Mechanistic analysis of the stereodivergent nitroalkane cyclopropanation catalyzed by nonheme iron enzymes

Abstract

BelL and HrmJ are α-ketoglutarate dependent nonheme iron enzymes that catalyze the oxidative cyclization of 6-nitronorleucine, resulting in the formation of two diastereomeric 3-(2-nitrocyclopropyl)alanine (Ncpa) products containing trans-cyclopropane rings with the (1′R,2′R) and (1′S,2′S) configurations, respectively. Herein, the catalytic mechanism as well as the stereodivergency of the cyclopropanses are investigated. Results suggest that the nitroalkane moiety of the substrate is first deprotonated to form the nitronate form. Spectroscopic analysis and biochemical assays with substrate and analogues suggest that an iron(IV)-oxo species abstracts the proS-H from C4 to initiate intramolecular C-C bond formation. The cyclopropanation reaction is unlikely involved a hydroxylation intermediate. Additionally, a genome mining approach is employed to discover new homologues of cyclopropanases that perform cyclopropanation of 6-nitronorleucine to generate cis-configured Ncpa products with the (1′R,2′S) or (1′S,2′R) stereochemistries. Sequence and structure comparisons of these cyclopropanases enable us to determine amino acid residues critical in controlling the stereoselectivity of cyclopropanation.

Introduction

Cyclopropanes are found in various natural products wherein the biological activities of these natural products, in some cases, arise from the reactions of the highly strained three-membered rings with biological molecules such as DNA, leading to the modulation of several biological processes.^1–3 Additionally, the molecular framework of the cyclopropane ring retains conformational rigidity, thus making it a useful pharmacophore with a specific stereo-orientation of substituents.⁴ Consequently, the stereochemistry of the cyclopropane rings is a critical determinant of biological functions. Despite their biological activities and abundance in natural products, the biosynthesis of cyclopropane-containing natural products is still poorly understood.

Belactosins and hormaomycins are peptide natural products containing (1′R,2′S)-3-(2-aminocyclopropyl)alanine (Acpa, 2) and (1′R,2′R)-3-(2-nitrocyclopropyl)alanine (Ncpa, 1) residues, respectively.^5–9 Previous research has shown that Ncpa (1) is biosynthesized from lysine via N-oxygenation catalyzed by the heme oxygenase-like enzymes BelK/HrmI, followed by cyclopropanation catalyzed by the α-ketoglutarate (αKG)-dependent nonheme enzymes BelL/HrmJ (Scheme 1).^10–12 The second biosynthetic step is particularly noteworthy since BelL and HrmJ dehydrogenate 6-nitronorleucin (3) to produce (1′S,2′S)-Ncpa ((1′S,2′S)-1) and (1′R,2′R)-Ncpa (1), respectively, with complete stereoselectivity. However, the plausible mechanism by which these enzymes catalyze the intramolecular C-C formation and control the stereoselectivity of the product remain unknown.

Scheme 1.

Stereodivergent cyclopropanation reactions catalyzed by BelL and HrmJ during biosynthesis of hormaomycin and belactosin A.

Results and discussion

Possible reaction mechanisms analysis

The cyclopropanation activities of BelL and HrmJ are unusual because αKG-dependent nonheme (Fe/αKG) enzymes typically catalyze hydroxylation at an unactivated carbon center.^13–17 As members of this enzyme family, it is hypothesized that the reaction of Fe(II), O₂, and αKG initially generates a high-valent Fe(IV)-oxo complex^18–21 that initiate the reaction through H atom abstraction. In addition to 3, because of the nitro group substitution, C6 of 3 may be deprotonated, in which 4 may serve as the substrate. Several possible pathways for the cyclization of 5 are shown in Scheme 2A. Following H atom abstraction at C4, The radical may intramolecularly add across the C–N double bond of the nitronate moiety of 5 to form 6, which has been suggested to be chemically feasible based on DFT studies (path a). The cyclopropane intermediate with the nitro radical anion (6) may be further oxidized by the ferric iron species to give 1. It is also possible that substate radical 5 may be hydroxylated via a hydroxyl rebound process (path b). The resulting hydroxylated intermediate 7 then undergoes cyclopropanation via a substitution reaction to yield 1. Alternatively, single electron transfer from 5 to Fe(III) may form a carbocation intermediate 8 (path c). The intermediate 8 is converted to 1 either via hydroxyl intermediate 7 or via direct ring closure via a polar mechanism.

