Introduction

Bicyclic compounds featuring saturated N-heterocycles fused to (hetero)aromatic units are highly prized in medicinal chemistry1,2,3,4, offering a combination of polar functionality, high Fsp3-content, and rigidly disposed groups on the aromatic core that can engage in key interactions with a protein target (e.g., H-bonds)5,6. Tetrahydronaphthyridines (THNs) are semi-saturated bicycles that ring-fuse a piperidine with a pyridine—these in turn being the two most popular N-heterocycles deployed in small-molecule pharmaceuticals7,8. Positioning of the two THN nitrogen atoms generates eight different structural isomers: four of which (1a–d) can be considered as CH → N bioisosteres of tetrahydroquinolines (THQs), and the remaining four (structures 2) as CH → N bioisosteres of tetrahydroisoquinolines (THIQs) (Fig. 1a). The substitution of CH units for N atoms in (hetero)aromatic systems can impart orders of magnitude improvements in key physicochemical (e.g., solubility) and pharmacological parameters9, and synthetic strategies that could provide facile access to any THN isomer (e.g., 1a–d) would be highly enabling. Without a trivial naming convention for THNs, we shall hereafter refer to structures 1 as “THNs” and their isomeric counterparts 2 as “THINs”10,11,12,13,14,15,16, by analogy to THQs and THIQs. Amongst other applications17,18, THNs have found use as guanidine mimetics of the arginine binding motif in RGD-binding integrin inhibitors (e.g., 3)19,20. Scaffold morphing of quinolines to THNs can also be an effective tactic to improve aqueous solubility, as exemplified during the development of the FGFR4 selective inhibitor Roblitinib (FGF401) 4 (Fig. 1b)21. Spirocyclisation of fused, semi-saturated N-heterocycles is also emerging as a powerful design strategy for medicinal chemistry. When compared to their flat, all-aromatic counterparts, partial saturation and installation of a spirocycle simultaneously increases Fsp3, reduces structural flexibility, and introduces alternative exit vectors for access to novel 3D chemical space22. In favourable cases, this can lead to greatly enhanced potency, selectivity, solubility, and metabolic stability23,24. For instance, Pfizer have exploited a spirocyclic THN as the core of their MC4R antagonist PF-07258669 5, which is currently in phase I clinical trials for the treatment of appetite loss (Fig. 1c)25. The spirocycle in 5 was rationally designed to enforce a cis-relationship between the N–H bond and the adjacent N(sp2) lone pair, which is the optimal geometry for target binding but opposite to the (trans) conformer favoured in solution for the non-constrained analogue.

Fig. 1: Tetrahydronaphthyridines and their importance.
figure 1

a Tetrahydronaphthyridine (THN) isomers. b THNs 1 in drug development. c Spirocyclic THNs 1 in drug development.

Full size image

Despite the many opportunities for structural and physicochemical tuning that THNs can offer, their widespread adoption in early-stage drug discovery has likely been hampered by their poor commercial availability, and the scarcity of THN synthesis approaches that are readily amenable to library generation. Other than the semi-hydrogenation of naphthyridines26,27,28, which can present regio- and chemoselectivity challenges, several routes to THNs have been devised based on the annulation of 2-aminopyridines. These processes tend to be relatively labour-intensive29,30, and whilst catalytic annulations do exist31,32, their functional group tolerance is low. Inverse electron demand, intramolecular, hetero-Diels-Alder reactions of tethered imidazolyl33 or alkynyl34,35,36 triazines sequenced with N2 extrusion are also on record for THN synthesis, but the substrate syntheses require multiple steps. Moreover, none of the above approaches are amenable to spirocyclic THN synthesis. Another distinct strategy is to form THNs via the N-arylative cyclisation of γ-pyridyl amines, either by intramolecular Pd-catalysed C–N coupling25,37,38, SNAr reactions37,38,39, or Chichibabin reactions40,41. γ-Pyridyl amines 6 can themselves be constructed via Sonogashira-hydrogenation sequences25,41, B-alkyl Suzuki-Miyaura coupling40, or the SN2 ring-opening of cyclic sulfamidates 9 with ortho-lithiated halopyridines39 (Fig. 2a). Whilst these approaches can enable access to spirocyclic THNs, the chemistry is not well suited to library synthesis, given the meagre commercial availability of α-(di)substituted propargylic amines (7), allylic amines (8), or γ-hydroxy amines as starting materials. Yu and co-workers have developed a Pd-catalysed γ-C(sp3)–H arylation of primary alkylamines that can access γ-pyridyl amines, and applied this to a single example of THN synthesis, but the amines amenable to this procedure are largely unfunctionalized and have limited commercial availability42. Recently, visible-light photoredox-catalysed approaches have been reported by ourselves and Gaunt et al., respectively, for the modular synthesis of γ-aryl primary amines by the C–C bond-forming coupling of readily available primary alkylamines 1037 or ketone-derived imines 1138 with styrenes. Between these two disclosures, four examples of spirocyclic 1,2,3,4-tetrahydro-1,8-naphthyridine synthesis were showcased, proceeding via the SNAr cyclisation of (isolated) γ-pyridyl amines from the photoredox step.

