Divergent photochemical ring-replacement of isoxazoles

Abstract

Isoxazoles, oxazoles, and other five-membered heteroaromatics are prevalent motifs in core structure of pharmaceuticals and agrochemicals. In early-stage drug discovery, it is common practice to prepare libraries of analogues featuring different heterocyclic cores and this generally requires a de novo synthesis for each scaffold. A valuable but currently unavailable strategy would involve the possibility for direct heterocycle “ring-replacement”. Here we report a photochemical platform for the selective conversion of isoxazoles into oxazoles, pyrazoles, pyrroles, and isothiazoles by exploiting excited-state reactivity. Starting from a successful isoxazole-to-oxazole transformation, we uncover position-sensitive reactivity that prompted computational investigation. These insights guide a systematic reactivity survey and reveal a solvent-controlled deconstruction–reconstruction pathway via α-ketonitrile intermediates. This approach enables scaffold diversification without de novo synthesis, affording access to five distinct azole classes under mild conditions. The method’s selectivity, functional group tolerance, and late-stage applicability suggest broad utility in heterocyclic library design for pharmaceutical research.

Isoxazoles, oxazoles, and other five-membered heteroaromatics are prevalent motifs in core structure of pharmaceuticals and agrochemicals. Here the authors report a photochemical platform for the selective conversion of isoxazoles into oxazoles, pyrazoles, pyrroles, and isothiazoles by exploiting excited-state reactivity.

Introduction

The ability to modify complex molecules at late stages, without the need for de novo synthesis, has become a central objective in modern synthetic chemistry¹. Considering (hetero)aromatic molecules, significant advances have been made in the way we perform peripheral modifications (e.g., cross-coupling and C–H activation)^2,3. However, direct replacement or remodeling of the core (hetero)aromatic scaffold remains a largely unsolved challenge, despite its potential to accelerate chemical space exploration in pharmaceuticals, agrochemicals, and functional materials^4,5.

Among five-membered heteroaromatics, isoxazoles and oxazoles occupy a privileged position, frequently appearing in marketed drugs and advanced materials (Fig. 1a)^6,7. These heterocycles can serve as bioisosteres for ketone, ester, and (hetero)aryl groups, often offering benefits such as improved metabolic stability, reduced lipophilicity, and enhanced potency^8,9. Their privileged status makes them invaluable tools in modern drug discovery and development.

Fig. 1. Relevance of oxazoles and other azoles in medicinal chemistry.

a Relevance of isoxazoles and oxazoles in pharmaceuticals. b Ring replacements in SAR studies towards the identification of a SETD2 inhibitor 1. c This work uses photochemistry to convert isoxazoles into other heteroaromatics.

A classic approach in early structure-activity relationship (SAR) studies involves the preparation of substrate libraries in which a specific heterocyclic system is replaced by closely related bioisosteres^10–12. A representative case is the development of the isoxazole-based SETD2 inhibitor 1, where oxazole, imidazole, and pyrazole analogs were individually synthesized and evaluated (Fig. 1b)^13,14. While effective for lead identification, such strategies typically necessitate the de novo synthesis of each variant, which becomes especially challenging when the heterocycle is embedded within the molecular core^15,16. Consequently, a late-stage method for heterocycle “ring-replacement” would offer significant value by enabling rapid access to bioisosteric analogs from a common lead derivative^17,18.

Isoxazoles offer a compelling platform for this strategy. Early reports demonstrated their capacity to rearrange photochemically to oxazoles and α-ketonitriles, but the reactions were typically unselective, low-yielding, and limited to a narrow set of derivatives^19–30. Due to these challenges, the photochemistry of isoxazoles has remained unexplored, and no general framework exists to harness their reactivity in a selective or synthetically useful way.

