Reduction of a Heme Cofactor Initiates N-Nitroglycine Degradation by NnlA

ABSTRACT

Linear nitramines are potentially carcinogenic environmental contaminants. The NnlA enzyme from Variovorax sp. strain JS1663 degrades the nitramine N-nitroglycine (NNG)—a natural product produced by some bacteria—to glyoxylate and nitrite (NO₂^–). Ammonium (NH₄⁺) was predicted as the third product of this reaction. A source of nonheme Fe^II was shown to be required for initiation of NnlA activity. However, the role of this Fe^II for NnlA activity was unclear. This study reveals that NnlA contains a b-type heme cofactor. Reduction of this heme—either by a nonheme iron source or dithionite—is required to initiate NnlA activity. Therefore, Fe^II is not an essential substrate for holoenzyme activity. Our data show that reduced NnlA (Fe^II-NnlA) catalyzes at least 100 turnovers and does not require O₂. Finally, NH₄⁺ was verified as the third product, accounting for the complete nitrogen mass balance. Size exclusion chromatography showed that NnlA is a dimer in solution. Additionally, Fe^II-NnlA is oxidized by O₂ and NO₂^– and stably binds carbon monoxide (CO) and nitric oxide (NO). These are characteristics shared with heme-binding PAS domains. Furthermore, a structural homology model of NnlA was generated using the PAS domain from Pseudomonas aeruginosa Aer2 as a template. The structural homology model suggested His⁷³ is the axial ligand of the NnlA heme. Site-directed mutagenesis of His⁷³ to alanine decreased the heme occupancy of NnlA and eliminated NNG activity, validating the homology model. We conclude that NnlA forms a homodimeric heme-binding PAS domain protein that requires reduction for initiation of the activity.

IMPORTANCE Linear nitramines are potential carcinogens. These compounds result from environmental degradation of high-energy cyclic nitramines and as by-products of carbon capture technologies. Mechanistic understanding of the biodegradation of these compounds is critical to inform strategies for their remediation. Biodegradation of NNG by NnlA from Variovorax sp. strain JS 1663 requires nonheme iron, but its role is unclear. This study shows that nonheme iron is unnecessary. Instead, our study reveals that NnlA contains a heme cofactor, the reduction of which is critical for activating NNG degradation activity. These studies constrain the proposals for NnlA reaction mechanisms, thereby informing mechanistic studies of degradation of anthropogenic nitramine contaminants. In addition, these results will inform future work to design biocatalysts to degrade these nitramine contaminants.

INTRODUCTION

Cyclic and linear (or aliphatic) nitramines are contaminants in soil and groundwater. Cyclic nitramines, such as hexahydro-1,3,5-trinitro-1,3,5-triazine, the high-energy compound RDX, are components of military-grade explosives. In addition to being products of RDX degradation, linear nitramines are by-products of some carbon capture technologies, formed when amines react with NO_x in the gas stream (1, 2).

RDX is a persistent soil contaminant at several explosives training facilities and manufacturing sites (3–5). Acute exposure to nitramines results in violent convulsions (6). Furthermore, the United States EPA lists RDX as an emerging contaminant and possible carcinogen. For these reasons, degradation pathways of RDX and other cyclic nitramines are well studied (7–13). One such pathway is initiated by a cytochrome P450 homolog, originally isolated from Rhodococcus rhodochrous strain 11Y, called XplA (14, 15). This enzyme reductively degrades RDX (Fig. 1A) (15–18). Under anaerobic conditions, the products are formaldehyde, nitrite (NO₂^–), and the one-carbon nitramine methylenedinitramine (MEDINA). Under aerobic conditions, RDX degradation produces 4-nitro-2,4-diazabutanal (NDAB) instead. A transgenic Arabidopsis strain expressing the RDX degradation enzyme, XplA, has been engineered as a promising soil bioremediation strategy for RDX (3, 19).

FIG 1.

Product distributions of nitramine degrading enzymes, including (A) aerobic degradation of RDX by XplA, and (B) degradation of NNG by NnlA. Reactants and products not yet verified are shown in square brackets.

Linear nitramines exhibit less acute toxicity than cyclic nitramines, but they have been shown to be skin and eye irritants (20, 21). Of greater concern, they are potential carcinogens (22–26). Nitramines often occur with their nitrosamine analogs, many of which are well-studied and potent carcinogens. Nitramines appear to be metabolized in a similar fashion as nitrosamines; specifically, their metabolism is initiated by hydroxylation by cytochromes P450 (22, 27–33). These hydroxylated products then degrade to form formaldehyde or alkylating agents, which are the proposed mutagenic agents. These nitramines are photostable (34); therefore, biodegradation appears to be their main environmental degradation pathway. Understanding the mechanisms of linear nitramine biodegradation is necessary to inform effective bioremediation strategies and design new biocatalysts for this purpose.

