Fourier transform infrared spectroscopy study of ligand photodissociation and migration in inducible nitric oxide synthase

Inducible nitric oxide synthase (iNOS) is a homodimeric heme enzyme that catalyzes the formation of nitric oxide (NO) from dioxygen and L-arginine (L-Arg) in a two-step process. The produced NO can either diffuse out of the heme pocket into the surroundings or it can rebind to the heme iron and inhibit enzyme action. Here we have employed Fourier transform infrared (FTIR) photolysis difference spectroscopy at cryogenic temperatures, using the carbon monoxide (CO) and NO stretching bands as local probes of the active site of iNOS. Characteristic changes were observed in the spectra of the heme-bound ligands upon binding of the cofactors. Unlike photolyzed CO, which becomes trapped in well-defined orientations, as indicated by sharp photoproduct bands, photoproduct bands of NO photodissociated from the ferric heme iron were not visible, indicating that NO does not reside in the protein interior in a well-defined location or orientation. This may be favorable for NO release from the enzyme during catalysis because it reduces self-inhibition. Moreover, we used temperature derivative spectroscopy (TDS) with FTIR monitoring to explore the dynamics of NO and carbon monoxide (CO) inside iNOS after photodissociation at cryogenic temperatures. Only a single kinetic photoproduct state was revealed, but no secondary docking sites as in hemoglobins. Interestingly, we observed that intense illumination of six-coordinate ferrous iNOS oxy-NO ruptures the bond between the heme iron and the proximal thiolate to yield five-coordinate ferric iNOS oxy-NO, demonstrating the strong trans effect of the heme-bound NO.


Introduction
Nitric oxide synthases (NOSs) are homodimeric heme enzymes that catalyze the oxidative degradation of L-arginine (L-Arg) to nitric oxide (NO) 1,2 . Three structurally similar NOS isoforms have been identified in endothelial cells (eNOS), neuronal tissues (nNOS) and in macrophages (iNOS) 3 . Different from eNOS and nNOS, iNOS is not expressed in resting cells but induced upon inflammatory and immunologic stimulation. Each NOS protomer consists of an oxygenase and a reductase domain. In the catalytic oxygenase domain (NOS oxy ), dioxygen (O 2 ) binds to a central heme prosthetic group, anchored to the polypeptide chain via a proximal cysteine residue ( Figure 1). Its thiol sulfur atom accepts a hydrogen bond from an adjacent tryptophan residue. The substrate, L-Arg, is accommodated directly on top of the heme plane in the distal pocket; the cofactor tetrahydrobiopterin, H4B, binds along the side of the heme 4-7 . L-Arg and H4B are linked through an extended hydrogen bonding network mediated by one of the heme propionate groups. The reductase domain, NOS red , binds flavin mononucleotide (FMN), flavin adenine dinucleotide (FAD), and reduced nicotinamide adenine dinucleotide phosphate (NADPH). It provides the electrons for the catalytic reaction proceeding in the oxygenase domain. In a first step, L-Arg is converted to N-hydroxy L-Arg (NOHA). Subsequently, NOHA is decomposed into citrulline and nitric oxide (NO). Electron transfer is enabled by calmodulin binding in the interface between the two domains 8 .
The NO molecule generated during enzymatic turnover can either coordinate directly to the heme iron or diffuse out of the protein into the environment. From there, it may again bind in a bimolecular process 9 . Formation of the very stable ferrous NO complex results in self-inhibition of the enzyme. The probability of forming this product depends on the dissociation rate coefficient of NO from the ferric heme, the likelihood of autoreduction of the ferric NO-bound form to the ferrous derivative with its much stronger NO affinity, and the probability of oxidizing the ferrous NO-bound species to the ferric form plus nitrate by O 2

10
. Deactivation of the enzyme may also occur via nitrosylation of the side chains of two cysteine residues coordinating a zinc ion in the dimer interface, which leads to irreversible dissociation into non-functional monomers 11-14 .