Scheme 2.

Possible mechanisms for the BelL and HrmJ-catalyzed cyclopropanation of 6-nitronorleucine (3)

Analysis of C6-deprotonation of the substrate

To investigate whether C6-deprotonation of 3 is an enzymatic process, we incubated 3 (1 mM) with BelL (140 μM) in D₂O buffer (>99% D, HEPES pH 7.5, 40 mM) and monitored the incorporation of the deuterium atom into the C6 position of 3 by ¹H NMR. The residual proton at C6 was determined by integrating the resonances of H-6 and H-3. In the absence of BelL, the initial rate of the D incorporation was 10 μM·min^–1, indicating that C6 of 3 can be deprotonated nonenzymatically. However, addition of BelL in the incubation mixture increased the rate of the deuterium incorporation to 21 μM·min^–1. These results suggested that BelL accelerates the rate of C6 deprotonation and implies that 4 is likely involved in the Fe/αKG enzyme catalyzed cyclopropanation.

To determine the protonation state of the substrate in the active site, we chemically synthesized 6,6-difuluoro-6-nitronorleucine (9) and tested it with BelL (Scheme 2B). Since the corresponding nitronate could not be formed, observation of hydroxylation or other outcomes of 9 with BelL would support the hypothesis that H atom abstraction by Fe(IV)-oxo species could occur prior to the deprotonation of C6. However, 9 was not accepted by BelL, implying that the substrate is likely bound to the active site in its deprotonation nitronate form, i.e., 4, and the development of a negative charge is crucial for substrate recognition. To further validate this hypothesis, 2-aminopimeric acid (10) was incubated with BelL (Scheme 2C). LCMS analysis of the reaction mixture after dansyl chloride derivatization indicated that the reaction generated an oxidized product with a mass consistent with oxygenation (+16, m/z 424.13). NMR analysis revealed that this oxidized product was Dns-2-amino-4-hydroxy pimelic acid (11) while the stereochemistry at C4 could not be determined. Since the carboxylic acid moiety of 10 is deprotonated under reaction conditions, the acceptance of 10 by BelL implies that C6 of the substate (3) is likely deprotonated at the enzyme active site. The result also highlights the importance of the nitro group for the cyclopropanation activity.

If 4 serves as the substrate, formation of 4 in solution which avoids C6 deprotonation might increase the reaction efficiency. To test this hypothesis, we tuned the distribution of 3 and 4 by adjusting the pH and observed the change using ¹³C-NMR. As shown in Figure 1a, through an increase of pH, the [6-¹³C]-3 shifts from 72.6 ppm to 116.1 ppm which is consistent with deprotonation of C6 of 3 (δ = 72.6 ppm)which leads to formation of 4 with an sp2 hybridization at C6 (δ = 116.1 ppm). At pH of 10.5, [6-¹³C]-4 is observed as the dominant species (> 95%). We then carried out enzymatic reactions at different pH conditions. Enzymatic reactions (0.22 mM enxyme (HrmJ, BelL, SscBelL or SrBelL), 0.20mM Fe(II), 1.0 mM of 3, and 2.0 mM of αKG) at different pH conditions were prepared (see SI for details). After 1 hour incubation, the reactions were derivatized using 1-fluoro-2–4-dinitrophenyl-5-l-alanine amide (FDAA) and analyzed by LCMS. As shown in Figure 1b, ~ 10-fold increase of product was detected when changing pH from 7.0 to 10.5. At a higher pH (pH 11), a slight decrease of reactivity is observed which likely associates with the protein instability. In HrmJ, BelL and SrBelL, a similar increase of product formation was also detected at pH 10.5. Herein, an increase of pH to facilitate the cyclopropanation is therefore established. A correlation between nitronate ratio and the product efficiency suggests that 4 is likely included in the reaction pathway.

Figure 1.