Compared with cyclic nitramines, there is far less known regarding the biodegradation pathways and mechanisms of linear nitramines. Some of these nitramines produced by carbon capture, particularly those with hydroxyl functionalities, were shown to be biodegraded by bacterial samples from soil and water (35). The products of these degradations were not reported. Biodegradations of NDAB by the fungus Phanerochaete chrysosporium and the bacterium Methylobacterium sp. strain JS178 have been reported (36, 37). In both cases nitrous oxide (N₂O) was produced. Initiation of the P. chrysosporium degradation was attributed to a manganese peroxidase. The mechanism of this degradation is unclear, but accumulation of N₂O suggests it is initiated by cleavage between the 4N-3C bond of NDAB (Fig. 1A).

The best characterized linear nitramine biodegradation pathway is an enzymatic pathway that degrades the natural product N-nitroglycine (NNG). This molecule is produced by some bacteria, including Streptomyces noursei (38, 39). The physiological function of NNG is unknown, but it exhibits toxicity toward Gram-negative bacteria and plants (38, 40). In addition, in vitro experiments have shown that NNG inhibits the ubiquitous succinate dehydrogenase enzyme (41). Variovorax sp. strain JS 1663 was enriched by its ability to use NNG as a sole carbon and nitrogen source (42). An NNG lyase (NnlA) was discovered to be essential for this phenotype. In vitro assays of NnlA showed it degraded NNG, producing glyoxylate and NO₂^– (Fig. 1B). A second nitrogenous product remains unidentified but was predicted to be ammonium (NH₄⁺). Previous characterization of NnlA showed no evidence for a redox cofactor, such as a heme or flavin.(43, 44) Therefore, NH₄⁺ as the final product was consistent with a redox-neutral degradation pathway and the apparent lack of a redox cofactor requirement. However, NnlA was shown to be homologous to Per-Arnt-Sim (PAS) domain proteins, which typically bind heme or flavin. Furthermore, initiation of activity for heterologously expressed, purified NnlA required addition of ferrous ammonium sulfate (Fe(NH₄)₂(SO₄)₂). This Fe^II may have reconstituted a nonheme iron-containing active site capable of activating dioxygen (O₂)—a well-known role for nonheme iron sites (45). An iron-dependent redox-mediated NNG degradation pathway by NnlA could be envisioned. To test this hypothesis, the role of Fe^II for NnlA activity needs to be clarified.

The purpose of this study was to identify the role of Fe^II in NnlA activity. En route to solving this question, we found that NnlA binds a heme cofactor. As with nonheme iron, heme cofactors are well-established to enable O₂ activation (46). Alternatively, heme may enable a reductive degradation pathway as observed for RDX degradation by XplA. We investigated the role of the NnlA heme cofactor, fully characterized the nitrogen mass balance of NNG degradation by NnlA, and the O₂-dependence of turnover. The cumulative data herein explains why Fe^II was needed to initiate activity in the prior work and shows it is not essential for activity. Finally, we provide evidence that NnlA shares several characteristics with heme-binding PAS domain proteins. The results provide insight into the mechanism of nitramine degradation and will aid in developing remediation strategies and engineering new enzymes for bioprocessing.

RESULTS

Characterization of NnlA and its heme cofactor.

Recombinant NnlA with an N-terminal protease-cleavable decahistidine tag (His₁₀-NnlA) was produced in Escherichia coli T7 Express cells. The His₁₀-NnlA protein was purified from the lysate by a Ni-NTA column. The theoretical mass of His₁₀-NnlA was calculated as 20,526 Da. While a monomer band was observed in the SDS-PAGE of protein purified by a Ni NTA column, several other higher molecular weight bands also appeared (Fig. S1A). These higher molecular weight bands have weights consistent with NnlA oligomers.

To determine the solution oligomerization state of NnlA, TEV protease-cleaved NnlA was analyzed using size exclusion chromatography in 10 mM Tris-HCl at pH 7.6 (Fig. S1B). The chromatogram of TEV-cleaved NnlA exhibited two peaks. The first peak elutes in the void volume suggesting a higher oligomer with apparent molecular mass of 200 kDa or greater. The second peak corresponds to a molecular mass of 44 kDa, which is consistent with an NnlA homodimer. There was no appearance of a monomer peak. The results suggested that NnlA is a homodimer in solution, with some of the NnlA appearing as a larger aggregate.

The UV-visible absorption spectrum of as isolated His₁₀-NnlA exhibited a feature centered at 412 nm consistent with the presence of heme (Fig. 2). Supplementing expression cultures with 5-aminolevulinic acid (5-ALA), a heme precursor, resulted in preparations with a 6-fold higher molar absorption coefficient at 412 nm. These results indicate that addition of 5-ALA increases the occupancy of heme in recombinant NnlA. These preparations routinely contained 0.87 ± 0.21 total Fe atoms per NnlA monomer (Table S1). In contrast, negligible nonheme iron was detected by ferrozine assay. UV-visible absorption of samples of NnlA and myoglobin treated with the pyridine hemochromagen assay were nearly identical (Fig. S2). These results are consistent with the binding of a b-type heme. Other transition metals were detected in His₁₀-NnlA samples by ICP-MS (Table S2). However, most of these were substoichiometric compared with the protein and are not expected to bind specifically to the protein. Nickel concentrations were high and likely due to the use of the Ni-NTA column during purification (47). As discussed above, amino acid sequence alignment shows that NnlA contains a PAS domain. The dimerization of the monomer and binding of a heme cofactor are common traits of these domains (43, 44, 48).

