Divergent Strain‐Release Amino‐Functionalization of [1.1.1]Propellane with Electrophilic Nitrogen‐Radicals

Abstract Herein we report the development of a photocatalytic strategy for the divergent preparation of functionalized bicyclo[1.1.1]pentylamines. This approach exploits, for the first time, the ability of nitrogen‐radicals to undergo strain‐release reaction with [1.1.1]propellane. This reactivity is facilitated by the electrophilic nature of these open‐shell intermediates and the presence of strong polar effects in the transition‐state for C−N bond formation/ring‐opening. With the aid of a simple reductive quenching photoredox cycle, we have successfully harnessed this novel radical strain‐release amination as part of a multicomponent cascade compatible with several external trapping agents. Overall, this radical strategy enables the rapid construction of novel amino‐functionalized building blocks with potential application in medicinal chemistry programs as p‐substituted aniline bioisosteres.


Introduction
The bicyclo[1.1.1]pentane (BCP) [1] is a rigid, linear, sp 3 -rich motif frequently used in medicinal chemistry programs as bioisosteric replacement for aromatic rings, alkynes and t-Bu groups. [2] Bioactive molecules incorporating this substructural unit usually benefit from improved pharmacokinetic properties like lipophilicity and passive permeability. [3] In particular, bicyclo[1.1.1]pentylamines (BCPAs) are of considerable interest as aniline bioisosteres for the preparation of novel small-molecule therapeutics with resistance towards metabolic clearance (Scheme 1A). [4] Since the pioneering work of Wiberg [5] and Szeimies, [6] few multi-steps approaches towards these building blocks have been developed, generally relying on the three steps conversion of [1.1.1]propellane 1 into 2 followed by Curtius rearrangement to give 3 and further elaboration (Scheme 1B, path a). [7] A breakthrough in the field has come in 2016 when Baran demonstrated that di-alkyl amines, upon prior deprotonation with i-PrMgCl•LiCl, react with 1 by ionic strainrelease thus enabling a straightforward construction of monosubstituted BCPAs (Scheme 1B, path b). [8] In contrast, the preparation of C-3-substituted BCPAs is much more difficult and still go through long multi-step synthesis. [9] In an effort to streamline the preparation of these building blocks, Gleason has very recently merged ionic strain-release amination with Cu(I)-catalysis using activated electrophiles to assemble C-3 alkylated and hydroxylated systems (Scheme 1B, path c). [10] This telescoped process requires the accumulation of the magnesiated-BCPA intermediates 4, prior addition of the Cu(I)/electrophile, thus restricting the source of N-partners to benzylic amines. Radical strain-release amino-functionalization would represent an attractive alternative to target the preparation of these building blocks. So far, the only approach for the direct preparation of BCPAs is based on the amination of C-3 BCP radicals with Nbased SOMOphiles as developed by Pfizer. [11] In this case, radical hydro-hydrazination of 1 using DEAD as the SOMOphile thus enabling upon further elaboration, the scalable preparation of building block 5 in three steps (Scheme 1B, path d). More recently Uchiyama has extended this approach to the preparation 3-functionalized BCPAs through a carbo-hydroazidation strategy using alkyl carbazates/aryl hydrazines as radical precursors and DEAD as radical trap (Scheme 1B, path e). [12] Also in this case, the C-N bond is formed after strain-release reaction with acyl/aryl radicals, and therefore only hydrazine-type substituents can be introduced. Albeit this multicomponent strategy has not been extended to any other class of radical traps, it has significantly shortened access to common the BCPAs intermediate 3, which can be further elaborated. The difficulties in accessing C-3-substituted BCPAs contrast with the ease of preparation of C-3-substituted 1-alkyl-BCPs. Indeed, carbon-, sulfur-and, to a lesser extent, silicon-and tin-radicals have been widely used in ATRA-based (atom-transfer radical addition) strategies to generate, in a single step, 1,3-disubstituted BCPs (Scheme 1C). [5,7,13] In this case however, the inherent mechanistic requirement for the chain carrying radical to be regenerated from the precursor in the final atom/group transfer step, does not offer option of divergency. Furthermore, despite the reach radical strain-release chemistry, no example of a nitrogen-radical able to engage 1 in direct strain-release amination has so far been reported. In view of our previous experience on the use of nitrogen-radicals in olefin amino-functionalization, [14] we recently questioned whether it would be possible to use these species in strain-release settings. Despite there is no example of nitrogen-radical addition to [1.1.1]propellane 1, we reasoned that its successful implementation had the potential to streamline access to important BCPA building blocks currently elusive from other approaches. Critical to our design would be the ability to intercept the C-3 