Introduction

The optimization of compound libraries remains a central task in drug-development campaigns. However, this process often suffers from challenges in terms of synthetic accessibility1,2. Ideally, late-stage interconversion of already-existing or easy-to-make libraries to other challenging candidates provides exceptional potential to accelerate drug discovery by avoiding cost- and labor-intensive de novo syntheses through nontraditional disconnections3. In this domain, skeletal editing has emerged as a promising strategy to precisely modify molecular skeletons at the atomic level without altering peripheral functional groups4,5,6,7,8,9. Many notable atom-insertion and -deletion reactions10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28, along with a handful of atom-swap reactions29,30,31,32,33,34,35,36,37, have been reported, providing improvements to the structural diversity and complexity of existing libraries, thereby enabling access to new biological space (Fig. 1a).

Fig. 1: Introduction.
Fig. 1: Introduction.The alternative text for this image may have been generated using AI.
Full size image

a Common skeletal editing paradigms. b Skeletal editing strategy in drug discovery. c Our working hypothesis to directly transform isoquinolines into naphthalenes via CN-to-CC transmutation. d This atom pair swap methodology induced by photocatalysis.

During the evaluation of drug candidates, a single-atom in the core skeleton can have a dramatic influence on its drug-target interactions, physicochemical properties and metabolic stability (Fig. 1b). For instance, Liu and co-workers discovered that replacing the nitrogen atom in a pyridine core with a carbon atom bearing a cyano group can efficiently modulate its hydrogen-bonding affinity with a receptor, reporting a 65-fold potency improvement in selective, orally active p38α MAP kinase inhibitors38. Obviously, conducting isosteric replacement skeletal editing (such as atom-swapping), without the need for multi-step, independent syntheses, is the most straightforward approach to preparing bioisosteres with minimal changes in the molecule’s topology, enabling rapid structure-activity relationship (SAR) analysis39. However, this is highly challenging due to the thermodynamic mismatches in atom-exchange processes, which require the precise cleavage and formation of multiple chemical bonds without disrupting peripheral substitutions.

Isoquinolines40 and naphthalenes41 are isosteres, both of which are of great significance in natural product chemistry, medicinal chemistry and material science. Conventional methods for their synthesis have relied on linear, iterative de novo synthesis. The presence of the nitrogen atom in isoquinolines endows the C-H groups around the core with different reactivities, thus making post-functionalization with precise regioselectivity control much easier. In sharp contrast, naphthalene is geometrically symmetric. Therefore, to obtain specific naphthalenes, a precursor with a correctly dispositioned functionality handle must be prepared in advance, much of which are difficult to access. Moreover, functionality sensitivity must be considered during the synthetic sequence. Accordingly, long- and/or low-yielding synthetic sequences are often required to obtain complex naphthalene derivatives.

The ubiquity and the importance of isoquinolines render them popular candidates for the development of CN-to-CC atomic-pair exchange reactions to directly prepare naphthalene derivatives (Fig. 1c). This offers the synthetic advantages that readily available precursors can be reshaped into synthetically challenging compounds for optimization libraries by obviating multi-step de novo syntheses. Accomplishment of this transformation necessitates orchestrating the sequential processes of breaking aromaticity, substituting from CN to CC, and restoring aromaticity. However, surgically performing and synchronizing these steps is a daunting task.

Herein, we report the successful implementation of such a CN-to-CC swap reaction. Upon visible-light irradiation, easily accessible isoquinolinium salts are directly transformed into the corresponding naphthalenes in the absence of any photosensitizer. Simultaneously, this strategy introduces a new target-binding site. This allows for rapid SAR analysis, providing the possibility to quickly recognize privileged compounds. The application of this highly functional group tolerable strategy is further demonstrated by late-stage functionalization of commercial drugs, thus expanding the biological space (Fig. 1d). Moreover, the desired products possess rich reactivity, allowing them to undergo various transformations to access useful polycyclic aromatic hydrocarbons (PAHs), ligands and drug analogies.

