Carbon-to-nitrogen single-atom transmutation of azaarenes

Abstract

When searching for the ideal molecule to fill a particular functional role (for example, a medicine), the difference between success and failure can often come down to a single atom¹. Replacing an aromatic carbon atom with a nitrogen atom would be enabling in the discovery of potential medicines², but only indirect means exist to make such C-to-N transmutations, typically by parallel synthesis³. Here, we report a transformation that enables the direct conversion of a heteroaromatic carbon atom into a nitrogen atom, turning quinolines into quinazolines. Oxidative restructuring of the parent azaarene gives a ring-opened intermediate bearing electrophilic sites primed for ring reclosure and expulsion of a carbon-based leaving group. Such a ‘sticky end’ approach subverts existing atom insertion–deletion approaches and as a result avoids skeleton-rotation and substituent-perturbation pitfalls common in stepwise skeletal editing. We show a broad scope of quinolines and related azaarenes, all of which can be converted into the corresponding quinazolines by replacement of the C3 carbon with a nitrogen atom. Mechanistic experiments support the critical role of the activated intermediate and indicate a more general strategy for the development of C-to-N transmutation reactions.

Single-atom changes to a molecule can have outsized effects, because molecular properties critical for function such as hydrogen bonding, polarity, metabolic stability, target specificity and solubility can all be modulated considerably by the replacement of one constituent atom by another^4,5. Of all the possible single-atom exchanges, the replacement of an aromatic carbon atom with a nitrogen atom is privileged in medicinal chemistry for its ability to productively tune these features. The prominence of this design strategy is reflected in the notion of a ‘necessary nitrogen atom’, a term of art arising from the sheer prevalence of nitrogen-containing heterocycles in approved medicines^6,7. Four specific examples from three US Food and Drug Administration (FDA)-approved pharmaceuticals wherein replacement of one or more carbons in a lead scaffold was critical for the ultimate clinical candidate’s success are shown in Fig. 1a (refs. 8–10).

Fig. 1 |. Introduction.

a, Illustrative examples of ‘necessary nitrogen atom’ effects in drug development, found by replacement of a carbon atom with a nitrogen atom in the discovery process. b, The philosophical framework of single-atom transmutation as summarized by a square scheme. c, Two main pitfalls of existing insertion–deletion approaches. d, C-to-N transmutation of azaarenes delineated in this work. DPP4, dipeptidyl peptidase IV; PDE5, phosphodiesterase-5; ETA/B, endothelin-A/B; Me, methyl; ^tBu, tertiary butyl.

Clearly, the ability to mutate a single nitrogen atom in a direct fashion, starting from its parent skeletal carbon atom, would be impactful for molecular optimization campaigns^3,11,12. Although in principle this reaction only requires a change of two nucleons, in practice one needs instead to cleave several strong C–C bonds and introduce a nitrogen in a fashion that is compatible with complex structures. Such skeletal editing of aromatic heterocycles poses a notable synthetic challenge, such that necessary nitrogen atoms are currently identified by iterative, arduous resynthesis of each new candidate during structure–activity relationship studies. Moreover, as the nitrogen content increases, the syntheses of these compounds often become more challenging. As a result, recent efforts among the synthetic chemistry community have been directed at expanding the toolbox of skeletal editing transformations^13–21; however, the ability to conduct a direct ‘transmutation’ of aromatic and heteroaromatic skeletons remains substantially limited.

A framework for developing a C-to-N transmutation can be constructed from the existing, sparse precedents in this space. Two potential strategies can be summarized by the ‘square scheme’ depicted in Fig. 1b: one can either delete a carbon first to generate a smaller (hetero) cycle followed by insertion of a nitrogen^22,23, or alternatively first insert the nitrogen into the aromatic system and then delete the carbon^24,25. Two critical pitfalls have limited the existing approaches (Fig. 1c). First, if the inserted nitrogen is not introduced with the same selectivity as the departing carbon atom, a skeletal rotation occurs. As a specific example, a hypothetical combination of our previously reported C2 deletion of quinolines with the Morandi laboratory’s N3-selective insertion would translocate the former C3 carbon of the quinoline to C2 of the product quinazoline, along with its associated substituent^19,22. Although potentially interesting, this rearrangement disrupts the orientation of substituent vectors in a manner that is unlikely to preserve the binding properties of the parent compound. Second, many insertion and deletion strategies (for example, aminoazepine approaches) simultaneously and necessarily introduce extra appendages to the starting skeleton as a consequence of their mechanisms, preventing the straightforward interrogation of necessary nitrogen effects²⁵.