FIG 2.

UV-visible absorption spectra of as isolated NnlA expressed in the absence (red trace) or presence (black trace) of 5-ALA. Both spectra were measured in 50 mM phosphate buffer at pH 7.2.

Reduced NnlA reacts with O₂, NO, and CO.

The role of the heme cofactor in PAS domains is often a sensor of redox environment or of gas molecules such dioxygen (O₂), or nitric oxide (NO) (48–50). Therefore, we characterized the reactions of paramagnetic gas molecules with reduced NnlA. Reduction of the NnlA heme was monitored by UV-visible absorption spectrophotometry (Fig. 3). The as isolated His₁₀-NnlA exhibited a Soret feature at 413 nm with broad absorbance in the 500- to 700-nm region, including poorly resolved peaks near 502 and 636 nm. The absorption spectrum is reminiscent of Fe^III hemoglobin or myoglobin (51). Therefore, the as isolated His₁₀-NnlA is hereafter termed Fe^III-NnlA. Reduction of the sample was achieved by titration of Fe^III-NnlA with dithionite until the absorbance at 413 nm no longer decreased. Alternatively, the protein could be reduced by addition of an excess of Fe(NH₄)₂(SO₄)₂ (Fig. S3). Under either reduction condition, the resulting reduced sample exhibited a Soret band at 437 nm and formation of a Q-band maximum at 560 nm. This spectrum of reduced NnlA resembles that of Fe^II myoglobin (51). Therefore, these features are hereafter attributed to Fe^II-NnlA and are achieved using dithionite as the reductant.

FIG 3.

Treatment of Fe^II-NnlA with (A) air, (B) NO, or (C) CO. Conditions before initiation of reaction by inversion of sample into gas headspace were 10 μM Fe^II-NnlA in either in 100 mM Tris-HCl pH 7.6 (panel A) or 100 mM tricine buffer at pH 8 (panels B and C).

The reaction of O₂, CO, or NO with Fe^II-NnlA was monitored by UV-visible absorption spectrophotometry. There were no noticeable changes in the UV-visible absorption spectra when O₂, CO, or NO were added to Fe^III-NnlA (data not shown). In contrast, each of these gases readily reacted with Fe^II-NnlA (Fig. 3). Exposure of 10 μM Fe^II-NnlA to air resulted in the loss of Fe^II-NnlA absorption features with a concomitant rise in Fe^III-NnlA absorption features over several minutes. The intensities of these features were decreased compared with those observed for the protein prior to reduction (Fig. 3A), suggesting that the observed oxidation of Fe^II-NnlA by O₂ destroys some of the heme. Optical features consistent with the Fe^II-O₂ were not observed, but it is expected to be an intermediate en route to heme oxidation.

Addition of NO gas to a deoxygenated sample of Fe^II-NnlA resulted in immediate appearance of new absorbance features at 421, 550, and 580 nm. This spectrum is consistent with that for the ferrous-nitrosyl ({FeNO}⁷ by Enemark-Feltham notation [52]) adduct of myoglobin (Fig. 3B) (51). Meanwhile, addition of CO to Fe^II-NnlA causes appearance of new absorption features at 425, 542, and 568 nm. This spectrum is similar to that of the CO adduct of myoglobin (53), thereby indicating that Fe^II-NnlA binds CO (Fig. 3C).

NnlA activity requires heme reduction and does not require O₂.

The NNG degradation activity of heme-occupied Fe^III-NnlA was first tested. Samples containing Fe^III-NnlA and NNG were analyzed with LC-ESI-MS to monitor for decomposition of NNG (m/z 119.0) and formation of glyoxylate (m/z 73.0). Extracted ion chromatograms (EICs) monitoring these molecular anions are shown in Fig. 4. These Fe^III-NnlA samples showed no evidence for NNG decomposition to form glyoxylate, suggesting Fe^III-NnlA needs to be activated to exhibit activity. As shown above, Fe(NH₄)₂(SO₄)₂—required to initiate activity in prior work—was able to reduce the NnlA heme. Therefore, we posited that NnlA activity is dependent on reduction of the heme instead of the presence of a ferrous iron source.

FIG 4.

Overlaid representative LC-MS EICs monitoring molecular anions of NNG (m/z 119.0) and glyoxylate (m/z 73.0) in samples containing 500 μM NNG and 10 μM dithionite (dithionite/NNG), 10 μM as isolated NnlA (Fe^III-NnlA), or 10 μM reduced NnlA (Fe^II-NnlA). Samples were incubated for 30 min at room temperature in deoxygenated 20 mM phosphate buffer, pH 7.2. The Fe^II-NnlA/O₂ sample was incubated in air-saturated buffer. Dashed and dotted gray lines indicate elution time of glyoxylate and NNG in standard solutions.

