A nitric oxide synthase–like protein from Synechococcus produces NO/NO3− from l-arginine and NAPDH in a tetrahydrobiopterin- and Ca2+-dependent manner

Nitric oxide synthases (NOSs) are heme-based monooxygenases that convert l-Arg to l-citrulline and nitric oxide (NO), a key signaling molecule and cytotoxic agent in mammals. Bacteria also contain NOS proteins, but the role of NO production within these organisms, where understood, differs considerably from that of mammals. For example, a NOS protein in the marine cyanobacterium Synechococcus sp. PCC 7335 (syNOS) has recently been proposed to function in nitrogen assimilation from l-Arg. syNOS retains the oxygenase (NOSox) and reductase (NOSred) domains present in mammalian NOS enzymes (mNOSs), but also contains an N-terminal globin domain (NOSg) homologous to bacterial flavohemoglobin proteins. Herein, we show that syNOS functions as a dimer and produces NO from l-Arg and NADPH in a tetrahydrobiopterin (H4B)-dependent manner at levels similar to those produced by other NOSs but does not require Ca2+-calmodulin, which regulates NOSred-mediated NOSox reduction in mNOSs. Unlike other bacterial NOSs, syNOS cannot function with tetrahydrofolate and requires high Ca2+ levels (>200 μm) for its activation. NOSg converts NO to NO3− in the presence of O2 and NADPH; however, NOSg did not protect Escherichia coli strains against nitrosative stress, even in a mutant devoid of NO-protective flavohemoglobin. We also found that syNOS does not have NOS activity in E. coli (which lacks H4B) and that the recombinant protein does not confer growth advantages on l-Arg as a nitrogen source. Our findings indicate that syNOS has both NOS and NO oxygenase activities, requires H4B, and may play a role in Ca2+-mediated signaling.

NADPH; however, NOS g did not protect Escherichia coli strains against nitrosative stress, even in a mutant devoid of NO-protective flavohemoglobin. We also found that syNOS does not have NOS activity in E. coli (which lacks H 4 B) and that the recombinant protein does not confer growth advantages on L-Arg as a nitrogen source. Our findings indicate that syNOS has both NOS and NO oxygenase activities, requires H 4 B, and may play a role in Ca 2؉ -mediated signaling.
Nitric oxide (NO) is a gaseous free radical involved in numerous biological processes; it is an intermediate in the denitrification pathway (1), a precursor to protein post-translational modification via S-nitrosylation (2), and is the activator of soluble guanylate cyclase in animals or H-NOX proteins in bacteria (3,4). In mammals, NO is the product of arginine oxi-dation by nitric oxide synthases (NOSs) 2 (5,6). The three mammalian isoforms, endothelial (eNOS), neuronal (nNOS), and inducible (iNOS) (6 -8), share a heme-containing oxygenase domain (NOS ox ) and a C-terminal reductase domain (NOS red ). NOS red , which functions to reduce NOS ox using NADPH, is composed of an FMN-binding domain and a ferredoxin-NADP ϩ -reductase (FNR)-like domain. NOSs function as N-terminal homodimers, whereby the NOS red of one subunit reduces the NOS ox of the opposite subunit (9). Electron transfer is activated by Ca 2ϩ -loaded calmodulin (CaM) (10) that binds at a conserved sequence between NOS ox and NOS red , and is also facilitated by the essential cofactor tetrahydrobiopterin (H 4 B) that acts to supply electrons to the NOS heme for oxygen activation (11). The mammalian NOS isoforms play key roles in many biological processes, such as vasoconstriction, immune response, and neuronal plasticity (12)(13)(14) and are also involved in several pathologies, including tumorigenesis, septic shock, and cerebral ischemia (15)(16)(17)(18).
Although NOS is ubiquitous in the animal kingdom, it is infrequently found in bacteria. The occurrence and purpose of bacterial NOS is highly species-dependent, ranging from recovery from UV damage (Deinococcus radiodurans nitric oxide synthase (drNOS)) (19) to signaling biofilm formation (Silicibacter nitric oxide synthase (siliNOS)) (20), protection from oxidative stress (Bacillus subtilis nitric oxide synthase (bsNOS)) (21), aiding pathogen virulence (Bacillus anthracis nitric oxide synthase (baNOS)) (22), and controlling oxygenbased respiration (Staphylococcus aureus nitric oxide synthase (saNOS)) (23,24). Although their heme domain structure and catalytic mechanisms are similar to that of mammalian NOS, most bacterial NOSs lack a dedicated reductase domain, instead relying on promiscuous cellular reductases (25,26). One NOS found in Sorangium cellulosum has an N-terminal reductase domain containing a 2Fe-2S cluster and ferredoxinlike FAD and NADPH domain, dissimilar to mammalian NOS red (27). No bacterial NOS with a covalently attached FMN/FNR reductase domain has been biochemically characterized thus far.
NOSs are also found in photosynthetic organisms. A mammalian NOS homolog was characterized from the algae Ostreococcus tauri (otNOS) (28), which is intriguing because NOS has not yet been identified in higher plants, despite NO having an undisputed role in plant signaling (29). Recently, a mammalianlike NOS with a C-terminal P-450 reductase domain was identified in the photosynthetic diazotroph Synechococcus sp. PCC 7335 (syNOS) (30). syNOS is the first prokaryotic NOS to contain a mammalian NOS red homolog; in addition, syNOS contains a somewhat unusual globin domain (NOS g ) N-terminal to NOS ox , as well as a 342-residue N-terminal region of unknown properties ( Fig. 1 and Fig. S1). The syNOS-harboring Synechococcus strain was shown to produce NO in an L-Arg-dependent manner, and this activity was inhibited by known NOS inhibitors (30). Based on genetic experiments in Synechococcus and heterologous expression experiments in Escherichia coli, syNOS was proposed to function in nitrogen utilization from L-Arg (30). Specifically, this model asserts that syNOS first converts L-Arg to NO with NOS ox and then from NO to NO 3 Ϫ with NOS g . Nitrate would then be assimilated back into reduced forms of nitrogen. Herein, we report the first recombinant expression, purification, and biochemical characterization of syNOS. The enzyme indeed acts as a bona fide NO synthase and also has strong NO dioxygenase (NOD) activity; however, it cannot utilize the general folate cofactor tetrahydrofolate as do other bacterial NOSs and instead requires tetrahydrobiopterin, like mammalian NOS. Although activation does not depend on CaM, it does strongly rely on Ca 2ϩ . Importantly, syNOS does not appear to aid in nitrogen utilization from L-Arg when recombinantly expressed in E. coli and also has minimal impact on NO detoxification.