Accordingly, we considered whether a distinct strategy for C-to-N transmutation could be devised in which a skeletal reconstruction would prime the system for simultaneous nitrogen insertion and carbon deletion. Such an approach addresses site selectivity innately, as the incoming and departing atom are necessarily modifying the same position. This direct pathway requires that the carbon atom is activated as a labile leaving group; at the same time, the activated intermediate should have the requisite electrophilic functionality to incorporate the incoming nitrogen atom^26,27. Provided that this latter electrophilic functionality is correctly selected, such an approach also avoids the introduction of vestigial functional groups. From a conceptual standpoint, this strategy is reminiscent of a restriction-enzyme-mediated ligation process, where our chemical ‘sticky ends’ would similarly serve to enable reclosure of the activated ring²⁸.

We proposed that such an intermediate could be prepared in situ from oxidative cleavage of a benzoxazepine, wherein antiaromaticity-driven deconjugation enables the enol ether moiety to behave as an independent alkenyl functionality (Fig. 1d). We previously reported that azaarene N-oxides (1) cleanly rearrange to benzoxazepines on irradiation by a mild and selective light-emitting diode (LED) light source²²; together these two operations yield an intermediate bearing two carbonyl groups that can act as the requisite ‘sticky ends’ (3). The excised carbon atom is contained in the imidic anhydride function, whose carboxylate subunit can be displaced as a leaving group (5) while activating the imidic carbon for attack by the incoming nitrogen atom. With ammonia serving as the nitrogen source, this outlined protocol renders transmutation a coherent process, accomplishing the direct synthesis of polyazaarenes such as quinazolines (2) from their parent heterocyclic precursors (4, Fig. 1d).

We investigated a number of potential oxidative cleavage conditions, ultimately finding ozonolysis (with standard reductive workup procedures) to afford the highest yields^29,30. Because hydrolysis of the imidic anhydride was problematic, and because we sought a one-pot protocol that would avoid further isolation operations, we explored addition of the ammonia nucleophile (in the form of ammonium carbamate) before ozonolysis³¹. In these experiments, we found a substantial improvement when applying Dussault’s pyridine-modified conditions, which bypass ozonide formation by catalysing decomposition of the carbonyl oxide intermediate (vide infra)³². In the event, 2-phenylquinoline 1-oxide was first converted to benzoxazepine by irradiation with a 390 nm LED, after which ammonium carbamate and pyridine were added to the crude benzoxazepine mixture before ozonolysis at −78 °C. On subsequent heating of this mixture to 90 °C, a 68% isolated yield of quinazoline (2a) was obtained. This protocol could also be streamlined with N-oxidation: initial peracid oxidation of the parent quinoline in chloroform and dilution with toluene, followed by the above protocol (carried out in the same pot) afforded quinazoline (2a) in an only slightly diminished yield (55% on a 1.0 mmol scale) compared with the two-pot protocol with purified N-oxide (68%).

In examining the scope of this method (Fig. 2), we found that a variety of C2-aryl quinazolines (2b, 2c and 2d) could be prepared in good yields, including heterocyclic substituents (2e, 2f and 2g). This could be extended to an N, N’-dioxide (1a), which was observed to cleanly rearrange to a bis(benzoxazepine) and could subsequently be converted to the biquinazoline (2h), by a direct double exchange of two carbon atoms. C2-ester (2j) and C2-alkyl (2k, 2l, 2m and 2n) quinazolines were similarly accessible by this method, although their polarity and aqueous solubility required an alternate, anhydrous workup procedure to avoid isolation losses (see Supplementary Information for details). A surprisingly broad scope of substituents on the excised C3 carbon atom was found. Quinoline N-oxides bearing hydrophobic and sterically demanding alkyl substituents (1b and 1c), aromatic substituents such as phenyl (1d), furan (1e) and isoxazole (1o), electron-withdrawing functionalities such as ester (1f), benzoyl (1g), nitrile (1h) and sulfonyl (1j) groups, and heteroatom substituents such as chloro (1i), methoxy (1k) and phthalimide (1l) could all be successfully converted into quinazolines by this method. The N-oxide moiety can also direct transition-metal catalysed C–H functionalization at the C8 position of quinolines^33–35. We used one such C–H iodination in combination with our protocol to afford quinazoline (2r), highlighting the potential synergy between peripheral and skeletal editing. Finally, the title reaction was found to be tolerant of various functionalities such as dioxane (2t), trifluoromethyl (2u), ethyl 2-oxyacetate (2z), methoxy (2y) and carbamate (2w), as well as useful synthetic handles such as bromo (2aa) and phosphonate (2x) at positions around the quinoline ring.

