Thiol catalyzed formation of NO-ferroheme regulates canonical intravascular NO signaling

Nitric oxide (NO) is an endogenously produced physiological signaling molecule that regulates blood flow and platelet activation. However, both the intracellular and intravascular diffusion of NO is severely limited by scavenging reactions with hemoglobin, myoglobin, and other hemoproteins, raising unanswered questions as to how free NO can signal in hemoprotein-rich environments, like blood and cardiomyocytes. We explored the hypothesis that NO could be stabilized as a ferrous heme-nitrosyl complex (Fe2+-NO, NO-ferroheme) either in solution within membranes or bound to albumin. Unexpectedly, we observed a rapid reaction of NO with free ferric heme (Fe3+) and a reduced thiol under physiological conditions to yield NO-ferroheme and a thiyl radical. This thiol-catalyzed reductive nitrosylation reaction occurs readily when the hemin is solubilized in lipophilic environments, such as red blood cell membranes, or bound to serum albumin. NO-ferroheme albumin is stable, even in the presence of excess oxyhemoglobin, and potently inhibits platelet activation. NO-ferroheme-albumin administered intravenously to mice dose-dependently vasodilates at low- to mid-nanomolar concentrations. In conclusion, we report the fastest rate of reductive nitrosylation observed to date to generate a NO-ferroheme molecule that resists oxidative inactivation, is soluble in cell membranes, and is transported intravascularly by albumin to promote potent vasodilation.


Introduction
Heme (iron protoporphyrin IX) is an iron-containing prosthetic group, ubiquitous in biology. In blood, heme is found primarily in red blood cell hemoglobin, along with several plasma molecules including hemopexin, apolipoproteins, and albumin. It is also soluble in membrane lipids, is present in the erythrocyte membrane, and is actively transported from cell to cell via a number of processes still being elucidated. [1][2][3][4][5] Nitric oxide (NO) is synthesized by nitric oxide synthase (NOS) and mediates canonical signaling via binding the reduced heme of soluble guanylate cyclase (sGC) with remarkably high selectivity and affinity (Kd = 5 x 10 -8 -4 x 10 -12 ). 6 Deoxygenating hemoglobin also produces NO and stimulates vasodilation via nitrite reduction, with observed NO formation as NO bound to the heme of hemoglobin (called iron nitrosyl hemoglobin) during physiological artery to vein deoxygenation. [7][8][9][10] Additionally, red blood cells (RBCs) themselves are now known to contain active endothelial NOS (eNOS). The enzymatic activity of RBC eNOS generates NO and leads to the build-up of iron nitrosyl hemoglobin and contributes an independent role in blood pressure homeostasis. 11,12 Antithetically, NO is also scavenged by hemoproteins such as oxygenated hemoglobin and myoglobin at almost diffusion-limited rates (~6-8 x 10 7 M -1 s -1 at 20°C), 13,14 and reacts with deoxygenated hemoglobin and myoglobin at similarly high reaction rates to form iron-nitrosyl globins. Both reaction processes limit NO bioavailability to activate sGC. [13][14][15] Although RBC hemoglobin has been shown to mediate NO signaling, 8,12,16,17 one of the great mysteries in the NO field over the last twenty years is how NO, generated via nitrite reduction by RBC hemoglobin or RBC eNOS, escapes the red cell and can avoid the extremely fast and irreversible NO scavenging by hemoglobin to signal vasodilation. These scavenging reactions also present a conceptual challenge for intracellular NO signaling in cardiomyocytes and skeletal muscle myocytes, which contain high concentrations of myoglobin. While we know that NOS is active in these cells, modeling reaction kinetics and relative NO and myoglobin concentrations suggest myoglobin is present at sufficient concentrations to completely scavenge and inactivate NO in these cells. 18,19 NO signaling in a hemoprotein-rich environment has been suggested to involve other species such as S-nitrosothiols [20][21][22] or formation of other nitrogen oxides such as N2O3, [23][24][25][26][27] as such species are relatively inert to autocapture and dioxygenation by hemoglobin and myoglobin. Our group and others have considered that a labile NO-bound ferrous heme or nitrosyl heme (NO-ferroheme) could serve as a possible intracellular signaling molecule and a way of protecting NO from autocapture and dioxygenation. 28,29 Interestingly, early in vitro observations by Ignarro's group indicate that nitrosyl heme can directly activate heme-free apo-sGC. 30 Further, labile heme itself, without NO, is now appreciated as a signaling molecule. [31][32][33][34][35] The details of intracellular heme trafficking are not completely understood, but molecules like glyceraldehyde-3-phosphate dehydrogenase (GAPDH) have been proposed to be involved, and such trafficking would play a role in signal amplification. 5,36,37 With the discovery of ubiquitous heme transporters found on the membranes of many cells, including red blood cells, [1][2][3]38,39 we hypothesized that a labile NOferroheme could transduce NO bioactivity in the vasculature. This hypothesis is supported by the observation that in humans, iron-nitrosylated hemoglobinmeasured selectively by both chemiluminescence and by electron paramagnetic resonance (EPR) spectroscopyforms in human blood endogenously during artery-to-vein physiological deoxygenation of blood, as nitrite is reduced to NO by deoxyhemoglobin and auto-captured as iron-nitrosylhemoglobin. 40,41 The levels of this species increase when humans breath NO gas and during infusions of sodium nitrite. 8,29 Importantly, we have previously observed unexpected artery-tovein gradients in iron-nitrosylated hemoglobin during the inhalation of NO gas in humans, suggesting the possible delivery of NO from the erythrocyte, an observation that we could never explain considering the slow off-rate of NO per se from ferrous hemoglobin. 40 Labile heme, taken to mean redox active heme weakly associated with proteins and thus readily exchangeable, is typically found in the ferric form due to its reduction potential, even in the presence of glutathione, though intercellular ferrous pools exist in high nanomolar concentrations. 1,4,42,43 NO does reversibly bind to ferric heme but with generally much lower affinity than for ferrous heme. 44 A nitrosyl-ferric heme with partial nitrosonium-ferrous heme character can then react with hydroxide to form nitrite and a ferrous heme, which rapidly binds any excess NO to form NO-ferroheme in a well-described process known as reductive nitrosylation. 45,46 The reduction step is rate-limiting and very slow, with a half-life of several minutes depending on pH, and requires two molecule of NO to form one NO-ferroheme. 45 This process is therefore an inefficient source of nitrosyl-ferrous heme in vivo as the reaction is slow 45 and involves NO binding to low affinity ferric heme. The current studies demonstrate an alternative mechanism for rapid formation of a stable iron-nitrosylated heme that is first order in NO. Small, abundant biological thiols such as glutathione or cysteine facilitate an unexpected, physiologically viable, and fast reaction between NO and ferric heme, rapidly generating a nitrosyl-ferrous heme (NO-ferroheme) and a thiyl radical. Moreover, NO-ferroheme produced in this manner is taken up readily by red blood cell membranes and serum albumin, is stable in the presence of oxyhemoglobin, and exhibits potent vasodilatory responses in vivo.