To test this hypothesis, Fe^II-NnlA was incubated with NNG. These samples were prepared in an anaerobic glove box to test the need for O₂ for the reaction. Anaerobic samples containing 10 μM Fe^II-NnlA and 500 μM NNG exhibited complete degradation of the NNG to form glyoxylate (Fig. 4). For comparison, samples containing only dithionite and NNG exhibited no degradation of NNG. This result showed that the NnlA heme needs to be in the Fe^II oxidation state to activate NNG decomposition activity. The need for reduction of the heme to initiate NnlA activity unambiguously showed that the heme cofactor is necessary for NnlA activity, and its appearance is not an artifact of recombinant expression. Furthermore, this activation could be achieved without the need for Fe(NH₄)₂(SO₄)₂.

These experiments also precluded the hypothesis that NNG degradation by NnlA proceeds by a reductive or oxidative pathway. First, 10 μM Fe^II-NnlA performed 50 turnovers under these conditions. This catalytic NNG degradation shows that electron transfer from the heme to the NNG is not necessary, and thereby eliminated the possibility of a reductive NNG degradation pathway. In addition, complete NNG degradation was observed in the absence of O₂. This result precluded the possibility that O₂ activation by the NnlA heme is required for NNG degradation. In fact, simultaneous addition of O₂ and NNG resulted in less NNG degradation, indicating O₂ inhibits the reaction. These combined observations strongly suggested that NNG degradation by NnlA is redox-neutral.

Determination of the nitrogen mass balance.

To verify that NNG degradation by NnlA is redox neutral, we determined the mass balance of the NNG degradation Table 1. In samples containing 10 μM Fe^II-NnlA and 1,000 μM NNG in 10 mM tricine buffer, pH 8 were prepared under anaerobic conditions and incubated at 20°C for 30 min. In parallel, NNG and glyoxylate were analyzed by LC-ESI-MS, NO₂^– was quantified by Griess assay or ion chromatography, and NH₄⁺ was assayed by a l-glutamate dehydrogenase coupled assay. The EICs showed that the NNG in these samples was completely degraded with concomitant appearance of 1,210 ± 290 μM glyoxylate, accounting for 100% of the carbon mass balance. Parallel analysis of nitrogenous products showed the appearance of 810 ± 40 μM NO₂^– and 1,050 ± 100 μM NH₄⁺ in these samples. Ion chromatography showed no evidence for the presence nitrate (Table S3), providing further evidence against NNG degradation proceeding by an oxidative pathway. The total nitrogen products accounts for 93% of the nitrogen mass balance. These results are consistent with the following reaction stoichiometry:

$$\mathrm{NNG}\hspace{0.17em}+\hspace{0.17em}{\text{H}}_2\mathrm{O}\hspace{0.17em}\to \hspace{0.17em}{\text{NH}}_4^{\hspace{0.17em}\hspace{0.17em}+}\hspace{0.17em}+\hspace{0.17em}{\text{NO}}_2^{\hspace{0.17em}\hspace{0.17em}–}\hspace{0.17em}+\hspace{0.17em}\text{glyoxylate}$$ (1)

TABLE 1.

Mass balance of NNG degradation by NnlA^a

[FeII-NnlA] (μM)	[NNG]initial (μM)	[NNG]final (μM)b	[NH4+]final (μM)b	[glyoxylate]final (μM)b	[NO2−]final (μM)b
10	1000	0 ± 0	1050 ± 100	1210 ± 290	810 ± 40 μM

^aSamples were incubated under anaerobic conditions in 10 mM tricine buffer at pH 8 and at room temp for 30 minutes.

^bValues and errors are reported as mean and one standard deviation of three trials, respectively.

This reaction stoichiometry of NNG degradation by NnlA is redox neutral.

NO₂^– oxidizes Fe^II NnlA.

Given that the reaction stoichiometry was consistent with a redox-neutral process, it was not expected that NNG would oxidize Fe^II-NnlA. However, samples containing 3 μM Fe^II-NnlA reacted with 133 μM NNG under anaerobic conditions resulted in oxidation of Fe^II-NnlA within several hours as monitored by UV-visible absorption spectrophotometry (Fig. 5A). The final spectrum of this reaction exhibited spectral features Q-band features consistent with formation of the heme {FeNO}⁷ shown in Fig. 3B This observation suggested NNG degradation resulted in formation of some NO, which subsequently bound to Fe^II-NnlA to form the observed heme-nitrosyl adduct. Heme centers are well known to reduce nitrite to NO (54); therefore, we posited that the product NO₂^– was responsible for oxidizing the heme center and not NNG.

FIG 5.

UV-visible spectra of reduced NnlA treated with either NNG (panel A) or NO₂^– (panel B) under anaerobic conditions. Reaction conditions: (panel A) 3 μM Fe^II-NnlA, 133 μM NNG, in deoxygenated 20 mM phosphate, pH 7.5 incubated for 5 h at room temperature; (panel B) 3 μM Fe^II-NnlA with 133 μM NO₂^– in deoxygenated 20 mM phosphate, pH 7.5 at room temperature.

