Main

Developing synthetic methodologies to construct complex, rigid, three-dimensional molecular architectures is a major focus in contemporary organic synthesis. This is because C(sp3)-rich polycyclic scaffolds have found widespread application in medicinal chemistry as replacements for planar aromatic rings, which can provide drug candidates with improved physicochemical and pharmacokinetic properties1,2,3. For example, structurally rigid bicyclo[n.1.1]alkanes have emerged as effective bioisosteres of benzene rings, with numerous reports of the use of bicyclo[1.1.1]pentane (BCP)3,4,5,6,7,8,9,10,11,12, bicyclo[2.1.1]hexane (BCHex)3,8,9,12,13 and bicyclo[3.1.1]heptane (BCHep)9,12,13,14 groups (Fig. 1a (A.1)). However, analogous isosteric mimetics of heteroaromatic compounds remain underexplored, despite the prevalence of nitrogen heterocycles in Food and Drug Administration-approved small-molecule drugs15,16,17. A notable study by Mykhailiuk and coworkers demonstrated that replacing the 3,5-disubstituted pyridine ring in the antihistamine rupatidine with a 3-azabicyclo[3.1.1]heptane (3-N-BCHep) improved solubility, lipophilicity and metabolic stability, underscoring the potential of this motif in medicinal chemistry18 (Fig. 1a (A.2)). Since then, substituted N-BCHep scaffolds have gained attention as bioisosteres of nitrogen heterocycles, with several studies focusing on 2-N-BCHep19,20,21,22,23,24 and 3-N-BCHep18,20,25,26,27,28,29,30,31 derivatives (Fig. 1a (A.2)). By contrast, there are limited examples of the use of substituted 6-N-BCHeps32,33,34, wherein the nitrogen atom is incorporated into the four-membered ring of the bicycle (Fig. 1a (A.3)). This modification reduces the basicity of the nitrogen and strengthens the adjacent C–H bonds33,35,36, which could offer distinct electronic properties and improved metabolic stability compared with other N-BCHep isomers. Furthermore, comparison of various geometric descriptors, including exit vectors, revealed good similarities between the 6-N-BCHeps and their aromatic counterparts (Fig. 1b). However, access to substituted 6-N-BCHeps is limited by a lack of general methodologies for their synthesis32, so there are scant data concerning their suitability as bioisosteres of nitrogen heterocycles33,34,37.

Fig. 1: Three-dimensional scaffold mimetics of heteroaromatic rings, showing state of the art and reaction design using anti-rule-of-five intramolecular aza-Paternò–Büchi reaction.
Fig. 1: Three-dimensional scaffold mimetics of heteroaromatic rings, showing state of the art and reaction design using anti-rule-of-five intramolecular aza-Paternò–Büchi reaction.
Full size image

a, Three-dimensional (3D) bioisosteres of substituted (hetero)aromatic rings. b, Exit vector analyses of N-heterocycles. D2-3, distance between carbon atoms 2 and 3; θ, angle between C2–C3 atoms and H1–C2; φ, angle between H4–C3 and H1–C2. c, Selectivity in intramolecular aza-Paternò–Büchi reactions39,47,48. d, DFT calculations: breaking the rule-of-five. Fsp3, fraction of sp3-hybridized carbon atoms; Ts, p-toluenesulfonyl; TS, transition state; k, rate constant.