(a) Distribution of 3 and 4 detected by ¹³C-NMR using [6-¹³C]-3 as the substrate and (b) SscBelL catalyzed reaction efficiency at various pH conditions.

Detection of an Fe(IV)-oxo species

To reveal reaction mechanism of cyclopropanation, we first identified that an Fe(IV)-oxo species serves as the key intermediate to effect C4-H activation using transient kinetics (via stopped-flow optical absorption (SF-Abs) spectroscopy) and Mössbauer spectroscopy. As shown in Figure 2, rapid mixing of the SscBelL•Fe(II)•αKG•3 quaternary complex with O₂ saturated buffer results in a rapid depletion of an optical feature centered ~ 520 nm, which is the typical metal-to-ligand-charge-transfer (MLCT) band of Fe(II)-αKG moiety in the ferrous quaternary complex of Fe/αKG enzymes. Its rapid depletion suggests the facile reaction between the ferrous quaternary complex with O₂. Concomitantly, a rapid formation of another optical feature in the near UV region (~300 – 350 nm) is also observed, which accumulated to a maximum at ~ 0.1 s and completely decayed after ~ 1 s. In the presence of [4,4-²H₂]-3, this optical feature exhibits a similar formation kinetics, but with a much higher accumulation level and a slower decay rate. The substantial decay is observed after ~ 30 s. The species is likely originated from a ferryl intermediate as demonstrated in several Fe/αKG enzymes. The difference of the kinetics using 3 and [4,4-²H₂]-3 most likely reflects the large H/D kinetic isotope effect (KIE). To validate the SF-Abs results, we carried out freeze-quench coupled Mössbauer analysis. When 3 is used, a Mössbauer species with parameters typical of a ferryl intermediate (isomer shift of 0.27 mm/s and quadrupole splitting of 0.80 mm/s) is observed, wherein it accumulates to a maximum (~20% of the total iron in the sample) at 0.1 s. It is fully decayed after 1 s. Differently, this intermediate accumulates to a larger extent (~ 60%) and is fully decayed after 30 s in the presence of [4,4-²H₂]-3. The kinetics of the ferryl intermediate observed in Mössbauer are fully consistent with the transient near-UV optical species observed in the SF-Abs experiments, revealing the reaction involves an Fe(IV)-oxo catalyzed C4-H bond activation. Together, these results provide direct evidence to support a mechanism that a ferryl species serves as the intermediate to trigger C4-H atom abstraction. Subsequently, the resulting radical or the subsequent C4-centered cation is captured by the double bond at nitronate of 3 to enable C-C bond formation.

Figure 2.

Spectroscopic evidence of C4-H activation by a ferryl intermediate during SscBelL catalyzed cyclopropanation. Top: selected spectra demonstrating the optical feature changes upon rapid mixing of the SscBelL•Fe(II)•αKG•3 with O₂ saturated buffer. The decay and reformation of the optical feature centered at ~520 nm is accompanied by the formation and decay of a broad UV feature (ca. 300–400 nm). The inset shows the absorption change at 330 nm with 3 or [4,4-²H₂]-3. Bottom: Mössbauer spectra of samples generated by freeze-quench technique on the selected time points from the reaction with 3 (left) or with or [4,4-²H₂]-3 (right). The grey solid lines represent the overall spectral simulations. The red solid lines represent the spectral component of the ferryl intermediate with isomer shift (δ) of 0.27 mm/s and quadrupole splitting (|ΔE_Q|) of 0.80 mm/s. See Table S2 and S3 for detailed simulation parameters.

C4-hydroxylation is unlikely involved in the cyclopropanation

Following C4-H atom abstraction, three pathways laeding to cyclopropanation can be envisioned (Scheme 2A). To investigate the possibility of utilizing a hydroxylated compound as an intermediate to facilitate the C-C bond formation. We used GlbB, an Fe/αKG enzyme that catalyzes C4-hydroxylation of L-Lys to generate the proposed hydroxylation intermediate. Incubating [6-¹³C]-3 with GlbB resulted in a peak with a chemical shift of 69.6 ppm. The up-field shift is originated from introduction of a heteroatom, e.g. hydroxylation, at the β-position to the 6-¹³C. In the corresponding LCMS, a peak with an m/z of +16 was detected. Next, we removed the GlbB through filtration and added SscBelL or SrBelL to the reaction mixture. If 7 serves as an intermediate for 1 formation, we anticipate that 7 could form in situ via C6-deprotonation and leads to 1. After 12 hour incubation, we did not observe formation of 1 by LC-MS. While we cannot completely rule out the possibility that C4-hydroxylation compound cannot enter the active site, our results are consistent with C4-hydroxylation is unlikely involved in the formation of 1 (Pathway b, Scheme 2).

