l-Arginine Binding to Nitric-oxide Synthase

Nitric-oxide synthase (NOS) catalyzes the oxidation of l-arginine to nitric oxide andl-citrulline. Because overproduction of nitric oxide causes tissue damage in neurological, inflammatory, and autoimmune disorders, design of NOS inhibitors has received much attention. Most inhibitors described to date include a guanidine-like structural motif and interact with the guanidinium region of thel-arginine-binding site. We report here studies withl-arginine analogs having one or both terminal guanidinium nitrogens replaced by functionalities that preserve some, but not all, of the molecular interactions possible for the –NH2, =NH, or =NH2 + groups of l-arginine. Replacement groups include –NH-alkyl, –alkyl, =O, and =S. Binding ofl-canavanine, an analog unable to form hydrogen bonds involving a N 5-proton, was also examined. From our results and previous work, we infer the orientation of these compounds in the l-arginine-binding site and use IC50 or K i values and optical difference spectra to quantitate their affinity relative tol-arginine. We find that the non-reactive guanidinium nitrogen of l-arginine binds in a pocket that is relatively intolerant of changes in the size or hydrogen bonding properties of the group bound. The individual H-bonds involved are, however, weaker than expected (<2 versus 3–6 kcal). These findings elucidate substrate binding forces in the NOS active site and identify an important constraint on NOS inhibitor design.

Nitric-oxide synthase (NOS) 1 catalyzes the two-step oxidation of L-arginine to L-citrulline and nitric oxide (NO). Oxygen and NADPH are co-substrates, and N -hydroxy-L-arginine (NOH-Arg) is a tightly bound intermediate (1,2). The enzyme is active as a homodimer, and each monomer is comprised of a heme-and tetrahydrobiopterin-containing oxygenase domain that binds and oxidizes L-arginine and a FAD-and FMNcontaining reductase domain that delivers electrons from NADPH to heme. Once reduced, the heme cofactor binds and activates O 2 , which in turn reacts with a terminal guanidinium nitrogen of the substrate L-arginine that is bound ϳ4 Å from the heme iron (3). That reactive, "proximal" nitrogen is first hydroxylated, forming NOH-Arg, and then oxidized further to NO. The other, previously equivalent guanidinium nitrogen is bound farther away from the heme cofactor and does not react; that "distal" nitrogen becomes the terminal -NH 2 group of the product L-citrulline.
There are three major isoforms of NOS in mammals (1,2,4). Two constitutive, Ca 2ϩ /calmodulin-regulated isoforms were initially identified in neurons (nNOS) and vascular endothelial cells (eNOS). An inducible, transcriptionally regulated isoform (iNOS) was initially identified in macrophages but can be expressed in response to inflammatory cytokines and endotoxin in many cell types. Neuronal NOS has a role in neurotransmission and/or neuromodulation (5,6), whereas eNOS produces NO that has an important role in controlling vasorelaxation and blood pressure (7,8). Nitric oxide derived from iNOS plays both regulatory and cytotoxic roles in the immune response (9,10). In addition to these physiological roles, NOS is known to contribute to several pathological processes, typically when nNOS is overstimulated or iNOS is induced inappropriately or in excess. For example, nNOS is implicated in stroke (11) and migraine headache (12), and iNOS is implicated in septic shock (13,14), inflammatory bowel disease (15), uveitis (16), and arthritis (17,18). The possibility of treating these and other conditions by inhibiting NOS has elicited intense efforts to identify or design NOS inhibitors, preferably isoform-selective NOS inhibitors. To date well over 100 inhibitors have been reported (19 -22). Almost all of these compounds contain a guanidine-like structural motif, and their initial binding is competitive with L-arginine, observations that suggest they are interacting with the guanidinium region of the L-arginine-binding site.
In the present studies we have designed and synthesized several novel L-arginine analogs ( Fig. 1) in order to "map" the guanidinium region of the substrate-binding site. We have confirmed and extended previous work showing that the binding site for the reactive (i.e. oxidizable) guanidinium nitrogen can accommodate a variety of alternative groups including those much larger than -NH 2 and ϭNOH. In contrast, the binding site for the non-reactive, distal guanidinium nitrogen does not accommodate larger groups (1,23,24). We have exploited this difference in size specificity to predictably direct the binding orientation of novel L-arginine analogs. We find that analogs bind poorly if the binding site for the non-reactive guanidinium nitrogen of L-arginine must be occupied by ϭO, ϭS, -CH 3 or larger groups. These results extend insights gained from recently reported mutagenesis and x-ray crystallographic studies of NOS and have implications for the design of NOS inhibitors. 1 H and 13 C NMR spectra were obtained using a Bruker AC 300 MHz spectrometer. High-resolution, electron impact (EI), chemical ionization, and FAB mass spectral analyses were generously carried out by Dr. Frank Laib at the Department of Chemistry, University of Wisconsin, Milwaukee, WI.