Expression, purification, and oligomeric state of syNOS
Full-length syNOS (residues 1-1468) was co-expressed with the chaperonin GroEL/ES in E. coli BL21 DE3 cells; excess chaperonin was necessary to produce consistently well-folded and active enzyme. The yield (ϳ3 mg/liter of culture) and activity were very similar when the protein was expressed from two different vectors (pET28a or pCW-LIC). Affinity chromatography with ADP-Sepharose targeting the reductase domain was more effective as a first purification step than with Ni-NTA resin targeting the His 6 tag. On size-exclusion chromatography (SEC), syNOS eluted in two major peaks presumably corresponding to monomer and dimer (and a minor amount of aggregate) ( Fig. 2A). Nonreducing SDS-PAGE of the trailing peak produced one band corresponding to the syNOS monomer at ϳ166 kDa, whereas the second leading peak produced two bands, representing the monomer and the syNOS dimer ( Fig. 2A, inset). SEC coupled with multiangle light scattering (SEC-MALS) confirmed formation of a syNOS dimer, and its sensitivity to factors known to influence NOS dimerization (Fig. 2, B and C). The measured molecular mass of the first peak in the elution trace, 333 Ϯ 4 kDa, equates to that of a syNOS dimer, and the second peak, 191 Ϯ 6 kDa, corresponds to the monomer. A small peak was also observed at intermediate mass between those of the dimer and monomer. The sample does not appear to suffer from contamination or degradation, and thus this third peak may represent a third syNOS species in rapid  A novel nitric oxide synthase from blue-green algae oligomeric exchange. The addition of substrate or calcium did not significantly affect the population of dimer; however, H 4 B modestly decreased the amount of dimer.
To isolate contributions from the two independent heme domains, variant proteins that removed key heme-binding ligands were also expressed and purified in the same manner as the WT enzyme. The NOS ox proximal cysteine (Cys-539) was identified by alignment to NOS sequences (Fig. S1) and was substituted for alanine (C539A). The NOS g -ligating histidine was identified as His-422 with sequence alignments to globins of known structure (Fig. S2) and was also substituted for alanine (H422A) in a separate variant. Heme content of the WT and each variant, measured with the pyridine hemochrome assay, indicated that the mutations substantially reduced heme binding in the targeted domains: syNOS, 1.00 Ϯ 0.10 M heme/M protein; H422A, 0.39 Ϯ 0.08 M heme/M protein; and C539A, 0.72 Ϯ 0.07 M heme/M protein (Table S2). A syNOS variant with both heme ligand substitutions (H422A and C539A) bound very little heme (Table S2 and Fig. S4). The sum of the heme content in the H422A and C539A variants approximately equaled that of the WT; therefore, 39% of the syNOS Soret was attributed to the NOS ox heme. The concentration of NOS oxbound heme and the Soret intensity at 415 nm were used to calculate an extinction coefficient for quantifying active protein in subsequent assays.

Spectroscopic properties
Purified syNOS has a Soret band at 415 nm, which is redshifted compared with the ferric heme absorption of typical globins (ϳ405 nm) and the high-spin thiol-ligated ferric heme of NOS ox (ϳ397 nm) (Fig. 3A). The Soret for the globin heme (the C539A variant) is observed at 413 nm (Fig. 3C), similar to a flavohemoglobin from M. tuberculosis (414 nm) (31,32), and the NOS ox heme Soret at 417 nm ( Fig. 3B) is more similar that of the NOS protein from S. cellulosum (416 nm) (27). Broad ␣-bands characteristic of ferric globin hemes are observed around 540 and 580 nm in all three proteins. A single band at ϳ550 nm, expected for a NOS-type heme, is not prominent in H422A, perhaps due to remaining globin heme and protonation or dissociation of the NOS ox proximal cysteine to form an inactive P420 state (33,34). After reduction with dithionite, the Soret shifts to 425 nm, and peaks at 530 and 560 nm are observed for syNOS and C539A; this is similar to spectra of hexacoordinate neuroglobin, known to oxidize NO to nitrate (35,36). These peaks are not observed for H422A, indicating that there is little globin heme bound in this variant. NOSs are thiolate-ligated P450-type heme proteins with a characteristic ferrous-CO Soret band at ϳ450 nm (37). For WT syNOS, this species was observed as a shoulder at 444 nm corresponding to the NOS ox heme and another intense absorbance at 420 nm corresponding to the ferrous-CO NOS g heme; however, the NOS ox heme in the P420 state may also contribute to the intensity at 420 nm. As expected, H422A exhibits greater P450 Soret intensity at 444 nm compared with WT syNOS; albeit, there is still significant absorbance at 420 nm, most likely caused by the inactive P420 species. The ferrous-CO complex of C539A has no Soret peak at 444 nm, and only a band from NOS g is A novel nitric oxide synthase from blue-green algae observed. These spectral features are evident in the Fe(II)CO Ϫ Fe(III) difference spectra (Fig. 3D).
Mammalian and bacterial NOSs primarily contain five-coordinate low-spin hemes that exhibit a shift to high spin (Soret band at ϳ390 nm) upon binding L-Arg; such a Soret shift was difficult to observe in WT syNOS. However, H422A undergoes a blue shift to ϳ391 nm upon the addition of excess L-Arg (Fig.  3E). A large amount of L-Arg (500 mM) is required for complete conversion. This far exceeds the observed Michaelis constant for L-Arg (101 Ϯ 12 M; Fig. S3A) but may reflect the fact that, without the globin domain, the protein is destabilized (the activity drops by a factor of 8; see Table 1), and the NOS heme at least partially converts to the P420 state. Large amounts of L-Arg may stabilize the protein fold and heme center in a nonspecific manner so that a substrate-induced transition to a high-spin state can be observed. No Soret shift is observed when L-Arg is added to the NOS ox heme-deficient C539A variant.