The most convergent route to substituted 2-N-BCHep and 3-N-BCHep derivatives is through intermolecular [3 + 2] cycloadditions of bridgehead-substituted bicyclo[1.1.0]butanes with nitrogen-containing 1,3-dipoles19,20,21,22,23,24,25,26,27,28,31 (Fig. 1a (A.2)). However, this method is unsuitable for the synthesis of 6-N-BCHeps because it requires replacement of the bicyclo[1.1.0]butanes with 2-azabicyclo[1.1.0]butanes, which have not previously been isolated due to their instability38. An alternative approach to azabicyclo[n.1.1]alkanes is intramolecular [2 + 2] photocycloadditions of acyclic unconjugated aza-dienes (aza-Paternò–Büchi) using energy transfer (EnT) photocatalysis. This method has been successfully applied to the synthesis of 5-azabicyclo[2.1.1]hexane (5-N-BCHex) rings from aza-1,5-dienes39 (Fig. 1c, top), but there are no reports of analogous reactions of aza-1,6-dienes providing 6-N-BCHeps. Intramolecular [2 + 2] photocycloadditions of unconjugated dienes are generally constrained by an empirical rule known as the ‘rule-of-five’, which dictates that the 1,2-biradical formed upon photoexcitation of the substrate undergoes kinetically favoured 5-exo-trig cyclization to form a five-membered ring40,41,42. This results in the generation of bridged BCHex products from 1,5-dienes, whereas 1,6-dienes selectively cyclize to give fused five-membered ring systems instead of bridged six-membered BCHeps. Although a limited number of reports in which geometric restrictions on the 1,6-diene substrates can bias the selectivity towards the BCHep scaffold43,44,45,46, no general methods for selectivity control or analogous syntheses of 6-N-BCHeps exist. Indeed, Schindler and coworkers observed that the intramolecular [2 + 2] photocycloadditions of aza-1,6-dienes (hex-5-enal oximes) afforded fused azetidine products, consistent with the rule-of-five47 (Fig. 1c, middle). Nonetheless, we recognized that using readily accessible aza-1,6-dienes as substrates in [2 + 2] cycloadditions could provide an efficient strategy for constructing multisubstituted 6-N-BCHeps and could also enable access to previously elusive substituted 3,6-diazabicyclo[3.1.1]heptane (3,6-N2-BCHep) and 3-oxa-6-azabicyclo[3.1.1]heptane (3-O-6-N-BCHep) products through simple substrate modification. Therefore, we sought to design an intramolecular photocycloaddition that defies the rule-of-five.

Achieving selective 6-N-BCHep formation in [2 + 2] photocycloadditions of aza-1,6-dienes 1 requires the alkene 1,2-biradical in 1* to react with the imine via 6-exo-trig cyclization at C1 instead of 5-exo-trig cyclization at C2 (Fig. 1c, bottom). This is challenging due to the disparity in rates of ring formation, with 5-exo-trig cyclization of non-stabilized alkyl radicals onto hydrazones reported to be more than 100 times faster than 6-exo-trig cyclization48 (Fig. 1c, middle). We hypothesized that a radical stabilization strategy could be used to overturn this inherent kinetic preference, wherein the introduction of a radical-stabilizing group at C2 of the alkene in 1 could reduce the rate of cyclization at C2 of the 1,2-biradical intermediate 1*, resulting in selective 6-exo-trig cyclization of the more reactive C1 position. In addition, the group at C2 would stabilize the resulting 1,4-biradical intermediate, thus providing a thermodynamic driving force for the formation of the 6-membered ring intermediate 1ʹ, before radical recombination yields the bridged bicyclic product 2. Support for this hypothesis was provided by a density functional theory (DFT) study of the transition state energies of two cyclization pathways of 1,2-biradical 3*, the precursor to 3,6-N2-BCHep 4 (Fig. 1d). Cyclization of C2-methylated substrate 3ʹ led to the expected 5-exo-trig cyclization (see Supplementary Section 5 for DFT results), whereas introducing a phenyl ring as a radical-stabilizing group at C2 kinetically suppressed this undesired pathway, thereby enabling the anti-rule-of-five 6-exo-trig cyclization. Interestingly, this same selectivity reversal effect is not observed for the analogous 1,6-dienes49,50,51, highlighting the importance of the presence of the oxime.