Syntheses of N 5 -Thioacyl-L-ornithines-N 5 -Thioacetyl-L-ornithine (TAO) was synthesized by reaction of an appropriately protected Lornithine with ethyl dithioacetate. Thus, N ␣ -(tert-butyloxycarbonyl)-Lornithine tert-butyl ester (2.9 g, 10 mmol) was dissolved in 50 ml of chloroform and added to a solution of 2.5 g of CaCO 3 and 1.20 ml of ethyl dithioacetate dissolved in 50 ml of water. The mixture was stirred vigorously for 20 h and filtered. The chloroform layer was separated, and the aqueous layer was extracted with chloroform (2 ϫ 50 ml). The combined chloroform extracts were dried with anhydrous Na 2 SO 4 and evaporated under reduced pressure. The residue was dissolved in a small amount of ethyl acetate (5 ml) and chromatographed over a silica gel column (3 ϫ 50 cm) eluted with ethyl acetate and petroleum ether (1:4). Fractions (5 ml) were collected and the product was identified by TLC on silica plates developed in the same solvent (R F ϭ 0.8). Productcontaining fractions were pooled and solvent was evaporated under reduced pressure to give N ␣ -(tert-butyloxycarbonyl)-N 5 -(thioacetyl)-Lornithine tert-butyl ester as an oily liquid in 25% yield. That intermediate was dissolved in ϳ10 ml of dioxane and added to ϳ20 ml of ice-cold dioxane containing 6 N HCl. After stirring on ice for 4 h and at room temperature overnight, evaporation of the solvent under reduced pressure yielded 0.4 g of TAO (m.p. 75-78°C): 1  N 5 -Thiobutyryl-L-ornithine (TBO) was synthesized by reaction of Lawesson's reagent with N 5 -butyryl-L-ornithine, which was obtained by reacting butyryl chloride with protected L-ornithine. Thus, butyryl chloride (2.13 g, 20 mmol) was added to a solution of N ␣ -(tert-butyloxycarbonyl)-L-ornithine tert-butyl ester (5.8 g, 20 mmol) in 100 ml of methylene chloride containing 2.0 ml of triethylamine. The reaction mixture was stirred at room temperature for 20 h, and the solvent was then evaporated under reduced pressure. The residue was chromatographed over silica gel as described for TAO. N ␣ -(tert-Butyloxycarbonyl)-N 5 -(butyryl)-L-ornithine tert-butyl ester (3.2 g, 45%) was obtained as an oily liquid. That product (8.0 mmol) was dissolved in benzene (50 ml), Lawesson's reagent (1.82 g, 4.5 mmol) was added, and the mixture was refluxed for 3 h. After cooling, the mixture was filtered, and the filtrate was washed with water (3 ϫ 50 ml). The benzene layer was dried over anhydrous Na 2 SO 4 and evaporated to give an oily liquid that was chromatographed over silica gel as described previously. Product-containing fractions were determined by TLC (R F ϭ 0.8) and evaporated to give N ␣ -(tert-butyloxycarbonyl)-N 5 -(thiobutyryl)-L-ornithine tert-butyl ester (2.5 g, 85%) as an oily liquid. Deprotection of that intermediate in dioxane-dry HCl as described above gave 1. Synthesis of N 5 -(1-Iminoalkyl)-L-ornithines-The following compounds, which are homologs of L-NIO, were synthesized by the general procedure reported previously for N 5 -(1-iminopropyl)-L-ornithine (methyl-L-NIO) (32). Synthesis of N 5 -Acyl-L-ornithines-These derivatives were synthesized by dioxane-HCl deprotection of the corresponding N ␣ -(tert-butyloxycarbonyl)-N 5 -acyl-L-ornithine tert-butyl esters which were in turn prepared from N ␣ -(tert-butyloxycarbonyl)-L-ornithine tert-butyl ester and the appropriate acyl chloride as described above for N ␣ -(tert-butyloxycarbonyl)-N 5 -butyryl-L-ornithine tert-butyl ester.

