Our initial investigation focused on the one-step sequential cycloaddition reaction of nitrone-containing allene 3a that was generated in situ from bisaldehyde 1a (prepared from commercially available 4-pentyn-1-ol in 5 steps. See section 2.3 of the Supplementary Information for details.) and N-benzylhydroxylamine hydrochloride (Table 1), with solvents, including ethanol, toluene, 1,4-dioxane, and TCE, examined in the absence of a base (entries 1–3; see Supplementary Table 1 for details). The use of TCE led to the highest yield of the desired racemic [5.5.5.5]diaza-dioxa-fenestrane 2a (entry 3, 89%). We also explored the impact of different bases, including Et3N, i-Pr2NEt, pyridine, and NaHCO3 (entries 4–7). Remarkably, the use of Et3N resulted in an excellent yield (91%) for 2a (entry 4). However, when i-Pr2NEt (as a stronger base) was employed, the yield of 2a dropped significantly (19%), likely due to undesired base-mediated reactions involving bisaldehyde 1a (entry 5). As expected, nitrone-containing allene 3a underwent sequential cycloaddition to afford diaza-dioxa-fenestrane 2a in high yield.Table 1 Investigating the one-step sequential (3 + 2)/(3 + 2) cycloaddition chemistry of a nitrone-containing alleneWe then investigated the one-step sequential cycloaddition reaction of nitrile oxide 6a, which was generated in situ from bisoxime 4a (prepared from bisaldehyde 1a in one step. See section 2.3 of the Supplementary Information for details.) (Table 2). Solvents, including CH2Cl2, ethanol, THF, and toluene, were examined in the presence of a 10% aqueous solution of NaOCl and Et3N. The use of CH2Cl2 resulted in a good yield of the desired racemic double-bond-containing [5.5.5.5]diaza-dioxa-fenestrane 5a (entry 1, 69%). The use of ethanol did not afford any of the desired product (entry 2), despite the high solubility of NaOCl in this solvent. The desired product 5a was obtained, in yields of 15% and 20% when THF and toluene was used, respectively (entries 3 and 4). Bases, including Et3N, pyridine, NaHCO3, and i-Pr2NEt were examined using CH2Cl2 as the solvent (entries 1 and 5–7); once again the use of Et3N afforded the highest yield (entry 1, 69%). The use of i-Pr2NEt afforded 5a in acceptable yield (entry 7, 59%) during the cycloaddition of bisnitrile oxide 6a, in contrast to the yield obtained using nitrone 3a. The desired product 5a was also obtained in an acceptable yield in the absence of the base (entry 8, 62%). At this point, we also examined the effect of temperature (0–100 °C) using haloalkane solvents (entries 9–13); the use of DCE at 80 °C afforded the highest yield (entry 12, 72%). Accordingly, we successfully developed a sequential cycloaddition-based approach using two types of allenes 3a and 6a containing nitrone and nitrile oxide, respectively. The structures of 2a and 5a were unambiguously determined by X-ray crystallography (see section 11 of the Supplementary Information for details); structural-analysis details are discussed below.Table 2 Investigating the one-step sequential (3 + 2)/(3 + 2) cycloaddition chemistry of a nitrile oxideThe substrate scope of the developed one-step sequential cycloaddition chemistry involving nitrones 3 was subsequently examined (Fig. 2a). While N-alkyl substituted fenestranes 2b–2d were obtained in good yields (64–76%), only a trace amount of N-t-butyl-substituted fenestrane 2e was obtained, while N-4-methoxybenzyl-substituted fenestrane 2f was obtained in moderate yield (48%). We next examined the functional-group tolerance of this reaction. Good yields (68–71%) were obtained in the syntheses of 2g containing Ph–Br bonds, 2h containing O–Si bonds, and 2i and 2j containing acid-labile THP and furyl groups, respectively. Fenestranes 2k and 2l containing either acid-labile Boc-carbamate or t-Bu-ester moieties, and either base-labile methyl ester or