Here, we report the successful implementation of this anti-rule-of-five intramolecular [2 + 2] photocycloaddition of aza-1,6-dienes using EnT photocatalysis with visible light, which provides a general strategy for the synthesis of substituted 6-N-BCHeps, 3,6-N2-BCHeps and 3-O-6-N-BCHeps. These structurally rigid motifs exhibit well-defined exit vectors and could therefore serve as bioisosteres for a range of nitrogen-containing heterocycles, including both aromatic and saturated systems. Moreover, the rich downstream chemistry offered by these compounds further underscores their synthetic utility in accessing structurally complex molecules. This allowed facile preparation of a 3,6-N2-BCHep analogue of a piperazine-based drug, whose physicochemical properties were tested to evaluate the potential of these scaffolds as bioisosteres in medicinal chemistry.

Results and discussion

Reaction design and optimization

To evaluate our hypothesis, aza-1,6-diene 3 containing an oxime and radical-stabilizing phenyl ring at C2 was chosen as the model cycloaddition precursor (Table 1). After evaluation of various parameters, we found that this reaction was indeed viable. Optimum conditions required the use of [Ir(dF[CF3]ppy)2(dtbbpy)]PF6 (1 mol%) as the photocatalyst under 427-nm light, in acetonitrile (0.1 M) at room temperature, which gave 3,6-N2-BCHep 4 in 68% isolated yield together with 13% of the fused product 4a (entry 1). This showed that the intramolecular [2 + 2] photocycloaddition of aza-1,6-dienes could undergo the anti-rule-of-five pathway through tuning the reactivity of the 1,2-biradicals (1*; Fig. 1c). Alternative photocatalysts with different redox potentials and triplet energies were tested (entries 2–5), and a correlation was found between the yield and triplet energy of the photocatalyst, but not with redox potentials, indicating that the reaction probably occurs through triplet EnT rather than a photoredox process52. The reaction was found to be relatively insensitive to solvent polarity (Supplementary Table 1), which again supports an EnT mechanism, where a neutral encounter complex rather than charge separation is involved53. Although thioxanthone, which was discovered to be similarly as effective at promoting the desired reaction of substrate 3, represents a cost-effective and practical alternative photosensitizer, it was deemed to be less general in its reactivity with a broad array of substrates. Finally, control experiments showed that both light and photocatalyst were necessary (entries 6 and 7).

Table 1 Optimization of the reaction conditions
Full size table

Substrate scope

Having established the optimized conditions, we then explored the substrate scope (Fig. 2). Variation of the aromatic alkene substituent demonstrated that electron-deficient (6–11) and electron-rich phenyl rings (12 and 13) as well as heteroaromatics (14–17) could be used, giving the corresponding 3,6-N2-BCHeps in moderate-to-good yields. The reaction was successfully scaled up to gram scale with lower catalyst loading (0.5 mol%), giving the product 4 in similar yield (61% versus 68%). The reaction was further extended to incorporate aromatic fragments of six marketed pharmaceuticals (18–23), highlighting the broad functional group tolerance of the photocycloaddition. In all these cases, the fused bicycle side-products were either isolated in low yields or not observed at all (see Supplementary Section 1.11 for details). Importantly, the [2 + 2] photocycloaddition was not limited to aromatic alkenes and could be extended to dienes (24–26)54, which displayed complete regioselectivity for the formation of the bridged bicycle over the fused bicycle. Enones could also be used (27 and 28)55, although they were lower-yielding and returned substantial quantities of starting material (20–40%). Modifications to the tether between the styrene and oxime were also explored, including other N-protecting groups (29 and 30), oxygen (31 and 32) and all-carbon linkers (33). Furthermore, 2,3-amide (34), 3,4-amide (35 and 37) and 2,3,4-imide (36 and 38–44) tethers could be used, delivering the bridged bicycles in high yields. Using amide and imide tethers also enabled the bridgehead position (C5) to be substituted with various (hetero)alkyl or aryl groups (37–44), with examples including pharmaceutically relevant heterocycles (42 and 43). In addition to O-methyl oximes, it was found that a broader range of oximes (47 and 48) and hydrazones (49–52) underwent successful cycloaddition. While the parent unsubstituted (N–OH) oxime reacted efficiently, the N–OH product 45 was found to be unstable and rearranged to the [3.2.1] bicycle 46. However, in situ O-acylation prevented this rearrangement and allowed isolation of the stable acetoxy derivative 47 in good yield.