Methods for Enzymatic Studies
Nitric-oxide Synthase Assays-Nitric-oxide synthase activity was routinely determined based on the oxidation of oxyhemoglobin to methemoglobin by NO (33). Sample cuvettes at 25°C contained in a final volume of 0. . After heating in a boiling water bath for 1 min, the samples were chilled and centrifuged. A portion (225 l) of the supernatant was applied to small Dowex 50 columns (Na ϩ form, 1 ml of resin), and the product L-[ 14 C]citrulline was eluted with 2 ml of water and quantitated by liquid scintillation counting. Where inhibition was determined to be competitive with L-arginine, K i values were estimated from measured IC 50 values using K i ϭ IC 50 ). For purpose of calculation, K m Arg values for nNOS, eNOS, and iNOS were estimated as 1.8, 3.6, and 12.5 M, respectively, based on the present and earlier (32) work.
Optical Difference Spectroscopy-Interaction of inhibitors with nNOS was determined spectrally using a Shimadzu or Perkin-Elmer dual beam UV/visible spectrophotometer (32). Typically 1.5-2.0 M nNOS in 0.5 ml of 50 mM Tris-HCl buffer, pH 7.5, 10% glycerol, and 0.1 mM EDTA was placed in the sample and reference cuvettes at 15°C, and the baseline was adjusted to zero. Inhibitor was then added to the sample cuvette, an equal volume of the same buffer was added to reference cuvette, and the difference spectrum was obtained. Spectra were normalized using the isosbestic point at 410 nm.

RESULTS
Binding Orientation of N -Substituted-L-Arginines-It has previously been shown that N -methyl-L-arginine (L-NMA), the prototypic NOS inhibitor (23), is a pseudo-substrate that is hydroxylated on the methyl-substituted guanidinium nitrogen (34 -36). It is also reported that asymmetrical N ,N -dimethyl-L-arginine, in which one guanidinium nitrogen is dimethylsubstituted, binds as an effective iNOS inhibitor, but that symmetrical N ,N Ј-dimethyl-L-arginine, in which each guanidinium nitrogen is monosubstituted, is a poorly bound, weak inhibitor (23,37). We have confirmed the latter results with nNOS (data not shown), and from these observations infer (i) that L-NMA binds with its unsubstituted guanidinium nitrogen in the distal, non-reactive guanidinium nitrogen pocket, (ii) that the binding region for the proximal, reactive guanidinium nitrogen can accommodate both monomethyl-and dimethylsubstitution, and (iii) that the distal pocket cannot accommodate even monomethyl-substitution.
As shown in Table I, N -monoalkyl-L-arginines with ethyl and propyl substituents are also moderately strong inhibitors, whereas the monobutyl derivative is less well bound, especially to iNOS. 2 These results suggest that the binding region for the guanidinium nitrogen near heme can accommodate substituents extending about 4 -5 Å from the guanidinium carbon. Note that the selectivity of N -propyl-L-arginine for inhibition of nNOS has been previously reported by Zhang et al. (38), although our results do not confirm the very large selectivity factors seen in their studies.
Binding Orientation of L-NIO-In L-NIO one guanidinium -NH 2 group of L-arginine is replaced by -CH 3 ; the resulting amidine side chain has a pK a of ϳ12, assuring its protonation at physiological pH. In principle, L-NIO could bind to NOS with either the ϭNH 2 ϩ or -CH 3 group in the binding region proximal to the heme cofactor. Based on our results with L-NMA homologs and the observation that inhibition by L-NIO is competitive with L-arginine ( Fig. 2A), we anticipated that L-NIO homologs containing alkyl groups larger than -CH 3 would be forced to bind with their alkyl groups near heme and their ϭNH 2 ϩ group in the sterically constrained distal guanidinium pocket. Binding affinity would be expected to fall off as the alkyl group became larger than 4 -5 carbons. As shown in Table  I, these expectations were realized; ethyl-L-NIO with a 3-carbon alkyl group inhibits nearly as well as L-NIO, but butyl-L-NIO with a 5-carbon alkyl group is poorly bound. In earlier studies we showed with all three NOS isoforms that methyl-L-NIO exhibits K i values that are intermediate between those seen with L-NIO and ethyl-L-NIO (32). This smooth progression in K i values as alkyl group size increases suggests that, as expected, L-NIO as well as its higher homologs binds with the amidine alkyl group positioned near heme and ϭNH 2 ϩ in the distal guanidinium nitrogen binding pocket.