Fig. 2: Prior art for THN synthesis and this work.
figure 2

a Selected synthetic routes to (spirocyclic) 1,8-THNs. b This work.

Full size image

In this work, we show that photoredox-catalysed hydroaminoalkylation43,44,45 (HAA) of halogenated vinyl pyridines, followed by intramolecular N-arylation via SNAr, can be sequenced in continuous flow46,47,48,49,50,51,52,53,54,55,56 to enable an automated synthesis of α-alkylated and spirocyclic 1,2,3,4-tetrahydro-1,8-naphthyridines (“1,8-THNs”) 13, in addition to their regioisomeric 1,6-THN analogues 14 (Fig. 2b). To access the corresponding 1,7- and 1,5-THN isomers—15 and 16, respectively—a photoredox-catalysed HAA step can be telescoped with a palladium-catalysed C–N bond formation. Altogether, this provides a highly modular approach to four THN isomers 13–16 from a common set of unprotected primary amine starting materials 10, using the same bond disconnections.

Results and discussion

Reaction optimisation

Photoredox-catalysed hydroaminoalkylation (HAA) of 2-fluoro-3-vinylpyridine 17 with cyclohexylamine 10a gives γ-pyridyl amine 18a in 97% yield, and subjection of this purified material to DIPEA (1.5 equiv) in DMF at 120 °C for 20 h delivers the corresponding THN 13a in 92% yield via intramolecular SNAr cyclisation37. In order to sequence these reactions together in continuous flow, we transferred the chemistry to a Vapourtec R-series flow system equipped with a Uniqsis PhotoSyn LED photoreactor (420 nm LEDs, ~260 W radiant output power, 5-ml reactor coil) and a high-temperature tube reactor (up to 250 °C). As per the batch procedure, the photoredox-catalysed HAA was initially carried out with 2,4,6-tris(diphenylamino)-3,5-difluorobenzonitrile (3DPA2FBN) as the photocatalyst, and tetrabutylammonium azide (Bu4N+N3–) as the hydrogen atom transfer (HAT) catalyst57,58. However, we found that Bu4N+N3– could be replaced with cheaper and far less hygroscopic sodium azide (NaN3), which is soluble in DMF at 0.06 M. For the SNAr step, a temperature of 180 °C with tR = 20 min proved sufficient for complete conversion (see Supplementary Table 1). By running both steps in sequence in continuous flow, an overall yield of 98% of spirocyclic THN 13a could be obtained from 2-fluoro-3-vinylpyridine 17 and cyclohexylamine 10a as feedstocks (in a 1:1 ratio). This corresponds to a productivity of 2.20 mmol h–1 (445 mg h–1).

Automated continuous flow synthesis of THNs from primary alkylamines

With an optimised continuous flow protocol in hand, we next sought to execute an automated library synthesis of ‘lead-like’5,59 THN products 13, using an autosampler to sequentially load different amine substrates into the Vapourtec flow system. The same autosampler also serves as a fraction collector, into which the steady-state solutions of each product 13 are dispensed (Fig. 3a and see Supplementary Data 1 for NMR spectra of all compounds). Each run using 1.50 mmol of the vinylpyridine substrate takes ~90 min, which corresponds to 16 compounds in a 24 h period, or 40 compounds total over 60 h if all rack positions are utilised. Cyclic primary amines 10a–d of varying ring sizes were well tolerated, and amine 10e bearing benzylic C–H bonds also participated smoothly. Various functionalities including free hydroxyl groups (13f, o), ethers (13g, k), thioethers (13h), carbamates (13i, j, p), and imidazoles (13q) proved compatible with the process. For amines bearing electronegative atoms attached to the β- or γ-carbon (10g–k, o, p), a slightly elevated temperature of 200 °C proved necessary in most cases to drive the SNAr step to completion within the 20 min residence time. Strained four-membered ring substrates 3-amino-N-Boc-azetidine 10j and 3-aminooxetane 10k proved especially challenging for the photoredox step, on account of their α-C–H bonds being strengthened by ring strain and inductive effects57; amine 10j for example gave only 49% yield of 13j, along with 46% of unreacted 10j. By increasing the stoichiometry of amines 10j and 10k to 3 equivalents, however, the valuable spirocyclic THNs 13j and 13k could be obtained in 80% and 61% yields, respectively. Non-spirocyclic THNs are also readily accessible via this methodology; isopropylamine 10l was used, for example, to generate α,α-dimethyl-substituted THN 13l in 75% yield. For α-monoalkylated amines (10m–q), it proved necessary to use 3.0 equivalents of the amine substrate, to mitigate against the formation of undesired dialkylated products during the photoredox α-C–H alkylation step37. As ethylamine 10m is a gas at ambient pressure, it was dispensed as a 2.0 M solution in THF, affording the simple α-methylated THN 13m in 36% yield. Ethanolamine 10o and N-Boc ethylenediamine 10p also proved to be effective substrates, generating α-hydroxymethyl- and α-aminomethyl-substituted THNs 13o and 13p, respectively.