Here we report a photochemical platform for the selective conversion of isoxazoles into other azoles, including oxazoles, pyrazoles, pyrroles, and isothiazoles (Fig. 1c). Beginning with the efficient isoxazole-to-oxazole rearrangement of 5-phenyl isoxazole (2a), we observed unexpectedly different behavior across closely related analogs. This prompted a mechanistic investigation using DFT, which revealed how small changes in the isoxazole substitution pattern drastically influence excited-state reactivity. Leveraging these insights, we constructed a comprehensive reactivity map encompassing 24 substituted isoxazoles and identified structural features that govern productive rearrangement. Beyond rearrangement, we discovered that solvent modulation could redirect excited-state reactivity away from C–C bond fragmentation, yielding α-ketonitrile intermediates that serve as branch points for further heterocycle formation. Using this principle, we developed a divergent deconstruction–reconstruction platform capable of generating five distinct heterocyclic families from readily available isoxazole precursors³¹.

Results and discussion

In developing a strategy for the conversion of isoxazoles into other heteroaromatics, we drew inspiration from the early photochemical experiments from Ullman and Singh^26,27. These pioneering studies demonstrated that simple photoexcited isoxazoles can undergo a series of structural rearrangements leading to either azirine intermediates or α-ketonitriles³². However, pioneering studies from Pavlik et al. demonstrated its synthetic potential in isoxazole-to-oxazole isomerization, although their conditions often led to complex reaction outcomes, low chemical yields, and mixtures of isomeric products^{19–25,28–30}. More recently, Baumann has translated the isoxazole-to-oxazole conversion under photo-flow settings using high-energy Hg-lamps, but this reactivity was applied to a small subset of derivatives that do not capture the full chemical space options around the isoxazole core³³. As a result, the synthetic scope and selectivity of isoxazole photochemistry remain essentially unknown.

We initiated our study with 5-phenyl isoxazole (2a) and explored its conversion into either the corresponding oxazole (2b) or α-ketonitrile (2c), depending on the reaction conditions (Fig. 2a). Under irradiation at λ = 310 nm, we observed some remarkable differences in the reaction outcome on the basis of the solvent and additive used. Specifically, irradiation in MeOH led to the selective formation of 2b in good yield (entry 1). Interestingly, other polar protic solvents lead to considerably lower conversions and significant decomposition (entries 2 and 3). We then evaluated a few additives, and while both standard acid and base did not improve the reaction efficiency (entries 4 and 5), the use of 20 mol% of 2,6-lutidine delivered 2b in 85% yield (entry 7)³⁴. These conditions could be scaled up in a single batch condition (8 mmol) using the same irradiating apparatus to provide 2b in 80% yield (see the Supplementary Information Section 7 for more details). The use of non-polar solvents (e.g., trifluorotoluene and EtOAc, entries 8 and 9) had a dramatic switch in the reaction, now providing nitrile 2c as the major product. Pleasingly, the use of DCE completely suppressed the formation of 2b and provided access to 2c in 72% yield.

Fig. 2. Isoxazole-to-oxazole conversion.

a Development of isoxazole-to-oxazole and isoxazole-to-α-ketonitrile photochemical conversions using 1a. b Extension of photochemical reactivity to 2a and 3a. c Mechanistic analysis of isoxazole photochemistry. ¹H NMR yields using 1,3-dinitrobenzene as an internal standard are reported.

Encouraged by these results, we explored how changes to the position of the Ph substituent on the isoxazole ring might influence reactivity (Fig. 2b). In general, all isoxazoles discussed below have been subjected to six photochemical conditions (A–F), and the results reported represent the highest yielding ones. Strikingly, the reactivity of 2a was not preserved on its constitutional isomers. Indeed, the 4-phenyl isoxazole (3a) did not undergo any reactivity when irradiated (λ = 310 nm) in the presence of 2,6-lutidine as the additive and MeOH as the solvent. However, partial conversion was achieved with more energetic irradiation (λ = 254 nm), yielding oxazole 3b in modest yield (23%, conditions E). Furthermore, the 3-phenyl analog (4a) proved photochemically inert under all tested conditions, even under prolonged irradiation.