Identification of cis-Ncpa-producing cyclopropanases

BelL and HrmJ exhibit 49% identity to each other. However, we could not identify protein residues that are critical for the selective formation of (1′S,2′S)-1 and (1′R,2′R)-1 based solely on the comparison between the amino acid sequences of BelL and HrmJ. To understand the correlation between the amino acid sequence and the stereochemistry in 1, we first attempted to obtain BelL/HrmJ homologues. Using BelL as a query in the BLAST analysis of the NCBI protein database, we found ca. 170 proteins that share more than 40% identity with BelL encoded by bacterial strains across two phyla Actinomycetota and Proteobacteria. To the best of our knowledge, whether these candidates also catalyze cyclopropanation of 1 is unclear (see Supporting Information). Nevertheless, ca. 80 of these proteins are encoded by the belL-like genes that are colocalized with the belK-like gene. Thus, we initially hypothesized that the sets of belL- and belK-like genes may encode the biosynthesis of 1 with either (1′S,2′S)- or (1′R,2′R)-stereoconfiguration. Although compounds 1 and 2 (after the nitro reduction) are incorporated into peptides during biosynthesis of hormaomycins and belactosins, the newly found belL- and belK-like genes are not necessarily clustered with genes encoding non-ribosomal peptide synthetases (NRPSs) and ATP-grasp enzymes. Thus, 1 could be either an end natural product or modified into distinct types of natural products in the belL-like gene-containing bacteria.

To test whether these homologues catalyze oxidative cyclization of 3, we cloned ten belL-like gene homologues from readily available bacteria and heterologously expressed the corresponding gene products in Escherichia coli. The selected bacterial strains and the protein codes are listed in Table S1. The recombinant BelL homologues (14 μM) were incubated with 3 (0.4 mM) in the presence of FeSO₄ (0.4 mM), ascorbate (2 mM) and αKG (0.4 mM). The enzyme reaction mixture was derivatized by Dns-Cl to achieve a product separation in LCMS analysis. It was found that the proteins completely consumed 3 and generated at least one product with a mass consistent with 1-Dns (Figure 3a). Unexpectedly, the products from the tested BelL homologues eluted at retention times different from those of (1′S,2′S)-1-Dns or (1′R,2′R)-1-Dns, which are the Dns-derivatized products of BelL and HrmJ, respectively. The dansyl-derivatized products of LaBelL, MspBelL, SaBelL, ScBelL, and PsBelL were identical and characterized as (1′R,2′S)-1-Dns with a cis configuration of the cyclopropane ring based on NMR analysis and X-ray crystal structure analysis. Although we could not obtain the crystals of the Dns-derivatized product from SrBelL or MiBelL for X-ray crystal structure analysis, the NMR analysis and retention time in HPLC analysis suggested that it is a new diastereomer of 1-Dns. The stereochemistry of the product was assigned as the other cis isomer (1′S,2′R)-1-Dns because we had characterized the other three diastereomers of 1-Dns among the four possibilities. To further support LC-MS results, we carried out the ¹³C-NMR assays of HrmJ, BelL, SscBelL (homolog to ScBelL which share 99 % sequence identity and SrBelL catalyzed reactions using [6-¹³C]-3 as the substrate (Figure 3b). While all reaction products have C6 of chemical shifts ~ 56.4–56.8 ppm, they show distinctive chemical shifts. It was noted that TbBelL and RpBelL catalyze the cyclopropanation to give (1′S,2′S)-1 and (1′R,2′S)-1 in ratios of 3 : 2 and 1 : 1, respectively, under the employed conditions although other BelL homologues exhibited near complete (>95% d.r.) diastereoselectivity. Therefore, this group of cyclopropanases can generate all four possible diastereomer of cyclopropane products from 3 as a common substrate.