Fig. 3: Automated continuous flow synthesis of THNs from primary alkylamines.
figure 3

aWith 3.0 equiv of amine. bSecond step carried out at 200 °C. cSecond step carried out at 220 °C. All reactions were carried out on 1.50 mmol of vinylpyridine 17 or 18. a Synthesis of 1,2,3,4-tetrahydro-1,8-naphthyridines 13. b Synthesis of 1,2,3,4-tetrahydro-1,6-naphthyridines 14.

Full size image

We next sought to extend our automated synthesis protocol to the formation of isomeric 1,2,3,4-tetrahydro-1,6-naphthyridines (“1,6-THNs”) 14, using 4-chloro-3-vinylpyridine 18 as a radical acceptor (Fig. 3b). Whilst the chlorinated compound 18 is far easier to access than its 4-fluoropyridine-derived60 counterpart, the decreased SNAr reactivity of the C–Cl bond necessitated that the temperature of the flow SNAr step be raised still further to 220 °C. Under these conditions, a small library of spirocyclic 1,6-THNs 14a–e could be prepared in 46–64% yield.

Gram-scale reaction and resolution of THN enantiomers

To demonstrate the scalability of our THN synthesis in flow, we executed the reaction of 4-aminopiperidine substrate 10i on gram scale on a 5-ml reactor coil, delivering 1.85 g of spirocyclic THN 13i in 87% yield (equating to a productivity of 600 mg h–1) (Fig. 4a). Whilst these reactions inevitably produce racemic materials, resolution of the THNs via chiral preparative HPLC provides convenient access to both enantiomers, as exemplified for THN 13n on a 520 mg scale (Fig. 4b and see Supplementary Figs. 4–8 for HLPC traces).

Fig. 4: Gram-scale reaction and resolution of THN enantiomers.
figure 4

a Gram-scale reaction in flow. b Resolution of chiral racemic THNs by prep HPLC.

Full size image

Access to THN derivatives with functional handles on the pyridine ring

Another important objective was to demonstrate further elaboration of the THN products on the pyridine ring. One strategy, which is especially useful for C(6) functionalisation, is to carry out electrophilic halogenation or catalytic C–H borylation61 reactions (i.e., 19–21) (Fig. 5a). In order to access THNs 23 and 25 halogenated ortho or para to the pyridine nitrogen, we utilised vinyl pyridines 22 and 24, respectively, with the necessary chloro handles preinstalled. Using this strategy, the C(7)-chloro THN 23 was isolated in 68% yield, and the C(5)-chloro THN 25 in 31% yield (Fig. 5b). The latter isomer was anticipated to be the most challenging, requiring the amine nucleophile to distinguish between a para-chloro and an ortho-fluoro site of attack during the SNAr step39. Taken together, these strategies enable vector growth from any ring position on the fused pyridine moiety of 1,8-THNs, which is likely to be of significant value for fragment-based drug discovery1,2,3,4.

Fig. 5: Access to THN derivatives with functional handles on the pyridine ring.
figure 5

a C–H functionalisation of the pyridine ring at C(6). b Access to other halogenated THN isomers.

Full size image

Stepwise synthesis of other THN isomers

Varying the position of the pyridine nitrogen atom in these spirocyclic THN scaffolds is another highly desirable objective from a medicinal chemistry standpoint9. Having already demonstrated an automated flow synthesis of 1,8- and 1,6-THNs 13 and 14 from primary alkylamine feedstocks, we were motivated to develop a practical catalytic solution to access 1,7- and 1,5-THN isomers, based on the same photoredox-catalysed HAA disconnection approach. With intramolecular N-arylation via SNAr no longer being feasible, we instead opted to carry out this key step using palladium catalysis. Following a flow photoredox HAA of amine 10i with 3-chloro-4-vinylpyridine 26, γ-pyridyl amine 27 was isolated in 25% yield. The low yield in this case was traced to extensive polymerisation side reactions, for which vinylpyridine 26 seems to be particularly prone. Subsequent cyclisation via a Buchwald-Hartwig C–N coupling then gave 1,7-THN 28 in 79% yield (Fig. 6a). An analogous sequence using 3-chloro-2-vinylpyridine 29 gave 1,5-THN in an overall 47% yield over the two steps (Fig. 6b).