Indeed, addition of 133 μM NO₂^– to 3 μM Fe^II-NnlA under anaerobic conditions oxidized the protein to Fe^III-NnlA (Fig. 5B). We note that in this experiment, there is no evidence for formation of an {FeNO}⁷ species. Nevertheless, the time course of oxidation of the Fe^II-NnlA center, monitored by the decrease in 437 nm, was nearly identical whether in the presence of NO₂^– or NNG (Fig. S4). These combined results indicated that NO₂^–, a product of NNG degradation, oxidized Fe^II NnlA to Fe^III NnlA, thereby inactivating the protein. In other words, the data show that NnlA was product inhibited.

Structural homology model of NnlA.

To date, there are no published crystal structure models of NnlA. However, the accumulated evidence suggests that NnlA is a PAS domain protein. To provide structural prediction of NnlA, we generated a structural homology model using SWISS-MODEL (Fig. 6). The model was selected by first allowing SWISS-MODEL to self-select recommended templates. Models were then generated on all structural templates containing a heme cofactor, consistent with the experimentally observed heme in NnlA. This method resulted in the generation of 17 structural models. Out of the 17 models, 16 were based on previously characterized PAS domains, including those from DOS, FixL, and Aer2 protein complexes. The best generated model threaded the NnlA sequence onto the crystal structure of the cyanide (CN)-bound PAS domain of Pseudomonas aeruginosa Aer2 (PDB ID: 3VOL) and had a GMQE score of 0.31. All other models exhibited a GMQE score between 0.03 and 0.18.

FIG 6.

Structural homology model of NnlA overlaid on the crystal structure of the CN^– complexed PAS domain from Pa Aer2 (PDB: 3VOL; gray). Peptide backbone of NnlA structural homology model shown as green cartoon and Pa Aer2 backbone shown as gray cartoon. Iron atom shown as orange sphere, oxygen and nitrogen atoms shown in red and blue, respectively. Structural homology model was generated by SWISS-MODEL.

We tested the homology model by generating an NnlA variant with decreased heme occupancy. In Pa Aer2, the heme is coordinated by an axial histidine ligand (55). Overlaying the NnlA structural homology model on the Pa Aer2 structure reveals that H73 of NnlA is in a similar position as the axial histidine coordinated to the Pa Aer2 heme (Fig. 6). To validate the structural homology model an H73A NnlA variant was generated by site-directed mutagenesis. The variant protein recombinantly expressed and purified as described for wild-type NnlA. The total iron concentration per monomer protein (0.35 ± 0.08 μM Fe/monomer) of the H73A NnlA was less than half that for wild-type NnlA (Table S1). Furthermore, there is a large amount of nonheme iron in the H73A NnlA samples, suggesting that the total iron assay overestimates the heme occupancy. Regardless, the iron analyses suggest that the H73A mutation adversely affects the binding of the heme cofactor. This conclusion is supported by UV-visible absorption spectra, which showed that the purified H73A NnlA exhibited an 8-fold lower molar absorption coefficient at 410 nm that that of the wild-type NnlA (Fig. 7). Finally, the variant protein lacks NNG degradation activity. There is no difference in the amount of NNG or glyoxylate in samples of Fe^II-H73A NnlA compared to Fe^III-H73A NnlA samples incubated with NNG for 30 min (Fig. S5). The combined results validate the proposed heme binding site from the structural homology model of NnlA.

FIG 7.

UV-visible absorption spectra of 20 μM wild-type or 17 μM H73A NnlA. All samples prepared in 100 mM tricine buffer, pH 8.

DISCUSSION

The conclusions of the data presented above are summarized in Fig. 8. A critical first conclusion is that NnlA contains a heme cofactor that is essential for activity. First, supplementing cultures expressing NnlA with the heme precursor 5-ALA resulted in NnlA preparations that bound one b-heme per subunit (Fig. 2; Table S1; Fig. S2). Reduction of the heme is required to initiate activity (Fig. 4). Finally, generation of the heme-depleted variant, NnlA H73A, results in loss of NNG degradation activity (Fig. S5).

FIG 8.

Summary of NnlA activity.

A key goal of this study was to understand the need for an Fe^II source in the previous study. Without culture medium supplementation with 5-ALA, preparations of NnlA in the prior study had low heme occupancies (Fig. 2). While samples had enough heme-bound monomers to turnover NNG, the low occupancy precluded detection of the cofactor by UV-visible absorption spectroscopy. With the observation that Fe(NH₄)₂(SO₄)₂ reduced the NnlA heme (Fig. S3), it is most likely that this Fe^II source reduced a small quantity of heme-bound NnlA, all of which was responsible for the observed activity. We conclude that initiation of NNG degradation by NnlA is dependent on reduction of the heme and does not require the presence of nonheme Fe^II.