The iNOS isoform has been implicated in the pathogenesis of various diseases; so there is a growing interest in developing potent and highly selective inhibitors 15,16 . Their targeted design requires detailed insights into the interactions between ligand, substrate and the surrounding protein matrix. Therefore, we have investigated ligand and substrate binding in the iNOS oxygenase domain, iNOS oxy , by using Fourier transform infrared (FTIR) spectroscopy of the stretching vibrations of carbon monoxide (CO) and NO as ligands rather than the physiological ligand O 2 . They are of similar size as O 2 , which suggests that ligand dynamics within the protein may be comparable for all three ligands. CO and NO both have excellent properties as infrared (IR) spectroscopic probes 17 . CO has proven to be an attractive heme ligand because the CO bond stretching vibration gives rise to strong mid-IR absorption bands that can be measured with exquisite sensitivity and precision 17,18 . The IR bands are fine-tuned by electrostatic interactions with the environment [19][20][21] ; therefore, CO is frequently utilized as a local probe of protein structure and dynamics 22 .
In the gas phase, CO absorbs at 2143 cm -1 23 . When bound to the central iron of a heme cofactor, the CO stretching frequency, ν CO , which is typically in the 1900 -2000 cm -1 spectral range, is susceptible to changes in the iron-ligand bond and the local electric field due to the vibrational Stark effect 24-29 . There are two major contributions to the heme iron-CO bond, i.e., σ-donation from a weakly antibonding 5σ MO of CO to the iron 4s and 3d z 2 orbitals and π-backbonding from the iron 3d z orbitals to the strongly antibonding CO 2π* orbital 30 . A positive charge located near the CO oxygen attracts electron density, causing a decrease in σ-donation and an increase in backbonding. Consequently, the C-O bond order is reduced and ν CO shifts to lower values [19][20][21] . A negative charge has the opposite effect.

Amendments from Version 1
We have made the following changes to the manuscript: 1. On page 6: "In substrate free iNOS oxy -CO, recombination is already maximal at 4 K and extends to ~70 K" was replaced by "In substrate free iNOS oxy -CO, the TDS signal is maximal at 4 K, indicating that there is substantial rebinding already at this low temperature. Rebinding extends up to ~70 K".
2. We also have added the following sentences on the same page: "The solid contours at ~2144 cm -1 in Figure 3f indicate a growth of this photoproduct population during the TDS measurement. Because data are taken in the dark, this can only occur via an exchange of photoproduct population from one band to another because of dynamics. Here, photoproduct population transfers from 2131 cm -1 to ~2144 cm -1 , and the underlying process is most likely a rotation of the CO by 180° so as to attain thermal equilibrium between the two states corresponding to opposite orientations of the CO 17 ".
3. The legends of Figure 3 and Figure 4 were corrected:

Protein expression
The iNOS oxy domain, with its first 65 residues deleted (Δ65 iNOS oxy , referred to as iNOS oxy in the following), was expressed essentially as described 44 . Briefly, iNOS oxy containing plasmids (pCWori) were transformed into competent Escherichia coli cells (strain BL21). The cells were plated on agar in the presence of 390 µM ampicillin (Carl Roth, Karlsruhe, Germany) and cultured overnight at 37°C. A single colony was added to 150 ml terrific broth (TB, Carl Roth) supplemented with ampicillin (390 µM) and agitated for 12 h at 37°C and 250 rpm. 10 ml of the overnight culture were added to 1.5 l TB, containing 390 µM ampicillin, and grown to an optical density of ~1 at 600 nm. Then, the temperature was lowered to 30°C and δ-aminolevulinic acid (44 µM, Sigma-Aldrich, St. Louis, MO, USA) and hemin (8 µM, Sigma-Aldrich) were added. iNOS expression was induced by adding isopropyl β-D-1-thiogalactopyranoside (IPTG, Carl Roth) to a final concentration of 100 µM. After 48 h (fresh ampicillin was added every 16 h), the cells were harvested by centrifugation for 20 min at 4°C and 2,000 rpm (swing-bucket rotor, 4-16 K, Sigma, Osterode, Germany). The cells were resuspended in lysis buffer (40 mM HEPES, 10% glycerol (vol.), 200 mM NaCl, pH 7.6, Carl Roth), mixed with 2 mg DNase (Sigma-Aldrich), and ruptured using a bead-beater (Biospec, Bartlesville, USA), filled with 0.1 mm (diameter) zirconia/silica beads (three treatments of 2 min each). The lysate was separated from the beads by a glass filter and loaded onto an immobilized-metal ion affinity column equilibrated with lysis buffer (Ni Sepharose 6 FastFlow, GE Healthcare). After washing with lysis buffer supplemented with increasing concentrations of imidazole (0, 10, 40 mM, Sigma-Aldrich), the protein was eluted with lysis buffer containing 160 mM imidazole. Appropriate fractions were pooled, dialyzed against water and concentrated by using Vivaspin Turbo 15 (cut-off 10 kDa) centrifugal concentrators (Sartorius, Göttingen, Germany). Finally, the protein was lyophilized and stored at -20°C.