Figure 3.

(a) LCMS analysis of the cyclopropanation reactions catalyzed by BelL homologues. EIC traces corresponding to [M + H]⁺ signals from 1-Dns (m/z = 408.1) are shown. (b) ¹³C NMR spectra of HrmJ, BelL, SscBelL and SrBelL catalyzed reactions using [6-¹³C]-3 as substrate.

Stereochemical analysis of H abstraction

To better understand the stereodivergency of the cyclopropanses, we analyzed the stereochemical courses of the cyclization. It was previously demonstrated using (4S)-[4-²H]-3 as a probe that both HrmJ and BelL abstract proS-H at C4;¹⁰ however, whether the cis-cyclopropane forming homologues exhibit the same stereoselectivity of H abstraction is uncertain. When SrBelL was incubated with a mixture of (4S)-[4-²H]-3 and (4S,5R)-[4,5-²H₂]-3 (36 : 64), LCMS analysis revealed the formation of 35% unlabeled Dns-1 (m/z 408.1 [M + H]⁺) and 64% monodeuterated 1-Dns (m/z 409.1 [M + 1 + H]⁺) with a trace amount (<1%) of dideuterated 1-Dns consistent with the loss of 4-proS-H(D) during the cyclization. This indicates the retention of the stereoconfiguration at C4. Other eight homologues also showed similar results. Thus, BelL and homologues abstract the same hydrogen, 4-proS-H, of the substrate regardless of the stereoconfiguration of the cyclopropane products. The stereochemical course at C6 was also analyzed using (6R)-[6-²H]-lysine and (6S)-[2,6-²H₂]-lysine. The reaction of (6R)-[6-²H]-lysine with HrmI and SrBelL for 20 hours, followed by Dns-derivatization, generated a mixture of 35% unlabeled Dns-1 (m/z 408.1 [M + H]⁺) and 65% mono-deuterated Dns-1 (m/z 409.1 [M + 1 + H]⁺). However, a complementary experiment with (6S)-[2,6-²H₂]-lysine revealed the production of 36% mono-deuterated Dns-1 (m/z 409.1 [M + 1 + H]⁺) and 60% di-deuterated 1-Dns (m/z 410.1 [M + 2 + H]⁺ Thus, it is proposed that SrBelL-catalyzed dehydrogenation of 3 exhibit no or little stereochemical selectivity consistent with the observed non-enzymatic C6 deprotonation of 3. Alternatively, the stereoselectivity of H removal from C6 of 3 may be almost completely masked with a large primary kinetic isotope effect of D incorporation at C6. Similar results were observed for other BelL homologues.

X-ray crystal structure analysis of ScBelL and comparisons of the enzyme homologues

Alignment of the tested BelL homologues is shown in Figure S5. We initially hypothesized that the active site residues around the iron-binding cavity may vary among the homologues to exhibit different stereoselectivities. To gain structural insights into the stereodivergency of the cyclopropanses, the apo structure of ScBelL was solved to 2.2 Å resolution. Its overall structure resembles those of l-isoleucine dioxygenase^{22, 23} from Bacillus thuringiensis and an uncharacterized dioxygenase²⁴ from Methylibium petroleiphilum PM1, both of which consist of a jelly roll folding with two anti-parallel β-sheets. However, the long loop (32 amino acids) between the β3 strand and the α2 helix is inserted into a pocket of another molecule of ScBelL and is partially disordered. This implies that the loop is flexible and functions as a lid for substrate binding as proposed for IDO²³ although attempts to obtain the complex ScBelL crystal structure with 3 were unsuccessful. To visualize a possible closed conformation of the loop for constructing the substrate binding cavity, the structure of ScBelL in the native sequence was predicted by AlphaFold2.²⁵ Comparison between the crystal and predicted structures of ScBelL revealed a RMSD value of 0.42 Å, suggesting that AlphaFold2 predicted a reasonable model, in which the loop adopts a closed conformation that creates the active site cavity with a volume of 157 ų. The structures of other nine BelL homologues were also predicted by AlphaFold2 that reveal only subtle deviations in the main chains.