BCPA radical resulting from strain-release amination with different SOMOphiles. Divergent radical functionalizations of [1.1.1.]propellane 1 have, to the best of our knowledge, not been reported, but their implementation would provide significant expansion in the structure and scope of the process. Herein, we demonstrate that amidyl radicals undergo efficient strain-release reaction with 1 and that this reactivity can be harnessed as part of a strategy leading to the multicomponent assembly of 3functionalized BCPAs (Scheme 1D). This approach displays significant synthetic complementarity to ionic strain-release amination as it enables, for the first time, the direct introduction of amide, carbamate and sulfonamide functionality across the BCP core. The fact that this process does not rely on a radical-chain propagation mechanism has enabled the diversification of the BCPAs with six different SOMOphiles that are not compatible with any of the previous amination technologies. Furthermore, as amidyl radicals display a distinct electrophilic character, this cascade reactivity represents an umpolung approach to ionic strain-release amination and, to the best of our knowledge, the first example of a radical process enabling divergent strainrelease-amino-functionalization. Scheme 1. A) Bioactive molecules containing the BCP motif. B) Previous radical and ionic approaches for the synthesis of BCPA building blocks. C) Formation of 1,3-disubstituted BCPs using ATRA-based systems. D) Proposed radical strategy for the divergent preparation of 3-functionalized BCPAs using nitrogen-radicals. Ac = acetyl, Bn = benzyl, Boc = tert-butyloxycarbonyl.

Reaction Design and Analysis
From the outset we recognized that the realization of a divergent strain-release amino-functionalization process with nitrogenradicals would require developing a three-component cascade compatible with external SOMOphiles. We therefore postulated a strategy based on a reductive quenching photoredox cycle (Scheme 2A). Under these settings, a visible-light excited photocatalyst (*PC) [15] would oxidize the carboxylate functionality of the radical precursor A, triggering two fragmentations (extrusion of CO2 and acetone) and forming the amidyl radical B. [16] As this species has a distinct electrophilic character, we were hopeful that it would be able to intercept electron rich 1 by cleaving its central and inverted sp 3 -sp 3 C-C bond. This polarized radical strain-release amination would assemble the key C-N bond and give the C-3 BCPA radical C. [17] This species would be then diversified by atom/group-transfer reaction (SH2) with a range of SOMOphiles (X-Y) providing the targeted building blocks D. At the end, the electron poor radical Y• would render the process redox-neutral by closing the photoredox cycle by SET with the reduced photocatalyst (PC •-). As mentioned before, as amidyl radicals are electrophilic species, this process represents the first example of an umpolung strain-release amination. This is mechanistically interesting because, while both nucleophiles and nucleophilic radicals can participate in strain-release chemistry, umpolung reactions can only be conducted via radical approaches owing to the known instability of BCP cations. [18] As strain-release reactions of nitrogen-radicals have not been reported before, we performed computational studies to evaluate the feasibility of the proposed reactivity (Scheme 2B). [19] Overall, our investigations demonstrated that these reactions share close similarities to the intermolecular additions of amidyl radicals to olefins. They are increasingly exothermic the more electrophilic is the nitrogen radical (determined using their electrophilicity indices, w + rc) and the barrier for strain-release decrease with increasing reaction exothermicity. Furthermore, there is a good correlation between both the exothermicity (DGº) and the barrier (DG ‡ ) of the reaction ( Figure 2B, chart i) and DGº and the electrophilicity of the incoming nitrogen radical ( Figure 2B, chart ii). Owing to the electron rich nature of 1 (IP = 9.7 eV), [20] there is strong  participation of polar effects in the transition state (|d TS |) due to the net charge transfer with the electrophilic amidyl radicals. Finally, there is also a linear correlation between bond formation distance [d(C-N)] and the reaction enthalpy, which is in agreement with the Hammond postulate (Scheme 2B, chart (iii)). [21] Overall, these computational findings suggest radical strain-release reactions of amidyl radicals are characterized by early transition states and the enthalpy governs to some extent the transition state geometry.

Scheme 2. A)
Proposed mechanism for strain-release with nitrogen-radicals. B) Computational studies on the strain-release reaction of nitrogen-radicals. C) Computational studies on the character of C-3 BCPA radicals. D) Comparison of energy profiles for the elementary steps in the radical cascade, compared with the side reactivity leading to staffane. E) Stern-Volmer analyses of 1 with common photocatalysts. Bz = benzoyl, Cbz = benzyloxycarbonyl.
