Development of a reactive oxygen species-sensitive nitric oxide synthase inhibitor for the treatment of ischemic stroke

Abstract

Ischemic stroke is caused by a blockage of cerebral blood flow resulting in neuronal and glial hypoxia leading to inflammatory and reactive oxygen species (ROS)-mediated cell death. Nitric oxide (NO) formed by NO synthase (NOS) is known to be protective in ischemic stroke, however NOS has been shown to ‘uncouple’ under oxidative conditions to instead produce ROS. Nitrones are antioxidant molecules that are shown to trap ROS to then decompose and release NO. In this study, the nitrone 5 was designed such that its decomposition product is a NOS inhibitor, 6, effectively leading to NOS inhibition specifically at the site of ROS production. The ability of 5 to spin-trap radicals and decompose to 6 was observed using EPR and LC-MS/MS. The pro-drug concept was tested in vitro by measuring cell viability and 6 formation in SH-SY5Y cells subjected to oxygen glucose deprivation (OGD). 5 was found to be more efficacious and more potent than PBN, and was able to increase phospho-Akt while reducing nitrotyrosine and cleaved caspase-3 levels. 6 treatment, but not 5, was found to decrease NO production in LPS-stimulated microglia. Doppler flowmetry on anesthetized mice showed increased cerebral blood flow upon intravenous administration of 1 mg/kg of 5, but a return to baseline upon administration of 10 mg/kg, likely due to its dual nature of antioxidant/NO-donor and NOS-inhibition. Mice treated with 5 after permanent ischemia exhibited a > 30% reduction in infarct volume, and higher formation of 6 in ischemic tissue resulting in region specific effects limited to the infarct area.

1. Introduction

Stroke remains one of the leading causes of death and disability in the United States, most notably ischemic stroke which accounts for over 80% of all stroke cases. Despite decades of promising research, clinical trials of potential therapeutics have not successfully led to novel treatments for the devastating effects of cerebral ischemia [1–3]. Ischemic stroke is associated with inflammatory and free radical-mediated cell death due to obstruction of blood flow to the brain. Reactive oxygen species (ROS) such as superoxide (O₂^•−) are generated under hypoxic and inflammatory pathologies by the mitochondrial electron transport chain (ETC), NADPH oxidase, and xanthine oxidase (XO). This oxidative burst causes direct DNA, protein, and cell membrane damage which ultimately leads to cell death in the core of the ischemic lesion. Additionally, O₂^•− is known to cause vasoconstriction independent of its effect on nitric oxide (NO) [4], which is detrimental to patient outcome. Increased levels of oxidative stress markers and decreased levels of antioxidant markers have a strong correlation with worse ischemic stroke outcome [5].

NO is a free-radical, paracellular signaling molecule produced by Ca²⁺-dependent endothelial NO synthase (eNOS) and neuronal NOS (nNOS), and inducible NOS (iNOS). Canonical NO signaling involves the increase in cGMP through activation of guanylyl cyclase to cause vasodilation through vascular smooth muscle relaxation. It has been recently discovered that NO is also involved in GABA [6] and 5-HT neurotransmission [7], neurogenesis [8,9], mitochondrial signaling [10], and post-translational protein modification [11]. During cerebral ischemia a dramatic increase in intracellular Ca²⁺ causes increased NOS activity through calmodulin binding concomitant with the generation of ROS and activation of pro-apoptotic pathways. The overall role of NO in ischemic stroke is still unclear; inhaled NO has shown pre-clinical efficacy in cerebral ischemia [12,13], and endogenous NO generated by NOS is observed to have similar beneficial effects [14], yet NOS can uncouple under ischemic conditions through direct oxidation or oxidation of cofactors to instead produce O₂^•−, which is scavenged by NO to form the highly reactive peroxynitrite (ONOO⁻). This phenomenon was initially observed with eNOS [15], but has since been detected from each NOS isoform [16,17]. Importantly, Guzic et al. were able to show the pan-NOS inhibitor Nω-nitro-L-arginine methyl ester (L-NAME) subdued O₂^{•− ·}production in tissues with NOS dysfunction [18]. L-NAME administered to tissue with functional NOS increased O₂^•− production however, which suggests a targeted approach to dysfunctional NOS inhibition would be preferable to total NOS inhibition. Isoform-specific inhibitors for nNOS or iNOS have shown reduced in vivo stroke volume in animal models, while pan-NOS inhibitors exhibit mixed results [19]. This may be due to the inhibition of functional NOS as well as uncoupled NOS which can cause global adverse effects that limit their protective effects in ischemia.

Antioxidant molecules that scavenge oxidative species such as Edaravone [20], Ebselen [21], and uric acid [22] have shown pre-clinical promise in ischemia, but have fallen short in clinical applications. The nitrone NXY-059 had advanced to phase III clinical trials, showing good safety and tolerability before it was deemed ineffective in the treatment of acute ischemic stroke [23]. Nitrones (Fig. 1A) are synthetic antioxidants that are shown to both scavenge ROS and produce NO upon decomposition [24] to yield neuroprotective effects under ischemic conditions. Linear, N-tert-butyl-α-phenylnitrone (PBN)-type nitrones have been extensively developed for use in pre-clinical models of stroke including innovative derivations such as a tetramethylpyrazine nitrone (TBN), which possesses anti-platelet activity [25], and a PBN-containing, PEGylated nanoparticle capable of extended exposure [26].