Sample preparation
To prepare CO-ligated iNOS oxy , 12 mg freeze-dried iNOS were slowly added to 40 µl cryosolvent (75%/25% glycerol/100 mM potassium phosphate buffer (v/v), pH 7.4, and, if so desired, supplemented with L-Arg and NOHA substrate (Sigma-Aldrich) or H4B cofactor (Sigma-Aldrich) to reach final concentrations of 200 mM and 100 mM, respectively) and stirred under 1 atm CO for 60 min. Subsequently, a two-fold molar excess of an anaerobically prepared sodium dithionite solution (Sigma-Aldrich) was added with a gastight Hamilton syringe, and the solution was stirred for another 10 min. To remove any undissolved protein, the solution was centrifuged for 10 min at 13,400 rpm (Minispin centrifuge, Eppendorf, Hamburg, Germany) before loading it into the sample cell. For an NO-ligated sample, ferric iNOS oxy was dissolved in cryosolvent and stirred under an N 2 atmosphere for 1 h. The gas phase above the sample was replaced repeatedly by N 2 to efficiently remove O 2 .
Finally, a few microliters of NO gas were added with a gas-tight syringe. NO ligation to the heme iron was confirmed by UV/vis absorption spectroscopy.

Experimental setup
A few microliters of the sample solution were sandwiched between two CaF 2 windows (diameter 25.4 mm) separated by a Mylar washer. The windows were mounted inside a block of oxygen-free highconductivity copper. The copper block was attached to the coldfinger of a closed-cycle helium refrigerator (model F-50, Sumitomo, Tokyo, Japan). The sample temperature was measured with a silicon temperature sensor diode and regulated in the range 3 -320 K by a digital temperature controller (model 336, Lake Shore Cryotronics, Westerville, OH). A continuous-wave, frequency-doubled Nd-YAG laser (Samba, Cobolt, Solna, Sweden), emitting up to 300 mW output power at 532 nm, was used to photolyze the sample. The laser beam was split and focused with lenses on the sample from both sides. Transmission spectra were recorded on a Vertex 80v FTIR spectrometer (Bruker, Karlsruhe, Germany) at a resolution of 2 cm -1 , using either an InSb detector (75 µm thick Mylar, 1,700 to 2,300 cm -1 ) or an MCT detector (<5 µm thick Mylar, 1,100 to 2,300 cm -1 ).

FTIR photolysis difference spectroscopy
The infrared absorption of CO and NO can be studied selectively by using photolysis difference spectroscopy, which involves measurement of IR transmission spectra, I(ν, T), before and after photolysis. The difference absorbance of the two spectra, ΔA(ν, T) = log(I dark /I light ), contains only features that are due to photodissociation of the ligand from the heme iron. The missing absorption of the heme-bound ligands (A bands) after photolysis and the corresponding absorption of the photolyzed ligands (photoproduct bands) are displayed with negative and positive amplitudes, respectively. Peak positions and fractional occupancies were determined by fits with Gaussian band shapes; they are compiled in Table 1. In the following, we use the Gaussian band positions (frequencies) at 4 K as a subscript to 'A' (denoting the heme-bound state) to distinguish the absorbance bands and also to refer to a particular substate of the protein.
Different illumination protocols were applied for photodissociation 17 . Before starting a TDS experiment, the sample was illuminated for 10 s at 4 K to trap the photolyzed ligand close to the heme iron at the so-called primary docking site B. Alternatively, under 'slow-cool' illumination, the sample was cooled from 160 to 4 K at a rate 0.3 K/min under constant laser illumination to enable the photodissociated ligands to sample alternative docking sites that may not be accessible upon photolysis at 4 K. In both protocols, 300 mW laser power at 532 nm was used. To monitor the photodissociation kinetics, the samples were continuously illuminated for 15,000 s at reduced laser power (0.3 mW or 10 mW), and transmission spectra were recorded continuously. For comparison, the photolysis yield was scaled with respect to complete photodissociation with full laser power (300 mW).
