Benchtop ¹⁹F Nuclear Magnetic Resonance (NMR) Spectroscopy-Optimized Knorr Pyrazole Synthesis of Celecoxib and Mavacoxib, 3-(Trifluoromethyl) Pyrazolyl Benzenesulfonamides, Non-Steroidal Anti-Inflammatory Drugs (NSAIDs)

Abstract

Fluorinated organic compounds have demonstrated remarkable utility in medicinal chemistry due to their enhanced metabolic stability and potent therapeutic efficacy. Several examples exist of fluorinated non-steroidal anti-inflammatory drugs (NSAIDs), including diflunisal, flurbiprofen, and trifluoromethylated pyrazoles celecoxib and mavacoxib. These trifluoromethylated pyrazoles, which are most commonly constructed through the cyclocondensation of a trifluorinated 1,3-dicarbonyl and an aryl hydrazine, are also found in numerous other drug candidates. Here, we interrogate the effects of solvents and the presence of Brønsted or Lewis acid catalysts on catalyzing this process. We highlight the utility of benchtop ¹⁹F NMR spectroscopy in enabling the real-time quantification of reaction progress and the identification of fluorinated species present in crude reaction mixtures without the need for cost-prohibitive deuterated solvents. Ultimately, we find that the reaction solvent has the greatest impact on the rate and product yield, and also found that the relationship between the keto-enol equilibrium of the dicarbonyl starting material pyrazole formation rate is highly solvent-dependent. More broadly, we describe the optimization of the yield and kinetics of trifluoromethylpyrazole formation in the synthesis of celecoxib and mavacoxib, which is made possible through high-throughput reaction screening on benchtop NMR.

1. Introduction

Since the initial success of fluorinated corticosteroids in the 1950s and fluoroquinolones in the 1980s, the number of FDA-approved fluorinated pharmaceuticals has steadily increased and now accounts for twenty percent of all FDA-approved pharmaceutical agents [1,2,3,4,5,6]. This can be attributed to the improved metabolic stability, bioavailability, and therapeutic profiles associated with fluorinated motifs [4,5,6,7,8,9,10,11]. Among these, trifluoromethylated pyrazoles have demonstrated particular pharmacological potency [12,13,14,15,16], being referenced over twenty-six thousand times in patents and peer-reviewed publications concerning the preparation and use of various therapeutics and agrochemicals as of 2024. FDA-approved celecoxib (Celebrex) and EU-approved mavacoxib (Trocoxil), examples of trifluoromethylated pyrazole-containing compounds, are non-steroidal anti-inflammatory drugs (NSAIDs) that act through the inhibition of cyclooxygenase (COX) enzymes that mediate the conversion of arachidonic acid into inflammatory prostaglandins [17,18,19,20,21,22] (Figure 1A).

Numerous methods have been described for the preparation of these and other fluorinated pyrazoles, including the fluorination of methylpyrazoles using electrophilic or radical fluorination reagents and the fluorination of aminopyrazoles using a Balz–Schiemann nucleophilic substitution involving a diazonium salt [13,23,24,25,26]. However, the most direct and common synthesis involves the Knorr cyclocondensation of 1,3-dicarbonyls and hydrazines [27,28,29,30] (Figure 1A). This Knorr synthesis has been industrially applied to the kilogram scale of celecoxib, prepared from the condensation of 4,4,4-trifluoro-1-(4-methylphenyl)butane-1,3-dione and 4-sulfonamidophenylhydrazine [31,32,33].

Lewis and Brønsted acid catalysts have demonstrated broad utility in accelerating various cyclocondensation constructions of heterocycles including Knorr pyrazole synthesis, likely through electrophilic activation by protonation or coordination to the 1,3-dicarbonyl [34,35,36,37,38,39,40,41,42]. Additionally, Brønsted acids may disrupt the intramolecular hydrogen bonding that stabilizes the less reactive enol tautomer and facilitates the protonation of the electrophilic carbonyl, enhancing its susceptibility to nucleophilic attack by the hydrazine [37,38,39,40,41].

Initial reaction condition screening performed by Reddy et al. has previously demonstrated that the efficiency and yield of the uncatalyzed Knorr pyrazole synthesis in the industrial-scale preparation of celecoxib can be significantly influenced by solvent effects [31]. Our laboratory has previously reported the utility of benchtop ¹⁹F NMR spectroscopy in high-throughput condition screening for the reaction optimization and quantitative reaction monitoring of fluorinated pharmaceutical agents [43,44,45], solvent effects on K_eq (keto/enol) of fluorinated dicarbonyls [44], and the mechanistic deconvolution of heterocycle-forming multicomponent cyclocondensation reactions, such as the Biginelli condensation reaction through the real-time quantification of even transient intermediates and products of complex multicomponent reactions that can be directly measured from crude reaction aliquots in non-deuterated solvents [45] (Figure 1B).

