The close proximity of the red blood cells (RBCs) to the site of NO^• production and the fast consumption of NO^• by oxyHb and deoxyHb observed in vitro [10,11] suggest that most NO^• generated by eNOS will be scavenged by the blood. In particular, as the values of the rate constants for the reactions of NO^• with oxyHb and with deoxyHb are in the same order of magnitude, one would expect oxyHb, present in about 99% in oxygenated blood, to be the main target for NO^•. Under physiological conditions, the reaction between NO^• and O₂ is too slow [15] to play a significant role. Indeed, the half-life of 100 nM NO^• in the presence of about 150 μM O₂ (the concentration found in arterial blood) is over two hours (t_1/2=1/[8×10⁶ M⁻² s⁻¹×100×10⁻⁹ M $\times 150\times {10}^{-6}\mathrm{M}]=2.3$ h). In contrast, the half-life of the reaction of 20 mM oxyHb (heme-based concentration present in the RBCs) with physiological amounts of NO^• (nM) is less than 1 μs (t_1/2=ln2/(8.9×10⁷ M⁻¹ s${}^{-1}\times 20\times {10}^{-3}\mathrm{M})=0.4$ μs, calculated with the value of the rate constant measured at pH 7.0 and 20 °C [11]). Thus, in principle most NO^• should be irreversibly converted to nitrate (Fig. 1). This hypothesis is supported by the strong correlation found, in venous blood of volunteers inhaling 80 ppm NO^• for two hours, between the time-dependent increase of the nitrate and the metHb concentrations [16].

However, under basal physiological conditions the biological activity of NO^• in the blood vessels is not completely lost. A substantial amount of NO^• reaches the smooth muscle cells, activates guanylyl cyclase, and thus leads to dilation of the blood vessels. This apparent paradox has prompted to a series of studies aimed to understand the detailed mechanism of the in vivo reaction between circulating hemoglobin and NO^•.

A number of theoretical and experimental studies have been carried out to identify the factors that determine the distance NO^• can diffuse from the site of its production in the endothelial cells [17–19]. Lancaster and coworkers showed that the rate of the reaction between NO^• and oxyHb is approximately 650 times slower when hemoglobin in encapsulated within the RBCs [20]. This result was explained with the assumption that the limiting step for the reaction of NO^• with hemoglobin within the RBCs is the diffusion of NO^• from the bulk solution to the RBCs membrane [20]. In contrast, to rationalize similar results Liao and coworkers proposed that it is the membrane that offers the main resistance to NO^• diffusion into the RBCs [21]. Nevertheless, a recent study has shown that NO^• uptake is primarily limited by extracellular diffusion resistance [22], conceptually represented by an undisturbed layer around the RBCs, where the NO^•-concentration is much smaller than the bulk concentration. By taking into account these results, one can explain the significant increase in blood pressure observed when extracellular hemoglobin-based blood substitutes are administered [23]: free hemoglobin, that is Hb not encapsulated within the RBCs, reacts extremely rapidly with NO^•, eliminates completely its bioactivity, and thus causes the blood vessels to contract.

A further factor that extends the lifetime of NO^• within the blood vessels is a RBCs-free (or RBCs-depleted) region created by intravascular flow close to the vessel wall [24]. This layer reduces the amount of NO^• scavenged by the RBCs as it increases the distance NO^• has to diffuse to reach them.

An alternative, conceptually different pathway has recently been proposed to explain how the bioactivity of NO^• is maintained within the blood vessels in the presence of high concentrations of hemoglobin (Fig. 2). Stamler and coworkers suggested that, despite the kinetic arguments discussed above, in vivo NO^• preferentially binds to the minor fraction of deoxygenated iron centers (about 1%) to yield partially nitrosylated hemoglobin, Hb(Fe^IINO)(FeO₂)₃ (reaction (5)) [25]. Difference absorption and EPR spectroscopy were utilized to show that addition of submicromolar concentrations of NO^• to oxyHb (in 10 mM phosphate buffer) leads primarily to the generation of nitrosyl Hb [25]. To account for these results, it must be assumed that binding of NO^• to partially oxygenated hemoglobin is cooperative. Specifically, NO^• should bind R-state Hb(Fe^II)(FeO₂)₃ at least 100 times faster than T-state Hb(Fe^II)₄ or Hb(Fe^II)(Fe^IINO)₃ [26,27].