The active site pocket in each homologue is constructed with side chains of 29 residues including 14 residues that are completely conserved among the curated BelL homologues. The other 15 other residues were varied and thus speculated to control the stereoselectivity of the cyclization (Figure 2B). Multiple alignment of the varied 15 residues in 11 proteins in hand suggested that they could be classified into two groups A and B. The proteins in the group A have Arg, Phe/Tyr, Ile/Val, Val, and Met/Leu at positions 40, 114, 116, 127, and 154, respectively. The proteins in the group B have Lys, Asn, Ser/Thr, Thr, and Met/Leu at positions 40, 114, 116, 127, and 154, respectively. Consistently, phylogenetic analysis of the BelL homologues generated two clades, suggesting that the class of cyclopropanses have evolved into two groups. Contrary to our expectation, however, the classification appears not to directly correlate with the stereoselectivity of the cyclopropanation because the enzymes in group A generate all four diastereomers whereas those in the group B generate (1′R,2′S)-1, (1′S,2′R)-1, and (1′S,2′S)-1. Thus, minor differences in protein structures may alter the stereoselectivity.

Examination of the substrate binding and the stereochemical control by using enzyme variants

To understand how the varied amino acid residues control the stereoselectivity of cyclization, we first examined the substrate binding mode by docking analysis. Recent structural analysis of IDO with norleucine substrate revealed a possible substrate binding pose, where the carboxylate group and amino group of the substrate interacts with Arg70 and Asp172 (which also coordinates to Fe).²³ These residues are conserved in the tested BelL homologues (Arg79 and Asp134). However, the binding mode of the substrate amino acid in BelL is likely different from that predicted in IDO because the R79A variant of BelL was still active toward 3 with only partially loss of the activity. Alternatively, another conserved Arg (Arg40) located next to Arg79 was found to be essential because substitution of this residue with Ala completely abolished the cyclopropanation activity. Glu129 near the iron binding site in BelL is also conserved in TqaL, which accepts l-valine as the substrate to perform aziridination.^26–28 The corresponding residue Glu175 in TqaL has been implicated as a key residue to interact with the amino group of l-valine. As expected, the E129A variant of BelL showed significantly diminished activity (<10%). Together, the amino acid moiety of 4 is likely anchored by Arg79 and Glu129 via electrostatic interaction. The hydrocarbon side chain of norleucine substrate in IDO protrudes toward the long loop functioning as a lid for the active side.²³ However, based on the BelL model predicted by AlphaFold2, there would not be enough space to accommodate 6-nitronorleucine (3) if it adopted the same conformation as the norleucine substrate in IDO. In line with this prediction, BelL did not accept norleucine, which is a good substrate of IDO.^{22, 23}

Interestingly, Glu174 in IDO is substituted with His136 in all BelL homologues except for HrmJ, which has Arg at the same position. Thus, it is proposed that the nitronate side chain of 4 is pointing toward His136, which may serve as a hydrogen bonding donor. This hypothesis was supported by the observation that the H136F variant was catalytically inactive toward 3.. Furthermore, in contrast to the wild type BelL, the H136F variant did not exhibit the rate enhancement of the C6-deprotonation (initial rate, 11 μM·min^–1) as determined by the deuterium wash-in experiment. These results are consistent with the proposed role of His136 in facilitating the substrate binding as in the nitronate form (e.g. 4). The conserved Tyr64 located near His136 may serve as an alternative hydrogen bonding donor; however, the Y64F was still catalytically active, excluding the possible involvement of the phenolic OH in Y64 in the catalysis. Importantly, with this substrate conformation, the C4–proS-H bond is in proximity with the Fe cofactor, which is fully consistent with the results of the stereochemical analysis described above.