Fmoc-carbamate moieties, were also obtained in good yields (73 and 62%). On the other hand, N-phenyl substituted fenestrane 2m was only obtained in moderate yield (35%), while C,N-alkyl(oxy)-substituted fenestranes 2n–2q were obtained in acceptable-to-good yields (56–70%) as mixtures of diastereomers. Although the diastereomers of 2o were inseparable, 2n, 2p, and 2q were readily separated into their diastereomers by silica-gel column chromatography. The stereochemistry of each diastereomer was determined via 1H NMR, COSY, and NOESY spectroscopy, along with DFT calculations (see section 8 of the Supplementary Information for details). [5.5.5.6]Diaza-dioxa-fenestrane 2r and [5.6.5.6]diaza-dioxa-fenestrane 2s were obtained in yields of 65 and 12%, respectively, while [5.7.5.7]diaza-dioxa-fenestrane 2t was not obtained.Fig. 2: Substrate scope.Substrate scope of our developed one-step sequential cycloaddition approach from a nitrone 3 and b nitrile oxide 6. c Derivatizing 2g using Pd-catalyzed cross-coupling reactions. d Reductive cleavage of the N–O bond in isoxazolidine 2a. e Mono- and bis-allylation of isoxazoline 5a using a Grignard reagent. aAddition of reagents at −78 °C and warmed to r.t. was repeated 5 times. Cy = cyclohexyl; TBDPS = t-butyldiphenylsilyl; THP = tetrahydropyranyl; Boc = t-butyloxy carbonyl; Fmoc = 9-fluorenylmethoxycarbonyl; Bn = benzyl; dppf = (diphenylphosphino)ferrocene; DMF = N,N-dimethylformamide.We next examined the substrate scope of the one-step sequential cycloaddition chemistry involving nitrile oxides 6 (Fig. 2b). C-Alkyl(oxy)-substituted fenestranes 5b and 5c were obtained in good yields (69 and 66%) as mixtures of diastereomers. While 5b was unable to be separated nor was its diastereomeric ratio able to be determined, 5c was readily separated into its diastereomers via silica-gel column chromatography. The stereochemistry of each diastereomer was determined by 1H NMR, COSY, and NOESY spectroscopy, along with DFT calculations (see section 9 of the Supplementary Information for details). To our delight, [5.5.5.6]diaza-dioxa-fenestrane 5d was obtained in moderate yield (36%), whereas [5.6.5.6]diaza-dioxa-fenestrane 5e was not obtained. While one-step sequential cycloaddition chemistry involving nitrones 3 enabled the synthesis of diaza-dioxa-fenestranes containing up to two six-membered rings, the developed chemistry involving nitrile oxides 6 enabled the synthesis of diaza-dioxa-fenestranes containing only one six-membered ring. The latter chemistry appeared to be more significantly affected by the ring size. Using the developed approach, fenestranes with different ring sizes were constructed through sequential cycloaddition.The prepared THP-, TBDPS-, Boc-, Fmoc-, t-Bu-, and Bn-protected diaza-dioxa-fenestranes can be readily derivatized via deprotection and subsequent chemical modification. In addition, the aryl-Br bond in fenestrane 2g can be directly activated in the presence of transition-metal catalysts for further derivatization. Accordingly, 2g was subjected to Suzuki–Miyaura, Sonogashira–Hagihara, and Mizoroki–Heck coupling, which afforded the desired products 7a–7d in acceptable-to-excellent yields (Fig. 2c, 57–96%). Moreover, the reactive N–O and C=N bonds in the diaza-dioxa-fenestranes were further derivatized; reductive cleavage of the N–O bond in isoxazolidine 2a afforded spirobicycle 8 in excellent yield (Fig. 2d, 93%). Spiro[4.4]nonane 8, which was densely functionalized by two amino groups and two hydroxy groups at the neopentyl positions, was obtained as a single diastereomer. Mono- and bis-allylated isoxazolidines 9a and 9b were selectively obtained by 1,2-addition using different amounts of an allyl Grignard reagent to isoxazoline 5a; both 9a and 9b were obtained diastereoselectively in acceptable yields (Fig. 2e, 58 and 51% yield, respectively). These results