Fig. 2 |. Scope of the C-to-N transmutation of azaarenes.

Conditions: 1 (0.3 mmol), toluene (0.06 M), 390 nm LED, 1–5 h at 25 °C, then ammonium carbamate (7.0 equivalents (equiv.)), pyridine (10.0 equiv.), O₃/O₂ bubbling for 5–20 min at −78 °C, followed by heating at 90 °C for 24 h. Isolated yields are given. ^aHeating for 48 h. ^b12.0 equiv. of ammonium carbamate. ^cAmmoniolysis was carried out with 12.0 equiv. of ammonium acetate in ethanol (0.1 M). ^dIsolated with alternative workup procedure. See Supplementary Information for the details. Ph, phenyl group; OMe, methoxy group; OEt, ethoxy group; mCPBA, meta-chloroperoxybenzoic acid; Tf, triflate; Cp*, 1,2,3,4,5-pentamethylcyclopentadiene; Boc, tert-butyloxycarbonyl.

Although the competitive hydride shift to form quinolones is only partially outcompeted by benzoxazepine formation in C2-unsubstituted quinoline N-oxides, quinazolines lacking C2 substituents could nonetheless be prepared by this method (2o, 2p and 2q), albeit in diminished yields^36,37. As noted in our previous report, heteroatom substituents at C2 are incompatible with either N-oxidation or with the subsequent photolysis, and pyridine N-oxides do not cleanly afford the corresponding oxazepines even under LED irradiation, instead forming a host of secondary photoproducts^22,38. Substrates with C4 substituents (2s, 2t and 2u) produce a ‘sticky end’ intermediate with a ketone rather than an aldehyde terminus, requiring more forcing ammoniolysis in ethanol in place of the parent protocol. When the C2 and C3 carbons were tethered by an eight-membered ring (1r), the ring could be ‘clipped’ to form the corresponding quinazoline 2ad while retaining the former C3 carbon as a pendant functional group³⁹. We propose that the intermediate carboxylate in this reaction is subsequently converted to the corresponding amide by the excess of ammonia used—this is more generally a liability in this transformation for acyl derivatives (acids and esters) that are prone to nucleophilic amidation (for example, 2v, for which the transamidation by-product is also formed). Further limitations include oxidatively sensitive functional groups that are not compatible with N-oxidation and/or ozonolysis (for example, alkenes), and 4-substituents that generate unstable intermediates on ozonolysis (for example, acid chlorides formed from 4-chloroquinolines) or when the intermediate resists condensation (for example, 4-aminoquinolines that generate amides).

Ozonolysis has been used to synthesize pharmaceuticals on an industrial scale^40–42, so we aimed to showcase our transmutation method on the gram scale in the synthesis of belumosudil (6), a drug approved by the FDA in 2021 (Fig. 3a)⁴³. The precursor quinoline 4b was prepared in two steps from 3-(quinolin-2-yl)phenol and, after N-oxidation and transmutation, 1.3 g of the quinazoline 2ae was obtained from 1.9 g of 1s. The synthesis of 6 was completed by a two-step C–H amination sequence. We further demonstrated the relevance of our method by applying it to the full set of isomeric [1,n]-naphthyridines, affording rare [1,3,n]-triazanaphthalenes in a straightforward fashion (Fig. 3b). When several basic nitrogen atoms are present on the same core skeleton, application of our method requires selective oxidation of a ‘quinoline-type’ rather than an ‘isoquinoline-type’ nitrogen. For 1,4- and 1,8-naphthyridines this is not an issue, and either a symmetric structure (2af) or one in which N-oxidation is selective owing to the 2-substituent (2ak) can be used. For 1,5- and 1,6-naphthyridines, our approach was to attenuate the nucleophilicity of the ‘isoquinoline-type’ nitrogen through either methoxy (2ah, 2ai) or chloro (2ag, 2aj) substitution. We also used the transmutation protocol to edit a synthetic intermediate in the preparation of talnetant, an FDA-approved TACR3 receptor antagonist; O-methyl talnetant N-oxide is converted to the corresponding quinazoline under our conditions⁴⁴. A methyl ester derivative of the dihydroorotate dehydrogenase inhibitor brequinar was similarly susceptible to our method to afford the corresponding quinazoline 2am (Fig. 3c)⁴⁵.