Reduced glutathione facilitates rapid formation of NO-ferroheme from NO and ferric heme in solution
To explore the reaction between ferric heme, reduced glutathione (GSH), and NO, we used a 5:1 mixture of methanol to phosphate buffered saline (PBS, pH 7.4)herein referred to as the MeOH:PBS bufferto fully solubilize free heme at concentrations amenable to UVvisible spectroscopic characterization. Addition of 1250 µM NO to a mixture containing 12.5 µM ferric heme under anerobic conditions at 23 °C results in slow (t1/2 = 340 s) formation of NOferroheme ( Figure 1A) with Q-band peaks at 543 and 568 nm, consistent with classic reductive nitrosylation reaction of two molecules of NO with ferric heme, yielding nitrite 45 followed by rapid NO binding to the resulting ferroheme (Eqs. 1 and 2-where PPIX is protoporphyrin IX).
[NO-Fe 3+ (PPIX) ↔ NO-Fe 2+ (PPIX)] Unexpectedly, addition of 125 µM glutathione with 12.5 µM ferric heme and only 125 µM NO results in rapid formation a product with the same spectroscopic signature, indicating NOferroheme formation ( Figure 1B, note markedly different time scale from 1A). This spectroscopic signature ( Figure 1C, red line) is also recapitulated in the absence of glutathione with the addition of 20-fold excess NO to 30 µM heme reduced with excess sodium dithionite ( Figure 1C, hashed line), though the glutathione-facilitated product lacks the characteristic dithionite absorbance at 315 nm. Addition of 300 µM GSH to 30 µM ferric heme in the absence of NO ( Figure 1D) resulted in no observable spectral changes, nor is the unliganded ferrous form observed (Figure 1E, hashed line), suggesting the glutathione does not coordinate or reduce heme under these conditions, respectively.  Varying the concentrations of NO or reduced glutathione under pseudo first order conditions allows for determination of the dependence of this reaction on either GSH or NO, yielding calculated second order rate constants of 7000 M -1 s -1 and 2300 M -1 s -1 , respectively ( Figure 1F). All spectra showed isosbestic points at 503 and 594 nm (e.g., Figure 1B, top), and traces (e.g., Figure 1B, bottom) fit well to single exponential kinetic analyses, together indicating a two-species rate limiting reaction without significant formation of intermediates to generate NO-ferroheme. The speed of this reaction was unexpected and to our knowledge is the fastest rate of reductive nitrosylation observed to date for a heme protein.
As discussed further below, the identity and stoichiometry of the reaction products supports a mechanism involving thiol-catalyzed reductive nitrosylation of ferric heme to form NO-ferroheme and a thiyl radical, which subsequently reacts with excess NO to form secondary S-nitrosoglutathione (GSNO). These nitrosyl species, NO-ferroheme and GSNO, were identified and quantified by reductive chemiluminescence with NO detection in the chemiluminescent nitric oxide analyzer ( Figure 1G, 'Untreated' bars). Specific detection of NO derived from NOferroheme (no NO release from GSNO) was achieved through chemical oxidation with potassium ferricyanide, which oxidizes ferrous nitrosyl species and liberates NO without degrading GSNO. 47 Figure 1A). Each injection corresponded to roughly 1 equivalent of NO (17.7 ± 2.2 µM) per equivalent of generated NO-ferroheme (Supplementary Figure 1C). Specific detection of NO derived from GSNO (no NO release from NO-ferroheme) was achieved using the well-described copper/cysteine (Cu/Cys, 2C) method for S-nitrosothiol detection (Supplementary Figure 1B), 48 Figure 1D). We anticipate a slightly substoichiometric yield of GSNO-derived NO due to oxidized glutathione (GSSG) formation and/or GSNO degradation, which occurs in the presence of the excess reduced thiol. 49 Taken together, these data suggest the formation of NO-ferroheme and GSNO in roughly a 1:1 stoichiometry.
A few possible reaction mechanisms account for the formation of NO-ferroheme and Second, generally, glutathione may undergo a one-electron oxidation to reduce a ferric heme nitrosyl to ferrous heme nitrosyl, followed by rapid reaction thiyl radical with an equivalent of NO (Eqs. 4 and 5), though this electron transfer may occur via an inner sphere or outer sphere electron transfer mechanism (vide infra). In the case of inner sphere electron transfer, the thiol may initially bind free ferriheme or pre-formed NO-ferriheme, though this order of addition cannot readily be determined from the reaction kinetics or stoichiometry.
These two mechanismstraditional reductive nitrosylation vs. thiol-catalyzed reductive nitrosylationyield the same products, though not necessarily in equivalent amounts. The traditional reductive nitrosylation mechanism does not yield a radical intermediate and, as it is base driven, should also result in a considerable amount of nitrite and a sub-stoichiometric amount of GSNO. Moreover, as described above, the observed reaction kinetics of NO-ferroheme formation are much faster than those observed for traditional reductive nitrosylation, suggesting that a thiyl radical-generating mechanism is operative.
To determine categorically that the thiyl-generating mechanism predominates in this system, we directly probed radical formation by reacting ferric heme, GSH, and NO in the presence of the radical trap 5,5-dimethyl-1-pyrroline-N-oxide (DMPO). It is worth noting that while NO itself is a radical, it does not react with DMPO. 50 We quantified the formation of NOferroheme and GSNO under these conditions from 20 µM hemin using the chemiluminescence