This sharp change in photochemical reactivity is difficult to rationalize based on the limited knowledge in the area, and therefore, we have run full reaction analysis by DFT. The proposed mechanism leading to the isoxazole-to-oxazole (A-to-E) isomerization or isoxazole-to-α-ketonitrile (A-to-G) ring-opening has generally been approached considering that irradiation populates the isoxazole π,π* singlet excited state (S₁-A) (step i), subsequently evolving to a vinylnitrene (B) via fast N–O bond homolysis (step ii) (Fig. 2c)^35,36. This high-energy intermediate can undergo either a ring-contraction by 2π-electrocycliczation to azirine C (step iii)^26,32,37 or, if R = H, a 1,2-H shift to ketenimine F (step vi)³⁸. Previous work has suggested that azirine C can undergo further photoexcitation to deliver upon isomerization and C–C fragmentation (step iv), the nitrile ylide D^38–40. At this point, ionic cyclization (step v) can complete the isoxazole-to-oxazole isomerization. Regarding the formation of α-ketonitriles G, this is most likely arising from a 1,3-H shift of F (step vii). The strong difference in reaction performance depending on the position of the Ph-group across the isoxazole core indicates a more complex reaction manifold than previously appreciated.

To understand the differences in photochemical reactivity observed among these three isoxazoles, we carried out computational studies on compounds 2a, 3a and 4a (Fig. 3). Upon photoexcitation to the singlets (S₁), all three species undergo N–O bond cleavage via rapid population of a πσ* state³⁵, independent of the position of the Ph substituent on the isoxazole core. Following internal conversion to the ground state (S₀), a singlet biradical is generated (L2–4). Previous analysis on this reactivity invoked the formation of a vinylnitrene B (see above). While B and L can be considered as resonance structures, our calculations suggest L as the best way of representing this intermediate on the basis of their spin density. From this intermediate, regeneration of the isoxazole or formation of the azirine can occur. For L4 (blue line), the pathway leading to isoxazole 4a regeneration is favored, with a significantly lower activation barrier (6.1 kcal mol⁻¹, via TS1) compared to that for azirine C3 formation (18.4 kcal mol⁻¹, via TS2). In contrast, for both L3 (yellow line) and L2 (red line) derivatives, azirine C3 and C2 formation is preferred, as the activation barriers for C–N bond development (C3: 2.0 and C2: 3.5 kcal mol⁻¹, respectively) are lower than those for isoxazole reformation (3a: 7.2 and 2a: 6.0 kcal mol⁻¹). The conversion of the azirine intermediates C2 and C3 to the corresponding oxazoles 2b and 3b via a thermal process is associated with prohibitively high activation barriers (TS3), supporting the hypothesis that this transformation is also photochemically driven. Indeed, our calculations demonstrate that photoexcitation of C2 and C3 to the singlet states (S₁) enables a facile conversion into the nitrile ylides N2 and N3 that can evolve into 2b and 3b via low-barrier processes. To account for the experimental difference in reactivity efficiency between the two starting isoxazoles 2a and 3a, we speculated that differences in the absorption profile of their corresponding azirines might impact the overall photochemical behavior. Notably, C3 exhibits minimal absorption in the λ = 200–300 nm range, while C2 shows strong absorption and should readily undergo photoexcitation. Overall, these computational findings might be used to rationalize the experimental outcomes: 4a (complete recovery of starting material) undergoes N–O cleavage upon photoexcitation, but the high barrier to azirine formation favors reversion to the starting isoxazole. 3a (~20% product, ~20% starting material, ~60% decomposition) and 2a (85% yield) undergo N–O cleavage, with azirine formation being slightly favored. However, due to poor UV absorption, the azirine coming from 3a, C3, cannot efficiently re-enter the photochemical manifold, leading to competing thermal degradation pathways. In contrast, the azirine intermediate coming from 2a, C2, strongly absorbs in the UV range, enabling a second photoexcitation that drives conversion to the oxazole product. The solvent-controlled conversion of 2a into either 2b (MeOH) or 2c (DCE) is more difficult to rationalize. We currently propose that the divergence arises from differential stabilization of reactive nitrile ylide intermediates, controlled by solvent polarity and proticity. A plausible explanation for this pronounced solvent effect involves the differential stabilization of the nitrile ylide intermediate. In methanol (MeOH), a polar protic solvent, the nitrile ylide is likely significantly stabilized, shifting the equilibrium toward its formation. This, in turn, facilitates an intramolecular cyclization, ultimately affording the oxazole product. In contrast, 1,2-dichloroethane (DCE), being an apolar aprotic solvent, likely disfavors the formation of the nitrile ylide and instead promotes two consecutive [1,n]-hydrogen shifts, resulting in the formation of the α-ketonitrile product. Overall, these results suggest that the observed experimental selectivity arises not only from the initial photoactivation but also from downstream differences in the photophysical behavior of high-energy intermediates.