Fig. 1.

(A) Structures of linear nitrones and NOS-inhibitor. (B) Proposed general scheme for the formation of 6 after the reaction of 5 with oxygen-centered free radicals (•OX).

Synthetic nitrones and NOS inhibitors have thus far failed to yield favorable clinical outcomes in diseases mediated by oxidative pathologies. This study aims to resolve their limitations through a pro-drug approach to the elimination and prevention of oxidative species. The molecule described herein, 5, is a PBN-type nitrone designed based on its radical adduct decomposition product. This decomposition product is a putative NOS inhibitor that will effectively be formed at the site of oxidative stress, selectively targeting dysfunctional, uncoupled NOS. In this study we use an in silico computational approach to design 5 and its product 6 before advancing to in vitro and in vivo models of ischemic brain injury.

2. Results

2.1. Thermodynamics of free radical addition to 5

It has been shown that oxygen-centered radical addition to PBN nitrones causes nitronyl bond cleavage to form a tert-butyl hydroxylamine and a benzaldehyde [27]. The amidine-containing NOS inhibitor 1400W exhibits strong selectivity for iNOS, while also showing inhibition of nNOS and eNOS at higher concentrations. For this proof-of-concept study, the PBN-type nitrone 5 was designed with the goal of forming the putative NOS inhibitor 6, modeled from 1400W, after reaction with oxygen-centered radicals (Fig. 1B). Amidine moieties typically exhibit an NH pKa > 10, thus the protonated forms of 5 and 6 were used for all in silico calculations. Multiple biologically relevant oxygen-centered free radicals were chosen for calculation of reaction thermodynamics. Fig. 2 shows calculated geometries of 5 and adducts formed from O₂^•−/HO₂^• (pKa 4.8), ^•OH, and ONOO⁻/ONOOH (pKa 6.8). Calculated ΔG_rxn for radical addition to 5 and PBN are summarized in Table 1.

Fig. 2.

(A-C) Optimized structures including bond lengths, and charge and spin densities (in parenthesis) for 5, its proposed free radical adducts, and calculated ΔG_rxn, formed from O₂^•−/HO₂^•, and HOONO (triplet products) at the PCM/B3LYP/6–31+G(d,p)//B3LYP/6–31G(d) level of theory. (D-E) Most energetically favored binding structures of 5 (D) and 6 (E) with nNOS (PDB: 1QWC) calculated by Swissdock.ch and rendered by UCSF Chimera.

Table 1.

Calculated free energies of reaction (ΔG_298K, kcal/mol) of oxygen-centered free radical species to form nitronyl spin adducts at the PCM/B3LYP/6–31+G(d,p)//B3LYP/6–31G(d) level of theory.

Radical Adduct	PBN	5
-O2−	19.8	1.6(12.6)a
-O2H	2.1	3.2
-OH	−32.7	−31.1
-OONOb	32.5	−6.7
-HOONOc	−23.4	−20.4

^aWith two explicit water molecules.

^bTriplet products of a O-3 cis-OONO adduct.

^cTriplet products of a O-3 trans-HOONO adduct.

Optimized geometries of O₂^•− addition to 5 calculated by Gaussian 09 predict proton abstraction of the amidine-NH to protonate the peroxyl anion adduct. While this resulted in a lower ΔG_rxn (1.6 kcal/mol), it is more likely that the peroxyl anion would abstract a proton from solution. Indeed, when the calculation was performed with two explicit water molecules, the ΔG_rxn of 5-O₂^•− was closer to that of PBN-O₂^•− (12.6 kcal/mol and 19.8 kcal/mol, respectively).

Peroxynitrite and its conjugate acid ONOOH are highly oxidizing species that have been shown to react with nitrones to yield radical products as detected by EPR [28]. The ΔG_rxn of potential triplet products were calculated, and predicts a favor able reaction of 5 with the relatively stable cis-ONOO⁻ to give NO₂, as compared to an unfavorable reaction with PBN (−6.7 kcal/mol and 32.5 kcal/mol, respectively). Overall, the effect of the H-bond donating amidine group is predicted to have favorable effects on the ΔG_rxn with anionic ROS. This is in line with previous observations with nitrones containing intramolecular H-bond donors [29,30].

2.2. Predictive binding of 5 and 6 with NOS isozymes

To predict if the nitrone 5 and its decomposition product 6 interact with the active site of each NOS isoform, docking using Swissdock.ch was performed. PDB structures of each NOS isoform bound with native ligand L-arginine, and those with bound 1400W were used (6 structures total). Results of predicted interaction with the orthosteric binding site are summarized in Table 2. The ΔG of ligand binding for 5 and 6 were found to be similar for each NOS isoform; the most favorable interaction with iNOS, and least favorable with nNOS. Both molecules gave similar predicted ΔG as that of L-arginine for each structure, with the exception of nNOS in which 5 and 6 had more favorable predicted binding for both nNOS structures. 1400W showed the most favorable binding for each structure, although seemingly similar to L-arginine for iNOS binding ΔG in spite of its high affinity (K_i = 7 nM). Upon calculating favorable interactions with NOS active sites and promising ΔG_rxn for free radical reactions, the compound 5 was advanced for synthesis.