N 5 -Thioacyl-L-Ornithines as NOS Inhibitors-L-TAO is a novel L-NIO analog in which ϭNH 2 ϩ is replaced by ϭS. Higher homologs of L-TAO were also prepared. With respect to alkyl chain length, L-TBO corresponds to ethyl-L-NIO and L-THO corresponds to butyl-L-NIO (Fig. 1). As shown in Fig. 2B, L-TAO-mediated inhibition of nNOS is competitive with respect to L-arginine, suggesting that interaction is with the L-arginine-binding site. The K i for L-TAO is 34.8 M, a value ϳ20-fold higher than that observed in similar studies with L-NIO ( Fig.  2A). As shown in Table I, with nNOS the IC 50 value for L-TBO is ϳ1.4-fold higher than that for L-TAO, and inhibition by L-THO is too weak to be usefully quantitated. With iNOS and eNOS, L-TAO and the higher N 5 -thioacyl-L-ornithines inhibit very poorly at accessible concentrations.

TABLE I Inhibition of NOS isoforms by arginine analogs
Inhibition of individual, purified NOS isoforms was determined using the L-[ 14 C]arginine assay as described under "Experimental Procedures." Final L-arginine concentrations were 20, 100, and 30 M for nNOS, iNOS, and eNOS, respectively. These values are each ϳ10ϫ the K m Arg for the individual isoform assayed (see "Experimental Procedures"), and it is thus possible compare values among isoforms as well as among L-arginine analogs. Inhibitor N 5 -Acyl-L-Ornithines as NOS Inhibitors-N 5 -Acetyl-L-ornithine, N 5 -butyryl-L-ornithine, and N 5 -hexanoyl-L-ornithine are analogs of L-NIO, ethyl-L-NIO, and butyl-L-NIO, respectively, in which the ϭNH 2 ϩ group of the NIO derivatives is replaced by ϭO (Fig. 1). When tested as NOS inhibitors in studies similar to those shown in Table I, none was an effective inhibitor and all exhibited IC 50 values Ͼ Ͼ1 mM (data not shown).
L-Canavanine as a NOS Inhibitor-L-Canavanine is an Larginine antagonist in which the ␦-methylene group is replaced by oxygen. The resulting hydroxyguanidine has a pK a of 7.0 and adopts the imino structure wherein the guanidinium double bond is orientated toward the N 5 nitrogen (Fig. 1) (39). In consequence, the L-canavanine side chain is not fully protonated at neutral pH and the N 5 nitrogen does not bear a proton. Although L-canavanine is reported to be an iNOS selective inhibitor (40,41), we find it inhibits all three isoforms (data for iNOS and eNOS not shown); Fig. 2C shows that inhibition of nNOS is competitive with L-arginine and characterized by a K i of 11.1 M at pH 7.4.
To assess the effects of L-canavanine side chain protonation on binding, we determined K i values for L-canavanine and L-NMA and also determined K m Arg as a function of pH (Fig. 3). As shown, affinity for L-arginine, L-canavanine, and L-NMA all decrease with decreasing pH. Although L-canavanine ranges from 24% protonated at pH 7.5 to 90% protonated at pH 6.0, its binding affinity closely tracks that of L-arginine and L-NMA, both of which remain Ͼ99.9% protonated over this pH range. At pH values Ͼ7.5, nNOS shows decreased affinity for L-canavanine but little change in affinity for L-arginine and L-NMA.
Possible Direct Interaction between Heme Iron and L-Arginine Analogs-As isolated, about 80% of nNOS contains highspin pentavalent heme iron (i.e. 4 bonds to the pyrrole nitrogens of protoporphyrin IX and 1 bond to the sulfur of Cys-415); the remainder of nNOS as isolated contains low-spin, hexavalent heme iron in which an unknown ligand occupies the sixth axial position that is occupied by O 2 during the NOS catalytic cycle (42,43). When L-arginine, L-NMA, and most amino acid inhibitors bind to NOS, they displace the unknown sixth axial ligand but do not themselves bind to heme iron; the transition to 100% high spin NOS is detected as a type I optical difference spectrum (42)(43)(44)(45). As shown in Fig. 4A, L-NIO binds in this manner, giving a type I difference spectrum. In contrast, other inhibitors interact covalently with heme iron, increasing the fraction of low-spin heme and producing a type II optical difference spectrum (46,47). One such inhibitor is L-thiocitrulline, for which one binding mode includes a bond between its thioureido sulfur and heme iron (47).