Results

Reaction development

Structurally, azaarenes have inbuilt π-conjugation systems, enabling them to serve as dienes to participate in Diels-Alder cycloadditions. However, unlike aromatics42,43, their moderate electron-deficiency makes them poor candidates, with only few intramolecular examples reported44. Generally, two activation modes are often adopted to drive the Diels-Alder reaction forward: one is transforming an azaarene into a more electron-deficient azaarene-based onium by quaternization, and the other is converting it into an electron-rich semi-saturated azaarene via a N/C-difunctionalization process (Fig. 2a). For instance, Kozmin and co-workers reported an inverse-electron-demand Diels-Alder reaction between quaternized isoquinoliniums and siloxy alkynes initiated by catalytic amount of Au or stoichiometric Ag45,46. Quite recently, the groups of Studer and Boswell independently pioneered polarity-inversion strategies to transform pyridines into 1,2-dihydropyridines, thus allowing for a normal (4 + 2) cycloaddition with arynes or electron-deficient alkynes47,48. Both strategies generate bridged intermediates, which are unstable and susceptible to rearomatization through the excision of a CN fragment from the core, thus achieving the CN-to-CC swap. Nevertheless, despite being highly efficient, they are constrained to polarity-matched two-electron processes, thus resulting in scope and product diversity issues.

Fig. 2: Design fundamentals and proof-of-concept studies.
Fig. 2: Design fundamentals and proof-of-concept studies.The alternative text for this image may have been generated using AI.
Full size image

a Previous reports for azaarene-to-arene conversion via polarity-matched cycloaddition process. b Our synthetic design: DFT calculations-directed photoredox radical chemistry to solve the mismatched polarity hurdle.

We reasoned that radical chemistry might overcome these electronic obstacles and facilitate thermally unattainable catalytic conversion. As shown in Fig. 2b, our synthetic design focused on initiating an electrophile-electrophile coupling through a polarity-inversion process under photoredox catalysis. Upon visible-light irradiation, isoquinolinium salts absorb photonic energy for excitation and incur charge-transfer from the anion to the isoquinolinium moiety. Then the electron-donor-acceptor (EDA) mechanism to reduce the isoquinolinium to a radical was proposed and calculated by density functional theory (DFT) method. Based on the screening studies on a series of bases through the UV/vis absorption spectroscopy and Stern-Volmer experiments, the isoquinolinium carbonate salt is demonstrated to be the effective photoactive complex (for details, see Supplementary Figs. S7–S14 in Supplementary Information). The TD-DFT calculations reveal this complex can be excited by irradiation at 462 nm. The electron-hole distributions in the first excited state (S1) clearly indicate charge transfer from the carbonate to the isoquinolinium through π-π interactions. Then the intersystem crossing process yields a triplet structure, releasing 19.4 kcal/mol of energy. The subsequent separation of the isoquinoline radical and carbonic radical anion (CO3-) liberates 7.0 kcal/mol of heat, mainly due to the effect of entropy increase. The spin density distributions in the isoquinoline radical show the largest population of the spin at the C(1)-atom. The reduced isoquinoline radical becomes electronically rich by the absorption of one electron, enabling cycloaddition with an electron-deficient alkyne through successive radical additions via transition states TS1 and TS2. As a result, (4 + 2) cycloaddition is accomplished, yielding the radical intermediate Int3. Regarding the regioselectivity, the computational results reveal higher energies for all corresponding isomers (TS1’, TS2’, Int2’ and Int3’) involved in the cycloaddition which is initiated by the isoquinoline radical adding to Cβ of the ethyl phenylpropiolate. Subsequently, the single electron transfer (SET) process occurs between intermediate Int3 and the carbonic radical anion (CO3-), affording the iminium cation intermediate Int4. Then the given CO32− is added back to Int4, yielding a much more stable intermediate Int5. The cascade SET and nucleophilic addition facilitates continuous energy release by totally 61.4 kcal/mol. Then the bridged structure is disrupted through the C–C bond cleavage via transition state TS3, with CO2 being released concurrently. The activation energy for this step is 23.8 kcal/mol, a reasonable value for the experimental conditions. Finally, the rearomatization proceeding via TS4 promotes release of the naphthalene product with simultaneous elimination of a N-benzylformamide anion. It is noteworthy that the N-benzylformamide was detected via GC-MS analysis (Supplementary Fig. S5 in supplementary information), offering experimental support for the rationality of the proposed mechanism. Therefore, the DFT calculations indicated that our photoredox radical strategy could be feasible. Although seemingly simple, unfavorable (4 + 2) cycloaddition between the phenyl rings in the isoquinoline core and alkynes should be avoided49. Moreover, all steps should be synchronized in a single reaction, genuinely enabling the direct late-stage functionalization of drugs.