Temperature derivative spectroscopy (TDS) TDS, an experimental protocol designed to study thermally activated rate processes involving enthalpy barrier distributions, has been described in great detail elsewhere 41-43 . Briefly, a non-equilibrium state is created in the sample at low temperature, e.g., by photolysis with visible light. The integrated absorbance, A, of a spectral band taken at the lowest temperature represents the total photolyzed population, N. Subsequently, thermal relaxation of the sample back to equilibrium is recorded while the sample temperature is ramped up linearly over a few hours in the dark. One FTIR transmission spectrum is taken for every 1-K temperature increase.
In the simplest analysis, we assume that any change in integrated absorbance is due to ligand rebinding and, therefore, proportional to a population change, ΔN, during acquisition of two successive spectra. TDS data are conveniently presented as two-dimensional contour plots, with solid lines indicating an absorbance increase and dashed lines a decrease. Contours are spaced logarithmically to emphasize small features.
Results and discussion 1. FTIR spectroscopy of iNOS oxy using CO as an internal probe In the following, we present 4-K FTIR photolysis difference spectra of iNOS oxy -CO and briefly discuss the influence of substrate, substrate intermediate and cofactor on the CO stretching vibration and rebinding. For additional information, we refer to Jung et al. 45 and Li et al. 46 .
Photolysis difference spectra at 4 K. The 4-K absorption difference spectrum of iNOS oxy -CO displays two broad, extensively overlapping A bands at 1945 and 1959 cm -1 , indicative of two active site subconformations with significant intrinsic structural heterogeneity ( Figure 2a). Adding the H4B cofactor induces only small changes; the resulting spectrum can be described by a dominant A band centered on 1951 cm -1 and a minor one at 1924 cm -1 ( Figure 2a). As H4B binds along the side of the heme 4 and, thus, not in the immediate vicinity of the heme-bound CO, it is not expected to modify ν CO to any significant extent. In contrast, the presence of L-Arg shifts the  (Figure 2a). The pronounced redshift of A 1904 arises from the electron-withdrawing effect of the terminal, positively charged NH 2 + moiety of the L-Arg side chain close to the bound CO 4 . The position of the A 1921 band is indicative of an electrostatic interaction of the CO dipole with a less pronounced positive partial charge, most likely the neutral terminal amino group of the L-Arg side chain.
If the reaction intermediate NOHA is present, three A bands at 1903, 1937 and 1956 cm -1 are discernable (Figure 2a). The crystal structure shows that NOHA binds in the same orientation in the active site as L-Arg, with the side chain pointing towards the heme iron 47 . Therefore, we suggest that, in those iNOS oxy molecules absorbing within the A 1937 band, a hydrogen bonding interaction exists between the CO ligand and the hydroxyl group of the NOHA side chain. A 1903 is most likely associated with iNOS oxy molecules, in which the terminal amine of the NOHA side chain is protonated (pK = 8.1 48 ) and points towards the heme-bound CO. The protonated NOHA has been suggested to be the catalytically active substrate intermediate 49,50 .
The absorption spectra of photolyzed CO are plotted in Figures 2b (brief illumination at 4 K) and 2c (slow-cool illumination); peak positions and relative areas are included in Table 1. For comparison, the integrated absorption in each spectral region was scaled to the same area. We note that the ratio of the integrated areas of the A and photoproduct bands is ~20 18 .
All photoproduct spectra obtained after 10-s illumination at 4 K have absorption bands in the 2120 -2130 cm -1 spectral range (Figure 2b). The spectrum of iNOS oxy -CO is composed of two stretching bands at 2124 and 2129 cm -1 . With H4B, photoproduct bands appear at 2124 and 2133 cm -1 , indicating that the cofactor has an effect on ν CO of the unbound CO. In the presence of L-Arg, the absorption bands are centered on 2120 and 2131 cm -1 , and there are two additional bands at 2144 and 2150 cm -1 . Their higher stretching frequencies suggest formation of a hydrogen bond between the ligand carbon and the terminal amine group of L-Arg 29 . Upon NOHA binding, the photoproduct bands are centered on 2122 and 2133 cm -1 . The minor absorption at 2145 cm -1 can be associated with CO ligands photolyzed from iNOS oxy /NOHA trapped in its A 1903 conformation.
The photoproduct spectra obtained after slow-cool illumination (Figure 2c) are similar to the ones recorded after 10-s illumination (Figure 2b), suggesting that it is not possible to populate additional docking sites to any significant extent. The greatest difference is seen for iNOS oxy -CO. Its photoproduct spectrum shows two well separated bands at 2124 and 2134 cm -1 rather than the non-separated doublet seen in Figure 2b. We also note that there is an additional shoulder at 2117 cm -1 for iNOS oxy -CO/NOHA.