Recombinant syNOS produces nitric oxide from L-arginine
NO production by full-length syNOS was first measured through the detection of its oxidized products, nitrate and nitrite, with the Griess assay. The specific activity of syNOS was 35.7 Ϯ 5 nmol/min/mg (Table 1), approximately half that of the nNOS control 64.0 Ϯ 2 nmol/min/mg, which is low compared with literature values (100 -400 nmol/min/mg (37)). The syNOS C539A variant had very little measurable activity, and the activity of syNOS H422A was attenuated by about a factor of 8 compared with WT, in keeping with the results above ( Table 1). The loss of NOS activity due to the globin substitution H422A likely reflects a general destabilization of the fulllength protein when the globin domain is disrupted.
syNOS activity requires L-arginine, H 4 B, calcium, and NADPH (Table 1). Unlike analogous mammalian NOS, syNOS activity was independent of Ca 2ϩ -calmodulin (bovine), perhaps not surprisingly given that the calmodulin binding site of mNOS is not conserved in syNOS (Fig. S1) and Synechococcus does not contain an obvious homolog of calmodulin. Remarkably, syNOS is substantially activated by calcium (Ͼ10-fold); in fact furthermore, activity was reduced in the presence of calmodulin, presumably due to competition for calcium. However, the apparent Michaelis constant for Ca 2ϩ activation is 228 Ϯ 9 M (Fig. S3), which may indicate that Ca 2ϩ serves as a proxy for another physiological factor that activates the enzyme at lower concentration. syNOS cofactor utilization also differs from other bacterial NOSs in that syNOS cannot substitute tetrahydrofolate (THF) for H 4 B. The NOS inhibitors L-NNA and L-NAA, which mimic the substrate L-Arg, completely inhibited syNOS. This is in keeping with previous observations that L-NAME inhibits syNOS in vivo (30), as L-NAME requires hydrolysis to L-NNA (typically by cellular esterases) for inhibition of NOS (38).
L-Citrulline, the byproduct of L-Arg-based NO production, was detected as the product of the syNOS reaction with analytical HPLC (Fig. 4). After derivatization with the fluorophore ortho-phthaldialdehyde (OPA), samples were applied to a reverse-phase column, and citrulline (8.58 min) was resolved from substrate L-Arg (13.34 min). The amount of citrulline detected by HPLC was roughly equivalent to the amount of NO 2 Ϫ ϩ NO 3 Ϫ measured by Griess, 0.95:1. To directly detect NO production from syNOS, NO was chelated by the spin-trap Fe 2ϩ -MGD and detected by continuouswave electron spin resonance (ESR) spectroscopy (Fig. 5). The NO-releasing small-molecule NOC-7 was used as a positive control. syNOS produced an NO signal identical to that of NOC-7. Moreover, the addition of the spin trap reacted with nearly all of the product NO and prevented conversion to NO 2 Ϫ or NO 3 Ϫ (Table S1).

NOS ox and NOS g are both directly reduced by NOS red
In mammalian NOS, NOS ox is reduced by NOS red and NADPH. To evaluate whether syNOS red can reduce syNOS ox and syNOS g independent of each other, the syNOS ox and syNOS g domains (residues 475-795 and residues 337-469, respectively) were cloned and expressed in isolation and then tested for their ability to accept electrons from NOS red (residues 856 -1468). In the case of NOS ox , the reduction experiment was carried out in the presence of CO to trap the reduced heme as a characteristic thiolate-ligated Fe(II)-CO (Soret band at 444 nm). In the presence of NADPH, NOS red produced some a Citrulline production as quantified with HPLC. b ND, no product detected.

A novel nitric oxide synthase from blue-green algae
reduction of NOS ox , as indicated by a small Soret shift to 444 nm (Fig. 6A). However, Ca 2ϩ addition substantially increased the reduced form relative to the inactive P420 form. Thus, either Ca 2ϩ facilitates NOS ox reduction by NOS red , or Ca 2ϩ attenuates the formation of the inactive P420 species through some means of NOS ox stabilization. Likewise, syNOS red and NADPH directly reduce syNOS g , as indicated by the Soret shift to 426 nm and ␣-bands at 530 and 560 nm (Fig. 6B). The reductase domains of flavoHbs usually contain binding sites for FAD and NADH, but not FMN (39). Thus, either the FAD-containing FNR domain or the flavodoxin-like FMN module of syNOS red directly reduced NOS g .