L-TAO is a structural analog of L-thiocitrulline as well as L-NIO (Fig. 1). Whereas steric constraints force the higher homologs of L-TAO (i.e. L-TBO and L-THO) to bind with their sulfur in the guanidinium-binding region distal to heme, L-TAO might bind in the reverse orientation with -CH 3 in the distal pocket and ϭS near heme. If so, the sulfur of L-TAO, like the sulfur of L-thiocitrulline, might bind as a sixth axial heme iron ligand and produce a type II optical difference spectrum. However, as shown in Fig. 4B, addition of L-TAO to nNOS produces a type I difference spectrum. The similarity in the spectra for L-TAO and L-NIO, which also has a -CH 3 group in the proximal guanidinium-binding region, is evident. This result is consistent with the view that L-TAO binds with its -CH 3 group rather than ϭS in the guanidinium-binding region near heme; the distal, non-reactive guanidinium pocket thus favors ϭS over -CH 3 .
L-Citrulline is an extraordinarily weak inhibitor of iNOS (46) and eNOS (K d Ͼ 200 mM) (48) but, when bound, it causes a type II optical difference spectrum (46,48). As shown in Fig. 5A, L-citrulline also causes a type II difference spectrum when bound to nNOS, a result suggesting that the ureido oxygen is bound to heme iron as a sixth axial ligand. Because N 5 -acetyl-L-ornithine can be viewed as a L-citrulline analog in which -CH 3 replaces -NH 2 (Fig. 1), we determined if exposure of nNOS to high concentrations of N 5 -acetyl-L-ornithine would cause a type II spectrum. As shown in Fig. 5B, type I spectra are obtained, suggesting that the analog binds with its -CH 3 moiety rather than ϭO near iron. Confirming this result, N 5butyryl-L-ornithine also gives a type I difference spectrum (Fig.  5C), and its K s (0.28 mM) is similar to that for the acetyl derivative (K s ϭ 0.36 mM).

DISCUSSION
High resolution x-ray crystallographic structures for the oxygenase domains of all 3 mammalian NOS isoforms have now been reported in publications (49 -51) or at meetings. As expected, the active sites show a high degree of homology, but there are subtle differences (49 -52). Consistent with conclusions from substrate specificity (20,37) and ENDOR (3) studies, the distal, non-reactive guanidinium nitrogen of L-arginine is found to bind Ͼ4 Å from heme iron in a sterically constrained pocket 3 ; a conserved enzymatic glutamate residue, identified earlier in mutagenesis studies (53,54), forms H-bonds to protons on the distal guanidinium nitrogen and the N 5 nitrogen of L-arginine. Unanticipated from earlier work, binding of the distal guanidinium nitrogen of L-arginine is also stabilized by an H-bond to the backbone carbonyl of a conserved tryptophan residue (Fig. 6) (49 -51). The present studies further elucidate

FIG. 3. Plot showing the K i values for L-canavanine and L-NMA and K m
Arg as a function of pH. K i values for L-canavanine and L-NMA were determined at the pH values indicated using the hemoglobin assay to measure NO formation. K i values and K m Arg were determined from Lineweaver-Burk plots of rate data obtained at the pH values indicated.
the importance of these interactions to substrate and inhibitor binding.
Although protonation state cannot be determined by x-ray crystallography, Crane et al. (49) plausibly propose that Larginine (pK a ϳ 12) initially binds as a protonated species; proton donation to heme-bound O 2 by the proximal, reactive guanidinium nitrogen of L-arginine then facilitates conversion of O 2 to water and the oxo-heme species required for substrate hydroxylation. The product formed, NOH-Arg, has a much lower pK a (ϳ7) and is presumed to remain unprotonated, thus facilitating reaction of the second heme-bound O 2 as a peroxo rather than as an oxo species (Fig. 6). Because the double bond in NOH-Arg is directed toward the hydroxylated nitrogen (39,49), the distal guanidinium pocket is occupied throughout the catalytic cycle by an -NH 2 or ϭNH 2 ϩ group rather than an ϭNH group; both hydrogens are involved in H-bonds.