CO rebinding in iNOS oxy .
To obtain more information on the photoproduct states, TDS measurements were started at 4 K immediately after illumination. Figure 3 displays the contour maps obtained after 10-s illumination at 4 K, with the absorption changes in the  A bands and the photoproduct bands in the left and right columns, respectively.
All iNOS oxy -CO samples display single-step CO rebinding. This observation indicates that there is only a single kinetic state of the photolyzed protein-ligand complex, and the presence of sharp photoproduct bands indicates that the photolyzed ligands are trapped in transient docking sites with well-defined orientations. In substrate-free iNOS oxy -CO, recombination is maximal at 4 K, indicating that there is substantial rebinding already at this low temperature.
Rebinding extends up to ~70 K (Figure 3a, b). Rebinding in the dominant A 1951 substate of iNOS oxy -CO/H4B peaks at 20 K; as in iNOS oxy -CO, the process extends to 70 K. Only the minor A 1924 subpopulation shows a focused rebinding peak at ~60 K ( Figure  3c). The photoproduct map does not yield additional information (Figure 3d). Binding of either L-Arg or NOHA in the active site shifts CO rebinding to higher temperatures, suggesting that the hydrogen bonding interaction stabilizes the ligands at the transient docking site against rebinding (Figures 3e and 3g). Maximal rebinding in iNOS oxy /NOHA, i.e., in A 1903 and A 1937 , occurs at 50 -60 K (Figure 3g). The corresponding photoproduct bands are centered on 2122 and 2133 cm -1 (Figure 3h). The contours at 1950 -1960 cm -1 (Figure 3g) represent rebinding in the NOHA-free A 1956 substate. With L-Arg anchored in the active site, CO ligands return to the heme iron also at ~50 -60 K (Figure 3e). The corresponding photoproduct map shows a concomitant loss of the photoproduct bands at 2150, 2144, 2131 and 2120 cm -1 , associating these bands with CO molecules trapped in the vicinity of the substrate (Figure 3f). A population transfer between photoproduct states due to CO rotation 32,51,52 is apparent from the mirror-imaged dashed and solid contours at 2131 and 2144 cm -1 at 12 K. The solid contours at ~2144 cm -1 in Figure 3f indicate a growth of this photoproduct population during the TDS measurement. Because data are taken in the dark, this can only occur via an exchange of photoproduct population from one band to another because of dynamics. Here, photoproduct population transfers from 2131 cm -1 to ~2144 cm -1 , and the underlying process is most likely a rotation of the CO by 180° so as to attain thermal equilibrium between the two states corresponding to opposite orientations of the CO 17 .
The TDS maps after slow-cool illumination (Figure 4) show only marginal differences to the ones obtained after brief 4-K illumination (Figure 3), which confirms that the photodissociated CO ligands populate only a single kinetic state. Notably, after slow-cool illumination, rebinding generally occurs at slightly higher temperatures than after brief 4-K illumination. The observed slowing may be attributed to small structural changes near the active site, causing an increase of the ligand binding barrier. A similar effect was also visible in MbCO upon extended illumination below 40 K 42 as well as in NO-and CO-ligated nitrophorin 4 35 .
In a typical globin protein involved in ligand transport or storage, the primary ligand docking site B is indispensable because it ensures efficient ligand binding to and release from the heme iron 53 . Incoming ligands are 'caught' in site B before the actual bond formation process occurs 32,54 . Upon thermal dissociation from the heme iron, ligands can remain unbound in site B for some time, which increases their probability to escape from the protein. Without this site, they would immediately recombine with the heme iron, as is, e.g., observed for NO-transporting nitrophorin 35 and modified cytochrome c 55 .
The catalytic reaction of iNOS requires sequential binding of two O 2 molecules and efficient release of the NO product. Therefore, the B site is likely to have dual functionality. On the one hand, it allows efficient O 2 binding to the heme iron. On the other hand, it ensures efficient release of the generated NO. Using CO as a ligand, we have shown that the B site is readily accessible for ligands photodissociated from the heme iron, both in the presence and absence of L-Arg or NOHA. The substrates stabilize the CO ligand at the transient site via hydrogen bonding. This stabilizing effect is also seen for the minor A 1924 subpopulation of iNOS oxy /H4B. Presumably, a small fraction of H4B molecules are positioned such that they can form a direct hydrogen bond.

