Synthesis of 1-azaspiro[4.4]nonan-1-oxyls via intramolecular 1,3-dipolar cycloaddition

Sterically shielded nitroxides of the pyrrolidine series have shown the highest resistance to reduction. Here we report the synthesis of new pyrrolidine nitroxides from 5,5-dialkyl-1-pyrroline N-oxides via the introduction of a pent-4-enyl group to the nitrone carbon followed by an intramolecular 1,3-dipolar cycloaddition reaction and isoxazolidine ring opening. The kinetics of reduction of the new nitroxides with ascorbate were studied and compared to those of previously published (1S,2R,3′S,4′S,5′S,2″R)-dispiro[(2-hydroxymethyl)cyclopentan-1,2′-(3′,4′-di-tert-butoxy)pyrrolidine-5′,1″-(2″-hydroxymethyl)cyclopentane]-1′-oxyl (1).


Introduction
Sterically shielded nitroxides are currently attracting much attention due to their high resistance to bioreduction [1,2]. It has been demonstrated that 2,2,5,5-tetraethylpyrrolidine nitroxides have the highest stability, sometimes exceeding that of trityl radicals [1]. Introduction of spirocyclic moieties has a smaller effect on the reduction rates of nitroxides than the introduction of linear alkyl substituents does; however, spirocyclic nitroxides may have much longer spin relaxation times at 70-125 K [3] and even at room temperature [4]. The latter effect may be useful for structural studies by means of PELDOR or DQC [5].
We recently reported the synthesis of sterically shielded pyrrolidine nitroxide 1 via a stereospecific consecutive assembly of two spiro-(2-hydroxymethyl)cyclopentane moieties. These procedures included the addition of pent-4-enylmagnesium bromide to the corresponding nitrone, oxidation to alkenylnitrone, intramolecular 1,3-dipolar cycloaddition, and isoxazolidine ring opening. Nitroxide 1 showed both an unexpectedly low reduction rate [6] and long relaxation times t 1 and t m at room temperature [4]. Unexpectedly high resistance of this nitroxide to chemical reduction results from the configuration of the hydroxymethyl groups, which are directed towards the nitroxide group, thereby making it more hindered. It is known that inductive effects of substituents can strongly affect the rate of nitroxide reduction [7,8]; therefore, one could expect that the removal of electron-withdrawing tert-butoxy groups at positions 3 and 4 of the pyrrolidine ring of 1 (Figure 1) should lead to a further decrease in the reduction rate of the nitroxide. Here we describe the synthesis of C 1 -symmetric racemic 3,4unsubstituted pyrrolidine nitroxides with only one spiro(2hydroxymethyl)cyclopentane moiety. The rates of reduction of the new nitroxides with ascorbate were measured.