syNOS globin oxidizes NO to nitrate
Upon assay of syNOS with the Griess reaction, it was found that the enzyme produces primarily nitrate with little to no nitrite formed, despite nitrite being the initial product of NO oxidation by air. Because related flavohemoglobins detoxify NO to nitrate, syNOS g may function as a NOD. Thus, we investigated the ability of syNOS to oxidize NO generated by NOC-7 (Table 2). NO was oxidized primarily to nitrate (74%) by syNOS in an NADPH-dependent manner during the time course of the experiment (NOS activity was stopped after 30 min, which equates to approximately three NOC-7 half-lives). Removal of the ligating cysteine from NOS ox (C539A) did not decrease nitrate production, confirming NO dioxygenation by the globin heme. In contrast, removal of the proximal histidine from NOS g did reduce NO 3 Ϫ production, but not completely. The H422A variant produced more nitrate than nonenzymatic oxidation by air, suggesting that the NOS ox domain also oxidizes NO to nitrate, as has been found for mNOS (40).
The rate constants for dioxygenation by syNOS and variants were measured with an NO-specific electrode ( Table 2). NOC-7-derived NO produced a measurable current on the order of microamperes. Upon the addition of syNOS, the NO signal decayed rapidly under first-order kinetics. Consistent with results from the Griess assay, the rate constant for NO oxidation by the C539A variant (0.6 Ϯ 0.4 s Ϫ1 nmol of heme Ϫ1 ) is approximately equal to that of WT syNOS (0.6 Ϯ 0.3 s Ϫ1 nmol Ϫ1 ), and the rate constant of the H422A variant is far less than either (0.10 Ϯ 0.06 s Ϫ1 nmol Ϫ1 ). These results not only reveal that NOS g is an efficient NOD, but confirm that NOS red directly reduces both the NOS ox and NOS g heme cofactors.

Activity of syNOS in E. coli cells
It was previously reported that syNOS enabled E. coli to use L-Arg as its sole nitrogen source and that expression of syNOS increased cell density when growing on L-Arg, compared with an empty vector (EV) control (30). Under our conditions, this benefit of syNOS was not observed; instead, we found that syNOS expression conferred no significant advantage for   Ϫ ϩ NO 3 Ϫ was measured by the Griess assay. The rate constants for NOD activity were measured with an NO electrode. The rate constants were averages of at least nine measurements, and a Q test was used to remove outliers. Student's t test indicated a significant difference between the rate constants of syNOS and H422A (p Ͻ 0.001) and between H422A and C539A (p Ͻ 0.001) but no significant difference between syNOS and C539A (p Ͼ 0.7). ND, not detected. Although syNOS should not be active as a NOS when recombinantly expressed in E. coli, it should retain NOD activity, which may mitigate the effects of nitrosative stress. To test the ability of recombinant syNOS to protect E. coli against NO, syNOS and the NOD-deficient variant H422A were expressed from the tac promoter (pCW-LIC) in the WT E. coli strain BW25113 as well as the flavohemoglobin-deficient strain JW2536. The absence of the flavohemoglobin gene (⌬hmp) renders E. coli more sensitive to NO (42), and the addition of a NOD should, in theory, allow growth at higher NO concentrations. However, this was not observed (Fig. 7, B and C). The ⌬hmp strain is more sensitive to nitrosative stress induced by the addition of DETA NONOate than the WT, but the added expression of syNOS actually increased sensitivity to NO. Additionally, the growth of cells containing the NOD-deficient H422A construct was indistinguishable from that of cells con-taining WT syNOS. Thus, syNOS cannot complement a hmp mutant of E. coli, indicating that either its NOD activity in this context is low or syNOS is detrimental in some other manner.