FTIR spectroscopy of iNOS oxy using NO as an internal probe
The NO stretching absorption is also very suitable as a local probe of the active site structure and of ligand movements within a protein 17 . Despite their similar sizes, the ligands may show different dynamics inside the protein 56 . For example, in myoglobin (Mb), a transient docking site on the proximal side of the heme is readily populated by CO but not at all by NO 56 . Such subtle differences could be relevant for the inhibitory effects of NO. Therefore, we have analyzed NO binding in ferric iNOS oxy using FTIR-TDS at cryogenic temperatures.
Photolysis difference spectra at 4 K. Figure 5 displays 4-K photolysis difference spectra of various ferric iNOS oxy -NO preparations. Most spectra show an A band at 1870 cm -1 associated with NO bound in an active site without bound cofactor or substrate ( Table 1). In the spectrum of iNOS oxy -NO, A 1870 is rather broad, suggesting significant conformational heterogeneity at the active site. The spectrum of iNOS oxy -NO/NOHA is very similar, dominated by the broad A 1870 band; the only clear change from iNOS oxy -NO is a shoulder at 1851 cm -1 . This comparison suggests that NOHA is bound only in a small subfraction reflected by the shoulder. In iNOS oxy -NO/L-Arg, A 1847 and A 1829 report the binding of L-Arg. A 1870 is still present due to incomplete saturation with substrate ( Figure 5). Interestingly, Rousseau et al. 2 could not identify any changes of ν Fe-N in their resonance Raman spectra upon binding of L-Arg and even hypothesized that L-Arg does not bind to ferric iNOS oxy -NO. With H4B anchored next to the heme, the A band is shifted to 1872 cm -1 , and another absorption band emerges at 1890 cm -1 .
Most of the observed spectral shifts can again be explained by backbonding 57 because the ferric NO-ligated ground state, which is best described as Fe II NO + , is isoelectronic to Fe II CO 58 . The heme-bound NO absorbs at 1870 cm -1 . L-Arg shifts ν NO to lower frequencies; the A 1829 and A 1847 bands indicate an interaction between the NO and the positively charged and neutral terminal amino groups of the L-Arg side chain. As already observed for CO, the effect of NOHA is less pronounced; its presence is visible via a shift of the A band to 1851 cm -1 . Interestingly, the NO stretching absorption is also affected by H4B. The band shifts slightly and, in addition, it becomes rather narrow, which is indicative of a more homogeneous active site environment or restricted dynamics of the heme-bound NO due to the bound H4B 35, 59 . In 2005, Rousseau et al. 2 reported that, upon H4B binding, a Raman band emerges that was assigned to the Fe-N-O bending mode, δ Fe-N-O , of the ferric adduct, indicating a more homogeneous bending of the bent NO. In thiolate-ligated Fe III NO adducts, NO is typically bound at an angle of 160°6 0-66 , and H4B binding next to the heme is not expected to modify this angle due to steric interactions. It may, however, restrict its librational dynamics around this angle, possibly because of the increased heme distortion caused by H4B 67,68 . The additional band at 1890 cm -1 may indicate partial occupancy of a water molecule in the active site 62 .
The photoproduct bands, displayed in Figure 5 with positive amplitudes, are in the 1810 -1830 cm -1 spectral range and, thus, redshifted by only ~50 cm -1 from those of the heme-bound NO (Table 1). For iNOSoxy-NO/L-Arg, the photoproduct and A bands even overlap. Their decomposition (details are discussed below) yields a narrow photoproduct band at 1814 cm -1 and a broad feature at 1822 cm -1 . iNOSoxy-NO and iNOSoxy-NO/NOHA show two photoproduct bands at 1814 and 1818 cm -1 . Interestingly, these bands are about as strong as the A bands, which strongly suggests that they do not represent unbound NO trapped in a transient docking site but rather hemebound NO with restricted librational freedom.
In contrast to all other samples, the iNOS oxy -NO/H4B photoproduct spectrum reveals only a very weak feature at ~1818 cm -1 . This finding may be explained by a photolyzed NO that cannot be trapped in well-defined orientations. As a result, the stretching absorption becomes extremely broad and hardly distinguishable from the background. A similar effect was observed for NO in the primary photoproduct site B of ferric Mb 56 .

NO rebinding in ferric iNOS oxy .