To test the feasibility of our mechanistic hypothesis, we began our investigation with isoquinolinium bromide 1 and ethyl 3-phenylpropiolate 2 as model substrates (Table 1). This reaction proceeded successfully under blue-light irradiation (430 nm, 15 W) in EtOH with Na2CO3 as the base, and the desired editing product 3 was obtained in 75% yield after 36 h (entry 1). Scaling the reaction up to 4.0 mmol delivered the desired product 3 in 62% yield (entry 2). Performing the reaction in DMSO resulted in a slightly lower yield (64%, entry 3). Replacing Na2CO3 with organic bases, such as NEt3, DABCO and DBU, gave inferior results (entry 4). An evaluation of solvents indicated that EtOH was optimal (entries 1 vs. 5). Isoquinolinium chloride was also amenable, albeit with lower efficiency (53%, entry 6). The reaction remained active under the irradiation with wavelengths ranging from 340 to 530 nm, with 430 nm emerging as optimal (entries 1 vs. 7). Addition of 5 Å molecular sieves was beneficial to yield (entry 8). The experiments were repeated three times, affording 3 in virtually identical yield each time (entry 8). Complete inhibition was observed when performing the reaction in the dark or in air (entry 9). The removal of Na2CO3 resulted in a dramatic decrease in yield, highlighting the crucial role of the base (entry 10). The addition of radical scavengers, such as TEMPO or BHT, to the reaction system did not afford the desired product (entry 11), and adducts between isoquinolinium and TEMPO or BHT were detected by HRMS analysis, confirming the radical nature of the reaction (for details, see supplementary information). Adding anthracene, phenanthrene, pyridazine, styrene or 2,5-dimethyl-2,4-hexadiene to the reaction system had no obvious inhibitory effect on the yields (entries 12-16), thus ruling out an energy-transfer process.

Table 1 Optimization of reaction conditionsaView full size image
Full size table

We performed a series of control experiments to further explore the mechanism of this polarity-mismatched skeletal editing strategy. UV–Vis absorption analysis showed that significant bathochromic shift was observed when mixing 1 with Cs2CO3 and the absorption extends into visible light region, confirming that isoquinolinium-CO32− is the light-active species (Fig. 3a). A series of Stern-Volmer quenching experiments indicated that isoquinolinium bromide 1 is efficiently quenched in a manner linearly correlated with the concentration of carbonate anion, supporting the formation of an EDA complex from isoquinolinium and carbonate anion (Fig. 3b). In the reaction system, both the bromine anion and the CO32− can serve as the reductant to transfer one electron to the isoquinolinium core. To clarify their roles, we performed some control experiments (Fig. 3c). Replacing Na2CO3 with 2,4,6-collidine, a base with poor reducing capacity, resulted in a decrease in yield (41% vs. 75%). Similarly, reduced yields were observed when changing the bromine anion with a non-redox active anion, such as BF4− and CF3CO2− (32% vs. 75%, and 54% vs. 75%). However, addition of TBAB to the reaction mixture offered a substantial improvement in yield. Not surprisingly, the reaction was completely suppressed when the counterion was CF3CO2− and the base was 2,4,6-collidine, while a 15% yield was obtained upon adding TBAB to the reaction system. These results clearly confirm that both the bromine anion and Na2CO3 behave as reducing agents to synergistically promote the skeletal-editing reaction. The light-on/off experiments indicated that consecutive irradiation is a prerequisite for product formation (Fig. 3D).

Fig. 3: Mechanistic investigations.
Fig. 3: Mechanistic investigations.The alternative text for this image may have been generated using AI.
Full size image

a UV–Vis absorption spectroscope. b Stern-Volmer quenching experiments to determine the interaction between 1 and CO32-. c Control experiments to evaluation the effect of the counterions and the bases. d “Light-on/off” experiments.