Table 2.

Predicted free energies of binding (ΔG_298K, kcal/mol) of ligands with NOS isozymes (L-arginine/1400 W X-ray crystallographic structures)^a.

Ligand	eNOSb	iNOSc	nNOSd
L-arginine	−10.32/−9.31	−10.64/−10.21	−7.78/−8.43
1400 W	−10.37/−10.34	−10.68/−11.56	−8.62/−10.26
5	−9.99/−10.11	−10.07/−10.75	−8.29/−9.88
6	−9.73/−9.88	−9.76/−10.09	−8.83/−9.77

^aDocking performed by Swissdock.ch [3,8].

^beNOS L-arginine PDB: 4D1O; 1400 W PDB: 1FOI.

^ciNOS L-arginine PDB: 1NSI; 1400 W PDB: 1QW5.

^dnNOS L-arginine PDB: 4D1N; 1400 W PDB: 1QWC.

2.3. Synthesis and characterization of 5

The four-step synthesis of 5 was carried out with a favorable overall yield of 52.5% (Fig. 3). The intermediates allow for multiple points of derivatization for the synthesis of a focused library. Activation of the boc-protected benzyl alcohol 1 allows for substitution with tert-butyl amine (3), then further deprotection permits substitution with an amidine (4). These two steps can be used for future derivatization to modulate activity of the nitrone and NOS-inhibitor aspects of the molecule, respectively. The final step is oxidation of the secondary amine to the nitrone 5 by NaWO₄-H₂O₂. While no oxidation of the amidine was observed, other elaborations may be sensitive to this oxidation step. Analysis by ¹H and ¹³C NMR, HRMS, and HPLC-PDA showed the successful synthesis of 5 at > 95% purity.

Fig. 3.

Synthesis of 5. Reagents: (a.) CBr₄, PPh₃, CH₂Cl₂, 0 °C→rt 30 h; (b.) tert-butylamine, K₂CO₃, DMF, 3 h; (c.) (i) TFA, CH₂Cl₂, 0 °C, 2 h; (ii) ethyl acetimidate, EtOH, rt, 12 h; (d.) H₂O₂, Na₂WO₄(H₂O)₂, MeOH, 0 °C→rt, 12 h.

2.4. EPR spin trapping of 5 and formation of 6

EPR characterization of radical adducts of 5 were carried out using conventional ROS systems. Spectra for PBN were obtained using the same systems, and matched previous observations (Supplementary Fig. S2). Resulting EPR spectra from addition of 5 to O₂^•−/HO₂^•, ^•OH/^•CH₃, and ONOO⁻/HCl systems are shown in Fig. 4. Solutions of O₂^•− and ^•OH generated oxygen and carbon-centered radicals whose spectra closely resemble that of PBN. There was no observable difference in signal intensity between 5 and PBN, which suggests a similar reactivity, and is in agreement with the in silico ΔG_rxn calculations above. Interestingly, an identical three-line EPR signal was observed for both 5 and PBN in the presence of ONOO⁻/HCl (no signal was observed for nitrone or ONOO⁻/HCl alone). Due to the absence of H_β splitting, this spectra is hypothesized to pertain to a tert-butyl nitroso formed after nitronyl N-C bond cleavage.

Fig. 4.

X-Band EPR spectra and simulations of 5 (35 mM) in the presence of (A) O₂^•−/HO₂^• (KO₂ in PBS/DMSO; HO₂^· adduct), (B) ^•OH/^•CH₃ (Fe²⁺/H₂O₂ in PBS/DMSO; ^•OH adduct, ^•CH₃ adduct), and (C) ONOO⁻/HCl (10 mM, pH~6 in PBS/DMSO; nitroxyl) hsfc’s in (SI Appendix Fig. S2). (D-F) HPLC-PDA analysis of 5 in corresponding radical-generating systems after 24 h incubation at 37 °C. The O₂^•−/HO₂^• and ^•OH systems (D, E) show substantial formation of 6, and the ONOO⁻ system (F) showed complete conversion.

Solutions of the 5-ROS systems used to acquire EPR spectra were replicated, and incubated at 37 °C for 24 h before analysis by HPLC-PDA at 254 and 289 nm. A new peak at 3.4 min was observed for each of the three radical systems tested (5 RT = 5.4 min). This more-polar compound was formed in varying amounts for each system, and was found to be the only peak in the 5-ONOO⁻/HCl system. Additionally, the λ_max of the new peak was lower than that of 5 (249 nm vs 289 nm), suggesting a loss of conjugation likely due to nitronyl N˭C modification or cleavage. Low-resolution LC-MS/MS was performed on the 5-ONOO⁻/HCl solution, and found a strong signal of an m/z matching the predicted mass of the aldehyde 6 (M + H⁺ = 177; M + NH₄⁺ = 195; M + K⁺ = 215). This m/z was used to create an MRM analysis by fragmentation. LC-MS/MS analysis of the other 5-ROS solutions using this MRM showed the presence of 6 in each. The ROS systems were also replicated with PBN, and the formation of benzaldehyde was similarly observed (Supplementary Fig. S3), agreeing with previous observations [27].