Heme cofactors often mediate redox-dependent chemistries. For example, reductive degradation of the cyclic nitramine RDX is mediated by the heme cofactor of XplA (15–18). Furthermore, heme cofactors are well-established to activate O₂, (46) presenting the possibility of a high-valent iron-oxo species participating in NNG decomposition. However, our experiments showed that activity does not require O₂, precluding the possibility that O₂ activation is required for NNG degradation (Fig. 4). In fact, O₂ inhibited the NnlA-catalyzed decomposition of NNG (Fig. 4). Furthermore, Fe^II-NnlA can perform at least 100 turnovers (Table 1), indicating that NNG is not reduced during degradation. Finally, complete determination of the nitrogen mass balance verified NH₄⁺ as the second nitrogenous product of NNG decomposition by NnlA (Table 1). The combined results are consistent with NNG degradation by NnlA being a redox-neutral process.

The finding of an essential heme for NnlA activity now requires further understanding of its specific role. As discussed above, NnlA is predicted to be structurally homologous with PAS domains. In addition, we have established that NnlA required a heme cofactor for activity. We characterized NnlA as a dimer (Fig. S1B), and PAS domain proteins are often dimers or tetramers in solution (44, 56). In addition, suggested templates for structural homology modeling by SWISS-MODEL were nearly all structures of PAS domain proteins. Finally, the NnlA heme is capable of binding CO and NO to form stable complexes (Fig. 3). The NnlA heme does not stably bind O₂ but is instead oxidized by O₂. However, the structural homology model predicts that a critical distal tryptophan residue that stabilizes O₂ binding in Pa Aer2 (55) is absent in NnlA and is replaced with a glutamine residue (Fig. 6). This may account for the differing reactivity of the NnlA heme compared with those of previously characterized heme-binding PAS domains that are O₂ sensors. Regardless, the accumulated evidence supports assignment of NnlA as a PAS domain protein.

Typically, PAS domains are redox or O₂ sensor domains (43, 44, 48). Activation of the domains by change in oxidation state or binding of O₂ triggers downstream processes, such as kinase activity or chemotaxis. Therefore, one possibility is that the NnlA heme is a redox sensor that triggers organization of an allosteric active site for NNG degradation in NnlA (Fig. 9). Most PAS domain-containing subunits do not also contain the catalytic site. However, our results indicate that the NnlA monomer contains both the regulatory PAS domain as well as a catalytic domain. Alternatively, the active site may reside at the subunit interface thereby requiring oligomerization for activity. By this hypothesis, NnlA activity would be regulated by the redox environment, either by the presence of O₂ or NO₂^–. As shown in Fig. 3, O₂ oxidizes Fe^II-NnlA, which inactivates the enzyme. This observation is consistent with the observation of less NNG degradation in oxic samples (Fig. 4). The oxidation of NNG by NO₂^– (Fig. 5) indicates that NO₂^– could also act in a regulatory fashion as a product inhibitor.

FIG 9.

Proposed roles of heme cofactor in NnlA.

A second hypothesis is that the Nnla heme is repurposed as an active site instead of a regulatory site, and, therefore, the heme directly participates in NNG degradation. It was previously proposed that a general base deprotonates the NNG α-carbon, thereby, promoting β-elimination of NO₂^– from NNG to form an imine (42). Subsequent hydrolysis of the imine intermediate results in formation of the degradation products NH₄⁺ and glyoxylate. By our second hypothesis, the Fe^II-heme could act as a Lewis acid to activate the nitro group and promote elimination of the NO₂^– (Fig. 9). Differentiating between these two hypotheses will be required to identify the active site and enable future engineering efforts.

While the physiological function and biosynthetic pathway of NNG is still unknown, the finding of an NNG-degrading enzyme suggests that it is present in some abundance in the environment. Several homologs of NnlA in diverse bacterial species were previously reported (42), which may support this hypothesis. However, the NNG degradation activities of these homologs remain untested, and, therefore, it still remains difficult to conclude if NNG degradation activity is widespread. It was previously shown that NNG is a potent inhibitor of the metabolic enzyme succinate dehydrogenase and exhibits some antibiotic activity toward E. coli (38, 41). Additionally, NNG is isolectronic with 3-nitropropionate, a potent toxin produced by some fungi and plants (57). It is reasonable to posit that NNG acts in some antibiotic function and that NnlA is a nitramine detoxification enzyme, albeit there is no evidence that NNG is toxic to Variovorax sp. JS 1663, the host organism of NnlA. Further studies of the role of this molecule and the ecological function of NnlA are warranted. Future studies may engineer NnlA to degrade nitramines of environmental concern, leveraging the mechanistic insight from this work.

MATERIALS AND METHODS

General reagents and protocols.

Isopropyl β-D-1-thiogalactopyranoside (IPTG) and 5-aminolevulinic acid (5-ALA) were purchased from GoldBio. NNG was purchased from AAblocks. General buffers and media components were purchased from Fisher Scientific or VWR. Stock dithionite concentrations were determined by UV-visible absorbance at 318 nm (ε₃₁₈ = 8000 M⁻¹cm⁻¹). The nitric oxide generator PROLI-NONOate was purchased from Cayman Chemicals. Stocks of PROLI-NONOate were prepared by dissolving approximately 10 mg of PROLI-NONOate in 10 mM NaOH and quantified by measuring the absorbance at 250 nm (ε₂₅₀ = 6,500 M⁻¹cm⁻¹) or 252 nm (ε = 8,400 M⁻¹cm⁻¹), respectively. NO gas was purified prior to use by bubbling into degassed 10 mM NaOH in a septum-sealed container. Buffers were degassed in septum-sealed glass bottles by 3x vacuum/N₂ gas purge cycles on a Schlenk line connected using a 22 G needle punctured through the septum. Water used for all solutions was of 18.2 MΩ·cm resistivity from a Barnstead Nanopure (Thermo Fisher Scientific). Solvents for LC-MS experiments were of at least HPLC grade and contained 0.1% vol/vol formic acid. Recombinant TEV protease was expressed and purified as previously described (58).