With the predicted substrate binding mode, we then investigated the reasons behind the altered stereoselectivity of BelL and HrmJ. These two cyclopropanses belongs to the group B and only exhibit minor differences of active site residues. A notable difference can be found at position 136, where BelL and HrmJ have His and Arg, respectively. When His136 in BelL was replaced with Arg, the stereochemistry of the product was completely switched from (1′S,2′S)-1 to (1′R,2′R)-1, although the enzyme activity was partially diminished. As a complementary experiment, Arg136 in HrmJ was also substituted with Ala. As expected, the resulting variant produced 25 % (1′R,2′R)-1 in addition to the 75 % (1′S,2′S)-1. These results indicated that His136 in BelL and Arg136 in HrmJ are critical in controlling the stereoselectivity of the cyclization. The His and Arg residues likely serves as H donors to interact with the nitro group of 3, while the differences in their size may lead to altered conformations that are suitable to produce either (1′S,2′S) and (1′R,2′R) isomer.. In addition, I33 and Y114 in BelL have minor impacts on the stereoselectivity because substitutions of these residues with the corresponding amino acids in HrmJ (Met and Phe, respectively) resulted in the production of 25% (1′R,2′S)-1, but not (1′R,2′R)-Y. Individual substitution of other active site residues varied between BelL and HrmJ did not significantly change the activity of the enzymes, implying that these residues cooperatively participate in the stereochemical control of the cyclopropanation.

Next, we investigated which residues are important to generate cis-configured cyclopropane products. After comparison of the sequences of BelL, TbBelL, and PsBelL, we speculated that the residue 206 (Ser in BelL and Ala in TbBelL) may be a critical factor to control the stereoselectivity because they are close to the putative substrate binding site in proximity to His136. As expected, the S206A variant of BelL generated 19% (1′R,2′S)-1 in addition to (1′S,2′S)-1. The result implies that the hydrophilicity of the Ser206 may be involved in the trans product formation by directly or indirectly interacting with the nitro group of 3 because the stereochemistry only at the 2′ position was altered. Further modification of the varied residues such as F70Y/V126T did not charge the ratio of the product stereoisomers. However, when Ile33 was replaced with Met, the resulting quadruple variant of S206A/F70Y/V126T/I33M produced (1′R,2′S)-1 as the major product (60%) (Figure 4B). Because the Ile33 residue is positioned closed to the substrate binding site, its substitution to Met may slightly decrease the cavity volume and force the substrate to adopt a cis-producing conformation, which should be more compact than trans-producing conformations.

Figure 4.

Comparison of residues around the active sites of the tested homologues. Indicated residue numbers are those of BelL. Stereochemistry of the main product(s) are also shown.

Conclusions

We have investigated the catalytic mechanism and stereodivergency of the nitroalkane cyclopropanases BelL and HrmJ. The assays with substrate analogues and the observed pH dependency suggested that the C6-deprotonation of the 6-norleucine substrate takes place for the cyclization activity. The transient spectroscopic analyses and LCMS using substrate isotope revealed the formation of an iron(IV)-oxo species that abstracts the pro-S H atom from the C4 of the substrate. Unlike canonical αKG-dependent nonheme iron hydroxylases, the resulting carbon radical is unlikely to undergo hydroxyl rebound. Instead, the radical may be directly cyclized through radical- or cation-mediated mechanisms (Scheme 2A).

The stereoselectivity of the cyclopropanase family is remarkable give that the conformationally unbiased liner nitroalkane substrate can be transformed into all four possible diastereomeric nitrocyclopropane products. The structural analyses and sequence comparisons revealed the highly conserved overall protein folding and the active sites with local amino acid variations crucial for the stereochemical control of the cyclopropanation. However, rational approaches involving multiple substitutions of the active site residues did not necessarily result in the complete alternation of the stereochemical outcomes, implying structural elements (such as point mutations or local changes in secondary structures) located away from the active sites may also influence the stereochemistry of cyclization. Thus, additional mechanistic and structural evaluations are anticipated to offer further insights into how nonheme iron enzymes perform cyclopropanation with high stereoselectivity.

Figure 5.

LCMS analysis of the cyclopropanation reactions catalyzed by BelL and HrmJ variants EIC traces corresponding to [M + H]⁺ signals from 1-Dns (m/z = 408.1) are shown in red.

Figure 6.

HPLC analysis of the reactions catalyzed by BelL variants. The reaction mixture was treated with dansyl chloride prior to the analysis. UV 340 nm chromatogram were shown.