To gain additional information on the peculiar, strongly absorbing NO photoproduct bands, TDS experiments were started immediately after illuminating NOligated samples at 4 K. Figure 6 displays the absorption changes in the A bands and in the photoproduct bands with solid and dotted lines, respectively. The contour maps obtained after slow cool illumination (not shown) are essentially identical, as for the CO-ligated samples.
In iNOS oxy -NO, NO rebinding in A 1870 starts already at the lowest temperatures ( Figure 6a) and extends to ~90 K. The decay of the photoproduct, however, occurs predominantly between 80 and 120 K, indicating that these bands cannot be associated with NO ligands photolyzed from the ferric heme iron, as reported by the A 1870 band. Apparently, laser illumination produces a photoproduct band from another NO species in the sample. The TDS map of iNOS oxy -NO/ NOHA (Figure 6d) shows essentially the same features. It is likewise evident that NO rebinding is complete below 80 K, whereas the strange photoproduct feature disappears in the temperature range 80 -120 K. In iNOS oxy -NO/H4B (Figure 6b), NO rebinding at the ferric iron also starts at 4 K. In a subpopulation, recombination peaks at ~65 K; absorption changes of photoproducts are too small to be detected. NO rebinding in the L-Arg-bound A 1829 and A 1847 substates occurs mainly below 30 K, concomitantly with the decay of the photoproduct (Figure 6c). The apparent maximum in the contours at 15 K and ~1850 cm -1 is artificial and results from the superposition of the A bands and the photoproduct bands (compare Figure 5). Recombination in the substrate-free A 1870 fraction of the sample is maximal at 4 K and extends out to ~70 K, consistent with the data shown in Figure 6a.
In summary, rebinding of NO to the ferric heme of iNOS oxy is a one-step process. The corresponding photoproduct bands, i.e., the absorption bands of NO photodissociated from the ferric heme, were not identifiable. Presumably, NO is bound only weakly within the protein, without any well-defined orientation and without any additional stabilization via hydrogen bonding interactions to the substrate or the cofactor. As a consequence, the NO has a broad stretching absorption that cannot be distinguished from the background. Note that, if the photoproduct bands were masked by the strong bands at ~1820 cm -1 , they should have become visible in the spectrum of iNOS oxy -NO/H4B ( Figure 5).

Identification of the iNOS oxy -NO photoproduct.
The TDS data in Figure 5 clearly prove that the strong absorption bands at ~1820 cm -1 are not generated by photodissociation of NO bound to ferric heme, absorbing at ~1870 cm -1 . To identify the corresponding pre-illumination states, we screened the 4-K FTIR photolysis difference spectrum of iNOS oxy -NO from 1,100 to 2,300 cm -1 and detected a band at 1616 cm -1 , which we tentatively associate with a six-coordinate (6C) ferrous NO adduct ( Figure 5, inset). This assignment is supported by the ν NO of 1591 cm -1 reported for 6C ferrous P450 cam -NO 69 . Praneeth et al. 70 also computed frequencies in this range, ν NO = 1617 cm -1 and ν NO < 1600 cm -1 for thiophenolate-and alkylthiolate-heme complexes, respectively, using density functional theory calculations on ferrous, thiolate-coordinated porphyrin model systems.  Raman measurements was sufficient to photodissociate the axial thiolate base trans to the NO 75 . This effect could be suppressed by lowering the temperature to 77 K and reducing the laser power. Accordingly, we have illuminated the iNOS oxy -NO/L-Arg sample at low laser intensity (0.3 mW at 532 nm). This power was still sufficient to photodissociate the NO from the 6C ferric heme adduct (dotted line in Figure 5), photoproduct bands at ~1820 cm -1 , however, did not emerge, confirming that the photoproduct was not formed. Therefore, we propose that illumination of 6C ferrous iNOS oxy -NO with sufficient laser power leads to rupture of the bond between the iron and the proximal Cys194 thiolate, leaving behind a 5C iNOS oxy -NO. Because the NO is still bound to the heme iron, the intensity of the IR bands at ~1820 cm -1 is comparable to that of other A bands 25,34 . The NO stretching frequency of the 5C adduct indicates that the ligand is coordinated to a ferric iron, so that the Cys194 sulfur is negatively charged after photodissociation. Similar NO stretching frequencies were reported for an isolated 5C ferric heme nitrosyl complex (ν NO = 1842 cm -1 76 ) and for NO-ligated porphyrins with phenyl (ν NO = 1825 cm -1 ) and pentafluorophenyl (ν NO = 1859 cm -1 ) substituents on the four meso positions 77 . If the laser power is sufficiently high (300 mW), it is even possible to photodissociate the NO from the 5C ferric iNOS oxy -NO, leaving behind a four-coordinate, 'naked' heme as a 'secondary photoproduct' (Figure 7a).