clearly demonstrate that our approach facilitates the creation of structurally diverse and complex heterocyclic compounds.As previously discussed, the flattened fenestrane quaternary carbon center, which is shared by four rings, has garnered much attention. The extent of flattening of such a quaternary carbon center can be evaluated from its two opposing angles (α and β in Fig. 3)4,5, these angles increase as the quaternary carbon center flattens. Keese et al. reported α and β values of 116.2°29 and 113.8°5 for c,c,c,c-[5.5.5.5]fenestrane (10) based on electron diffractometry and semi-empirical calculations, respectively (Fig. 3a). The flattest reported fenestranes 11 and 12 have significantly larger angles (Fig. 3b, α = 134.9°, β = 119.2°30; Fig. 3c, α = 129.2°, β = 128.3°31) than non-distorted quaternary carbon (109.5°, Fig. 3d).Fig. 3: Fenestrane structures.Opposing angles: a in c,c,c,c-[5.5.5.5]fenestrane 10, b, c in the flattest previously synthesized fenestrane 11 and 12, and d for a non-distorted quaternary carbon center. e–g Experimentally determined and calculated opposing angles in c,c,c,c-[5.5.5.5]fenestrane 2a containing isoxazolidine rings, c,c,c,c-[5.5.5.6]fenestrane 2r containing isoxazolidine rings, and c,c-[5.5.5.5]fenestrane 5a containing isoxazoline rings.The racemic [5.5.5.5]- and [5.5.5.6]fenestranes 2a and 2r, respectively, containing isoxazolidine rings and the [5.5.5.5]fenestrane 5a containing isoxazoline rings were analyzed using X-ray crystallography, which revealed α and β values consistent with those of the most stable conformers determined by DFT at the B3LYP32/6-31 G + (d,p)33,34,35,36 level of theory (Fig. 3e–g). A comparison of the quaternary-carbon angles in fenestrane 10 with those in diaza-dioxa-fenestrane 2a (Fig. 3a, e) reveals that replacing the carbon atoms at the non-bridgehead positions in the fenestrane skeleton with nitrogen and oxygen atoms results in slight flattening of the quaternary carbon center. A comparison of the angles in [5.5.5.5]diaza-dioxa-fenestrane 2a with those in [5.5.5.6]diaza-dioxa-fenestrane 2r (Fig. 3e, f) reveals that ring expansion reduces the degree of flattening of the quaternary carbon center, which is consistent with the previously reported tendency5. In addition, a comparison of the angles in [5.5.5.5]fenestrane 2a containing isoxazolidine rings with those in [5.5.5.5]fenestrane 5a containing isoxazoline rings (Fig. 3g) reveals that the introduction of double bonds at the bridgehead positions results in flattening of the quaternary carbon center, which is also consistent with the previously reported tendency5. However, the observed angles in 5a (Fig. 3g, α = 134.7°, β = 114.9°) are very large and comparable to those of fenestrane 11 and 12, which are among the flattest fenestranes known (Fig. 3b, c). Compound 5a contains the flattest quaternary carbon center among heteroatom-containing fenestranes discovered thus far.We performed a conformation search for fenestranes 2b and 5a. To reduce the calculation cost, fenestrane 2b with methyl groups was used instead of 2a with benzyl groups. The four most stable conformers 1–4 of 2b and the two most stable conformers 1 and 2 of 5a are shown with relative energy levels and α and β values in Supplementary Table 9 of the Supplementary Information. The chemical structure of 2a experimentally observed via X-ray crystallographic analysis (α = 117.4°, β = 117.0°) was consistent with the calculated most stable conformer 1 of 2b (α = 117.6°, β = 116.8°). Although the conformers 2–4 of 2b with more flattened quaternary carbon centers were found in the conformation search (Supplementary Table 9), they were less stable. Only two conformers with almost consistent α and β values and similar structures were found in the case of 5a. These results indicated that the compound 5a has a very rigid structure.