The 5-ONOO⁻/HCl system was performed in 100% PBS at a larger scale to produce sufficient quantity of 6 for isolation and characterization. The reaction was extracted with CH₂Cl₂, then the aqueous fraction was lyophilized to give a white solid containing 6 and various salts. Addition of ACN to this solid followed by filtration allowed for the isolation of 6. ¹H NMR showed the absence of the tert-butyl group, and a new peak at 10.0 ppm which is characteristic of a benzaldehyde (Supplementary Fig. S1). LC-MS/MS analysis of the isolated 6 yielded the same parent mass and fragmentation pattern as for 5-ONOO⁻/HCl above. These data demonstrate the ability of 5 to trap ROS and decompose to the putative NOS inhibitor 6.

2.5. In vitro neuroprotective and anti-inflammatory properties of 5 & 6

5 was tested in an in vitro model of neuronal ischemia/reperfusion, oxygen-glucose deprivation (OGD), to assess its neuroprotective potential. A dose-finding study was first performed to determine the optimal concentration. SH-SY5Y cells exposed to 1.5 h OGD treated with 5 or PBN (0.05–100 μM) were assayed for viability and plotted as ratio to vehicle control (Fig. 5A). A bi-phasic trend was observed for 5 treatment, with increasing viability from 0.1 to 1.0 μM 1.21–2.57 times greater than vehicle; p < 0.05, p < 0.01 respectively), giving way a decline to baseline at 100 μM. PBN trended in a similar fashion, but did not approach significance at any concentration tested. Additionally, 1.0 μM of 5 showed significantly greater neuroprotection than PBN at the same dose (p < 0.05). 1.0 μM was thus determined to be the optimal dose of 5, and was used in further in vitro studies.

Fig. 5.

(A) In vitro dose-response plot of SH-SY5Y viability after oxygen-glucose deprivation (OGD) and 24 h treatment relative to vehicle control. 5 showed the highest neuroprotection at 1.0 μM, which was significantly higher than that of 1.0 μM PBN (n = 6, p < 0.05). (B) Griess assay of LPS-stimulated SIM-A9 cells following 24 h treatment indicates anti-inflammatory potential for 6 and 1400W, but not 5 (n = 6). (C-E) Western blot analysis of SH-SY5Y cells exposed to OGD and 24 h treatment. 5 treatment (1.0 μM) was found to increase pAkt/Akt ratio (n = 6), and decrease total 3-NT and cleaved-caspase 3 compared to vehicle control (n = 3). Data represented as mean ± SEM, *p < 0.05, **p < 0.01, and ***p < 0.001, from respective vehicle controls, One-way ANOVA followed by Newman-Keuls post-hoc test.

The mechanisms into the neuroprotection afforded by 5 were investigated by Western blotting. SH-SY5Y cells subjected to OGD and treated with 1.0 μM 5 showed a dramatically increased phosphor-Akt (pAkt)/total Akt ratio (4.1 times higher, p < 0.05). This increase was also observed for PBN, and the combination of PBN and 1400W (1.0 μM), albeit this increase did not reach significance from vehicle. Treatment with 5 also caused a reduction in OGD-induced protein nitration. 3-Nitrotyrosine (3-NT), formed due to direct nitration of protein tyrosines by ONOO⁻, was increased by OGD 1.5-fold in vehicle groups, and ameliorated by 48% upon treatment with 5 (p < 0.01). Similarly, the pro-apoptotic cleavage of caspase-3 was induced by OGD, but was significantly reduced with 5-treatment compared to vehicle (47% reduction, p < 0.01).

Microglia express iNOS and convert to an inflammatory phenotype upon stimulation with LPS. 1400W is a potent iNOS inhibitor that has been previously shown to exhibit anti-inflammatory properties [31,32]. To investigate the anti-inflammatory potential of 5 and/or 6, SIM-A9 microglia were stimulated with LPS (100 ng/mL), and treated with vehicle or drug for 24 h. Nitrite, as measured by Griess assay, was increased as expected after LPS, and attenuated by 1.0 μM 1400W (Fig. 5B). 5 was not observed to lower NO production at concentrations as high as 50 μM, however 1.0–50 μM 6 exhibited a dose-dependent decrease in nitrite. This indirect measurement of NO production suggests a reduction in iNOS activity due to 6 treatment, which indicates that 5 needs to be converted to 6 to allow NOS inhibition.