Fig. 3. Computational studies.

a Formation of oxazoles 2b–3b from isoxazoles 2a–4a (Gibbs free energies in kcal mol⁻¹ are given related to the corresponding singlet biradical intermediate Ln). b Calculated spin densities for intermediates L2–L4. c Calculated absorption profiles for intermediates C2 and C3. Computational method: (TD-)CAM-B3LYP/cc-pVTZ/SMD(MeOH)//CAM-B3LYP/cc-pVDZ/SMD(MeOH).

Given the strong dependence of reactivity on changes in substitution, we recognized the need for a comprehensive and systematic evaluation of isoxazole substitution patterns (Fig. 4). We sought to explore relevant mono- and disubstituted isoxazoles across a diverse range of aryl (Ph), alkyl (Me), and electron-withdrawing (CO₂Me) groups. We therefore selected a set of 24 isoxazole derivatives categorized into five structural classes: (1) Ph-monosubstituted (2a–4a); (2) ester-monosubstituted (5a–7a); (3) Ph,Me-disubstituted (8a–13a); (4) ester,Me-disubstituted (14a–19a); and (5) Ph,CO₂Me-disubstituted (20a–25a). Each derivative was subjected to six photochemical conditions (A–E) varying solvent (MeOH vs 1,2-DCE), additive (with or without 2,6-lutidine), and light source (λ = 254 vs 310 nm). In Fig. 4, only the result leading to the highest yield/lowest decomposition was reported for each substrate; the remaining results can be found in the Supplementary Information Section 4.

Fig. 4. Scope analysis of differentially substituted isoxazoles.

E = CO₂Me/CO₂Et. Isolated yields are reported (see Supplementary Information Section 7 for details on recovered starting material for each scope entry).

Our systematic analysis revealed a highly sensitive reactivity profile dictated by the nature and position of substituents on the isoxazole ring. In many cases, small structural changes dramatically altered photochemical outcomes, ranging from clean isomerization to complete decomposition or photostability, highlighting our inherent low understanding of isoxazole photochemistry. We have performed detailed ¹H NMR spectroscopy analyses of the crude mixtures for representative low-yielding examples. These spectra reveal that aside from the desired oxazoles, the major components typically consist of decomposition products manifesting as broad, unresolved signals in the aromatic and aliphatic regions, consistent with polymeric or oligomeric byproducts. Unfortunately, these byproducts could not be cleanly isolated or fully characterized due to their complexity and instability. In some cases, minor amounts of starting isoxazole were also detected.

For instance, within the Ph-substituted series, only the 5-Ph isomer (2a) underwent efficient and selective rearrangement. The 3-Ph (4a) was photostable, while 4-Ph (3a) gave a low yield of the corresponding oxazole. For ester-substituted isoxazoles, the 3- and 4-substituted variants (5a and 6a) were unstable and decomposed across all tested conditions, precluding rearrangement. In contrast, the 5-substituted isomer (7a) cleanly afforded oxazole 7b in 38% yield, underscoring the privileged reactivity of the 5-position. These findings support the hypothesis that productive rearrangement correlates with the favorability of azirine formation and subsequent re-excitability, as seen in our computational studies.