Despite these encouraging features, it is important to note that the success of the strategy does not hinge solely on the feasibility of the strain-release reaction of amidyl radical B (1 + B g C) but also on the following atom/group-transfer reaction (C + X-Y g D). In fact, this step is required to be efficient in order to avoid the known and facile oligomerization of BCP-radicals with other molecules of 1 that leads to common staffane by-products. Indeed, while exploring the reactivity of S-radicals, Scaiano demonstrated that despite these species undergo facile strain-release reaction with 1, their reactivity could only be unlocked in systems benefitting from very efficient ATRA-based chain propagations to stop staffane formation. [22] As our proposed reactivity relies on a multicomponent radical cascade where the amidyl radical generation is dissected by the product formation, there is a fundamental mechanistic requirement for the final atom/grouptransfer step to be kinetically preferred over BCP-radical oligomerization. [5] In general, BCP radicals display enhanced reactivity towards SOMOphiles compared to tertiary radicals (e.g. t-Bu•) because of their (i) permanent pyramidal arrangement, (ii) reduced steric shielding and (iii) increased SOMO s-character. [23] In our case however, we were concerned that due to the very short C1-C3 cross-cage distance, the electronegative amide functionality could retard the C-3 BCPA-radical reactivity with polarized SOMOphiles by inferring a through-space negative kinetic polar effect that could instead favor the polarity matched oligomerization. [23][24] Indeed, our calculations have demonstrated that C-3 BCPA radicals C are significantly less nucleophilic than 1-alkyl-BCP ones and that their electrophilicity closely mirrors the one of the incoming amidyl radicals B (Scheme 2C, chart iv). This supports the presence of strong cross-cage interaction leading to a remote through-space modulation of their radical philicity. Nevertheless, Cl-transfer from N-chlorosuccinimide (NCS) was calculated to have a slightly lower barrier than staffane formation, thus supporting the feasibility of the overall cascade reactivity (Scheme 2D). Before starting the optimization of the aminofunctionalization strategy, we decided to evaluate the behavior of 1 in the presence of several photocatalysts. As this species is electron rich and has also demonstrated to undergo energy transfer (EnT) with triplet benzophenone, [25] we wanted to evaluate the ease of formation of 1 •+ /1*, that suffer fast decomposition. These Stern-Volmer studies demonstrated 1 undergoes appreciable bimolecular quench only with the strongly oxidant Fukuzumi's acridinium (*Eox > 2.1 V vs SCE) (Scheme 2E). Importantly, the deprotonated amidyl radical precursors A undergo luminescence quenching at significant higher rates with all photocatalyst evaluated, [16,21] which ensure the overall feasibility of our proposed photoredox manifold.

Reaction Development and Scope
The amino-functionalization strategy was first evaluated using 7 as the amidyl radical precursor, NCS (6a) as the SOMOphile, Cs2CO3 as the base and Fukuzumi's acridinium (PC-1) as the photocatalyst (*Eox >2.1 V vs SCE) [26] under blue-light irradiation (l = 420 nm) in CH2Cl2 solvent (Scheme 3A). The first challenge to address was the preparation of 1. Baran has demonstrated its large-scale synthesis as stock solution in Et2O, [8] however, as amidyl radicals are very electrophilic, they undergo efficient Hatom transfer (HAT) [27] from this solvent. Indeed, all efforts to optimize the proposed reactivity using 1 as 1.0 M Et2O solution resulted in quantitative formation of N-methyl-benzamide 8 (entry 1). We therefore had to attempt the preparation of 1 in different solvents to exclude the presence of hydridic H-atom donors. Pleasing, when the reaction was run using 1 as a 0.25 M solution in perfluoro-n-hexane, 3-Cl-BCPA 9 was obtained in 40% yield albeit in combination with several side products arising from staffane-formation and inefficient Cl-transfer (entry 2). Crucially, by co-distilling 1 in benzene (entry 3) and changing the photocatalyst to the less oxidative Ir[dF(CF3)ppy2](dtbpy)(PF6) (PC-2) (*Eox = 1.69 V vs SCE) (entry 4), we progressively improved the yield to 78%. Interestingly, when the reaction was completely run in benzene, 9 was obtained in lower efficiency, thus pointing to a favorable solvent effect played by CH2Cl2 (entry 5). Co-distillation of 1 with CH2Cl2 was attempted but found not feasible as this solution underwent decomposition over just 1 h at room temperature. We therefore evaluated the preparation of 1 in various benzene:CH2Cl2 mixtures assessing their stability over time. Eventually we developed a procedure for its reproducible and large-scale preparation as a 0.5 M solution in a 3:1 benzene:CH2Cl2 mixture. [21] This stock solution can be stored for over a month at low temperature (-5 ºC under nitrogen in the dark) and displays sufficient stability at room temperature for handling (>1 day). The use of this propellane source was critical for the implementation of the strain-release cascade and 9 was obtained in 90% yield (entry 6). Control experiments confirmed the requirement for both photocatalyst, base and continuous bluelight irradiation (entries 7-9).