Discussion
Full-length syNOS proved to be a challenging protein to express recombinantly in E. coli. Although soluble protein with heme and flavin cofactors bound could be produced under several conditions, many attempts at purification produced protein with little or no synthase activity. Furthermore, active and inactive syNOS share the same spectroscopic characteristics and elute similarly on SEC. Co-expression with the chaperonin GroEL/ES was essential to consistently produce active protein.
E. coli encodes GroEL/ES and several other chaperones in its genome; however, their basal level of expression was insufficient to reliably correct syNOS misfolding.
Biochemical and spectroscopic results confirm that syNOS is a genuine NOS and NOD. The NOS ox ferrous-carbonmonoxy species is observed at 444 nm, as expected of a P450 type heme. In the absence of the NOS ox heme, nitric oxide is not produced, and in the absence of the NOS g heme, the amount of NO dioxygenation is attenuated. syNOS shares the same substrate, products, and activating cofactors expected of a mammalian NOS; however, there are several unusual facets of syNOS enzymology.
A feature that distinguishes syNOS from animal NOSs is the absence of the calmodulin-binding sequence and auto-inhibitory loop. All three mammalian NOS isoforms require Ca 2ϩ -CaM,andineNOSandnNOS,thecalciumconcentrationdependence is dictated by the presence of an auto-inhibitory loop in the FMN domain (43,44). Although syNOS does not bind mammalian calmodulin (and no protein in its genome has significant similarity to calmodulin), the addition of calcium increases NO turnover 10-fold. Moreover, NOS ox reduction by NOS red is substantially enhanced by Ca 2ϩ , and thus Ca 2ϩ alone may be playing a similar role in syNOS as Ca 2ϩ -CaM does in mNOS. No other NOS has been reported to be activated by calcium independent of calmodulin. The Swiss Institute for Bioinformatics ScanProsite tool was used to search for possible EF-hand calcium-binding motifs (Prosite accession numbers PS50222 and PS00018) in syNOS, but no such sites were identified. It is possible that calcium may have a structural role, perhaps at the dimer interface, similar to zinc in mammalian NOS (45,46), or at interdomain contacts to facilitate electron transfer to the NOS ox heme. However, the measured activation constant for calcium was high, 228 Ϯ 9 M, far exceeding biological concentration ranges (hundreds of nM to 10 M (47, 48)). The mechanism of calcium binding and activation of syNOS is currently unknown; additional cofactors or proteins may be required for efficient calcium use, or Ca 2ϩ may serve as a proxy for another factor in vivo.
Cofactor utilization also differentiates syNOS from other bacterial NOSs. All bacteria produce THF, but few produce H 4 B, which differs from THF in its dihydroxypropyl side chain. All bacterial NOSs characterized to date can utilize both cofactors; thus, the preference of syNOS for H 4 B over THF agrees with its mammalian-like NOS domain architecture. Genome analysis suggests that Synechococcus sp. PCC 7335 can produce A novel nitric oxide synthase from blue-green algae H 4 B. The H 4 B biosynthetic pathway requires GTP cyclohydrolase I (GTPCH I), 6-pyruvoyl tetrahydropterin synthase (PTS), and sepiapterin reductase (SR) (49,50). Both GTPCH and PTS are found in the genome of Synechococcus sp. PCC 7335 and are highly homologous to the mammalian enzymes (Ͼ50% identity); however, no gene in Synechococcus is annotated as an SR. SR belongs to the short-chain dehydrogenase/reductase (SDR) family of oxidoreductases, a large family of proteins found in all kingdoms of life (51). Synechococcus encodes many genes belonging to the SDR family. Although none share high sequence similarity (greater than 30% identity) with mammalian SR, one gene annotated as an SDR (CDS YP_002711555.1) is immediately adjacent to a gene encoding GTPCH I. Additionally, of the 15 photosynthetic prokaryotes containing gene sequences highly similar to syNOS (Ͼ60% identity), 14 also encode an SDR adjacent to a GTPCH I. Thus, it is highly likely that Synechococcus has the enzymatic machinery to produce H 4 B.
Mammalian NOSs cannot use THF because an N-terminal ␤-extension, known as the N-terminal hook, occludes the long THF p-aminobenzoyl-glutamate side chain (52). In syNOS, this region is replaced by a short linker (18 residues) to the globin domain. This raises questions concerning not only the manner of selective H 4 B binding, but also the manner of syNOS ox dimer formation and coupling to NOS g . The NOS ox motifs located at the dimer interface in other NOS, the helical lariat and helical T (52), are conserved in syNOS; however, the close proximity of NOS g suggests that it could also play a role in stabilizing the NOS ox subunit, as well as the NOS ox dimer. In support of a tight coupling between NOS ox and NOS g , the H422A substitution in NOS g appears to also affect the stability of NOS ox and/or its affinity for L-Arg.
In addition to structural implications, the NOS g domain adds a layer of complexity to syNOS chemistry and physiology. Sequence alignments assign this domain to the globin superfamily of proteins. In particular, flavoHbs catalyze reduction of nitrite to nitric oxide and reduction of nitric oxide to nitrous oxide (53) but most commonly carry out the oxidation of NO to nitrate (39,54). syNOS catalyzes the oxidation of NO to nitrate, and this activity depends on a functional globin domain. Removal of the NOS ox heme (the C539A variant) did not hinder NO oxidation; however, the NOS ox domain was also capable of NO dioxygenation because nitrate production was still observed in the absence of the NOS g heme (H422A). NO dioxygenase activity by NOS enzymes has precedent; mammalian NOSs are also capable of NO dioxygenation, and chimeras composed of iNOS ox and nNOS red exhibit increased NOD activity (40). By producing chimeras that coupled the fast heme reduction (k r ) of nNOS with the slow NO dissociation (k d ) of iNOS, as well as the addition of a V346I substitution that further slowed the NO release (55), NO dioxygenation by mNOS was substantially accelerated (40). Note that in syNOS, the equivalent position of Val-346 in the distal heme pocket contains Ile natively, typical of bacterial NOS. The NO dioxygenase activity of syNOS has two implications: 1) syNOS is the first NOS whose final product can be nitrate and not NO, and 2) the reductase domain of syNOS can reduce both the NOS ox and NOS g directly. It is also worth noting that the spin trap com-pound intercepted nearly all NO from syNOS before it could be oxidized to NO 3 Ϫ by NOS g . Thus, NOD activity in syNOS is independent from NOS activity, with any NO produced by NOS ox released to the solution before reaction with NOS g .
Our biochemical results confirm that syNOS oxidizes L-Arg to nitrate (30); however, we are unable to replicate the finding that syNOS allows E. coli to use L-Arg as the sole nitrogen source. The growth of E. coli transformed with empty vector is indistinguishable from that transformed with syNOS. Both strains are capable of growth on L-Arg, which is not surprising given that E. coli already contains the arginine succinyltransferase pathway to derive reduced nitrogen from arginine (56,57). Although Synechococcus does not contain the arginine succinyltransferase pathway, it does contain alternate L-Arg salvage pathways that rely on enzymes such as arginase and deaminating L-amino acid oxidases (58,59). Furthermore, syNOS is not expected to be active in E. coli, as the third enzyme in the H 4 B biosynthesis pathway, sepiapterin reductase, is absent from its genome (41), and syNOS cannot substitute H 4 B with THF. The proposed role of syNOS in nitrogen assimilation is somewhat questionable given the environmental conditions in which the cyanobacteria are found. Synechococcus sp. PCC 7335 is a marine organism, where the concentrations of dissolved nitrates (tens of M (60 -62)) far exceed that of arginine (tens of nM (63)(64)(65)). Additionally, this organism is capable of nitrogen fixation (66 -68). It is unclear what biochemical merits result from expending reducing power (NADPH) to oxidize L-Arg to nitrate, only to then reductively assimilate nitrate back to ammonia.
Owing to the lack of H 4 B, syNOS ox should be inactive in E. coli. However, syNOS g and syNOS red do not depend on specialized cofactors, and thus syNOS may function as an NO dioxygenase in E. coli. However, the flavohemoglobin-deficient strain (JW2536) containing syNOS is actually more susceptible to damage by NO compared with the untransformed control. If syNOS were functioning as a flavoHb we would expect the opposite. Thus, either the protein does not exhibit NOD activity in E. coli because of interfering cellular factors, or it has other additional activities that are detrimental to growth that overcome any benefit of NO oxidation to nitrate.
The dual functionality of syNOS as both a NOS and a NOD is mysterious. As syNOS is actively expressed in growing cyanobacteria (30), some regulatory mechanism or "on/off switch" may be necessary to control NO production. NOS g may participate in such a function. Globin-based regulation of NOS has precedent in animals; eNOS binds to and stabilizes ␣-globin at the myoendothelial junction so that ␣-globin can regulate NO signaling by oxidizing NO to nitrate (69). Additionally, in vitro experiments found that full-length eNOS was able to reduce ␣-globin to the active ferrous state at a faster rate than the methemoglobin reductase cytochrome B5 reductase (69).
In S. aureus, NOS is proposed to play a role in the transition from aerobic respiration to nitrate respiration under microaerobic conditions (23). This control is mediated by the combined action of NOS and flavoHb; at high oxygen concentrations, NOS-derived NO is detoxified by flavoHb, whereas under microaerobic conditions, flavoHb cannot bind oxygen as substrate and NO is free to inhibit cytochrome oxidase, thus inhibiting oxygen reduction and favoring nitrate respiration. Although syNOS would genetically link NOS and NOD activity for such a purpose, the NOS-containing Synechococcus species does not respire nitrate (as it lacks ccNIR and associated nrf genes). Finally, a feature of syNOS activity that may provide clues as to its biological function is its reliance on calcium, which is well-known to be a tightly regulated signaling molecule in cyanobacteria (47,70). Interestingly, Ca 2ϩ is used in cyanobacteria as a signal to convey changes in nitrogen utilization. Increased levels of calcium in Synechoccocus elongatus accompany acclimation to nitrogen starvation (71), and in Anabaena sp. PCC 7120, elevated calcium levels are necessary for heterocyst differentiation (72). Thus, syNOS may be poised to respond to these signals. The high Ca 2ϩ threshold that we observe in our assays is not without note but may be a consequence of the ex vivo conditions.
In conclusion, we demonstrate that syNOS has both NOS and NOD activities. The protein represents a bacterial NOS enzyme with properties closely related to its mNOS counterparts, especially with respect to reductase coupling and cofactor utilization. However, the NOS function is not activated by Ca 2ϩ -CaM and instead appears to require only Ca 2ϩ for activity. The enzyme's reliance on H 4 B calls into question any NOS activity when recombinantly expressed in E. coli, and likewise the protein is unable to aid a flavoHb null strain in tolerating nitrosative stress. The properties of syNOS in the context of Synechococcus metabolism suggest that it is unlikely to be solely involved in nitrogen utilization from arginine and may serve a more typical NOS function in signal transduction. That said, the coupling of NOS and NOD activity in a single protein indicates a genetic link between these respective activities that is beneficial to cyanobacteria.