Fig. 6: Stepwise synthesis of other THN isomers.
figure 6

a Synthesis of 1,2,3,4-tetrahydro-1,7-naphthyridines. b Synthesis of 1,2,3,4-tetrahydro-1,5-naphthyridines.

Full size image

Application to the synthesis of Pfizer’s MC4R antagonist PF-07258669 5

Finally, we sought to apply our methodology to a concise synthesis of the spirocyclic THN core (35) of Pfizer’s MC4R antagonist PF-07258669 5, which was previously synthesised in 15 total steps (11 steps LLS) (Fig. 7a)25. In our case, starting from commercially available 3-amino N-Boc pyrrolidine 32, a photocatalytic HAA reaction with vinylpyridine 33 in continuous flow gave γ-pyridyl amine 34 (427 mg) in 79% yield. Attempted thermal SNAr cyclisation of 34 at 220 °C in a high-temperature tubular reactor (tR = 20 min) gave only 22% yield of THN 35, indicating that the methyl substituent α- to the pyridine nitrogen deactivates this pathway. Fortunately, an intramolecular, palladium-catalysed Buchwald-Hartwig N-arylation process (as used in the Pfizer route) proved more efficacious, delivering the spirocyclic THN core 35 in 84% yield (Fig. 7b). Taking into account a 3-step synthesis of vinyl pyridine 33, the longest linear sequence is five steps. The industrial route, whilst 11 steps in the longest linear sequence, is enantioselective, compared to a racemic synthesis in our case. Nevertheless, this illustrates how dramatically the synthesis of complex spirocyclic amines can be streamlined when using a photoredox annulation strategy from unprotected amines57.

Fig. 7: Application to synthesis of the THN core of Pfizer’s melanocortin MC4 receptor antagonist (5).
figure 7

a Prior art: Pfizer’s enantioselective synthesis of core (35) of melanocortin MC4 receptor antagonist 5. b Our (racemic) synthesis of core (35) of melanocortin MC4 receptor antagonist 5.

Full size image

Conclusion

In summary, we have developed an automated, continuous flow synthesis of α-alkylated and spirocyclic 1,2,3,4-tetrahydro-1,8-naphthyridines (“1,8-THNs”), in addition to their regioisomeric 1,6-THN analogues, from abundant primary amine feedstocks. An annulative disconnection approach based on photoredox-catalysed hydroaminoalkylation (HAA) of halogenated vinylpyridines is sequenced in combination with intramolecular SNAr N-arylation. To access the remaining 1,7- and 1,5-THN isomers, a photoredox-catalysed HAA step is telescoped with a palladium-catalysed C–N bond formation. Altogether, this provides a highly modular access to four isomeric THN cores from a common set of unprotected primary amine starting materials, using the same bond disconnections. The simplifying power of the methodology is illustrated by a concise synthesis of the spirocyclic THN core (35) of Pfizer’s MC4R antagonist PF-07258669 (5).

Methods

A general procedure for the flow chemistry protocol described in Fig. 3 can be found in Supplementary Methods (pages S4–5), plus photographs and schematics of the setup in Supplementary Figs. 1–3.

Representative procedure for the automated continuous flow synthesis of 1,2,3,4-tetrahydro-1,8-naphthyridine (13a): following the General Procedure (pages S4–5), 5 ml of reagent feed A [2-fluoro-3-vinylpyridine 17 (185 mg, 1.50 mmol, 1.0 equiv) and 3DPA2FBN (9.6 mg, 15.0 μmol, 1 mol%) in anhydrous DMF], 5 ml of reagent feed B [cyclohexylamine 10a (149 mg, 1.50 mmol, 1.0 equiv) and NaN3 (19.5 mg, 300 μmol, 20 mol%) in anhydrous DMF], and 10 ml of reagent feed C [DIPEA (291 mg, 2.25 mmol, 1.5 equiv) in anhydrous DMF] were reacted in flow, setting the high-temperature tube reactor to 180 °C. The steady-state mixture (10 ml) was collected and concentrated in vacuo on an Asynt spiral evaporator. Purification via automated flash column chromatography on SiO2 gel (12 g) in 40–60 °C petroleum ether (5 CV) then 100:0 → 0:100 40–60 °C petroleum ether–EtOAc (over 20 CV) then EtOAc (5 CV) gave 13a as a colourless, crystalline solid (149 mg, 98%, productivity = 2.20 mmol h–1).