NO-ferroheme formation in hemoglobin-depleted red blood cell membrane white ghosts
Hemin (ferric heme) is hydrophobic and must be solubilized in organic solvents like methanol. In vivo, heme is solubilized in cell lipid bilayers and is particularly abundant in the erythrocyte membrane, though this has predominantly been investigated in the context of free heme toxicity and lipid peroxidation. 51,52 To demonstrate that thiol catalyzed reductive nitrosylation occurs in a physiologically relevant context, we employed suspended red blood cell "white ghost" membranes, prepared as previously described, 53 to solubilize heme ( Figure   2A). 54 NO-ferroheme formation in membrane suspensions was measured using a UV-Vis spectrophotometer with an integrating sphere detector. Consistent with observations in MeOH:PBS solution, a slow reaction characteristic of reductive nitrosylation was observed in ghost membrane suspensions bearing ferric heme and NO in the absence of thiol ( Figure 2B).
However, the reaction rate of NO-ferroheme formation was greatly enhanced by the addition of a stoichiometric excess of GSH to ferric heme, resulting in formation of a product spectroscopically analogous to that observed in the MeOH:PBS system ( Figure 2C, here 250 µM GSH). Importantly, the use of RBC membrane ghosts enabled characterization of the NOferroheme product by low temperature electron paramagnetic resonance (EPR) spectroscopy.
Both reactions (in the presence and absence glutathione) give rise to a rhombic EPR spectrum characteristic of a five-coordinate NO-ferroheme species with g-values at 2.081, 2.054, and 2.012 for gmax , gmid , and gmin , respectively, and a three-line hyperfine splitting centered on gmin ( NO A = 47 MHz,, Figure 2D). 55,56 As described earlier, DMPO reacts with a glutathionyl radical to form a stable adduct that is detectable by EPR spectroscopy. To directly probe formation of the glutathionyl radical under physiologically relevant conditions, we generated NO-ferroheme solubilized in red blood cell ghosts in the presence and absence of excess DMPO. While the heme nitrosyl EPR signal intensity remains proportional to the square root of the microwave power, the signal derived from the DMPO-glutathionyl adduct saturates at higher power. 57 Addition of ferric heme, excess NO, and excess glutathione in membranes without DMPO results in the typical ferrous nitrosyl spectrum overlapping at all three power levels after normalization to the square root of these powers ( Figure 2E, top). In the same reaction with the addition of 50 mM DMPO, a signal characteristic of the DMPO-glutathionyl spin adduct, centered at 3350 G, saturates as power increases ( Figure 2E, bottom). We quantified GSNO quantification using Cu/Cys based reductive chemiluminescence in the NO analyzer with and without DMPO in this aqueous RBC membrane ghosts system; GSNO production was significantly inhibited by DMPO ( Figure 2F). The spectra overlap after normalization (dividing raw intensity by the square root of the power). Bottom, This same reaction was carried out in the presence of 50 mM DMPO and EPR spectra recorded at the same three powers. Here, however, an organic radical saturates at higher powers, demonstrating formation of the DMPOglutathionyl radical. EPR spectra were collected at 110 K. F) Chemiluminescent measurement of GSNO using the 2C assay of the reactions in Figure 1G show blunting of S-nitrosothiol formation in the presence of DMPO. All reactions were carried out under anaerobic conditions.