Mutagenesis of nnlA.

Site directed mutagenesis of the nnlA gene was performed to produce the variant protein NnlA H73A. Primers were designed for an “Round-the-horn” mutagenesis (Table 2) (59). The standard protocol was performed using Phusion polymerase and an annealing temperature of 63°C. The ligated reaction mixture was transformed into competent E. coli DH5α cells and plated on terrific broth (TB) agar plate containing 0.1 g/mL ampicillin. Colonies were selected and used to inoculate 5-mL Luria broth (LB) cultures with 0.1 g/mL ampicillin, which were grown overnight at 37°C. DNA was extracted from these cultures and analyzed by DNA sequencing (GeneWiz) to verify the mutation.

TABLE 2.

H73A mutant primers

Variant	Forward primer	Reverse primer
H73A	GCCCGTCGTGCAATCACGAAT	GCGATTATGAATTTCATCCCCAAT

Expression and purification of NnlA and variant.

Gene expression and affinity purification of NnlA protein using the pDG708 expression vector was performed as previously described (42). Alternatively, NnlA could be expressed from T7 Express cells transformed with pDG750, which contains the nnlA gene with an N-terminal His₁₀-tag, which is codon-optimized for expression in E. coli. NnlA with high heme occupancy was expressed by growth of these transformants in 4 × 1-L flasks of TB with 0.1 g/L of ampicillin at 37°C. At an OD₆₀₀ of 2, the temperature was decreased to 20°C and NnlA expression induced with 100 mg/L of isopropyl ß-D-1-thiogalactopyranoside (IPTG), 1 g/L of ferric ammonium citrate (FAC) and 0.8 g/L 5-aminolevulinic acid (5-ALA). The cultures were grown for another 24 h. Cells were pelleted by centrifuge at 6,353 g, yielding 17 g/L of wet cell mass. The pellet was either lysed immediately or stored frozen at −60°C.

Resuspended cells in a 1:2 (m/v) ratio of cells to Ni-buffer A (100 mM Tris-HCl, 100 mM NaCl at pH 7.6) were lysed by sonicating at 4°C. The sonication program used a 20% amplitude for 10 min (10 s pulse, 10 s pause). The low amplitude required three 10-min cycles for complete lysis. The cell debris was pelleted in a centrifuge at 53,000 g yielding a clear cherry-brown lysate. His₁₀-NnlA was loaded on a Ni-NTA HTC column (GoldBio), washed with 2 column volumes of Ni-buffer A, and eluted using a 6 column-volume gradient of 0% to 100% Ni-buffer B (100 mM Tris-HCl, 100 mM NaCl, 1 M imidazole at pH 7.6). Reddish-brown fractions were evaluated for relative heme incorporation using the ratio of 412 nm to 280 nm based on UV-visible absorption spectra. We found that fractions with a ratio A₄₁₂/A₂₈₀ > 1.0 had the highest heme occupancies. Therefore, these fractions were evaluated for purity using SDS-PAGE. Pooled fractions were buffer exchanged into Ni-buffer A. Concentrated protein was aliquoted into microcentrifuge tubes and stored at −60°C. NnlA H73A was expressed and purified as described above.

NnlA characterization and spectroscopy.

NnlA was quantified using a bicinchoninic acid assay (Thermo Scientific). Total heme and nonheme iron was quantified using an iron assay that allows for release and subsequent detection of heme-ligated iron (60). Nonheme iron was specifically quantified using a ferrozine assay (61). Differentiation of the bound heme type was determined using the pyridine hemochromagen assay (62). Dissolved metal concentrations were determined using a Thermo Fisher Scientific iCAP-Qc inductively coupled plasma mass spectrometer (ICP-MS) with QCell technology and operated in kinetic energy discrimination (KED) mode of analysis with helium as the collision gas. Calibration, internal, and quality control standards (Inorganic Ventures) were prepared in 2% (vol/vol) HNO₃ and calibration standards were prepared at concentrations of 1 to 1,000 μg L⁻¹ (ppb).

To characterize the oligomer state of NnlA, the His₁₀-tag was cleaved from NnlA using recombinant TEV protease as previously described (58). The His₁₀-NnlA was incubated with TEV protease at 4°C for 72 h. The digested protein was passed over a 2-mL Ni-HTC (GoldBio) gravity column to separate the tag-free NnlA, remaining His₁₀-NnlA, His₇-TEV, and cleaved His₁₀-tag. The brown flowthrough containing tag-free NnlA was concentrated, and 500 μL was then passed over a Superdex TM200 10/300 GL analytical gel filtration column. The column was preequilibrated with 100 mM Tris-HCl, 100 mM NaCl at pH 7.8 at a flow rate of 0.1 mL/min. A standard mass curve was generated by passing a premade Bio-Rad Gel filtration standard (#1511901) and blue dextran.