As shown in Fig. 6, all of the tightly bound L-arginine analogs can be oriented in the active site so as to place -NH 2 or ϭNH 2 ϩ in the distal guanidinium nitrogen pocket (e.g. the shorter N -alkyl-L-arginines and L-NIO derivatives all have IC 50 values indicating binding comparable to or tighter than L-arginine (Table I) and L-thiocitrulline binds to nNOS and iNOS with K i values of 0.06 and 3.6 M, respectively (47)). The presence or absence of charge is apparently of little consequence since both the cationic amidines (i.e. L-NIO and its derivatives) and neutral L-thiocitrulline bind tightly. Similarly, cationic L-arginine and neutral NOH-Arg have similar K m values with iNOS and, more importantly, k cat /K m is only ϳ50% higher for L-arginine than for NOH-Arg (55). At the typical assay pH of 7.4, Lcanavanine possesses a mostly uncharged side chain (pK a ϳ 7.0) and binds somewhat less tightly than L-arginine despite being isosteric and able to form all of the H-bonds characteristic of the distal guanidinium pocket (Fig. 6). However, presence of oxygen adjacent to N 5 is known to orient the guanidinium double bond into the side chain (39), and consequent loss of the N 5 proton abolishes the H-bond between N 5 and the enzymatic glutamate residue. 4 The modestly decreased binding affinity of nNOS for L-canavanine is attributed to loss of this H-bond. Comparison of the K i value of L-canavanine with K m Arg (viewed as a "kinetic" binding constant) suggests that the difference in free energy of binding between these species is ϳ1.1 kcal/ mol. 5 This value is much smaller than the average H-bond energy measured in vacuo (3-19 kcal/mol (57)), but is within the range of values typically observed in enzyme-substrate and enzyme-inhibitor interactions where H-bond formation to a ligand is balanced by the loss of H-bonds with water. Such bonds typically contribute 0.5-1.5 kcal/mol of binding energy when uncharged groups are involved and 3-6 kcal/ mol when at least one charged group is involved (58). Because the H-bond lost in L-canavanine involves a presumably unprotonated, anionic glutamate residue (49) (Fig. 6), the observed change in binding affinity is ϳ2 kcal/mol less than predicted, but we are reluctant to invoke specific compensating factors with the limited data available. Interestingly, at lower pH the K i values for L-canavanine and L-NMA and K m Arg all increase, but there is no improvement in the relative affinity of L-canavanine (Fig. 3). This result suggests that protonation of L-canavanine does not restore the missing H-bond between N 5 of the inhibitor and the enzymatic glutamate residue but rather, as predicted by Boyar and Marsh (39), L-canavanine protonates on a terminal nitrogen. Furthermore, the similarity in the shapes of the K i or K m versus pH profiles for L-canavanine, L-NMA, and Larginine at low pH suggests that the enzymatic glutamate residue does not protonate even at pH 6 to restore an H-bond 4 Absence of the N 5 proton and H-bond has been independently described based on an x-ray crystallographic structure of L-canavanine bound to the eNOS oxygenase domain (C. S. Raman, H. Li, P. Martá sek, V. Krá l, B. S. S. Masters, and T. L. Poulos, submitted for publication). 5 Free energy changes were estimated from the van't Hoff equation, ⌬G ϭ -2.3 RT log(K Eq 1 /K Eq 2 ). In most cases 1/K i were used as surrogate K Eq values, but comparable results were obtained using 1/K s or, for L-arginine, 1/K m . It has earlier been shown the K s Arg is similar to K m Arg (0.7 versus 1.8 M, respectively) (J. C. Salerno, personal communication). We also limit our analysis to comparison of reasonably isosteric amino acids where differences in binding energy can be confidently attributed to the presence or absence of specific H-bonds. Other classes of inhibitors (e.g. the non-amino acid isothioureas) include compounds that bind with high affinity and yet form few H-bonds (50,56). The affinity of NOS for such compounds depends on interactions not present in the amino acids studied here, and differences in binding affinities cannot therefore be attributed to gain or loss of specific H-bonds. with N 5 of L-canavanine. 6 The apparent loss of affinity for L-canavanine at pH Ͼ 8 is distinct from what is seen with L-NMA and L-arginine and is not presently understood in terms of specific interactions.
Although earlier mutagenesis studies identified Glu-371 of iNOS (53) and Glu-361 of eNOS (54) as important for L-arginine binding, those results could not be unambiguously interpreted until x-ray crystallographic studies established that H-bonds to that residue stabilize the ␣-amino, distal guanidinium and N 5 -nitrogens of L-arginine (49 -51). For both iNOS and eNOS, replacement of the glutamate residue with alanine (iNOS) or leucine (eNOS) completely abolished L-arginine binding (53,54), a result indicating the critical importance of the H-bonds made to that residue. 7 In the present studies we have modified the substrate rather than the enzyme and have been able to selectively probe the importance of specific H-bonds. Our results suggest that H-bonds to the distal guanidinium nitrogen or to the N 5 -nitrogen are likely to be important in establishing guanidinium group orientation, but they are relatively weak (Ͻ2 kcal) and therefore not individually essential to binding per se. Thus, by taking advantage of the unique ability of the proximal, reactive guanidinium nitrogen-binding site to accommodate n-alkyl substituents as large as 3-4 carbons, we have designed L-arginine analogs that can bind only if the distal guanidinium pocket is occupied by groups other than -NH 2 or ϭNH 2 ϩ . We have shown that some of those analogs bind despite being unable to form H-bonds to the enzymatic glutamate and tryptophan residues identified by x-ray crystallography (49 -51). Because we have spectral data only with nNOS, we limit our detailed analysis to that isoform and then comment briefly on differences with iNOS and eNOS.