Reaction of NO, heme, and glutathione with albumin forms albumin-bound nitrosyl heme under physiological conditions
As in the RBC ghost membrane system, glutathione-catalyzed NO-ferroheme formation occurred in the presence of plasma serum albumin. Albumin is an abundant plasma protein that can solubilize a number of hydrophobic compounds. In particular, heme binding to albumin is well-characterized, 58,59 and heme-nitrosyl species have been observed in human plasma, especially after the inhalation of NO gas. 60 We observe thiol catalyzed NO-ferroheme formation using serum albumin to solubilize ferric heme and NO-ferroheme product ( Figure 3A). Indeed, analogous to both our MeOH:PBS system and red blood cell ghosts system, glutathione accelerates the rate of NO-ferroheme formation and NO in the presence of serum albumin ( Figure 3B). Here, we solubilized ferric heme in 500 µM serum albumin, approximating the mammalian plasma serum albumin concentration, 61   The experimental data are shown together with the theoretical simulation used to obtain g-values as described in the methods. D) Transfer of NO-ferroheme from membranes to serum albumin. 25 µM ferric heme, 50 µM glutathione, and 50 µM NO were added to red blood cell membrane ghosts under deaerated conditions. Addition of 75 µM serum albumin results in the hashed black line spectrum, which still exhibits typical light scattering due to turbidity from insoluble membranes. To confirm that the NO-ferroheme was transferred from membranes to albumin, the mixture was centrifuged at 30,000g for 2 hours resulting in complete membrane precipitation and pelleting, leaving behind NO-ferroheme in the albumin (red spectrum). E) Transfer of NO-ferroheme from albumin to apo-myoglobin in 1:1 ratio of heme to apomyoglobin using a 2 mm cuvette. Nitrosyl-myoglobin is formed given the distinct spectral absorbances at the indicated wavelengths, with isosbestic points from the NO-ferroheme in albumin (red spectrum) over time, indicating direct NO-ferroheme transfer. Inset: formation of nitrosyl-myoglobin over time following the Soret formation at 421 nm at 23°C with a half-life of ~12 min under these conditions.