L-Arg binding in the active site lowers the yield of ferric 5C iNOS oxy -NO upon laser illumination (Figures 7a and c) and favors reformation of the iron-sulfur bond as soon as the laser is switched off (Figures 7b and d). This effect may result from the competition between the NO ligand and the thiolate for σ charge donation to the heme iron; the higher the donation, the stronger the bond to the donor and the weaker the bond to the opposing heme ligand. The σ donor strength of the thiolate is altered by hydrogen bonding interactions to the sulfur atom 66 . Using sulfur K-edge x-ray absorption spectroscopy and density functional theory calculations, Dey et al. 78 showed that each hydrogen bond reduces the electron-donating power of the thiolate sulfur. The NO electron donor ability and, therefore, its repulsive trans effect can be reduced by interactions that draw electron density away from the NO 79,80 , here by the hydrogen bonding interaction with L-Arg, so that the axial iron-sulfur bond is stabilized.
We also note that 6C ferrous iNOS oxy -NO is not stable in the presence of H4B but spontaneously oxidizes to the ferric form 46 . Consequently, the yield of the 5C adduct is negligible, as is indicated by the low intensity of the absorption bands ( Figure 5).

Ferric 5C iNOS oxy -NO.
In view of the competition between the NO ligand and the thiolate for σ charge donation to the heme iron, one should expect ν NO of the 5C photoproduct lacking the thiolate ligand to be blue-shifted with respect to ν NO of the 6C adduct because the repulsive trans effect of the thiolate has been removed. Experimentally, however, the opposite behavior is observed ( Figure 5). To resolve this apparent discrepancy, one has to consider that the 5C ferric form originates from a 6C ferrous species, in which the NO is typically bound at an angle of ~140°. In the corresponding 6C ferric derivatives, the Fe -N -O angle is normally ~160°. At cryogenic temperatures, the dynamics of the protein matrix is completely arrested 39,40 .
Consequently, the NO is held in the strongly bent (lower angle) orientation of the 6C ferrous form. Based on DFT calculations, Linder et al. 81 reported that reducing the angle from 160° to 150° shifts ν NO in 5C model porphyrins from 1897 to 1857 cm -1 . Therefore, we suggest that the low ν NO of the 5C form is caused by NO binding at a small angle. We note that the similar ν NO in 5C and 6C ferric iNOS oxy -NO/L-Arg implies that the bound substrate controls the angle at which the NO binds. Apparently, steric constraints override the bending induced by the trans effects.
Finally, we point out that, in contrast to the photo-induced 6C ferric → 5C ferric transition observed in the FTIR experiments at cryogenic temperatures, the spontaneous conversion of the 6C ferric NO-bound iNOS oxy derivative at physiological temperatures involves two NO molecules and yields a 5C ferrous species 71,72,82 . After binding the first NO, the ferric 6C iNOS oxy -NO reacts with a second ligand to yield 6C ferrous iNOS oxy -NO. This complex immediately converts to the 5C form and a nitrosonium ion (NO + ). The ion may diffuse towards the zinc binding site and nitrosylate one of the Cys residues involved in coordinating the zinc. Detailed information on the dataset can be found in the text file "Raw data legend".

Conclusions
FTIR spectroscopy at cryogenic temperatures, especially in combination with sophisticated illumination and data acquisition temperature protocols, provides quantitative data on protein-ligand interactions. Our FTIR-TDS studies on iNOS oxy have shown that CO and NO rebinding involve only a single transient state in iNOS oxy . The CO is stabilized in well-defined orientations at the docking site by hydrogen bonding interactions and, therefore, gives rise to rather narrow photoproduct bands. In contrast, photoproduct bands associated with the photolyzed NO cannot be resolved. The NO appears to be trapped in less specific orientations, which may favor the release of this ligand. Under physiological conditions, release of the generated NO from the protein is facilitated.
Upon illumination of 6C ferrous iNOS oxy -NO at cryogenic temperatures, a 5C ferric NO adduct was identified, providing direct evidence for light-induced breakage of the iron-thiolate bond. Future studies along these lines are likely to contribute to a better understanding of functional processes in which the NO ligand is involved.