L-TAO is a weak competitive inhibitor of nNOS; its K i is 34.8 M, 21-fold higher than K m Arg . As shown in Fig. 6, L-TAO might bind in either or both of two conformations. Conformation B, in which -CH 3 occupies the distal guanidinium nitrogen pocket and ϭS occupies the region near heme, is analogous to the preferred binding mode of L-thiocitrulline (47,49,59). Because L-thiocitrulline (K i ϭ 0.06 M (47)) binds much more tightly than L-arginine (K m ϭ 1.8 M), L-NIO (K i ϭ 1.7 M (this work and Ref. 32)), or L-citrulline (K s ϳ 25 mM (this work)), it is clear that occupancy of the region near heme by ϭS rather than -NH 2 , -CH 3 , or ϭO is energetically favorable. That said, optical difference spectroscopy with L-TAO (Fig. 4B) provides no evidence for the sulfur-heme interaction seen with L-thiocitrulline. Although that finding argues against binding in conformation B, we note that sulfur-heme bonds are seen for only 10 -20% of the L-thiocitrulline bound to nNOS and are not seen with L-homothiocitrulline, which is nonetheless thought to bind analogously (47,59). It is thus possible that some L-TAO binds in conformation B but with its sulfur atom sufficiently far from heme iron to prevent covalent interaction. If that is the mode of binding, the loss of binding energy attributable to placing -CH 3 rather than -NH 2 in the distal guanidinium nitrogen pocket can be calculated from the difference in K i values between L-TAO and L-thiocitrulline (34.8 M versus 0.06 M (47)). The ⌬G difference is ϳ3.8 kcal/mol, somewhat lower than the 6 -12 kcal/mol expected for the loss of two H-bonds involving charged groups (Fig. 6).
Conformation A for L-TAO (Fig. 6) places ϭS in the distal guanidinium nitrogen pocket and -CH 3 near heme; in this conformation L-TAO can be viewed as an analog of L-NIO in which ϭS replaces ϭNH 2 ϩ . If the observed loss of binding affinity is fully attributed to this structural difference, then comparison of K i values (34.8 M versus 1.7 M) suggests a ⌬G difference of ϳ1.8 kcal/mol. This is again a surprisingly small value considering two H-bonds involving charged groups are lost (i.e. the enzymatic glutamate is thought to be ionized and the amidinium group of L-NIO is certainly cationic). Thus, independent of binding conformation, the L-TAO results suggest that the H-bonds involving residues in the distal guanidinium pocket are relatively weak.
For the higher homologs of L-TAO, there is no ambiguity in binding conformation; the size of the alkyl substituent precludes conformer B-type binding. Calculating K i values from the data in Table I (Table I) or 5.3 M (32)) suggests that binding energy decreased ϳ1.05 kcal/mol in L-TBO. This value is again substantially smaller than would be expected for loss of two H-bonds involving charged groups. However, the relatively close similarity between this value and ϳ1.8 kcal/mol estimated for L-TAO binding in conformation A supports the view that conformation A does, in fact, best represent L-TAO binding. Although we cannot rigorously exclude the possibility that some L-TAO binds in conformation B, the simplest interpretation of our results is that it is energetically more favorable to place ϭS rather than -CH 3 in the distal guanidinium binding pocket. This selectivity is unlikely to be attributable to the small difference in van der Waals radii between the groups (1.85 Å for ϭS versus 2.0 Å for -CH 3 (60)), but may reflect the much greater polarizability of sulfur and its ability to form bonds based on dispersion forces. Electron spin resonance studies by Salerno et al. (61) showed that L-citrulline binds as a sixth axial ligand to heme iron in nNOS; the major interaction is through ϭO but a minority species shows iron bonded to the -NH 2 group. That result, coupled with the poor overall affinity for L-citrulline (K s ϳ 25 mM (Fig. 5A)), suggests (i) that the binding energy for L-citrulline bound with ϭO near heme is similar to the binding energy with ϭO in the distal guanidinium binding pocket, and (ii) that ϭO is very poorly bound at either site. N 5 -Acetyl-L-ornithine is a L-citrulline analog and, like L-citrulline, is weakly bound. However, N 5 -acetyl-L-ornithine binding apparently favors ϭO in the distal pocket since we observed a type I optical difference spectrum with the compound (Fig. 5B). Confirming