Transfer of NO-ferroheme from membranes to albumin and albumin to apo-myoglobin
As labile heme (and thus potentially NO-ferroheme) can be found in membranes, we qualitatively characterized the transfer of NO-ferroheme formed in our RBC membrane ghosts to serum albumin, the most abundant protein in blood. Ferric heme (25 µM), 50 µM glutathione, and 50 µM NO were added to red blood cell membrane ghosts under anaerobic conditions at 37°C (Figure 3D). This spectrum exhibits substantial light scattering due to the relatively large membrane ghosts. The absorption spectrum after adding 75 µM serum albumin (hashed black line) likewise shows considerable light scattering due to the presence of the membrane ghosts. To confirm that NO-ferroheme was transferred from the membranes to the albumin, the mixture was centrifuged at 30,000 x g for 2 hours resulting in complete membrane precipitation and pelleting, and leaving behind NO-ferroheme solubilized in the albumin (solid red line) accompanied by loss of the membrane-ghost dependent turbidity.
One putative biologically relevant mechanism of action of NO-ferroheme presupposes not only an ability for NO-ferroheme to be ferried from cell membranes and hydrophobic spaces to serum albumin in the plasma for protection and transport, but to then further transfer from albumin to heme binding protein targets. To assess and model such transfer, NOferroheme in albumin was added to an equivalent of apo-myoglobin in an anaerobic solution at room temperature ( Figure 3E). Though somewhat slow under these conditions with a half-life of roughly 12 min ( Figure 3E, inset), these data clearly show NO-ferroheme transfer from the NO-ferroheme in albumin (red spectrum) to directly form nitrosyl-myoglobin with signature absorbances of 421, 549, and 581 nm. 62 Critically, the isosbestic points at 397, 522, and 593 nm indicate direct transfer without formation of any other species or intermediates.

Stability of NO-ferroheme in oxyhemoglobin
To perform a signaling function, NO-ferroheme must be stable to oxidation and premature NO release under physiological conditions. To measure the stability of NOferroheme in albumin under oxygenated conditions, we added aliquots of albumin-solubilized NO-ferroheme to solutions of oxyhemoglobin and followed the formation of methemoglobin spectroscopically (Supplementary Figure 3). While free NO is rapidly oxidized by oxyhemoglobin to form nitrate (NO dioxygenation reaction), 13

In vitro signaling properties of NO-ferroheme with albumin
NO is a well-established inhibitor of platelet reactivity via activation of sGC. 66,67 We hyothesize that generation of NO-ferroheme acts as a physiologically relevant species that shields NO from scavenging until in the proximity of sGC; sGC must then be stimulated, either via NO release or directly binding to apo-sGC, [68][69][70] although our experiments incubating NOferroheme with albumin and oxyhemoglobin indicate NO release is slow. To establish NOferroheme signaling in vitro, we isolated platelet-rich plasma (PRP) and triggered platelet activation with 2 µM adenosine diphosphate (ADP, Figure 4). 71 In the absence of glutathione, addition of 2.5 µM ferric heme and 2 µM NO in 7.5 µM albumin did not significantly inhibit platelet activation (Figure 4A, triangles). However, when solutions containing these same reagents with 25 µM glutathione were added to the activated platelets, significant inhibition was observed (Figure 4A, hexagons), suggesting that NO-ferroheme indeed inhibits platelet activity. Glutathione by itself, ferric heme alone, or heme with albumin or with albumin and glutathione have no effect on activated platelets (Supplementary Figure 4).
S-nitrosoglutathione (GSNO) has been shown to inhibit platelets in vitro. 72,73 As GSNO is generated by the reaction, we employed an alternate method to generate NO-ferroheme without GSNO by first reducing hemin with 1.6 fold excess sodium dithionite and then adding an equivalent of NO to simply bind the pre-formed ferrous heme at a final concentration of 2 µM ( Figure 4B). Ferrous heme, generated by chemical reduction with dithionite in the absence of NO did not inhibit platelet activation; in fact, it seemed to activate them slightly more than ADP alone (Figure 4B, upside down triangles). However, NO-ferroheme, generated using dithionite and not glutathione, significantly inhibited platelet activation in the presence of absence of glutathione ( Figure 4B, diamonds and hexagons). We note that addition of glutathione to the dithionite-prepared NO-ferroheme albumin does not generate GSNO via the 2C assay (data not shown). Taken together, these in vitro data suggest that NO-ferroheme in albumin functions as a canonical NO signaling molecule to directly activate sGC in activated platelets. Figure 4 Effects of NO-ferroheme in albumin on platelet activation. Platelet-rich plasma (PRP) was diluted sevenfold with anaerobic PBS. Addition of 2 µM adenosine diphosphate (ADP) to platelets stimulates activation (squares) and was added to all experiments except the vehicle controls (circles) for 10 min. ADP addition was 5 minutes after adding heme albumin samples to PRP. Each symbol within a bar represents data from a different blood donor on a different day. A) 2.5 µM heme and 2 µM NO in 7.5 µM albumin were reacted for 35 min resulting in traditional reductive nitrosylation and thus slow and incomplete NO-ferroheme formation. Upon addition to activated platelets, little abrogation of activation was observed (triangles). However, addition of 25 µM glutathione to the reaction results in thiol catalyzed NO-ferroheme formation; significant inhibition of platelet activation was observed upon addition of this sample to platelets (hexagons, p = 0.0002), and significantly more than without GSH in the reaction mixture (triangles, p = 0.0031). B) Using the same concentrations as A, NO-ferroheme albumin was synthesized using 40 µM sodium dithionite (Na2S2O4). Dithionite did not inhibit platelet activation alone (upside down triangles). However, NO-ferroheme generated in this manner significantly inhibited platelets with and without glutathione present (diamonds, p = 0.004, and hexagons, p = 0.0082, respectively). Statistics were completed using Fisher's least squared difference test. Platelet experiments were performed at 37 °C