Fig. 2: Substrate scope of the 6-N-BCHep synthesis.
Fig. 2: Substrate scope of the 6-N-BCHep synthesis.
Full size image

Reactions were performed with a mixture of E/Z oximes (or hydrazones) as the substrates, using 1 mol% [Ir(dF[CF3]ppy)2(dtbbpy)]PF6 in MeCN (0.1 M) under blue LED (427 nm) irradiation for 1–4 h. Yields are of isolated products. aThe reaction time is 20 h. b0.5 mol% [Ir(dF[CF3]ppy)2(dtbbpy)]PF6 was used. cThioxanthone (10 mol%) was used as the photocatalyst under 390-nm LED irradiation. d5 h reaction time, and 22% of the substrate was recovered. e24 h reaction time, and 41% of the substrate was recovered. fCH2Cl2 (0.1 M) as the solvent, then acylation using Ac2O. gDMSO (0.1 M) as the solvent. i-Pr, iso-propyl; Bn, benzyl; Ac, acetyl; DMSO, dimethyl sulfoxide.

Synthetic transformations and applications

Having shown the broad range of 6-N-BCHeps that could be accessed using this anti-rule-of-five photocycloaddition of aza-1,6-dienes, we were interested in extending the diversity of accessible products through further derivatisation. Vinyl 3,6-N2-BCHep 26 is a valuable substrate for diversification because of the synthetic utility of the vinyl group (Fig. 3, left). Importantly, the synthesis of 26 was scalable, with an 85% yield obtained when the cycloaddition was performed on a gram scale. We subsequently demonstrated that the vinyl group in 26 could be hydrogenated to give alkyl-substituted 3,6-N2-BCHep 53 and subjected to hydroboration/oxidation to give alcohol 54. Ozonolysis followed by amination, reduction or oxidation provided oxime 55, alcohol 56 and carboxylic acid 58, respectively. In addition, methylated product 57 could be accessed by sulfonylation and reduction of alcohol 56. Finally, derivatization of carboxylic acid 58 by amidation and reductive decarboxylation allowed the generation of hydroxamic acid 59 and the unsubstituted 3,6-N2-BCHep skeleton 60.

Fig. 3: Product diversification.
Fig. 3: Product diversification.
Full size image

Diversification of products 26 and 61 using a wide variety of conditions. aPd/C, H2. b9-Borabicyclo(3.3.1)nonane (9-BBN), then H2O2, NaOH. cO3, Me2S, then NH2OBn·HCl, NaOAc. dO3, then NaBH4. e4-(Trifluoromethyl)benzenesulfonyl chloride, Et3N, 4-dimethylaminopyridine (DMAP), then NaBH4. fO3, Me2S, then NaH2PO4, 2-methyl-2-butene, NaClO2. gOxalyl chloride, then NH2OH·HCl, K2CO3. hN-hydroxyphthalimide, N,N′-diisopropylcarbodiimide, DMAP, then [Ir(ppy)2(dtbbpy)]PF6, Et3N, 1,4-cyclohexadiene, 456-nm blue LEDs. iIron powder, NH4Cl. jPhenyl isocyanate. kPhenyl isothiocyanate. lp-Toluenesulfonyl chloride, Et3N, DMAP. mN,N-dimethylsulfamoyl chloride, Et3N, DMAP. nDi-tert-butyl dicarbonate, DMAP. op-Toluoyl chloride, Et3N, DMAP. Tol, 4-methylphenyl; Boc, tert-butoxycarbonyl.

