the same amino acid sequence as the lolO2 gene from Ep. uncinata e167 (GenBank accession no. AY723750.1)[10] but codon-optimized for expression in Escherichia coli was synthesized and inserted into pET28a between the NdeI and BamHI restriction sites by GeneArt (Regensburg, Germany). This construction puts lolO2 under control of the T7 promoter and appends an N-terminal His6 metal ion affinity tag to the protein. The plasmid was used to transform BL21 (DE3) cells (Invitrogen, Carlsbad, CA) for protein expression. The cells were grown at 37 °C in Luria-Bertani (LB) medium supplemented with 50 μg/mL kanamycin until an OD600 of ∼0.7 was reached and then cooled to 16 °C. Expression was induced by addition of isopropyl β-d-1-thiogalactopyranoside to a final concentration of 0.5 mM, and cells were shaken overnight at 16 °C before being harvested by centrifugation.

Purification of Recombinant LolO

All purification steps were performed at 4 °C. The cell paste was resuspended in 3 volumes of lysis buffer containing 100 mM Tris-HCl (pH 7.5) with 5 mM imidazole and 1 mM phenylmethanesulfonyl fluoride. The cells were lysed by being passed through a microfluidizer (M110EH-30, Microfluidics, Newton, MA) at ∼20,000 psi for 4 min (approximately two cycles of bursts). The cell debris was then removed by centrifugation, and the supernatant was loaded onto a Ni-NTA agarose column (approximately one-third the volume of the cleared lysate). The column was washed twice with 3 column volumes of the same buffer, and the protein was then eluted by application of 100 mM Tris-HCl (pH 7.5) containing 250 mM imidazole. Samples from each step were analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) visualized by Coomassie blue staining. Eluted fractions containing the target protein were pooled and concentrated with a centrifugal concentrator with a 10 kDa molecular weight cutoff filter (Pall Corp., Port Washington, NY). The concentrated protein was then dialyzed against 50 mM sodium HEPES buffer (pH 7.5) containing 5 mM EDTA to remove iron. Two additional rounds of dialysis with buffer lacking EDTA removed the metal chelator. The enzyme, assessed by SDS–PAGE with Coomassie blue staining to be >95% pure, was quantified by UV absorption at 280 nm by assuming a molar absorptivity (ε280) of 76,235 M–1 cm–1 (http://ca.expasy.org).

LC–MS Analysis of LolO Reactions

Constitution of the enzyme reactions is described in the appropriate figure legends. Each reaction solution was passed through a Nanosep centrifugal filter (0.2 μm, Pall Corp.) before being injected onto the LC–MS instrument (Agilent 1200 series LC system coupled to an Agilent 6410 QQQ mass spectrometer). A 2 μL aliquot of each sample was injected, and compounds were examined by electrospray ionization MS in positive mode. For chromatography, a Kintex 2.6 μm HILIC (150 mm × 2.1 mm) column (Phenomenex, Inc., Torrance, CA) was used at a flow rate of 0.3 mL/min with 50% acetonitrile in water as mobile phase A and 95% acetonitrile as mobile phase B. Both solvents were supplemented with 10 mM NH4CH3CO2 (pH 5.8). The elution program was: 100% B for 2.5 min, from 100 to 50% B from 2.5 to 12 min, 50% B to 22 min, from 50 to 100% B from 22 to 30 min, and 100% B for 10 min (to prepare for the next injection). Standards of AcAP and NANL for the LC–MS assay were extracted from seeds of Elymus canadensis infected with Epichloë canadensis e4815.[6]

Isolation of 2-endo-OH-AcAP

An anoxic solution containing 1.5 mM LolO, 1.5 mM FeII [from Fe(NH4)2(SO4)2], 3.0 mM (±)-AcAP, and 1.2 mM 2OG (conditions determined empirically to accumulate the greatest quantity of the hydroxylated intermediate) in 50 mM sodium HEPES buffer (pH 7.5) was mixed with an equal volume of cold, O2-saturated buffer. The reaction was allowed to proceed for 5 min at 4 °C, and the mixture was then passed through a centrifugal concentrator with a 10 kDa molecular weight cutoff filter. The flow-through was pooled and lyophilized. This sample was dissolved in a small volume of water and fractionated by preparative HPLC using a Kintex 5 μm HILIC (150 mm × 21.2 mm) column (Phenomenex, Inc.) at a flow rate of 40 mL/min with the same mobile phase and elution method as described above. Fractions were collected every 0.2 min, and their composition was determined by LC–MS with a reverse-phase Agilent Extend-C18 column (4.6 mm × 50 mm, 1.8 μm particle size). The C18 column was eluted isocratically with 0.1% formic acid in water at a flow rate of 0.3 mL/min. Compounds were examined by electrospray ionization MS in positive mode. The fractions with >95% OH-AcAP were pooled and dried by N2 gas, yielding a transparent solid that was used for HRMS and NMR analyses.

Stopped-Flow Absorption and Freeze-Quench Mössbauer Experiments

These experiments were performed essentially as previously described.[34] Stopped-flow experiments were performed at 5 °C with an Applied Photophysics Ltd. (Leatherhead, U.K.) SX200 stopped-flow spectrophotometer installed in an anoxic chamber (MBraun, Peabody, MA). The instrument was configured for single mixing with an optical path length of 1 cm. The ferryl absorption feature was acquired by using monochromatic 320 nm light and a photomultiplier tube (PMT) detector. The ΔA320-versus-time traces were simulated according to a kinetic model with two irreversible steps by using the KinTek Explorer software package (KinTek Corp., Snow Shoe, PA). Fits to the data for the different substrate isotopologs were optimal when the difference in molar absorptivity between the reactant (or product) complex and the ferryl intermediate was taken to be 2000 M–1 cm–1. The error estimates for the kinetic constants were obtained by comparing results of two independent experiments, each with at least three replicates.

To prepare the FQ samples, an O2-free solution containing 2.3 mM LolO, 1.8 mM 57FeII, 12 mM 2OG, and 3.6 mM (±)-AcAP or 2,2,8-[2H3]AcAP in 50 mM sodium HEPES buffer (pH 7.5) with 20% glycerol was rapidly mixed with an equal volume of the same buffer saturated with O2 (1.8 mM O2). After being passed through an aging hose of a length appropriate to give the desired reaction time, the solution was rapidly frozen by being injected into cold (−150 °C) 2-methylbutane cryosolvent. Packed samples were stored in liquid N2 until they were analyzed. Mössbauer spectra were measured on a spectrometer from WEB Research (Edina, MN) equipped with a SVT-400 cryostat (Janis, Wilmington, MA) to maintain the temperature at 4.2 K. The isomer shifts quoted are relative to the centroid of the spectrum of a metallic foil of α-Fe at room temperature. Simulations of the Mössbauer spectra were performed with the WMOSS spectral analysis software package (www.wmoss.org, WEB Research).