In vivo signaling properties of nitrosylated-heme with albumin
To assess whether NO-ferroheme in albumin elicits a vasodilatory response in vivo, we Similarly, using freshly prepared GSNO, 74 we made fresh controls for the maximum amount of GSNO theoretically generated by this synthesis (300 µM GSNO, 3 mM glutathione, and 500 µM albumin). Following an analogous procedure, GSNO also did not elicit a large change in MAP    Figure 5A, under 10% oxygenated breathing conditions and administered L-NAME, a mouse was administered NO-ferroheme albumin prepared by hemin reduction by 10% excess sodium dithionite (teal line, n=5). By itself, this did not have the same vasodilatory activity observed in NO-ferroheme albumin solution prepared via glutathione catalyzed reductive nitrosylation (red line, n=9). However, addition of the 3 mM glutathione to this solution restored such activity (black line, n=5). These regimens were administered in doses 10 minutes apart giving estimated blood concentration in the mouse of 7.5 nM, 75 nM, 0.75 µM, and 7.5 µM of each preparation. B) ΔMAPmax for each injected species described in A at a given concentration. Statistics were completed using Tukey's multiple comparisons test; only significant interactions are shown.

Discussion
The presented studies identify a kinetically fast chemical reaction pathway to form a Mechanistically, glutathione serves as a reducing partner for the ferric heme, and accelerates the formation of the ferrous nitrosyl. 26 However, glutathione does not directly reduce the labile ferric heme, 43 suggesting that NO association is required for glutathionecatalyzed heme reduction. Kinetic fits under pseudo-first order conditions (where NO and glutathione are at least in fivefold excess compared to heme) yield single exponentials, and the spectral changes exhibit clean isosbestic points with no intermediate species observed. Substoichiometric equivalents of thiol also accelerate the reaction (data not shown). Under biological conditions, thiols oxidized in the process of generating NO-ferroheme would be rapidly re-reduced by cellular reductases, making the process catalytic. Together these observations suggest a sequential mechanism where NO binds ferriheme, forming a ferric nitrosyl that is subsequently reduced by glutathione (Scheme 1), though the possibility that glutathione binds first cannot be excluded. Such a mechanism has been considered before, 26,75,76 however, thiyl radical formation was never confirmed and rates not quantified. In our study, one electron reduction of the ferric heme-NO by glutathione and concomitant formation of a transient thiyl radical was confirmed using excess of the radical spin-trap DMPO.
Taken together, an electron transfer mechanism is implicated (Eqs. 4 and 5) rather than generation of an electrophilic nitrosonium that is typical of reductive nitrosylation (Eqs. 3 and 2), though the one-electron reduction may proceed via either an inner sphere (I.S.) or an outer sphere (O.S.) reaction (Scheme 1).