Further product diversification was possible through reductive cleavage of the oxime-derived N–O bond and functionalization of the resulting secondary amine (Fig. 3, right). While reduction of the N–OMe bond in 4 was not possible due to the sensitivity of the aryl-azetidine group, the more polarized N–O bond in acetoxy derivative 47 was readily cleaved under mild reducing conditions (iron powder)56, providing amine 61 in 85% yield. This enabled high-yielding derivatization to generate urea 62, thiourea 63, sulfonamide 64, sulfamide 65, carbamate 66 and amide 67, demonstrating the broad range of functional groups that could be incorporated onto the 3,6-N2-BCHeps.

To determine whether the 3,6-N2-BCHep scaffolds are suitable for use in bioactive molecules, 68 was prepared as an isosteric analogue of the matrix metalloproteinase inhibitor DB04232 and subjected to analysis57 (Fig. 4). The physicochemical and pharmacological properties of 68 were assessed in comparison with DB04232. Interestingly, although the isosteric analogue has a greater molecular weight (MW) than DB04232, it has reduced lipophilicity (logP). This counterintuitive reduction in lipophilicity has been observed previously and ascribed to an increase in solvent-exposed polar surface area resulting from conformational constraints58. Apart from this positive attribute, 68 exhibited similar intrinsic clearance (CLint) in human liver microsomes (in vitro human CLint, ml min−1 kg−1) and had a similar half-life (T1/2) compared with DB04232. These findings indicate that 68 is at least as metabolically stable as DB04232, demonstrating the potential for application of 3,6-N2-BCHep analogues as isosteres of frequently used heterocycles in drug candidates. However, further studies on these systems, including efficacy analysis and comparisons with unsaturated analogues, are required before the prospective utility of these systems can be fully evaluated.

Fig. 4: Metabolic profile of DB04232 3,6-N2-BCHep analogue 68.
Fig. 4: Metabolic profile of DB04232 3,6-N2-BCHep analogue 68.
Full size image

aDefined as logP at pH 7.4; determined by high-performance liquid chromatography (HPLC) analysis. bDefined as metabolic half-life. cDefined as microsome intrinsic clearance. dDefined as hepatic intrinsic clearance. eDetermined at 60 min in the presence of human liver microsomes with NADPH (reduced nicotinamide adenine dinucleotide phosphate). Cbz, benzyloxycarbonyl.

Conclusions

In this study, we have introduced a visible-light-mediated intramolecular [2 + 2] cycloaddition of aza-1,6-dienes, a synthetic approach that defies the long-standing rule-of-five in radical cyclization. By exploiting differential radical stabilization, we demonstrated that this strategy can be used to access a diverse range of substituted 6-azabicyclo[3.1.1]heptane (6-N-BCHep) scaffolds, a previously challenging to access class of bridged nitrogen heterocycles. The methodology demonstrates broad functional group tolerance and enables selective substitution at multiple positions, expanding the chemical space for medicinal chemistry applications. Moreover, preliminary studies on the newly synthesized 6-N-BCHeps show their potential as bioisosteres, offering similar or enhanced pharmacokinetic properties compared with their saturated counterparts. This work not only provides a synthetic tool for accessing complex heterocycles but also contributes to the ongoing exploration of three-dimensional chemical space, paving the way for future developments in medicinal chemistry and beyond.

Methods

General procedure for intramolecular [2 + 2] cyclization

An oven-dried round-bottom flask containing aza-1,6-dienes (1.0 equiv.) and [Ir(dF(CF3)ppy)2(dtbbpy)](PF6) (1 mol%) was sealed and subjected to three N2 evacuation–refill cycles before anhydrous solvent (0.1 M) was added. The flask was placed at a distance of approximately 1 cm from a 40 W KSPR160L-427 nm (or 390 nm) Kessil light (100% intensity), and the reaction was stirred under continuous irradiation under a nitrogen atmosphere for 1–4 h until complete (judged by thin-layer chromatography analysis). The solvent was removed in vacuo, and the crude product was purified by flash column chromatography.