In a subsequent step, the ‘NO group’ has been proposed to be transferred intramolecularly from HbFe^IINO to the conserved cysteine residue β93, and to form the so-called S-nitroso-hemoglobin (SNOHb) [28]. It is presumed that this reaction proceeds via oxidation of the iron center by O₂ to produce HbFe^IIINO ↔ HbFe^II(NO⁺), which can subsequently undergo a formal NO⁺-transfer to Cysβ93 (reaction (6)) [27,29]. It has been further suggested that, when the RBCs reach O₂-depleted tissues, partial deoxygenation of hemoglobin facilitates the release of the ‘NO group’ from SNOHb. A thermodynamic argumentation has been applied to support this reaction step: crystal structures of SNOHb [30] have shown that the S-nitrosocysteine is significantly more stable in oxyHb (R-state), as the S-atom is directed inward and is not exposed to the solvent. Upon dissociation of O₂, when the RBCs reach hypoxic tissues, hemoglobin is converted to the T-state (deoxyHb). This transition is coupled with a reorientation of the S-nitrosocysteine toward the solvent. The exposed SNO-group can thus readily undergo transnitrosation reactions with other thiols within the RBCs. Initially, it was proposed that SNOHb reacts with GSH, present in millimolar concentration within the RBCs, to generate GSNO [31]. However, analysis of the S-nitrosated components of the RBCs suggested that SNOHb rather undergoes a transnitrosation reaction with cysteine residues in the Hb-binding cytoplasmic domain of the band-3 anion-exchange membrane protein [32]. Finally, to reestablish the bioactivity of NO^•, the ‘NO group’ should be transported out of the RBCs (possibly with the involvement of extracellular GSH [33]), reduced to regenerate NO^•, which should diffuse into the smooth muscle cells, bind and activate guanylyl cyclase. The mechanism of these last reaction steps is still unidentified [32,34].

One of the central issues of the hypothetical mechanism presented above is that binding of NO^• to the heme of partially oxygenated hemoglobin (R-state) is cooperative and depends on the oxygen saturation of hemoglobin. However, Kim-Shapiro and coworkers [35] have recently shown by EPR spectroscopy that, upon mixing of hemoglobin with NO^•, there is a linear correlation between the oxygen saturation of hemoglobin and the yield of HbFe^IINO. Comparable results were also obtained with blood samples [35]. In another work [36], the same group has reported that the rate constant for NO^• binding to R-state hemoglobin, measured by photolyzing HbFe^IICO in the presence of NO^•, is (2.1±0.1)×10⁷ M⁻¹ s⁻¹. This value is nearly identical to that reported for NO^• binding to T-state hemoglobin (2.6×10⁷ M⁻¹ s⁻¹ [10]). Taken together, these results clearly indicate that the rate of NO^• binding to hemoglobin is not cooperative and is independent of oxygen saturation.

The pathway suggested for the transfer of the ‘NO group’ from nitrosyl Hb to Cysβ93 (reaction (6)) is also highly unlikely. Reaction of nitrosyl Hb with O₂ has been shown to yield metHb and nitrate [37]. The analogous reaction with myoglobin has been proposed to proceed via the intermediate N-peroxynitrito-complex MbFe^III(N(O)OO) (reaction (7)) [38].