Here, we sought to determine the role of solvents and catalysts on the rate and efficiency of trifluoromethylated pyrazole formation and to understand how perturbations of the keto-enol tautomerization equilibrium of the trifluoromethylated 1,3-dicarbonyl starting materials influence the efficiency of pyrazole formation. In this study, we utilize benchtop ¹⁹F NMR spectroscopy as a high-throughput condition screening platform to identify the optimal catalyst, solvent, and mechanistic trends in the preparation of two trifluoromethylated pyrazoles: celecoxib and mavacoxib. The presence of the trifluoromethyl motifs conveniently serves as spectroscopic handles with which to quantitatively monitor reaction kinetics [46,47,48,49,50]. With this workflow, we screened for potential applications of nine common Lewis and Brønsted acid catalysts across a representative selection of protic and aprotic solvents in the preparation of fluorinated pyrazoles from 1,3-dicarbonyl substrates. Through this reaction screening, we identified that pyrazole formation occurs the fastest in dimethylformamide (DMF), irrespective of the presence of Lewis or Brønsted acid catalysts, suggesting that solvents play the most critical role in reaction efficiency. Additionally, we show that the direct effect of keto-enol equilibrium on the reaction rate is most pronounced in polar protic solvents. More broadly, we show the applicability of benchtop ¹⁹F NMR spectroscopy as an analytical screening platform that provides rapid, quantitative mechanistic insight into tautomerization and cyclocondensation kinetics without the necessity of deuterated solvents, and which cannot be easily or as rapidly observed through other means. This enables the condition optimization of these and potentially other fluorinated heterocycles.

2. Materials and Methods

2.1. General Synthesis Procedure (Micro-Scale)

To a standard 5 mm NMR tube an initial solution of trifluoromethylated 1,3-dicarbonyl (0.20 mmol/mL), catalyst (0.02 mmol/mL), and solvent (0.25 mL) was added. The reaction mixture was placed into a 60 °C water bath and was allowed to equilibrate over a 30 min period. Another separate solution of 4-sulfonamidophenylhydrazine HCl and solvent (0.20 mmol/mL, 0.25 mL), as well as the addition of α,α,α-trifluorotoluene (0.10 mmol/mL) directly into the reaction mixture as a chemical shift reference and integration standard due to its compatibility with reaction conditions [45], was then added. An initial 60 MHz ¹⁹F NMR was taken and the reaction mixture was placed into a 60 °C water bath, and ¹⁹F NMR spectra were taken at 30 min intervals over a 4.5 h period with a Nanalysis NMReady 60-Pro benchtop nuclear magnetic resonance spectrometer.

2.2. Kinetic Parameters

Kinetics were run in a water bath heated to a temperature of 60 °C and maintained on a thermocouple heat plate. Data points are reported as the K_eq of a single data experiment at which the concentrations of each tautomer reach equilibrium. NMR spectra were taken with a 200 ppm spectral width and a 1.0 s scan delay with 32 scans.

3. Results and Discussion

3.1. Initial Rate

The initial keto-enol equilibration of the fluorinated 1,3-dicarbonyls 2 and 3 under different reaction conditions was found to be dependent on both the solvent and catalyst. Consistent with expectation and with previous reports, solvents capable of hydrogen bonding gave predominantly the ketone tautomer (¹⁹F δ = −80 to −84 ppm, depending on the solvent, and assigned by the corresponding ¹H NMR (Figure S28)), whereas the enol tautomer (¹⁹F δ = −71 to −77 ppm, as assigned by the corresponding ¹H NMR (Figure S28)) was observed to dominate the equilibrium in aprotic polar solvents (DMSO, NMP, and DMF). When dicarbonyl 2 or 3 is subjected to reaction with 4-sulfonamidophenylhydrazine 1, direct conversion into the trifluorinated pyrazole celecoxib 6 or mavacoxib 7 was observed by aliquot reaction ¹⁹F NMR (Figure 2); the identity of the product peak by NMR was deemed to be identical to an authentic sample of celecoxib or mavacoxib and verified by liquid chromatography–mass spectrometry (LC-MS) (Figure S27). From these initial solvent investigations, we observed that the formation of both celecoxib and mavacoxib is most rapid in DMF compared to other solvents, while the slowest rates of conversion are observed in ethanol (Figure 2B). Remarkably, there is no generalizable correlation between the solvent-driven solution phase equilibrium constant of the dicarbonyl and the initial rate of pyrazole formation (Figure 3A), though specifically among the two protic solvents screened (methanol and ethanol) a greater K_eq in methanol versus ethanol (0.67 vs. 0.24) resulted in a much faster rate of conversion (8.83 μM/s in MeOH, as opposed to 3.33 μM/s in EtOH). Subsequently, we screened a number of common Brønsted acids, including hydrochloric acid, p-toluenesulfonic acid, and acetic acid, along with five transition metal Lewis acid catalysts, including LaCl₃, Yb(OTf)₃, Sc(OTf)₃, In(OTf)₃, Er(OTf)₃, and Cu(OTf)₂. Within trials performed in protic solvents, this general positive correlation is retained between the K_eq and the initial reaction rate (Figure 3B). However, in aprotic solvents, no consistent positive trend is observed consistently across all five solvents (Figure 3C).