In the Ph,Me-disubstituted series, most isomers were accessible and evaluated, with the exception of 11a, whose synthesis remains elusive, highlighting the synthetic limitations even for apparently simple heterocycles. Here, introducing a Me group at C4 (8a) enabled reactivity from an otherwise inert 3-Ph core (see the results for 4a), forming 8b in moderate yield. However, the 5-Me (9a) and 3-Me (10a) analogs decomposed entirely. Notably, 12a and 13a underwent efficient rearrangement to 12b and 13b, respectively, with wavelength sensitivity influenced by the methyl group’s position (310 nm for 12a vs 254 nm for 13a). The ester,Me-disubstituted isoxazoles, posed greater challenges. Several substrates (14a, 16a, and 19a) decomposed under irradiation, while 15a, 17a, and 18a yielded the corresponding oxazoles with variable efficiency. In the case of 18a, careful reaction timing (2 h irradiation time) was key to maximizing yield.

Finally, in the Ph,CO₂Me-disubstituted series, both 3-Ph derivatives (20a and 21a) successfully underwent rearrangement to give oxazoles 20b and 21b in moderate yields. Although 22a and 23a were synthetically inaccessible, the 5-Ph analogs 24a and 25a rearranged smoothly to 24b and 25b in good yields.

Taken together, this survey identified 5-substituted isoxazoles bearing aryl or electron-withdrawing groups as privileged substrates for selective ring-replacement, while reactivity at other positions proved highly variable and substrate-specific. We therefore focused subsequent synthetic efforts on this class. In terms of synthetic potential, it is interesting to note that this photochemical strategy uses inexpensive and easy-to-make isoxazoles to access 2,5-disubstituted oxazoles (12b, 15b, and 24b) that in turn are generally difficult to make or very expensive (see the Supplementary Information Section 9 for more details).

We next evaluated the generality of the transformation using a broad range of 5-aryl isoxazoles, which our initial screen identified as robust substrates (Fig. 5a). The reaction conditions (typically λ = 310 nm, MeOH, 2,6-lutidine) were broadly applicable, with minor tuning required in isolated cases. Overall, the transformation proved tolerant to diverse functional groups.

Fig. 5. Photochemical isomerization of isoxazoles to oxazoles.

a Scope of the process. b Concomitant isoxazole-to-oxazole conversion and 6π electrocyclization. c Different photochemical behavior of 12a. Isolated yields are reported (see Supplementary Information Section 7 for details on recovered starting material for each scope entry). Reaction conditions: ^A 1,2-DCE, r.t., 24–48 h, λ = 254 nm; ^B 2,6-lutidine, 1,2-DCE, r.t., 24–48 h, λ = 254 nm; ^C 1,2-DCE, r.t., 24–48 h, λ = 310 nm; ^D MeOH, r.t., 24–48 h, λ = 254 nm; ^E 2,6-lutidine, MeOH, r.t., 24–48 h, λ = 254 nm; ^F 2,6-lutidine, MeOH, r.t., 24–48 h, λ = 310 nm.

The method tolerated a wide range of aryl substituents, including electron-donating (Me, OMe, thioether, and NH₂; see 26a–29a) and electron-withdrawing (halides, CF₃, CN, and B(pin); see 30a–34a) groups. Remarkably, sensitive functionalities such as vinyl and C(sp²)–Br groups (see 31b and 35b) remained intact, highlighting the method’s chemoselectivity.

Meta- and ortho-substituted arenes (36a, 37a) as well as polyfunctional systems (38a–40a) also performed well, demonstrating broad positional tolerance. While the reaction for 41a was initially low-yielding under standard conditions, extending the reaction time to 72 h at λ = 254 nm improved the yield to 59%. Importantly, heteroaromatic groups were compatible, including benzoxazole (42a), benzothiazole (43a), thiophene (44a), and furan (45a) derivatives, which underwent clean isomerization with no damage to the auxiliary heterocycles.

We also examined polysubstituted isoxazoles (46a–52a), which required individual optimization. Trisubstituted systems featuring combinations of aryl and Me groups rearranged selectively to their oxazoles, provided that substitution patterns aligned with those favoring productive excited-state behavior. The high yield observed for 49b (from 3-Ph,5-Me isoxazole) contrasts with the decomposition seen for simpler analogs (9a, 20a), again underlining the complex interplay between sterics, electronics, and photophysical properties.