that view, N 5 -butyryl-L-ornithine, which for steric reasons must bind with ϭO in the distal pocket, also gives a type I spectrum (Fig. 5C), and the K s values for the acetyl and butyryl compounds are comparable (0.37 and 0.28 mM, respectively). Comparison of these values to the K i values for L-NIO and ethyl-L-NIO (or the K s for L-NIO) indicates that binding affinity is decreased by 2.7-3.2 kcal/mol (N 5 -acetyl-L-ornithine versus L-NIO) or 2.4 kcal/mol (N 5 -butyryl-L-ornithine versus ethyl-L-NIO). These are again unexpectedly low values considering two H-bonds involving charged groups were lost. Taken together with the similar findings for the corresponding sulfur derivatives (TAO and TBO), our results suggest that the two H-bonds involving 6 That is, if the glutamate residue were to protonate as the pH decreased from 7.5 to 6.0, then the missing H-bond could be restored, forming between the glutamate proton and the unprotonated N 5 of L-canavanine; affinity for L-canavanine would be expected to improve at lower pH relative to L-NMA or L-arginine. This is not observed. Similarly, if the glutamate residue were protonated even at neutral pH, the missing proton on L-canavanine would not affect the number of H-bonds stabilizing binding, and the affinity for L-canavanine would be expected to match that for L-arginine. It does not. 7 In eNOS replacement of Glu-361 with glutamine also abolished activity (54). Since glutamine could form as many H-bonds as glutamate, we tentatively attribute loss of activity in the E361Q mutant to a conformational misalignment of the glutamine residue. Loss of charge interaction with the protonated guanidinium group of L-arginine is an alternative explanation favored by P-F. Chen et al. (54), but, as discussed above, native NOS binds both cationic and neutral L-arginine analogs and we think this explanation is less likely. It would be interesting to know if the E361Q mutant binds N -nitro-L-arginine, a tightly bound, but uncharged inhibitor. the distal guanidinium nitrogen of L-arginine are much weaker (0.9 -1.5 kcal/mol each) than would have been predicted for such bonds (3-6 kcal/mol each). Such relative weakness may reflect an unfavorable H-bond length or very favorable interactions of glutamate and tryptophan with water or other residues that are then lost on binding L-arginine or its analogs. Although bond lengths are not yet reported for nNOS, the corresponding bonds in eNOS are 3.19 Å (Glu-363 . . . . H 2 ϩ Nϭ) and 2.88 Å (Trp-358 . . . . H 2 ϩ Nϭ) 8 ; at least the former is too long for maximum strength.
Taken together our results indicate that the distal guanidinium nitrogen binding pocket of nNOS has highest affinity for -NH 2 and ϭNH 2 ϩ , moderate affinity for ϭS, substantially less affinity for ϭO, and very little affinity for -CH 3 . The binding region near heme (i.e. the proximal guanidinium nitrogen site) has highest affinity for ϭS (and -S-alkyl (45,62)), very little affinity for ϭO, and moderate affinity for alkyl, -NHalkyl, and -NH 2 or ϭNH 2 ϩ (approximately in that order). The results shown in Table I suggest iNOS and eNOS differ from nNOS in several respects. Most dramatically, iNOS is less able to bind N-alkyl groups larger than ethyl near heme, and eNOS and, to a lesser extent, iNOS do not accommodate ϭS in the distal guanidinium pocket as well as nNOS does. Fan et al. (52) similarly concluded from resonance Raman studies that the nNOS L-arginine-binding site is more open than that of iNOS and eNOS (52).
Finally, we note that most NOS inhibitors designed or discovered to date bind competitively with L-arginine and are therefore presumed to occupy at least part of the L-argininebinding site. With the exception of certain aromatic compounds (e.g. 7-nitroindazole), almost all such inhibitors include Structure I as a structural motif. The present results strongly suggest that the ϭNH 2 ϩ moiety of that motif is bound in the distal guanidinium pocket. Although that binding region will accommodate other groups as long as they are not larger than ϭNH 2 ϩ , even the best of the surrogates (e.g. ϭS for nNOS) is bound with substantially less affinity. We conclude that inhibitors targeting the guanidinium region of the L-arginine-binding site should include a guanidine, amidine, thiourea, or isothiourea motif.