Scheme 1
The formation of NO-ferroheme in red cell membranes is consistent with the hydrophobic nature of labile heme. One would expect that if NO-ferroheme was formed in red blood cells, it would occur on or in lipophilic compartments such the plasma membrane.
Further, red blood cells are known to release heme into plasma, 77 though the mechanisms of heme cellular export are not as well established. However, a recently discovered heme transport protein, feline leukemia virus subgroup C receptor 1 isoform a (FLVCR1a), has been found in the plasma membrane of red blood cells. 1,39,77,78 As the biochemical mechanism is not fully established, 1 FLVCR1a, or other yet to be established transporters, may provide a means for cellular export of NO-ferroheme into the plasma, where it can be picked up by serum albumin as we have observed in vitro; albumin binds significantly more labile heme overall than the heme-chelating specific plasma protein, hemopexin. 79,80 Another known heme exporter found in erythrocytes and other cells is ATP-binding cassette super-family G member 2 (ABCG2). This transporter is suspected to export and transfer heme directly to serum albumin, 2,42,81,82 and therefore may act as a functional transporter for NO-ferroheme as well. At any given time there is 1.5 -50 µM heme-albumin found in human circulation. 59 Some of this may be NO-ferroheme, and specifically detecting NO-ferroheme in the vasculature requires further study. Alternatively, NO-ferroheme may also be readily formed in albumin itself in the presence of glutathione and NO from ferric heme. A well-defined heme binding pocket has been characterized in serum albumin, which axially coordinates heme via Tyr161, though heme binds non-specifically in other lipophilic areas as well. 58,83,84 However, the EPR signal of NO-ferroheme in albumin in our studies suggests a single five-coordinate iron site with only axial NO bound. 55,56 Combined with the observed transfer of NO-ferroheme from albumin to apomyoglobin to directly generate nitrosyl-myoglobin, these results suggest that NO-ferroheme can be generated and shuttled, both from membranes to albumin and albumin to heme-free hemoproteins.
The observed vasodilation and platelet inhibition suggest NO derived from NOferroheme, or more probably NO-ferroheme itself, directly binds and activates sGC in smooth muscle and platelets, respectively. We know that NO-ferroheme in albumin is protected against Injection of an equivalent of dissolved NO under otherwise analogous conditions exhibited little response related to red blood cell hemoglobin scavenging. As an equivalent of GSNO should be produced in our synthesis of NO-ferroheme via our determined mechanism, barring other protein/small molecule reactions to consume the generated thiyl radical, we tested purified GSNO in this concentration regime. Similar to dissolved NO, GSNO under these conditions had no effect, consistent with prior published data that GSNO does not significantly effect mean arterial pressure in rodents at concentrations less than 300 µM. 89 Interestingly, NO-ferroheme produced without thiol using sodium dithionite to reduce the heme showed a weaker vasorelaxation response at all tested concentrations in vivo. This observation contrasts with our experiments in vitro inhibiting platelet activation, where both the thiol catalyzed NO-ferroheme and the dithionite-generated NO-ferroheme without glutathione were equally effective inhibitors. However, addition of the thiol after NOferroheme formation via dithionite rescued the in vivo vasodilatory effect. Taken together, these data suggest that addition of thiol either stabilizes the NO-ferroheme albumin or facilitates transport of albumin to smooth muscle. The thiol may affect the NOferroheme/albumin equilibrium by altering the redox environment or may otherwise trigger mixed disulfide formation effecting protein dynamics and perhaps cellular import. Alternatively, the possibility cannot be excluded that NO-ferroheme with GSH axially ligated is triggering vasorelaxation, although we observe no spectroscopic evidence of such a species in vitro.
A possible candidate for import of NO-ferroheme bound albumin is the transferrin receptor (CD71), which is known to specifically endocytose heme-albumin into cells and use it as an iron/heme source in several cell types, promoting proliferation in the absence of transferrin. 79 Moreover, CD71 is expressed on the surface of many cell types, including vascular smooth muscle cells, endothelial cells, and cardiomyocytes. 90,91 Albumin is also known to cross endothelial cells in a processes called transcytosis via albumin activation of endothelial surface glycoprotein gp60, providing another potential route for albumin-solubilized NO-ferroheme to reach smooth muscles cells and other tissues. 92 NO could be released from NO-ferroheme bound albumin at the primary binding site once imported; this site is allosterically controlled by binding of other macrocycles at 'Sudlow's sites,' specifically pushing His146 to replace the NO to make a hexacoordinate heme with Tyr161. 59,85 However, as originally hinted at by the Ignarro group, and consistent with observed NO-ferroheme transfer from membrane to albumin and subsequently from albumin to apo-myoglobin, NO-ferroheme itself, once released in the cytosol, may directly activate heme-deficient apo-sGC. 28,30 The details of NO/NOferroheme release of release to stimulate vasorelaxation are the subject of our future studies.

Conclusions
In this manuscript, we have shown previously unexplored chemistry in which physiologically abundant reduced glutathione facilitates the rapid reductive nitrosylation of ferric heme, resulting in the formation of NO-ferroheme and a thiyl radical, which in the presence of excess NO yields a S-nitrosothiol. We suspect NO-ferroheme represents a novel signaling molecule that helps shield NO from autocapture and dioxygenation on the way to signaling sGC. Indeed, even low-to-mid nanomolar concentrations in circulation of NO-ferroheme in serum albumin engenders a strong vasodilatory response. While numerous questions remain regarding transport as well as cellular import and export, formation of NOferroheme may provide an answer in the NO signaling field as to how relatively ephemeral NO survives highly reactive species such as oxyhemoglobin in the red cell and oxymyoglobin in cardiac and skeletal muscle cells.