A particularly impactful application was the two-step synthesis of oxazole 50b, a derivative of a friulimicin B lipopeptide, previously accessed via an eight-step route⁴¹. Our method reduced this to a simple esterification followed by photochemical rearrangement, showcasing the synthetic streamlining potential of this approach.

Finally, we demonstrated late-stage applicability by converting two pharmaceutically relevant isoxazoles—parecoxib (51a) and a Hepatitis C antiviral (52a)—into their oxazole analogs (51b, 52b). These transformations illustrate the direct replacement of heterocyclic cores without de novo synthesis, which is otherwise required for SAR studies.

Interestingly, certain substrates exhibited divergent or tandem reactivity, opening new avenues for discovery. Isoxazoles bearing two aryl groups at C4 and C5 (53a–55a) underwent both oxazole formation and 6π-electrocyclization, yielding fused polycyclic products with potential applications in organic electronics (Fig. 5b). The selectivity and modularity of this cascade merit further exploration in materials-oriented contexts^42,43.

Even more striking was the behavior of 12a, which could be selectively diverted to three different products depending on conditions (Fig. 5c). While conditions B (DCE, 2,6-lutidine, λ = 254 nm) gave oxazole 12b, reactions in MeOH yielded 12c via trapping of a ketenimine intermediate⁴⁴, and lower-energy irradiation in DCE afforded the azirine 12 d. This conditional divergence from a single scaffold underscores the unique power of photochemical control and further validates the need for detailed reactivity mapping.

As discussed above, while investigating the isoxazole-to-oxazole rearrangement, we discovered that solvent tuning can redirect the reaction pathway toward C–C bond cleavage, yielding α-ketonitrile intermediates. Notably, α-ketonitriles are valuable intermediates in condensation chemistry and can also be accessed by conventional routes. However, their direct formation from isoxazoles offers a streamlined and synthetically powerful “ring-replacement” opportunity to directly convert them into pyrroles (H), pyrazoles (I), and isothiazoles (K).

It is important to note that, at present, the photochemical isoxazole-to-α-ketonitrile conversion is only accessible using 5-Ar-substituted derivatives. The inclusion of other substituents on the isoxazole core leads mostly to the formation of oxazoles, photostability, or photodecomposition (see Figs. 3 and 4).

Recognizing the synthetic potential of α-ketonitriles, we designed a modular platform for isoxazole-based scaffold remodeling, converting these intermediates into five heterocyclic frameworks via one-pot transformations (Fig. 6a, b). Specifically, in order to demonstrate the applicability of this approach in generating a large library of heteroaromatic derivatives, we identified conditions for converting 5-aryl-isoxazoles into five distinct heterocyclic systems: amino-pyrazoles, pyrazoles, pyrroles, amino-isoxazoles, and isothiazoles. This methodology was applied to a selection of seven commercially available starting materials (2a, 27a, 29a, 35a, 37a, 45a, and the 4-Cl-Ph not represented in Fig. 6), delivering a library of 34 new derivatives without resorting to de novo individual syntheses.

Fig. 6. Direct and divergent isoxazole “ring-replacement”.

a Photochemical deconstruction–reconstruction strategy for divergent “ring-replacement of isoxazoles”. b Development of a library on isoxazole “ring-replacement. Reaction conditions: hν = 1,2-DCE, r.t., 24–48 h, λ = 310 nm. (a) PhSO₂NHNH₂, NBS, EtOH, 80 °C. (b) DIBAL-H, Et₃N, THF, r.t. then NH₂NH₂•H₂O, i-PrOH, 70 °C. (c) TsCl, Et₃N, CH₂Cl₂, r.t., then diethyl aminomalonate, EtONa, EtOH:THF, r.t. (d) NH₂OH•HCl, MeOH, rt. (e) Mukaiyama reagent, Et₃N, CH₂Cl₂, r.t. then Na₂S, ClNH₂, EtOH:H₂O, 70 °C. Isolated yields are reported (see Supplementary Information Section 7).