Materials
Synechococcus PCC 7335 (ATCC 29403) was purchased from the American Type Culture Collection. E. coli strain JW2536 was purchased from Dharmacon, and BW25113 was from the Coli Genetic Stock Center at Yale University. NOC-7 and N-(dithiocarbamoyl)-N-methyl-D-glucamine (MGD) were purchased from Santa Cruz Biotechnology, Inc. Nitrate reductase was purchase from Roche Applied Science. L-NNA, L-NAA, L-citrulline, N-hydroxyarginine, and OPA were purchased from Sigma-Aldrich. H 4 B was purchased from Cayman Chemical. 2Ј,5Ј-ADP-Sepharose 4B and Superdex 200 resins were purchased from GE Healthcare. Ni-NTA was purchased from Thermo Scientific.

Genomic DNA extraction and cloning
Genomic DNA extraction was performed following the method of Singh et al. (73). Synechococcus (50 ml) was grown for 1 week and then harvested by centrifugation at 2,000 ϫ g. Cells were resuspended in 400 l of lysis buffer (4 M urea, 0.2 M Tris, pH 7.4, 20 mM NaCl, 0.2 M EDTA) supplemented with 50 l of 20 mg/ml proteinase K. The sample was incubated for 1 h at 55°C and mixed by gentle inversion every 15 min. One ml of the extraction buffer (3% cetyltrimethylammonium bromide, 1.4 M NaCl, 20 mM EDTA, 0.1 M Tris, pH 8.0, 1% Sarkosyl) heated to 55°C was added to the sample. The resulting sample was subsequently incubated at 55°C for 1 h with gentle inversion every 10 min. Once the sample cooled to room temperature, two volumes of chloroform/isoamyl alcohol (24:1) solution were added, and the sample was mixed by inversion. The sample was centrifuged for 5 min at 10,000 ϫ g, and the upper aqueous phase was removed. Two volumes of ethanol ϩ 0.1 volume of 3 M sodium acetate (pH 5.2) were added to the aqueous phase. This solution was incubated at Ϫ20°C for 1 h and then centrifuged for 3 min at 10,000 ϫ g. The pellet was washed with 500 l of cold 70% ethanol. After evaporating the ethanol, the DNA was dissolved in 50 l of water, and the purity was assessed by the absorbance ratio 260 nm/280 nm.
Full-length syNOS (residues 1-1468, NCBI Protein database accession number WP_006458277) was cloned from bp 486,069 to 490,475 (NCBI Nucleotide database accession number NZ_DS989905) and inserted into the following vectors: 1) pET28 (Novagen) using the restriction sites NdeI and EcoRI and 2) pCW-LIC (a gift from Cheryl Arrowsmith, Addgene plasmid 26098) using BamHI and KpnI. Point mutations were constructed by primer overlap extension PCR.
Each domain of syNOS was subcloned into expression vectors by PCR. NOS g (residues 337-469), NOS ox (residues 475-795), and NOS red (residues 856 -1468) were inserted into the vector pET28 using the restriction sites NdeI and XhoI.