Reaction scope

With the optimized conditions in hand, the substrate scope of this photoinduced polarity-mismatched skeletal-editing strategy was explored. Initially, we investigated the effect of substituents on the aromatic rings of alkynes. As shown in Fig. 4a, a wide range of substituents at varied positions and with different electronic characters were found to be compatible in this transformation, enabling the synthesis of 4–23 in 46–97% yields. Among them, the iodo substituent remained untouched under visible-light irradiation (10 and 19). The phenyl substituent could be extended to fused aromatic rings (24 and 25) and heteroaromatic rings (26, 27, and 28). A di-alkynoate was also a competent substrate with the two C-C triple bonds participating in the reaction successfully to construct a more complex molecule (29), which have been difficult to access by conventional methods. In examining the scope of the electron-withdrawing groups on the alkynes (Fig. 4b), we found that the additional unsaturated bonds on the ester groups did not interfere with the skeletal-editing processes, leaving these reactive functional groups untouched (30, 31, 32, and 34). Furthermore, a cyano group was also tolerant (33). The broad compatibility of this method prompted us to investigate its application to complex alkynoates with bioactive natural products and drugs, and the results indicated that they show good reactivity with excellent functional group tolerance and complete stereochemical retention (34–39).

Fig. 4: Substrate compatibility toward various alkynes.
Fig. 4: Substrate compatibility toward various alkynes.The alternative text for this image may have been generated using AI.
Full size image

a Scope of aryl groups. b Sope of EWGs. Reaction conditions: 1 (0.4 mmol), alkynes (0.2 mmol), Na2CO3 (0.2 mmol), EtOH (1.5 mL), argon, blue LEDs (430 nm, 15 W), 35 °C. Asterisk indicates 100 mg of 5 Å molecular sieves added. Isolated yields obtained by silica gel column chromatography.

Subsequently, synthetic tolerance toward a wide range of isoquinolinium salts was studied (Fig. 5a). In general, easily accessible isoquinolinium salts, irrespective of their substitution positions and electronic natures, displayed good reactivity and successfully participated in this reaction, providing an efficient entry to otherwise difficult-to-access multi-substituted naphthalenes (40–52). Indole substitution was also well-accommodated (45 and 51). For a 5-methoxy substituted isoquinolinium salt, the reaction occurred regioselectively at the azaheterocycle rather than the benzene ring, regardless of the steric repulsion between the methoxy group and the incoming phenyl group during the cycloaddition step, to produce 52 in 58% yield. A 4-cyano substituted isoquinolinium was also applicable, but with distinctly site-selective CN deletion (53) in which the C1 carbon atom was cut off. Norharman, a neuroactive β-carboline alkaloid natural product50, was directly transformed into the synthetically useful 9H-carbazole (54) through a simple CN-to-CC swap. A broad scope of substituents on the excised C3 carbon atom was also observed. Isoquinoliniums bearing sterically demanding phenyl and methyl groups could all be used (3, 46, and 49).

Fig. 5: Substrate scope with respect to various isoquinoliniums.
Fig. 5: Substrate scope with respect to various isoquinoliniums.The alternative text for this image may have been generated using AI.
Full size image

a Scope of isoquinoliniums. b Editing of drugs. Reaction conditions: isoquinoliniums (0.4 mmol), alkynes (0.2 mmol), Na2CO3 (0.2 mmol), EtOH (1.5 mL), argon, blue LEDs (430 nm, 15 W), 35 °C. Asterisk indicates 100 mg of 5 Å molecular sieves added. Isolated yields obtained by silica gel column chromatography.

To further underscore the applicability of this approach, we demonstrated its capacity for late-stage functionalization of complex medicinal molecules (Fig. 5b). SGC707 is a potent, selective and cell-active PRMT3 allosteric inhibitor51. Using our method facilitated rapid conversion to its naphthalene analog (55). Fasudil, a clinical drug for the treatment of cerebral vasospasm52, was also a good candidate to test the capacity of our method, smoothly affording 56 in 30% yield. These investigations clearly showcase the potential application of this CN-to-CC transmutation strategy, in which a single operation from readily available materials could provide a robust entry to products that are otherwise difficult to accessible, in drug discovery and development.