Results and Discussion
Aldonitrones 5b,c were prepared similarly to the well-known synthesis of 5,5-dimethyl-1-pyrroline-1-oxide (DMPO, 5a) [9,10] from nitrocyclohexane and 3-nitropentane (Scheme 1). In brief, the reaction of nitrocyclohexane and acrolein in a CH 3 ONa/CH 3 OH solution afforded the corresponding nitroaldehyde 3b with a 70% yield. Of note, the Michael addition of 3-nitropentane to acrolein was accompanied by remarkable tarring and gave a much lower yield of nitroaldehyde 3c (25%). The reactive aldehyde groups were protected via 1,3dioxolane assembly, and the resulting compounds were treated with Zn dust and NH 4 Cl in a water-THF solution to reduce nitro groups. The resulting hydroxylamines were treated with hydrochloric acid to hydrolyse dioxolane moieties, and careful basification resulted in intramolecular cyclisation givnig 5b,c with a yield of 51% and 42%, respectively.
Nitrones 5a-c readily react with 4-pentenylmagnesium bromide. Quenching of the reaction mixtures with water under aerobic conditions leads to partial oxidation of resultant N-hydroxypyrrolidines 6a-c to corresponding nitrones 7a-c. Therefore, this conversion was finalised via bubbling of air into the solution in the presence of Cu(NH 3 ) 4 2+ (Scheme 2). Samples of resulting 2-(pent-4-enyl)nitrones 7a-c remarkably deteriorate during storage under aerobic conditions with dark tar formation. A possible pathway of decay may include oxidation of the ene-hydroxylamine tautomeric form to vinylnitroxide; similar compounds are prone to various dimerisation reactions (Scheme 3) [11]. It is worth noting that in the mass spectrum of 7c, the [M − 1] + ion was observed instead of the molecular ion. The easy loss of a hydrogen atom is consistent with the susceptibility of 7c to oxidative decay (Scheme 3).
Intramolecular cycloaddition of similar nitrones is known to lead to hexahydro-1H-cyclopenta[c]isoxazoles [6,12]. Indeed, heating of 7а-с at 145 °C in toluene for 30-60 min in a microwave oven produced 8a-c (racemic mixtures; Scheme 2). The structure assignment was performed on the basis of 1 H and 13 C NMR spectra and 1 H, 1 H and 13 C, 1 H correlations (see Sup- porting Information File 1); the spectral data are in agreement with the literature on similar systems [6].
To decrease the tarring, the reaction was carried out in the presence of 2,2,6,6-tetramethylpiperidine-1-oxyl. It should be noted that heating of alkenyl nitrones 7a and 7b gives the corresponding cycloadducts with yields close to quantitative, whereas for nitrone 7c, the complete conversion could not be achieved either at 145 °C or at a higher temperature. According to the 1 Н NMR spectra, the 8c/7c ratio never exceeded 3:1. Heating of a pure sample of 8c under similar conditions caused the emergence of signals at 5.70, 4.92, and 4.87 ppm in the 1 H NMR spectra. These signals were attributed to the protons of the terminal vinyl group of 7c. 1,3-Dipolar cycloaddition of nitrones to alkenes is known to be reversible [13,14]. We recently re-ported a similar reversibility of the intramolecular cyclization of sterically hindered pent-4-enylnitrone of the 2H-imidazole series [15].
We have previously reported that oxidation of secondary amines with a spiro(2-hydroxymethyl)cyclopentane moiety at the α-carbon with the H 2 O 2 /WO 4 2− system is ineffective whereas conversion of these amines to the corresponding nitroxides can be easily performed using m-chloroperbenzoic acid (m-CPBA) [6,15]. Treatment of 9a with m-CPBA in dry chloroform at −10 °C afforded a nitroxide, which was isolated as orange oil with a yield of 73% (Scheme 4). An infrared spectrum of the isolated compound showed a strong absorption band at 1725 cm −1 typical for carbonyl compounds and no absorption in the region 3100-3500 cm −1 , suggesting that the hydroxymethyl group was affected in the reaction. The mass spectrum featured the molecular ion [M + ] = 196.1335 corresponding to the molecular formula C 11 H 18 NO 2 , which matches element analysis data. These results allowed us to assign the structure 15 to this nitroxide. Indeed, oxidation of amines with peracids is known to proceed via oxoammonium cation formation [17,18], and the latter can oxidize alcohols to carbonyl compounds [19]. The close proximity of the hydroxymethyl group to the oxoammonium one favours the reaction. Treatment of 15 with NaBH 4 in EtOH caused quantitative reduction of the aldehyde group to the hydroxymethyl one, thus yielding 12a identical to that prepared by the alternative method (see below). To prevent oxidative reactions in the side chain, the hydroxymethyl group in 9a-c was protected via acylation. Heating of 9a-c with an excess of acetic anhydride in chloroform quantitatively afforded the corresponding esters 10a-c. The products of acylation of the sterically hindered amino group were not detected in the reaction mixture.
Oxidation of 10а-c with m-CPBA yielded the desired nitroxides 11a-c as orange oils. IR spectra of the new nitroxides contained no absorption bands in the region of 3000-3600 cm −1 and did not have an intense band at 1740 cm −1 characteristic for the ester C=O group. To confirm the structure of the nitroxides, alkoxyamines 16a-c were prepared by Matyjaszhewski's method (Scheme 5) [20].  13 C-HSQC spectra allowed us to assign these signals to an isolated CH 2 -CH 2 system. All these findings unambiguously support the assignment of structure 18 to the isolated alkoxyamine and structure 17 to the corresponding nitroxide. Therefore, 17 is formed due to hydrogen abstraction in the spirocyclopentane ring. To the best of our knowledge, similar transformations have never been observed before. Presumably, formation of 17 occurs due to the close proximity of the N + =O group (in the intermediate strained oxoammonium cation) to the hydrogen atom of an adjacent methylene group of the cyclopentane ring (Scheme 7).