Protein expression and purification
E. coli BL21 DE3 cells were cotransformed with syNOS (in either pCW-LIC or pET28) or its heme domain truncations, plus the chaperonins GroEL/ES in pACYCDuet. The reductase domain truncation was transformed into BL21 DE3 without GroEL/ES. Lysogeny broth Miller was inoculated with an overnight culture and incubated at 37°C until the A 600 nm reached ϳ0.6. Protein expression was induced by the addition of 25 g/ml isopropyl ␤-D-1-thiogalactopyranoside (IPTG), 80 g/ml ␦-aminolevulinic acid, 4 g/ml hemin, 2 g/ml FAD, and 2 g/ml FMN and incubated at 17°C overnight. Hemin and ␦-aminolevulinic acid were excluded from NOS red expression, and flavins were excluded from the expression of either heme domain. Cells were harvested 18 h after induction by centrifugation at 5,000 ϫ g and then frozen at Ϫ20°C.
The cell pellet from a 2-liter growth was resuspended in 50 ml of lysis buffer (200 mM NaCl, 50 mM Tris, pH 7.5, 10% glycerol) with 174 g/ml phenylmethylsulfonyl fluoride, 1.5 g/ml pepstatin A, and 1 g/ml leupeptin. Cells were lysed on ice by sonication using a Fisher Scientific Sonic Dismembrator 550 (amplitude of 7, pulsed 2 s on, 2 s off) for a total sonication time of 5 min. Lysate was centrifuged for 1 h at 48,000 ϫ g at 4°C. For the full-length constructs, the soluble portion was applied in batch to a 2Ј,5Ј-ADP-Sepharose resin pre-equilibrated with lysis buffer and then incubated at 4°C for 2 h with gentle rocking. The resin was collected and washed with five column volumes of lysis buffer and then eluted with lysis buffer plus 5 mM NADPH. Domain truncations were lysed in the same manner, with 5 mM imidazole in the lysis buffer. After centrifugation, the soluble portion was applied to Ni-NTA, and the resin was washed with five column volumes of lysis buffer with 20 mM A novel nitric oxide synthase from blue-green algae imidazole and then eluted with lysis buffer plus 200 mM imidazole.

Multiangle light scattering
MALS was performed using a Phenomenex BioSep SEC-s3000 column (5 m, 290 Å, 300 ϫ 7.8 mm) with a Phenomenex SecurityGuard guard column, connected to an Agilent 1200 series HPLC with G1314D variable wavelength detector. Light scattering was detected using a Wyatt DAWN HELEOS-II, and differential refractive index was measured by a Wyatt Optilab T-rEX refractometer. The mobile phase contained 100 mM Tris, pH 7.5, 200 mM NaCl with either 5 mM arginine, 5 mM CaCl 2 , or 250 M H 4 B with 1.5 mM DTT. A solution of BSA monomer (5 mg/ml) was used as the standard to control for peak alignment and molecular weight calculations. Data were collected for 30 min at a flow rate of 1 ml/min at 25°C. ASTRA V software was used to analyze the molecular weight and polydispersity of each peak.

UV-visible analysis of purified syNOS
Heme content was measured using the pyridine hemochrome method. Twenty microliters of protein at 1 mg/ml were diluted to 100 l in 20% pyridine 0.2 M NaOH. An Agilent 8453 UV-visible spectrophotometer was blanked with this solution, and then ϳ0.5 mg of solid dithionite was added, and the absorbance difference at 557-573 nm was recorded (extinction coefficient 32.4 mM Ϫ1 cm Ϫ1 ). Protein concentration was measured using the Bradford assay (Bio-Rad protein assay dye) and the calculated extinction coefficient of the denatured protein (⑀ 280 nm ϭ 0.24053 M Ϫ1 cm Ϫ1 ). The protein was denatured in 4 M urea, and then unbound cofactors were removed by concentrating in a spin column and diluting in 4 M urea, which was repeated three times. The protein concentration calculated from the absorbance at 280 nm agreed with that of the Bradford assay.
As syNOS does not appear to be fully loaded with heme, the concentration of active protein was estimated based on the amount of NOS oxy heme bound in the H422A variant. The heme concentration of a syNOS sample was measured as stated previously, and the Soret intensity at 415 nm was recorded. Thirty-nine percent of the measured heme concentration was presumed to originate from the NOS oxy heme and thus represents the concentration of active protein. This concentration and the Soret intensity were used to calculate an extinction coefficient that was used to quantify active protein in subsequent assays.
The spectra for the ferrous and ferrous-carbonmonoxy syNOS were taken after the ferric enzyme was sparged with argon and reduced with ϳ0.5 mg of solid dithionite, and then carbon monoxide gas was gently bubbled into the solution. Difference spectra were constructed by subtracting the spectrum of the ferric species from that of the ferrous-carbonmonoxy species.

NOS enzymatic reaction
NOS activity was assayed following the method of Moreau et al. (74), using 250 M H 4 B, 5 mM arginine, 5 mM CaCl 2 , 1.5 mM DTT, and 1 mM NADPH in 100 mM Tris, pH 7.5, 200 mM NaCl. Where specified, THF was used at 250 M, and CaM was used at 10 g/ml. The reaction proceeded for 30 min at room temperature and was stopped by rapid heating to 60 -70°C for 5 min. Samples were centrifuged at 16,000 ϫ g for 5 min to remove any insoluble debris.

Nitrate and nitrite measurement by Griess assay
Nitrate was measured by adding 75 l of 0.2 units/ml nitrate reductase, 1 mM NADPH, and 0.1 mM FAD to 150 l of sample and then incubated for 2 h at 37°C. Nitrite was quantified similarly but in the absence of nitrate reductase. NADPH was removed by zinc acetate precipitation (75,76). To each sample, 100 l of 0.5 M zinc acetate in 50% ethanol was added, and the samples were vortexed. Subsequently, 100 l of 0.5 M sodium carbonate was added, followed by vortexing. After 5 min of centrifugation at 16,000 ϫ g, 150 l of each sample were plated in duplicate, and 25 l of 2% sulfanilamide, 1 M HCl were added, followed by 25 l of 0.2% naphthyl-ethylenediamine. The absorbance at 540 nm was recorded immediately afterward using a Biotek Synergy HT plate reader. A standard curve was prepared from nitrate and nitrite standards under the same conditions as the experimental samples. The standard curve was used to convert the absorbance at 540 nm to concentration.