Synthetic applications

The ester groups in our products not only provide a target-binding site, they also offer valuable transformable functionality, undergoing various chemical conversions. By harnessing their rich reactivity, we prepared a lot of synthetically valuable PAHs and ligands (Fig. 6a). For instance, 3 could undergo a Friedel-Crafts acylation reaction with the aid of H2SO4 to afford useful benzofluorenone 57 in 79% yield. Condensation of the ester in 27 with (S)−2-amino-2-phenylethan-1-ol followed by intramolecular cyclization provided an approach to chiral oxazoline ligand 58. The ester groups could be effectively transformed into aldehydes through a sequence of DIBAL-H reduction and PCC oxidation. The fruitful reactivity of aldehydes further extended the application of this skeletal-editing strategy. For example, a sequential process comprising a Wittig reaction and photoinduced carbocyclization afforded tetraphene 59. Moreover, its analogue benzo[j]phenanthridine 60 could be attained by replacing the ylide reagent with hydroxylamine. Condensation with (S)-N-phenylpyrrolidine-2-carboxamide yielded 61 in 98% yield. Ligand 62 could also be efficiently prepared through a sequence of aldehyde-to-cyano transformation and ensuing condensation with L-valinol. Notably, the resulting 2-naphthaldehydes could undergo a classical Dakin reaction to be converted into synthetically useful 2-naphthols, which underwent self-coupling or cross-coupling with isoquinolines to prepare BINOLs (63–67) and QUINOLs (68 and 69), respectively. In view of the synthetic capacity of this method to construct various differently substituted naphthols, our approach can largely enrich the structures of resulting BINOLs and QUINOLs53. The synthetic utility of this protocol was further explored to prepare an adapalene analog. As shown in Fig. 6b, the isoquinoline precursor could be easily synthesized via a Pd-catalyzed Suzuki coupling from commercially available materials. Then, by employing our skeletal editing strategy, the desired adapalene analogue 70 was afforded.

Fig. 6: Synthetic applications.
Fig. 6: Synthetic applications.The alternative text for this image may have been generated using AI.
Full size image

a Synthetic transformations to access various PAHs and liangd. b The use of our skeletal editing strategy to prepare adapalene analog. Isolated yields obtained by silica gel column chromatography.

Discussion

The common-sense paradigm that electron-deficient isoquinolines, without cumbersome activation, will not serve as efficient dienes to undergo (4 + 2) cycloaddition with electron-deficient dienophiles may have been leading chemists to ignore important reactions. We have herein developed a photoinduced photosensitizer-free skeletal-editing strategy to transform easily accessible isoquinolines into synthetically challenging naphthalene derivatives. Directed by DFT calculations, the (4 + 2) cycloaddition between the two polarity-mismatched species was accomplished successfully. The pivotal factor to this success was utilizing the isoquinolinium carbonate as an EDA complex to harness photonic energy and achieve polarity inversion for the isoquinolinium cation. This strategy featured simple conditions, wide applications in late-stage functionalization, and a robust synthetic capacity to access valuable PAHs, various ligands and drug analogues. We believe that this study will have an important impact on developing more skeletal-editing strategies to shorten drug-discovery campaigns.

Methods

General procedure for the photoinduced skeletal editing of isoquinoliniums into naphthalenes

To an oven-dried 10 mL vial sealed with rubber septa charged with a magnetic stir bar was added isoquinolinium salts (0.40 mmol, 2.0 equiv.), alkynes (0.20 mmol, 1.0 equiv.) and Na2CO3 (0.20 mmol, 1.0 equiv.) at room temperature. Then, EtOH (1.5 mL) was added. Then, the reaction mixture was degasified using argon for 15 min under constant purging and monitoring. The reaction mixture was stirred at 35 °C under irradiation at 430 nm LEDs (15 W) for the specified time. After completion (monitored by TLC), the crude mixture was concentrated under reduced pressure, then purified using silica gel column chromatography (silica gel 200–300 mesh size, using petroleum ether/ethyl acetate as eluent). (Note: for the formation of products 21, 24, 27, 28, 33, 43, 49–54 and 56, 100 mg of 5 Å MS were added.).

Experimental procedures

Full details on all experimental procedures can be found in the Supplementary Information.