It is possible that this behavior of the fluorinated dicarbonyls 2 and 3 in protic solvents (MeOH, EtOH) is attributed to the increased availability of the reactive keto tautomer, which is aligned with the proposed reaction mechanism, which proceeds through an initial nucleophilic attack of the keto tautomer while the enol tautomer remains unreactive at the desired trifluoromethyl ketone center. However, polar aprotic solvents, such as DMF and DMSO, do not strictly follow this trend, demonstrating that the rate of reaction is not solely dependent on the K_eq. We attribute this inconsistency in the initial reaction rate to a solvent’s ability to stabilize the reaction intermediates by acting as a hydrogen bond acceptor, such as in DMF. Some notable conditions include Er(OTf)₃ and being uncatalyzed in DMF, which have the fastest initial formation of product in celecoxib and mavacoxib, respectively. Additionally, the lack of a keto tautomer present in all polar aprotic solvents, such that K_eq is almost zero, highlights the lack of catalyst effects on keto-enol tautomerization and the determinant factor being the solvent.

The lack of these observable intermediates by ¹⁹F NMR suggests that the initial nucleophilic attack of hydrazine 1 to the trifluoromethylated ketone of 2 or 3 is rate-determining, and that subsequent cyclization through intramolecular condensation from hydrazone 4 or 5 is a much faster process. This is consistent with first-order kinetics observed for both hydrazine and 1,3-dicarbonyl.

3.2. Final Conversion

Irrespective of the presence of a catalyst, methanol conditions resulted in the greatest yield of both compounds (celecoxib: 96%, mavacoxib: >99%), notably Cu(OTf)₂ and being uncatalyzed for the formation of celecoxib and mavacoxib, respectively. Consistent with expectation, solvents that favor the keto tautomer result in greater formation of the desired product, which can be attributed to the more efficient formation of the product (Figure 4B). However, despite DMSO and ethanol being able to stabilize the keto tautomer at a higher rate than NMP and DMF, these conditions have shown the least conversion into celecoxib and mavacoxib, respectively (Figure 4A). The accelerated conversion of intermediate 5 into mavacoxib can be ascribed to its heightened reactivity with the phenylhydrazine starting material. This is likely due to the electron-withdrawing effects of the para-fluorine substituent, which increases the electrophilicity of the 1,3-dicarbonyl, thus promoting nucleophilic attack by the hydrazine.

3.3. Regioisomer

Previously, the undesired 1,3-aryl regioisomers of 1G have been reported as a byproduct (¹⁹F δ = −58.85 ppm in methanol) [31], most likely resulting from a nucleophilic attack of the hydrazine to the less electrophilic aryl ketone rather than the trifluoromethyl ketone (Figure 5). We observed the formation of this regioisomer exclusively when the reaction was conducted in methanol. The formation of the regioisomer is attributed to the competing nucleophilicity of solvents such as methanol and ethanol; however, this phenomenon is not observed in ethyl or isopropyl alcohol solvents, presumably due to their greater steric bulk, resulting in decreased favorability of the formation of these solvates. Therefore, we attribute the formation of the regioisomer to methanol being added to the 3-dicarbonyl, forming a dimethyl acetal, and thereby reducing electrophilicity at that site, forcing a nucleophilic attack by the hydrazine on the 1-dicarbonyl 1A (Figure 5). Additionally, the formation of the regioisomer is not catalytically affected and not consistent across both compounds (

Table 1. Percent of the regioisomer observed after 4.5 h.

Catalyst	Celecoxib	Mavacoxib
Uncatalyzed	4.5%	0.0%
LaCl3	3.2%	2.8%
Yb(OTf)3	2.2%	3.0%
Sc(OTf)3	3.9%	4.0%
In(OTf)3	4.4%	3.2%
Er(OTf)3	3.4%	3.3%
Cu(OTf)2	2.5%	2.7%
Acetic acid	4.4%	3.3%
p-toluenesulfonic acid	3.9%	3.7%
Hydrochloric acid	2.7%	3.6%

).

4. Conclusions

In summary, we demonstrate through benchtop ¹⁹F NMR that the kinetics of the Knorr pyrazole cyclocondensation of trifluorinated diketones and aryl hydrazines towards the preparation of celecoxib and mavacoxib are heavily influenced by reaction solvents over the presence of Lewis or Brønsted acid catalysts. In the protic solvents screened in this study, the rate of pyrazole formation was found to be directly dependent on the relative concentration of the 1,3-dicarbonyl ketone tautomer in solution. We observed that the aprotic solvents DMF and NMP favor the initial formation of celecoxib and mavacoxib, while the protic solvent methanol results in greater overall conversion. Additionally, we found minimal to no consistent catalytic effect of both Lewis and Brønsted acids on the initial rate and total product formation. This trend might be attributed to the differences in each solvent to stabilize key reaction intermediates.

Ultimately, we demonstrated the utility of benchtop ¹⁹F NMR for real-time kinetic analysis for high-throughput condition screening in the preparation of two trifluoromethylated phenylpyrazoles of pharmaceutical importance. Beyond the preparation of celecoxib and mavacoxib, the workflow for high-throughput solvent and catalyst screening described here is broadly applicable to the synthesis of other heterocyclic constructions containing key fluorinated motifs.