HPLC product detection
The NOS reaction was performed as stated above, with the substitution of 50 mM HEPES (pH 7.5) for Tris. HPLC detection of derivatized citrulline was performed using the method of Davydov et al. (77) with the following modifications. Using an Agilent 1100 series HPLC equipped with an Agilent 1260 fluorescence detector, an Agilent Eclipse Plus reverse phase column (150 ϫ 4.6 mm; equipped with a Supelguard LC-18-DB guard column) was equilibrated with 50 mM TCA (pH 4.0) and 15% acetonitrile at 1 ml/min. The derivatization agent OPA was dissolved in methanol (8 mg/ml), and then 100 l of the OPA reagent was added to 900 l of 100 mM sodium borate (pH 10.0) and 6 l of ␤-mercaptoethanol. Then 20 l of OPA was added to 10 l of filtered sample. Following 3 min of incubation at room temperature, the mixture was injected onto the column. After injection, the concentration of acetonitrile was increased to 25% over 20 min, and fluorescence was detected at ex ϭ 360 nm and em ϭ 455 nm. Derivatized citrulline eluted at 8.58 min, followed by NOHA at 12.01 min and L-Arg at 13.34 min. A standard curve was prepared from citrulline standards under the same conditions as the experimental samples. The standard curve was used to convert the peak area to concentration.

Fe-MGD spin trap and ESR detection
The NOS reaction was performed as stated above, with the addition of 0.7 mM Fe(II) sulfate and 2.7 mM MGD (78). After reacting for 30 min at room temperature, glycerol was added to A novel nitric oxide synthase from blue-green algae the sample to 15% (w/v). The sample was immediately transferred to an X-band ESR tube and flash-frozen in liquid nitrogen. Continuous-wave ESR spectra were acquired using a Bruker Elexsys E500 CW ESR spectrometer with an ER4131VT variable temperature unit at 150 K and 9.4 GHz, with a modulation amplitude of 1.5 G and modulation frequency of 100 kHz.

NO detection by electrochemistry
NO oxidation was measured using an Innovated Instruments amiNO-2000 electrode and a CH Instruments Electrochemical Analyzer CH1630B potentiostat. The electrode was submerged in buffer (100 mM Tris, pH 7.5, 200 mM NaCl, 1 mM NADPH) while stirring, and a 0.85-V potential was applied for about 5 min to prime the electrode. Data collection was initiated, and the current was measured at a sampling interval of 2/s. The baseline current was recorded for about 5 min. Upon baseline stabilization, 10 M NOC-7 was added, and the current was allowed to plateau (about 5 min). syNOS was then added to initiate NO oxidation, and data collection continued for approximately 10 min. Signal decay was fit to a monoexponential equation using Matlab, and the rate constants were extracted.

Reduction of NOS g and NOS ox by NOS red
All solutions were sparged with argon and then degassed in an anaerobic COY chamber for 6 h. NOS ox was added to 2 molar eq of NOS red and 1 mM NADPH, with or without 5 mM CaCl 2 . NOS g was added to 2 molar eq of NOS red with 1 mM NADPH. Samples were transferred to an anaerobic cuvette, and the absorbance spectra were recorded before and after sparging with carbon monoxide.

Minimal medium growth assay
The growth assay described by Lamattina and co-workers (30) was carried out with the following modifications. E. coli BL21 (DE3) pLysS cells were transformed with either pET-28a containing syNOS or the empty vector. Ten milliliters of LB were inoculated with 100 l of an overnight culture, with 50 g/ml kanamycin and 30 g/ml chloramphenicol, and the culture was incubated at 37°C until the A 600 nm was ϳ0.3-0.4. Protein expression was induced with 0.1 mM IPTG, 500 M ␦-aminolevulinic acid, and 1 mM arginine and incubated at 37°C for 1.5 h. Cells were pelleted by centrifugation at 2,000 ϫ g for 10 min and then washed three times with 5 ml of minimal medium (5.44 g of KH 2 PO 4 , 2 g of glucose, and 6 ml of salt solution dissolved in 1 liter of distilled water, pH 7.4; the salt solution contained 10 g of MgSO 4 ⅐7H 2 O, 1.0 g of MnCl 2 ⅐4H 2 O, 0.4 g of FeSO 4 ⅐7H 2 O, and 0.1 g of CaCl 2 ⅐2H 2 O dissolved in 1 liter of distilled water). The cells were resuspended in minimal medium and supplemented with 0.1 mM IPTG, 500 M ␦-aminolevulinic acid, and 1 mM arginine in an effort to maintain the conditions specified by Lamattina and co-workers (30). The culture was diluted 1:100 in minimal medium containing either 0.2% (w/v) arginine, 0.3% NH 4 Cl, or 0.018% NH 4 Cl and incubated at 37°C. (Washing with minimal medium was strictly required; the 1:100 dilution in minimal medium contained enough nitrogen to allow for significant growth.) A 600 nm was recorded using an Agilent 8453 UV-visible spectrophotometer to monitor cell density.

NO minimum inhibitory concentration
The vector pCW-LIC containing syNOS was transformed into cells of the flavohemoglobin-deficient E. coli strain JW2536 or of the WT strain BW25113. LB containing 100 g/ml ampicillin (and 50 g/ml kanamycin for strain JW2536) was inoculated with an overnight culture (1:200 dilution) and incubated at 37°C until the A 600 nm was ϳ0.6. Cells were diluted to 10 6 cfu/ml (A 600 nm 1 ϭ 10 8 cfu/ml) in LB supplemented with antibiotics and 0, 6.25, 12.5, or 25 g/ml IPTG. DETA NONOate was diluted in 10 mM NaOH, and 20 l of each dilution was added to 180 l of each solution of cells in a 96-well plate. Final concentrations of DETA NONOate were 2.7, 1.35, 0.675, 0.338, 0.169, 0.084, 0.042, and 0 mM. Microplates were wrapped with parafilm to prevent excess evaporation and incubated at 37°C for 18 h. The A 600 nm was recorded using a Biotek Synergy HT plate reader.