2.6. Effect of 5 on cerebral blood flow (CBF)

Due to the complex interplay of O₂^•−-scavenging, NO donation, and NOS inhibition, the effects of 5 on CBF were investigated. Doppler flowmetry measurement over the middle cerebral artery (MCA) for 20 min after intravenous administration was performed on anesthetized, C57BL/6 mice. A known vasodilator, isoamyl nitrate, was found to increase CBF over 20 min. Paradoxically, the lower dose of 5 (1 mg/kg) caused a significant increase in CBF (126% ± 4.6% baseline, p < 0.001), while a higher dose (10 mg/kg) produced baseline readings (102% ± 0.7% baseline, Fig. 6A). LC-MS/MS analysis confirmed that the animals in the higher-dose group were exposed to larger doses of 5 in plasma (1681 ± 881 nM for 10 mg/kg; 171.7 ± 51.8 nM for 1 mg/kg) and brain (17.97 ± 2.23 nM for 10 mg/kg; 6.41 ± 2.03 nM for 1 mg/kg). A similar effect was previously observed by Inanami and Kuwabara [33] where administration of the NOS-inhibitor L-NAME abrogated an increase in MCA blood flow after PBN administration in rats.

Fig. 6.

(A) Doppler CBF measurement after intravenous injection of isoamyl nitrate or 5 to anesthetized mice over 20 min. The lower dose of 5 (1 mg/kg) elevated CBF, whereas a higher dose of 5 (10 mg/kg) showed no change in CBF from baseline (n = 3, One-way ANOVA followed by Newman-Keuls post-hoc test). (B-D) Neurobehavioral performance of mice treated with vehicle (n = 10) or 5 (n = 12) after pMCAO. No significant differences were observed for rota rod performance (C), but treatment with 5 improved grip strength (D, Two-way ANOVA followed by Bonferroni multiple comparisons test), and reduced neurological deficit scoring (B, unpaired t-test). (E-F) Infarct volume analysis of vehicle (top) vs 5-treated (bottom) mouse brains 72 h after pMCAO by TTC staining (unpaired t-test). Data represented as mean ± SEM, *p < 0.05, **p < 0.01, ***p < 0.001 from vehicle or baseline.

The average brain:plasma ratio after 25 min for the two treatments of 5 was found to be 0.029 ± 0.010 (n = 6). 6 formation was detected by LC-MS/MS using the MRM method attained previously, and was observed to be present at higher levels in the brain than plasma (24.5% and 12.7%, respectively, expressed as percent of 5 AUC within the same run).

2.7. In vivo neuroprotection in a mouse model of permanent ischemia

Animals were subjected to permanent distal MCA occlusion (pMCAO) followed by a bi-phasic dosing regimen of 5 to assess in vivo neuroprotection and effects on neurobehavioral parameters. Vehicle or 5 (10 mg/kg at 3 and 6 h after pMCAO, 1 mg/kg twice daily thereafter) were administered intravenously by lateral tail vein, and neurobehavioral assessments were conducted at 24, 48 and 72 h. Both doses were well tolerated, and there were no observed adverse effects at any time point.

No effect on rotarod performance was observed for both treatment groups, which remained near baseline levels after ischemia through 72 h (Fig. 6). Vehicle treated animals showed a decline in grip strength after pMCAO, while the 5-treated group performed significantly better than vehicle at each time point. Neurological deficit scoring (NDS) was conducted prior to sacrifice at 72 h, in which 5-treated animals showed a less-severe deficit than vehicle treated animals (4.2 ± 0.4 and 5.3 ± 0.4, respectively, p < 0.05, unpaired t-test).

Infarct volume was measured by TTC staining of 2-mm thick coronal brain slices immediately after sacrifice at 72 h post-pMCAO (4 h from last dose). Mice treated with 5 were found to have a 37% decrease in infarct volume compared to vehicle (p < 0.05, unpaired t-test). Directly after imaging, sections of brain regions were rapidly frozen pertaining to ischemic (no TTC staining), penumbra (mild staining), and contralateral (full staining) areas. Preliminary LC-MS/MS analysis of these regions for the levels of 5:6 show decreasing levels of 6 and increasing levels of 5 from the core of the ischemic area outward towards the contralateral area (Fig. 7). This data supports the hypothesis that 5 would be converted to 6 at higher levels in tissues undergoing oxidative stress.

Fig. 7.

LC-MS/MS analysis of 5 and 6 levels in ischemic (A) penumbra (B) and contralateral (C) brain regions of mice 72 h after pMCAO. Brains were rapidly removed after sacrifice by CO₂, sectioned into 2 mm coronal slices, and stained by TTC to visualize the ischemic area. 6 was observed to be formed at higher levels in ischemic tissues, giving way to higher levels of the parent compound, 5, moving outward from the infarct area.

3. Discussion

3.1. Design of a ROS-sensitive NOS inhibitor

NOS presents a unique target in the treatment of ischemia due to seemingly conflicting results from NOS inhibition or genetic knockouts, which have shown both beneficial [19,34–36] and deleterious effects [37–39], and may additionally have sex-specific outcomes [40–42]. Indeed, the precise design criteria for potent and selective NOS inhibitors has seen significant exploration to minimize off-target effects [43]. The co-administration of NOS inhibitors with nitrones for treating neuronal pathology has been previously explored [44], as well as administration of a nitrone to NOS knockout animals in cardio myocytes [45], each of which show complimentary beneficial effects.