A further matter of debate is the concentration of SNOHb present in the blood. If NO^• was released cooperatively in the capillaries after O₂ dissociation, an arterial-venous gradient should exist for the SNOHb-concentration. Indeed, Stamler and coworkers found SNOHb-concentrations of 311±55 nM and 32±14 nM in arterial and venous rat blood, respectively [39]. Nevertheless, Gladwin and coworkers recently developed an accurate and sensitive method to measure very small SNOHb concentrations also in the presence of larger amounts of nitrite and nitrosyl Hb [40]. These authors found that the levels of SNOHb in the basal human circulation, including RBCs membrane fractions, are 46±17 nM in arterial RBCs and 69±11 nM in venous human RBCs [40]. In addition, the same group showed that SNOHb is intrinsically unstable in the reductive environment of the RBCs [40]. This feature further explains the extremely small SNOHb concentrations found in vivo and suggests that SNOHb cannot accumulate in the RBCs to form a reservoir of bioactive NO^•.

Finally, the postulated allosterically controlled O₂-concentration dependent release of the ‘NO group’ from SNOHb was sustained by kinetic and thermodynamic arguments [41]. However, Patel et al. demonstrated that the rate of the reaction between GSH and SNOHb is rather slow and does not depend on the oxygenation state of hemoglobin [42]. Transnitrosation reactions between thiols in proteins are usually slower than reactions between low molecular weight thiol compounds. Thus, the proposed transnitrosation between SNOHb and the band-3 anion-exchange protein is likely to be too slow to take place within the time taken for a given RBCs to pass through the precapillary vasculature.

We [43] and others [44,45] have recently shown that, because of the extremely fast reaction between NO^• and oxyHb, addition of a small volume of a concentrated NO^• solution to a larger volume of a diluted oxyHb solution may lead to artefactual generation of significant amounts of SNOHb. In addition, in agreement with Zhang et al. [46], we observed similar effects for the reaction of NO^• with oxyMb in the presence of GSH [43]. These findings imply that some of the high SNOHb yields reported in previous in vitro studies [25,28] may be artefacts arising from the chosen experimental conditions.

To understand the mechanism of this artefactual S-nitrosation of hemoglobin, one has to consider the half-life of the reactions that could take place in the system before the concentrations of all reagents have reached homogeneity. When a saturated (2 mM) NO^• solution is mixed with a 50-μM oxyHb solution, at the interface of the two solutions, the half-life of the reaction is so short (t_1/2=ln2/(8.9×10⁷ M⁻¹ s${}^{-1}\times 2\times {10}^{-3}\mathrm{M})=3.9$ μs) that metHb is generated before NO^• has reached a uniform concentration in the reaction mixture. Thus, NO^• could also react with metHb and produce the nitrosating species HbFe^II(NO⁺) (t_1/2^(β)=ln2/(6.4×10³ M⁻¹ s${}^{-1}\times 2\times {10}^{-3}\mathrm{M})=54$ ms and t_1/2^(α)=ln2/(1.7×10³ M⁻¹ s${}^{-1}\times 2\times {10}^{-3}\mathrm{M})=203$ ms). Finally, before NO^• will be dispersed in the solution, the reaction of NO^• (2 mM) with O₂ (in the air saturated oxyHb solution, 220 μM) [15] could also occur ($t_{1/2}=1/[8\times {10}^6{\mathrm{M}}^{-2}$ s⁻¹×2×10⁻³ M$\times 220\times {10}^{-6}\mathrm{M}]=280$ ms). The reaction of NO^• with O₂ in aqueous solution is believed to proceed via the formation of NO₂ and N₂O₃ [15]. Taken together, these simple calculations suggest that upon reaction of NO^• with oxyHb nitrosating species such as NO₂, N₂O₃, and HbFe^II(NO⁺) can be generated in high local concentrations and thus lead to artefactual production of SNOHb.