Scheme 8:
Optimisation of the synthesis of nitroxide 1. Table 1: Parameters of the EPR spectra (hyperfine coupling constants, a N ; peak-to-peak linewidths, ΔB p-p ; and g-factors), second order rate constants of reduction with ascorbate and partition coefficients octanol-water (K p ) for nitroxides 1 and 12a-c. The ester groups in 11a-c were easily cleaved in an aqueous-methanol solution of ammonia. Nitroxides 12a-c were isolated as orange compounds moderately soluble in water.
Overall yields of these radicals via the acylation-oxidation-deprotection pathway were in the range of 54-75%. Of note, the use of a similar three-step procedure for the synthesis of 1 from 19 increased the yield of this nitroxide from 20% [6] to 60% (Scheme 8).
The EPR spectra of nitroxides 12a-c and 1 acquired in a deoxygenated buffer revealed a significant difference in line widths (see Table 1 and Supporting Information File 1, Figures  S7-S10), with the broadest lines expectedly being shown by 12b. Introduction of spirocyclohexane moieties to α-carbons of pyrrolidine nitroxides was found to cause strong broadening of lines in the EPR spectra, presumably owing to unresolved hfc on the hydrogens at positions 2 and 6 of the cyclohexane ring [22]. It has been reported that EPR spectra of pyrrolidine or imidazolidine nitroxides with pair(s) of geminal ethyl groups at α-carbon atoms may feature large doublet hyperfine splittings [23][24][25]. For imidazolidine nitroxides, these splittings were unambiguously attributed to hfc on one of four methylene hydrogens of each pair of ethyls [23]. The difference in apparent hyperfine coupling constant a H on these hydrogens is due to a substituent at position 3 or 4 of the ring. This substituent hinders rotation and affects the population of conformations of neighbouring geminal ethyl groups, thereby preventing averaging. In agreement with this conclusion, there are no large splittings in the EPR spectrum of 12c, and the line width is a bit greater than that in the spectra of similar nonspirocyclic 3,4-unsubstituted 2,2-diethylpyrrolidine nitroxides [26], implying free rotation of ethyl groups. In the EPR spectra in toluene the nitroxides 11a-c show 0.035-0.04 mT lower hfs constants on nitrogen atom compared to nitroxides 12a-c, presumably due to intramolecular hydrogen bond formation. The difference in a N between 11a-c and 12a-c is almost one order of magnitude lower for EPR spectra in water (0.005-0.006 mT), obviously because strong intermolecular hydrogen bonds are formed due to solvation.
The initial rates of the EPR signal decay were used to obtain the rate constants of the nitroxide reduction by ascorbic acid (see Figure 2 and Supporting Information File 1).
The rate constants for all these nitroxides are remarkably higher than this constant for nitroxide 1. Even nitroxide 12c, which is the most stable among the new pyrrolidine nitroxides has a ca. 6-fold higher reduction rate than 1 does, and both are less resistant to reduction than 2,2,5,5-tetraethyl-substituted pyrrolidine nitroxides are (k = 2 × 10 −3 to 3.3 × 10 −4 ) [1,25]. The rate of reduction of 12a is close to that of 3-carboxy-2,2,5,5-tetramethylpyrrolidine-1-oxyl [24]. Obviously, a single spiro-(2hydroxymethyl)cyclopentane moiety cannot provide higher reduction resistance than geminal ethyl groups can. Thus, estimation of the steric effect of neighbouring substituents cannot account for the high reduction resistance of 1. Presumably, the symmetric structure with bulky substituents at positions 3 and 4 is an important factor for nitroxide stability. Due to the steric repulsion of trans-oriented tert-butoxy groups and spiro cyclopentane moieties in the symmetric structure of 1, the (2-hydroxymethyl)cyclopentane groups tightly embrace the nitroxide group making it less accessible for reductants. It is also noteworthy that the symmetric repulsion from both sides of the pyrrolidine ring favours a planar nitroxide group and desta- bilises the corresponding hydroxylamine with the sp 3hybridised nitrogen. Recently, we observed a similar effect for (3S(R),4S(R))-2,2,5,5-tetraethyl-3,4-bis(hydroxymethyl)pyrrolidine-1-oxyl, which manifested the highest resistance to reduction among known nitroxides [25]. In contrast, structures 12a-c are asymmetric, and corresponding hydroxylamines could be additionally stabilised by hydrogen bonding between the nitrogen atom and the proton of the hydroxymethyl group.

Conclusion
In this study, we again demonstrated feasibility of the general synthetic approach to sterically hindered spirocyclic nitroxides based on an intramolecular 1,3-dipolar cycloaddition reaction in alkenylnitrones followed by isoxazolidine ring opening. The resulting asymmetric pyrrolidine nitroxides have unexpectedly high rates of reduction with ascorbate. These results lead us to the assumption that symmetric structures with bulky substituents at positions 3 and 4 should be favoured for achieving higher resistance to reduction.

Supporting Information
Supporting Information File 1 Full experimental details and analytical data (UV, IR, 1 H NMR, 13