The active constituent of NXY-059 used in the SAINT-I/SAINT-II clinical trials for acute ischemic stroke may not have been the drug itself, but a decomposition product, N-tert-butylhydroxylamine [46], formed after the oxidation of the nitrone, and cleavage of the N-C bond. It has been repeatedly shown that the nitrone PBN decomposes to release NO and benzaldehyde under oxidative conditions [27], and that this mechanism may account for much of its therapeutic effects [47]. Bi-functional nitrones are actively being developed to treat ischemic stroke, including a tetramethylpyrazine- functionalized nitrone that was shown to reduce ADP-induced platelet aggregation [25]. However, until now there has not been an attempt to design nitrones from their decomposition products as pro-drug molecules that are activated under oxidative conditions.

With the goal of forming a putative NOS inhibitor, and knowing the oxidative decomposition product would result in formation of a benzaldehyde, the molecule 6 was designed to structurally mimic the NOS inhibitor 1400 W. Amidine-containing NOS inhibitors such as 1400W have been shown to be irreversible NOS inhibitors through a hemeoxygenase mechanism that does not cause breakdown of the inhibitor [48,49]. It was then hypothesized that this benzaldehyde would be formed from the nitrone 5. The structure of 5 and its synthesis are such that future derivatives can be rapidly made to separately modulate ROS reactivity and NOS affinity and selectivity. The screening of future derivatives can begin in silico as outlined herein, by predicting reactivity to biologically relevant ROS, and interaction with NOS isozyme active sites. These approaches have been taken separately for the design of nitrones [29] and NOS inhibitors [43], and yielded predictions that agree with subsequent observations.

3.2. The hit molecule 5 MN and ROS-induced formation of 6

Protein docking of both 5 and 6 to each NOS isoform suggests a similar affinity for the protein active site for both compounds. While this needs confirmation in an in vitro assay, future compounds will ideally possess reduced affinity of the nitrone prodrug for NOS compared to its decomposition product. After synthesis of 5, EPR spectra were acquired of radical adducts which confirm its antioxidant potential. Direct NO measurement was not performed in the ROS systems; however, NO formation from PBN was confirmed to be due to nitrone decomposition in similar systems as well as both in vitro and ex vivo by EPR using ¹⁵N-PBN [50]. Furthermore, analysis of these systems by HPLC-PDA and LC-MS/MS strongly suggests the formation of 6 in each of the ROS systems. The development of an MRM method by LC-MS/MS fragmentation of 6 allowed for the sensitive detection of its formation in vitro and in vivo. The decomposition product 6 was detected in vitro by LC-MS/MS in the supernatant and cell lysate of SH-SY5Y cells exposed to OGD. After OGD, levels of 6 were observed to be similar in supernatant fractions (7.9% and 18.6% of 5 AUC for sham and OGD, respectively), but significantly higher in the lysate of OGD-exposed cells (129% vs. 29% of 5 AUC in sham, respectively, p < 0.001).

The antioxidant properties of 5 were further confirmed through the neuroprotection afforded to SH-SY5Y cells exposed to OGD. The viability increase was observed to reach a maximum at 1.0 μM of 5 before declining back down to baseline at higher concentrations. That 5 was significantly more neuroprotective at 1.0 μM than PBN suggests the amidine substitution of 5 provides some additional beneficial effect, potentially through NOS inhibition. In the SIMA9-LPS system, while iNOS is induced, nitrones like PBN are shown to reduce the expression of iNOS under such inflammatory conditions, but not affect its activity. 5 alone did not affect nitrite production, while 6 showed a dose-dependent reduction, which strongly suggests iNOS inhibition. Thus while in silico calculations predict similar iNOS affinity for 5 and 6, only the latter was observed to exhibit inhibition. This provides support to the hypothesis that 5 needs to be converted by ROS to act as a functional NOS inhibitor. LPS can also stimulate ROS production in microglia [51], however, due to concomitant upregulation of the far-superior radical scavenger MnSOD [52], it is likely that insufficient amounts of 5 were reacted to form 6. Another possible explanation for the lack of nitrite reduction by 5 treatment could be that any conversion of 5 to 6 releases NO which would add to the overall nitrite measured. Further studies will aim to delineate the in vitro antioxidant and NOS inhibition effects of 5 & 6 to elucidate the mechanisms of neuroprotection and anti-inflammation.

Additional evidence of NOS inhibition by 5 and/or 6 was observed through the measurement of CBF by Laser Doppler Flowmetry over the MCA of treated mice. Intravenous injection of 10 mg/kg 5 to anesthetized mice gave no change in CBF from baseline, yet a lower dose, 1 mg/kg, caused a significant increase in CBF over 20 min. When looking for literature precedent, a study by Inanami and Kuwabara [33] showed increased MCA CBF in rats with PBN alone, which was lowered to baseline upon co-administration with L-NAME. It is unclear if the CBF lowering effects herein are due to excess 5, given its predicted affinity for NOS, or the formation of 6, which was detected in plasma by LC-MS/MS. Additionally, levels of 6 were higher than expected after only 25 min exposure, which indicates a short half-life of 5. For this reason, a higher loading dose of 5 was chosen for the pMCAO study, followed by a lower dose- a regimen that has found success for experimental nitrones in ischemia [53].