To get a better understanding of these mixing effects, we have recently carried out a systematic study of the influence of different factors on the amount of SNOHb produced from the reaction of NO^• with oxyHb [43]. Our results showed that the relative SNOHb yields (expressed relative to the amount of NO^• added) increase with decreasing amounts of NO^• added. Moreover, when the NO^• solution is added very slowly (within 1–2 min) from a diluted solution (equivalent volumes of the NO^• and the oxyHb solutions) the SNOHb yields are larger than those obtained by adding a very small volume of a 2 mM NO^• solution (at identical final NO^•- and oxyHb-concentrations). Interestingly, Han et al. [44] found that when NO^•-donors were employed to slowly deliver small concentrations of NO^•, oxyHb was converted exclusively to metHb and nitrate, without formation of detectable amounts of SNOHb. In contrast, Joshi et al. [45] reported that the SNOHb yields were nearly identical when NO^• was added as a bolus or with an NO^•-donor. These contrasting results may arise from the different NO^•-donors used by the two groups (DEA [44] vs. MHMA [45]), and thus suggest that NO^•-donors are not always good substitutes for NO^•.

As mentioned above, N₂O₃ may be one of the possible species responsible for SNOHb generation upon mixing of NO^• with oxyHb (reaction (11)). This hypothesis is supported by our observation that, in some cases (by slow addition of diluted NO^• solutions), the SNOHb yields are slightly lower when the reaction is carried out in more concentrated phosphate buffer (0.1 M vs. 10 mM) [43]. The decrease of the SNOHb production is due to the concurrent hydrolysis of N₂O₃ (reaction (12)), which has been shown to be catalyzed by inorganic phosphates [47].

To evaluate the nitrosating efficiency of the iron(III) nitrosyl heme complex, we studied the reactions of NO^• with metHb and with a mixture of metMb/GSH [43]. We found that the amount of SNOHb formed from the reaction of NO^• with metHb is larger than the GSNO yields obtained under identical conditions from the reaction with metMb/GSH [43]. Systematic studies indicated that maximal relative SNOHb yields (expressed relative to the amount of NO^• added) are generated when substoichiometric amounts of NO^• are added to metHb. Under these conditions a larger fraction of NO^• is bound to the β-subunit of hemoglobin [13,29] and may lead to higher SNOHb yields via an intramolecular nitrosation of Cysβ93. In fact, this residue is located on the proximal side of the heme, at a distance of only ∼10 Å from the iron center [30]. Studies with myoglobin have shown that photolyzed heme ligands can diffuse to the proximal side and partly occupy a hydrophobic pocket close to Cysβ93 in hemoglobin [48,49]. Thus, HbFe^II(NO⁺) may react with H₂O, generate H₂NO₂⁺ or HNO₂, which may rapidly diffuse to the proximal side of the heme, and react with Cysβ93. This intramolecular pathway, possibly facilitated by the hydrophobic environment of Cysβ93, may explain why this residue is nitrosated more effectively than GSH.

Finally, we have recently proposed that an additional pathway may be responsible for SNOHb formation. Specifically, we found that addition of 1 or 0.1 equiv of NO^• to the heme-blocked metHbCN (under strictly anaerobic conditions), followed by degassing of the reaction mixture, and subsequent exposure to air, led to the generation of SNOHb (0.9±0.2% and 4.0±0.3% relative to the amount of NO^• added, respectively) [43]. It has recently been shown that NO^• can also occupy non-covalent binding sites in hemoglobin, possibly trapped by aromatic amino acid residues in hydrophobic pockets [50]. As discussed above, the environment of Cysβ93 is largely hydrophobic. Moreover, two aromatic amino acids, Tyrβ145 and Pheβ103, are located at a distance of 4–5 Å from the sulfur atom of Cysβ93. Because of its lipophilic nature, NO^• preferentially partitions into lipid membranes and hydrophobic protein environments [51]. Taken together, these observations suggest that high local amounts of NO^• and O₂ could be trapped close to Cysβ93 and may therefore lead to their efficient nitrosation.