Materials and solution preparation
All

UV-Visible Spectroscopy and Kinetics
All individual spectra and pseudo-first order kinetics were performed using a

Preparation of NO-ferroheme in methanol and PBS solution
Hemin stock (5 mM) was prepared regularly in 20 mN anaerobic NaOH solution, sealed in a septum capped vial. The concentration of hemin was precisely determined by absorption spectroscopy using the extinction coefficient at 385 nm of 58.4 mM -1 cm -1 . 93

Preparation of NO-ferroheme in albumin solution (heme-nitrosyl albumin)
Stocks of hemin, glutathione, and NO were prepared as described above. An albumin stock (0.5 -1 mM) was made in phosphate buffered saline (PBS) and the concentration verified by absorption spectroscopy using an extinction coefficient of 43.8 mM -1 cm -1 at 280 nm. Stock solutions were used to produce mixtures on a Schlenk line and transferred using Hamilton syringes with concentrations given in the results section. Solutions were made directly or indirectly and transferred, with no differences between products of each preparation. The direct method involved adding ferric heme solution and glutathione directly to albumin in an argon or nitrogen filled septum-sealed vial, followed by NO addition. Within seconds, the solution changes from a brownish color to a deep red color, with no trace of turbidity. Solutions were used after five minutes. The indirect method involved NO addition to heme and GSH in PBS before then adding albumin after five minutes. The NO-ferroheme was then incubated with albumin for thirty minutes.

Preparation of red blood cell membrane ghosts
Red cells were separated from the whole blood by sedimentation via centrifugation at 1000 g for 10 minutes. These packed red cells were then hemolyzed using 5 mM phosphate buffer, pH 8 (4.674 mM Na2HPO4 + 0.326 mM NaH2PO4) in the volume ratio of 1:40 for an hour. The mixture was spun at 35,000 g for 30 minutes at 20°C and washed with the buffer repeatedly until the white ghosts were produced. The absence of hemoglobin was confirmed by taking the absorption of the ghosts in the range of 300 -700 nm.

EPR spectroscopy
Electron paramagnetic resonance (EPR) spectra were recorded on a Bruker EMX spectrometer operating at 9.4 GHz, 5-G modulation, 10 milliwatt power, 328-ms time constant, and 164-s scan over 600 G at 110 K or 10 K as described previously. 95 Two scans were taken and averaged for each sample. Concentrations were calculated after double integration using heme-nitrosyl basis spectra with known concentrations. For power saturation measurements, spectra were taken with 0.1, 1, and 10 mW power (33, 23, 13 decibels

Platelet activation
Samples were prepared using heme (25 µM), NO (20 µM), albumin (75 µM) with and without GSH (250 µM) under normoxic or anoxic conditions. Fresh blood was drawn into sodium citrate tubes with the first tube being discarded and platelet rich plasma (PRP) collected after sedimentation of the red blood cells. Platelet activation was measured as described previously using platelet agonist ADP. 97 First, PRP was diluted 1:7 with oxygenated or deoxygenated PBS buffer, then albumin samples were mixed with this diluted PRP in 1:10 ratio. After 5 minutes of incubation at 37˚C, 2 μM ADP was added to all test samples except negative vehicle controls, followed by 10 additional minutes incubation. 20 μL of each sample was subsequently added to PAC-1 FITC (labels activated platelets) and CD61 (labels all platelets) antibodies and incubated at room temperature in the dark for 20 minutes and then fixed in 1% buffered formaldehyde.
Platelets were sorted using a BD FACSCalibur Analyzer. Gating strategy was based on side and forward scattering compared to red blood cells and then also for activated platelets. Platelets were labelled along the x-axis and activation on the y-axis. Activation was taken as the upper right quadrant divided by the sum of the two right quadrants (Supplementary Figure 5).

In vivo vasodilation experiments
Male C57BL/6 mice (the Jackson Laboratory), age between 12 to 14 weeks, were anesthetized by isoflurane (2.5%). Body temperature was maintained at 37°C by a heating pad connected to heat pump (Androit). Catheters were implanted for carotid artery and jugular vein. Arterial blood pressure was continuously recorded through a blood pressure transducer (MLT699, ADInstrument) connected to an 8-channel Powerlab (ADInstrument). A tracheal tube was introduced and connected to a rodent ventilator (Model 849, MidiVent). Upon completion, in some animals, the fraction of inspiratory O2 was decreased from 21% to 10% to induce hypoxia.
75µL of L-NAME was administered intravenously in a bolus at a dose of 10 mg/kg bodyweight.
10 minutes after L-NAME injection, animals were injected with 50 µL of NO-ferroheme or normal saline (NS) for 4 times at 10-minute intervals. An NO-ferroheme in albumin stock was prepared via a concentrated version of the direct method as described above: 300 µM ferric heme was mixed anaerobically on a Schlenk line with 3 mM GSH and 600 µM NO in 500 µM albumin using air-tight Hamilton syringes. This stock was diluted further in air with albumin solution before injection by 1000-, 100-, and 10-fold. Mean arterial pressure (MAP) values were obtained by processed the recorded arterial pressure data with LabChart 7.3.8 (ADInstruments).