3.3. 6 forms preferentially in the penumbra and ischemic lesion

Animals treated with 5 exhibited a > 30% reduction in infarct volume 72 h after pMCAO as measured by TTC, coupled with significant improvements in neurobehavioral assay performance. Western blotting was used to investigate the signaling mechanism of the neuroprotection afforded by 5 in vivo and in vitro. The pro-survival pAkt (Ser473) was found to be significantly increased in 5-treated SH-SY5Y cells exposed to OGD. This pathway has been shown to be affected in neurons by both antioxidants [54] and NO-donors [55] in ischemia. PBN was also observed to elicit an increase in Akt phosphorylation, although not significant at the dose selected, and was unaffected by co-treatment with 1400W. 3-NT and cleaved caspase-3 were decreased by 1.0 μM 5, which illustrate reductions in ONOO⁻ production and apoptosis induced by OGD. Correction of the ROS/NO imbalance by 5 & 6 after ischemia and reperfusion is suggested to provide the neuroprotection observed in vitro and in vivo.

NXY-059 was found to increase pAkt/Akt levels in brains of animals after transient MCAO specific to the infarct and penumbra areas [56]. While the authors attributed this to cell surface interactions, it is more likely to be due to increased reactivity of the nitrone in those areas, leading to ROS neutralization and NO release. This hypothesis is supported by the data herein which shows greater decomposition of 5 under oxidative conditions in vitro and in vivo analogous to that of the ROS chemical systems. Utilizing this site-specific property can unlock new avenues of drug targeting in ischemia and various pathologies characterized by oxidative stress.

3.4. Future directions – nitrones as CNS-permeable, ROS-sensitive prodrugs

Despite being charged at physiological pH, 5 and 6 were both detected in the CNS by LC-MS/MS of brain tissue. Future analogues will aim to increase permeability while fighting the prototypical hydrophilicity of NOS-inhibitors and increasing isoform specificity. However, the molecule described here, 5, demonstrates the potential of nitrones to act as more than antioxidants, but as carrier molecules that are able to provide an additional, site-specific action. NOS was explored as one such target due to its unique dual-role under oxidative pathology, however other molecular pathways could be targeted using the above methodologies.

4. Materials and methods

4.1. Computational methods

Ab initio structures of nitrones, radicals, and radical adducts were generated using Avogadro [57] at the MMFF94 level. Optimized geometries were then determined by density functional theory (DFT) [58] at the B3LYP/6–31G* level of theory as previously described [29] yielding no imaginary vibrational frequency. All calculations were performed using Gaussian 09 [59] at the Ohio Supercomputer Center and visualized by GaussView 5.0 software. B3LYP/6–31G* geometries were corrected using a scaling factor of 0.9806 [60] for the zero-point vibrational energy (ZPE). Solvation effects on the gas-phase calculations were determined using the PCM [61], and spin and charge densities were assigned using natural population analysis (NPA) [62] at the PCM/B3LYP/6–31 + G** level of theory. All doublet and triplet calculated minima yielded negligible spin contamination (0.75 < 〈S2〉 < 0.76). ΔG_{rxn,298 K} (kcal/mol) of adduct formation was determined by ΔG_Adduct – ΔG_Reactants.

Protein docking studies were carried out using Swissdock.ch [3,8] and visualized by UCSF Chimera software [63]. X-ray crystallographic structures of each NOS isoform with L-arginine or 1400W bound were used to determine the free energy of interaction with the enzyme active site. Docking was limited to ± 10.0 Å of the native bound ligand (L-arginine or 1400W) in the X, Y, and Z directions.

4.2. Synthesis and characterization

Full synthetic details and EPR and HPLC characterization data are provided in Supplementary material.

4.3. Cell culture

Human neuroblastoma SH-SY5Y cells and SIM-A9 mouse microglia (ATCC) were separately cultured on poly-D-lysine coated plates in 1:1 Dulbecco’s Modified Eagle Medium / Ham’s F12 (DMEM/F12) medium containing 5% fetal bovine serum, 5% horse serum and penicillin-streptomycin. Cultures were stored in a 37 °C incubator with a 5% CO₂ atmosphere. Cells were allowed to adhere overnight prior to subjection to experimental systems.

4.4. Oxygen-glucose deprivation (OGD)

Growth medium was removed from SH-SY5Y cells and replaced with sterile HBSS (140 mM NaCl, 3.5 mM KCl, 0.4 mM KH₂PO₄, 5 mM NaHCO₃, 1.3 mM CaCl₂, 1.2 mM MgSO₄, 20 mM HEPES, pH 7.4, bubbled with 95%/5% N₂/CO₂), and placed in an air-tight container subsequently purged three times with 95%/5% N₂/CO₂. The container was then placed in a 37 °C incubator for 1.5 h. After OGD, the cells were returned to a normoxic environment, and medium was replaced with growth medium and vehicle or treatment for 24 h to simulate reperfusion. Sham cells were left in growth medium under a normoxic environment. Viability was measured by adding 10% PrestoBlue^® Cell Viability Reagent (Thermo Fisher) for 30 min, and reading fluorescence at 560 nm/590 nm excitation/emission. Viability is reported as a ratio to vehicle-treated control.