Each substrate was first subjected to the optimized photochemical conditions (DCE, λ = 310 nm), generating the corresponding α-ketonitriles, which were immediately used in subsequent condensation reactions.

For instance, treatment with benzenesulfonyl hydrazide in EtOH followed by heating under reflux enabled isoxazole-to-3- amino-pyrazole “ring-replacement” (56a–56g). The 5-amino-pyrazole motif is highly versatile, with widespread applications in drug discovery, particularly in anticancer, antibacterial, antimalarial, and anti-inflammatory agents^7,45.

Alternatively, DIBAL-H reduction of the α-ketonitriles, followed by condensation with H₂N–NH₂•H₂O, converted 5-aryl-isoxazoles into 3-aryl-pyrazoles 57a–57g⁴⁶. This transformation represents a formal oxygen-to-nitrogen transmutation, and while related strategies are known, they typically require nickel catalysts and high-pressure hydrogenation, making our method a milder alternative^47–49.

Further heterocyclic diversity was achieved by treating the crude α-ketonitriles with tosyl chloride and then diethyl aminomalonate to form 3-amino-pyrroles (58a–58g)⁵⁰.

Additionally, refluxing the crude α-ketonitriles with HO–NH₂•H₂O in EtOH led to the conversion of 5-aryl-isoxazoles into 3-aryl-5-amino-isoxazoles (59a–59g)⁵¹. This transformation effectively moves the aromatic substituent from C5 to C3 while introducing an amino group at C5.

Lastly, the isoxazole-to-isothiazole conversion was realized as a formal O-to-S atom transmutation. This was achieved by converting the α-ketonitriles into an alkyne using the Mukaiyama reagent, followed by reaction with sodium sulfide and chloramine-T, furnishing the desired isothiazoles (60a–60g) in moderate yields⁵².

In summary, we reported a photochemical strategy that transforms isoxazoles into structurally distinct heterocycles, including oxazoles, pyrazoles, pyrroles, and isothiazoles. This platform exploits both ring-contraction and deconstruction–reconstruction pathways, enabled by selective control of excited-state reactivity. The approach provides a rare example of scaffold editing for aromatic heterocycles and offers a practical route to heterocycle diversification without de novo synthesis. Its operational simplicity, broad functional group tolerance, and successful application to drug-like scaffolds suggest strong potential for adoption in medicinal chemistry.

Methods

General procedure for the permutation of isoxazole

A dry tube equipped with a stirring bar was charged with the corresponding isoxazole (1.0 equiv.). The tube was capped with a Supelco aluminum crimp seal with septum (PTFE/butyl), evacuated and refilled with N₂ (×3). The corresponding anhydrous and degassed solvent and the corresponding additive (0.2 equiv.) were added. The tube was placed into a Helios photoreactor. The photoreactor and a fan were switched on, and the mixture was stirred under irradiation for the specified time. The solvent was evaporated, and the residue was purified by column chromatography on silica gel to give the desired product. All modifications regarding the solvent and/or additive used are detailed in the Supplementary Information in Section 2, page S6.

General procedure for the preparation of α-ketonitrile

A dry tube equipped with a stirring bar was charged with the corresponding isoxazole (0.2 mmol, 1.0 equiv.). The tube was capped with a Supelco aluminum crimp seal with septum (PTFE/butyl), evacuated and refilled with N₂ (×3). Degassed DCE (4 ml) was added. The tube was placed into a Helios photoreactor equipped with 310 nm lamps, and the fan was switched on. The mixture was stirred under irradiation for 24 h. Upon completion, the solvent was evaporated under reduced pressure, and the crude product was used without any further purification.

Author contributions

D.L., A.R., and B.R. designed the project. Y.X., L.P., and B.R. run all the synthetic experiments; E.M.A. run all the computational studies. All authors discussed the results and wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks Xiangyang Chen and Mohamed Salem, who co-reviewed with Rubal Sharma, and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

Open Access funding enabled and organized by Projekt DEAL.

Data availability

All data is available in the main text or the Supplementary Information. All data are available from the corresponding author upon request. The source data are provided with the article. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68960-w.