Fig. 3 |. Synthetic applications of C-to-N transmutation of azaarenes.

a, Scalable synthesis of belumosudil. b, Synthesis of rare triazanaphthalenes using single-atom transmutation logic. c, Editing of talnetant and brequinar. d, Alternative oxidative cleavage mediated by photoexcited nitroarene. Isolated yields from N-oxide substrates. See Supplementary Information for the detailed conditions. BOP, benzotriazol-1-yloxytris(dimethylamino) phosphonium hexafluorophosphate; DBU, 1,8-diazabicyclo[5.4.0]undec-7-ene; TFA, trifluoroacetic acid; TACR3, tachykinin receptor 3; DHODH, dihydroorate dehydrogenase.

As some chemists do not have access to an ozone generator, a supplemental oxidative cleavage protocol was developed using a photochemical nitroarene-promoted method, recently developed contemporaneously by Parasram and Leonori^46,47 (Fig. 3d). Nitroarene (7) was added to the benzoxazepine generated by LED photolysis in acetonitrile at −30 °C, and continued photolysis with the same LED followed by addition of ammonium carbamate and trifluoroacetic acid resulted in a 63% yield of 2a, along with a small scope of other quinazolines. On the basis of previous reports, we suspect the oxidative cleavage proceeds via the 1,3,2-dioxazolidine 8, but its subsequent breakdown and ammoniolysis is probably more complex than the ozonolytic protocol, and in general we have found that the nitroarene approach is more limited in scope than our parent method.

To gain further insight into the mechanism of this transformation, we prepared ¹³C-labelled quinoline N-oxide 1u (Fig. 4a). Monitoring of the reaction by ¹³C nuclear magnetic resonance spectroscopy enabled monitoring of the ¹³C label, starting at 122.70 ppm for 1u, followed by photorearrangement to 9 with a concomitant shift of the label to 145.01 ppm. If ozonolysis is conducted in the absence of pyridine, a complex mixture is formed, containing a new resonance at 119.04 ppm, which we assign to the intermediate secondary ozonide 11^48,49. However, in the presence of pyridine, this species is substantially depressed, and instead a single main peak at 163.36 ppm is observed, consistent with the formyl carbon of the imidic anhydride of 3a^50,51. This species can also be observed at 25 °C when the ozonolysis is conducted with ammonium carbamate present (as in the parent protocol), and subsequent heating of this mixture to 90 °C results in formation of the quinazoline 2a. The crude organic layer after thermolysis does not contain any significant ¹³C-labelled products; however, dissolution of the precipitate in D₂O/DMSO-d₆ allows the identification of both ammonium formate 5a (167.48 ppm) and formamide 5b (171.79 pm). On the basis of our control experiments, we suspect that 5b forms from transamidation of 5a on heating. As noted earlier, attempted isolation of 3 was invariably accompanied by hydrolysis to the 2-amidobenzaldehyde products 10. In control experiments (Fig. 4b), these species did not afford the corresponding quinazoline under the standard reaction conditions, including when extra formic acid or formamide were added to the reaction mixture. We note that addition of both formamide and formic acid does induce partial conversion of 10 to the quinazoline, but with lower conversion on the same timescales compared with the anhydride intermediate 3. Although this may represent a rescue mechanism for the hydrolysis products, given that these experiments use the maximal concentrations of the additives that can form under the reaction conditions, the relevance of such a pathway remains unclear^52,53. Collectively, these results underscore the key of the ‘sticky end’ intermediate 3—it not only activates C3 as a carboxylate leaving group, but also facilitates the condensation of ammonia by activating C2 for substitution as an imidic anhydride.

Fig. 4 |. Mechanistic experiments.

a, Tracing of a ¹³C label during the course of the reaction. b, Control experiments for assessing the reactivity of hydrolysed intermediates 10a and 10b.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41586-023-06613-4.

Data availability

All data are in the Supplementary Information or are available from the corresponding author upon reasonable request.