4.5. Western blotting

SH-SY5Y cells exposed to 1.5 h OGD or ONOO⁻ (100 μM) followed by 24 h vehicle or treatment were harvested and lysed by RIPA buffer. Equivalent amounts of protein as determined by Bradford assay were loaded onto 10–20% tris-glycine gels and separated by electrophoresis. Proteins were transferred to a PVDF membrane, blocked with 5% BSA-TBST, and probed with primary antibodies in blocking buffer (anti-phospho-Akt(Ser473) 1:2000; anti-Akt 1:1000; anti-Nitrotyrosine 1:1000; anti-cleaved caspase-3(Asp175) 1:1000, Cell Signaling; anti-GAPDH 1:2000, Thermo Fisher) followed by secondary antibody (HRP goat anti-rabbit IgG 1:10,000, Jackson) in 5% milk-TBST. Images were acquired using a BioRad ChemiDoc^™ XRS+, and densitometry was analyzed using ImageJ software (NIH) normalized to GAPDH loading control.

4.6. Griess assay

SIM-A9 cells were treated with 100 ng/mL lipopolysaccharide (LPS, Sigma) along with vehicle or drug treatment for 24 h. Conditioned medium was then diluted 1:1 with Griess reagent (0.4% naphthy-lethylenediamine dihydrochloride, and 4% sulphanilamide in 10% phosphoric acid), and analyzed by colorimetry at 540 nm.

4.7. Animals

All animal protocols were approved by the University of Toledo Health Science Campus Institutional Animal Care and Utilization Committee, and NIH guidelines were followed. Male C57BL/6 mice, 6–8 weeks old at 23–25 g (Charles River) were housed with a 12 h light/dark cycle at 22 ± 1 °C.

4.8. Permanent middle cerebral artery occlusion (pMCAO)

The distal part of the MCA was permanently occluded as previously optimized by our lab [64,65]. Briefly, mice were anesthetized with 1% isoflurane, the site was aseptically cleaned, and a small 1.0 cm incision was made between the left eye and ear. The temporal muscle was moved aside to view the MCA under the temporal bone, and a 2 mm hole was drilled with a dental drill directly over the distal part of MCA. The artery was directly occluded with a bipolar coagulator, and the incision was sutured. The animals’ body temperature was maintained at 37.0 ± 5 °C throughout the procedure and after surgery until recovery. Animals were intravenously injected by lateral tail vein with 100 μL normal saline or 10 mg/kg 5 in vehicle at 3 and 6 h after pMCAO. Either vehicle or 1 mg/kg 5 were then dosed twice daily thereafter, spaced 8 h apart, until sacrifice by CO₂ inhalation at 72 h post-pMCAO.

4.9. Laser doppler cerebral blood flow (CBF)

Animals were anesthetized and the incision was performed as above to expose the temporal bone, through which the MCA was visible. The end of a fiber optic cable fitted to a MoorVMS-LDF Laser Doppler Monitor (Moor Instruments) was positioned in place over the MCA. The baseline CBF as flux was monitored for 5 min prior to intravenous injection by lateral tail vein of normal saline vehicle, isoamyl nitrate (20 mg/kg), or 5. Measurements were taken every 5 min in triplicate for each animal, and reported as ratio to baseline. Animals were sacrificed at 25 min by CO₂ inhalation, blood was collected into K₂EDTA coated tubes (BD Microtainer^®), and brains were rapidly dissected out and frozen on dry ice for LC-MS/MS analysis.

4.10. Neurobehavioral assays

Mice were trained on neurobehavioral paradigms for 3 days prior to pMCAO, and baseline readings were recorded 24 h prior to surgery. For rotarod analysis, mice were placed on a moving rod (Columbus Instruments) programmed to rotate at 1 rpm and accelerate by 1 rpm every 10 s until the animal falls from the rod. The latency to fall was recorded manually and reported as ratio to baseline. Grip strength analysis was performed using a grip strength meter (Columbus Instruments) fitted with a pull bar assembly. Forelimbs of mice were placed on the bar and peak force until release when pulled by the tail was displayed on the digital display and manually noted. Neurological deficit scoring (NDS) was conducted 72 h after pMCAO prior to sacrifice, and evaluated on a 28-point score pattern optimized by our lab [65,66]. The total score was determined by the sum of the seven criteria graded from 0 to 4, with higher scores indicating more severe deficits.

4.11. Infarct volume

Brains of animals sacrificed by CO₂ inhalation 72 h after pMCAO were rapidly dissected out, sliced into five 2 mm-thick coronal sections, and incubated in warm 1% triphenyltetrazolium chloride (TTC, Sigma) in normal saline. Infarct volumes were estimated by measuring rostral and caudal sides of each section in conjunction with the thickness, and expressed as a percentage of the volume of the contralateral hemisphere. After imaging, portions of the ischemic lesion, penumbra, and contralateral tissue were quickly frozen on dry ice for LC-MS/MS analysis.

4.12. LC-MS/MS

Full methods provided in Supplementary material.