UV-visible absorption spectra of NnlA oxidation states and gas-bound forms were collected in an anaerobic glove box (Vacuum Atmospheres Co.) atmosphere using an Ocean Optics USB2000+ UV-visible absorption spectrometer. Samples containing 10 μM NnlA in degassed buffer were titrated with stock sodium dithionite until the 430-nm absorbance of Fe^II NnlA no longer increased. The reaction of Fe^II-NnlA with O₂, CO or NO was performed by adding air, CO gas, or purified NO gas to the evacuated headspace of an anaerobic sample of 10 μM Fe^II NnlA contained in a septum sealed quartz cuvette. The reaction was initiated by inverting the cuvette 3 to 4 times to dissolve gas in the protein solution. Spectra of Fe^II NnlA reacted with NNG or NO₂^– were collected after reaction of 2.7 μM Fe^II NnlA with 133 μM NNG or nitrite. Single wavelength traces were collected by monitoring 415, 417, or 437 nm absorbance every 4.8 s for the duration of the reaction.

General LC-MS methods.

LC-MS analysis was performed using an Agilent 1260 LC stack equipped with a Zorbax RX-C18 column (5 μm, 4.6 × 150 mm) and connected to an Agilent 6230 TOF mass spectrometer with electrospray ionization (ESI). Analyses used an isocratic mixture containing 65% water, 25% acetonitrile, and 10% isopropanol at a flow rate of 0.5 mL/min. The mass spectrometer was run in negative ion mode with a probe voltage of 4,500 V and fragmentation voltage of 175 V. To monitor NNG and glyoxylate, extracted ion chromatograms were obtained at m/z 119.0 and m/z 73.0, respectively. Concentrations of NNG and glyoxylate in samples were determined using calibration curves of standards containing 65, 125, 250, 500, and 750 μM NNG or glyoxylate in 10 mM tricine buffer.

Nitrogen assays.

Ammonium concentrations were determined using a glutamate dehydrogenase assay (Sigma-Aldrich) kit using the manufacturer’s instructions. Nitrite concentrations were determined by reacting 100 μL aliquots of reaction sample with 50 μL of deoxygenated Griess reagent R1 (Cayman Chemical, 1% sulfanilamide in 5% H₃PO₄) followed by addition of 50 μL of deoxygenated Griess reagent R2 (Cayman Chemical, 0.1% napthylethylenediamine dihydrochloride in water). The NO₂^– concentration was determined by using the molar absorption coefficient of the Griess-treated sample (ε₅₄₂ = 50 mM⁻¹cm⁻¹) or from a standard curve generated from reactions with known concentrations of NO₂^–.

Samples were analyzed by ion chromatography for nitrite and nitrate using a Dionex Integrion High-Pressure Ion Chromatography (Thermo Scientific) equipped with a 4-mm anionic exchange column (IonPac AS20), suppressor (Dionex ADRS 600 Suppressor) and a conductivity detector, operated at constant voltage (4.0 V). The sample loop was 20 μL and was first degassed in an internal oven at 30°C and then carried through the column by 35 mM NaOH (ultrapure, carbonate free, Acros Organics). The elution times under the conditions studied were 4 min and 4.9 min for nitrites and nitrates, respectively.

Preparation of NNG decomposition stoichiometry samples.

Stoichiometry samples were prepared in an anaerobic glove box using degassed buffers. Samples containing 10 μM as-isolated (NnlA_as-iso) or Fe^II NnlA were mixed with 500 μM NNG as needed for each sample. The dithionite control was prepared by mixing 500 μM NNG with 10 μM Na₂S₂O₄. Reduced samples contained 10 μM Na₂S₂O₄, 10 μM NnlA_as-iso, and 500 μM NNG in 10 mM tricine buffer, pH 8. Samples reacted with O₂ were prepared in two parts: 250 μL of 1 mM NNG remained outside the glove box under oxic conditions, while 250 μL of reduced protein was prepared inside the anaerobic glove box by mixing NnlA and Na₂S₂O₄ to a final concentration of 20 μM each. The 250 μL aliquot of protein was removed from the box and added quickly to the 250 μL aliquot of oxygenated NNG in buffer with a final concentration of 20 mM potassium phosphate and pH 7.2. Every sample prepared was incubated for 30 min regardless of conditions.

Homology modeling.

Structural homology modeling was performed using SWISS-MODEL (63). Models were ranked by the Global Model Quality Estimate (GMQE) score, which assesses the expected accuracy of the model based on target-template alignment and the structure. The GMQE score ranges from 0 to 1, for which 1 is the best score.

Data availability.

The accession number for the Variovorax sp. JS1663 has been previously deposited in GenBank with accession no. LYMK01000012 (42). The locus tag for NnlA is A8M77_12345.