Having identified optimal conditions for strain-release aminochlorination, we next explored the possibility to render this strategy divergent using other SOMOphiles (Scheme 3B). While NBS lead to a complex reaction mixture, Br-CCl3 6b provided the amino-bromination product 10 in 20% yield albeit in the presence of staffane by-products. We rationalized this outcome on the basis of a more challenging Br-transfer reaction (calculated barrier: DG ‡ = 11.2 kcal mol -1 ) that can be outcompeted by the oligomerization. Phthalimide-based reagents 6c-e were efficient for the introduction of S-and Se-functionalities (11-13). Building block 11 was further modified by simple oxidation providing access to BCPAs 14 and 15 that contain a C3-sulfone and sulfoxide respectively. Finally, several HAT catalysts were evaluated to achieve the formation of amide-containing BCPA 16 that cannot be accessed by ionic strain-release. While the commonly used methyl thioglycolate resulted in complex reaction mixtures, 2-Phmalononitrile [28] 2f provided 16 in good yield. The amidyl radical substitution pattern was evaluated next using NCS 6a as the SOMOphile. As shown in Scheme 3C, N-Mebenzamides containing both electron donating (17) and electron withdrawing (18) substituents were tolerated without loss in reaction efficiency. Moreover, the presence of heteroaromatic (19 and 20) and alkyl (21) substituents did not have a deleterious impact on the reaction performance. The N-substituent was also modified and we successfully obtained 22 and 23 that contain sterically demanding Ncyclohexyl and N-t-Bu groups. Primary BCPAs (e.g. 24) currently represent a limitation of the methodology as the reaction with NHamidyl radicals gave consistently complex mixtures. The strategy was then extended to more electrophilic carbamoyl radicals based on commonly used protecting groups like N-Cbz (25), N-Boc (26) and N-Ts (27). BCPA 28 contains a N-Boc,N-Bn substituent and can be orthogonally deprotected and modified. Pleasingly, we also succeeded in engaging nitrogen-radicals containing two carbonyl substituents (29-32). Their ability to successfully take part in the strain-release reactivity is noteworthy as these species are known to undergo facile intramolecular addition to aromatics (e.g. benzene). [29] This side reactivity became particularly problematic when extending the chemistry to   CF3C6H5-CH2Cl2. This electron poor aromatic co-solvent suppressed the unwanted N-arylation and enabled the assembly of 34 albeit in moderate yield. [30] An interesting structural feature of BCP-based building blocks is that despite their rigid and compact nature, the C1-C3 nonbonded distance changes depending on the electronic nature of their substituents. In particular, electronegative atoms/groups inductively remove electron density from the bridgehead carbons rear-lobe orbitals and, by decreasing orbital repulsion, shorten their distance. [31] As the impact of a nitrogen substituent on a 3substituted-BCP-core has never been assessed, we prepared BCPAs 35-41 containing a p-Ph-benzamide that enabled growing of crystals suitable for X-ray analysis (Scheme 3D). These substrates display progressively shorter C1-C3 cross-cage distances upon increasing electronegativity of the C3 substituent up to a very short 1.828 Å in the case of 36. A good correlation was found between these distances and the field/inductive parameters (F) for the various C3-substitutents, which supports the presence of strong through-space C1-C3 interactions. [21,32]

Conclusions
In conclusion, we have demonstrated that electrophilic nitrogenradicals are powerful intermediates to achieve multicomponent strain-release amino-functionalization reactions of [1.1.1]propellane 1 using external SOMOphiles. To the best of our knowledge, the process reported here represents the first example of a divergent radical strain-release-functionalization strategy. The ability to engage a range of different SOMOphiles has enabled the assembly of novel building blocks with potential application for the preparation of p-substituted aniline bioisosteres. DFT calculations have given an insight into the interplay of both enthalpic, polar and through-space effects operating in the strainrelease reaction of nitrogen-radicals and the following functionalization of the resulting C3-BCPA radicals. Furthermore, we have demonstrated that inter-molecular H-atom transfer, a common side reactivity of electrophilic radicals, can be effectively overcome with the preparation of [1.1.1.]propellane 1 in CH2Cl2benzene solution. We hope that these insights might lead to the rational design of related strain-release functionalizations using other classes of electrophilic radical species.