Kinetics of Electrophilic Fluorination of Steroids and Epimerisation of Fluorosteroids

Abstract Fluorinated steroids, which are synthesised by electrophilic fluorination, form a significant proportion of marketed pharmaceuticals. To gain quantitative information on fluorination at the 6‐position of steroids, kinetics studies were conducted on enol ester derivatives of progesterone, testosterone, cholestenone and hydrocortisone with a series of electrophilic N−F reagents. The stereoselectivities of fluorination reactions of progesterone enol acetate and the kinetic effects of additives, including methanol and water, were investigated. The kinetics of epimerisation of 6β‐fluoroprogesterone to the more pharmacologically active 6α‐fluoroprogesterone isomer in HCl/acetic acid solutions are detailed.


Introduction
In the history of the development of fluorine-containingd rug substances, fluorosteroids made significant early contributions when Fried and Sabo discovered that the introductiono fa single fluorine atom into ac orticosteroid 1 increased its potency tenfold. [1] Since this observation, numerous fluorinated steroids have been marketedf or the treatment of various disease classes, including cancers and inflammation. [2] In particular, fluorosteroids bearingaf luorine atom at the 6-position, such as flurandrenolide 2 and fluticasone 3,continue to be commercially significant (Figure1). Fluticasone therapeutics were ranked in the top 200 drugs prescribed in the USA in 2017. [3] The introductionofafluorine atom at the 6-position is normally carried out by reactiono fasteroid enolate derivativew ith an electrophilic fluorinating agent. Early exampleso ft his transformationincluded the use of perchlorylf luoride (ClO 3 F) [4] and trifluoroacetyl hypofluorite (CF 3 COOF), [5] but due to the hazardous natures of these reagents, they were not suitable for large scale use. In recent years, several electrophilic fluorinating reagents of the NÀFc lass, such as N-fluoropyridinium salts, [6] NFSI, [7] Selectfluor, [8] and Accufluor [9] have been used for the fluorination of steroid enolate derivatives. Within the NÀF class of reagents, Selectfluor has been used in larger scale applications,s uch as in the manufacture of fluticasone 3. [10] Indeed, an estimated 80 %o fc ommerciallya vailable fluorosteroids are synthesised using Selectfluor. [11][12][13][14] The introductiono faf luorine atom using an NÀFr eagent has also been achieved at other positions within the steroid; for example, Lectka et al. recently described the photocatalytic fluorination of steroidsa tt he 15-position using Selectfluor. [15] 6-Fluorosteroids are generally formeda sm ixtures of 6a-a nd 6b-isomers, where the former is usually pharmacologically active. [16,17] The ratio depends upon the fluorinating reagent employed,s teroid structure, temperature and timescale of the reaction. These factors were explored by Herrinton et al. [18] using severalN ÀFr eagents,w here Selectfluor was determinedt ob ethe most efficient fluorinating agent.
To our knowledge,t here have been no kinetics studies on the electrophilicf luorination of steroidal enolate systems, despite their therapeutic and commercial importance.H owever, Figure 1. Examples of biologically-active fluorinated corticosteroid drugsc urrently on the market.
[a] Dr.N . Rozatian recent studies by Nelsone tal. [19] reported the kinetics of fluorination of ar elatedt etralone system using Selectfluor.T hese studies establishedt he mechanistic pathway of fluorination using Hammett correlations, concluding that an S N 2r eaction occurred rather than SET.Furthermore, the kinetics and mechanisms of acid-mediated epimerisations of fluorosteroid mixtures to their more pharmacologically active a-isomers remain underexplored.
General studies on the mechanismsa nd reactivities of NÀF reagentstowards carbon nucleophile systems have also recently been performed. Mayr et al. [20] used ah eterogeneousr ange of C-nucleophile systems to providee vidence for an S N 2m echanism and quantitative electrophilicity values, based on the Mayr-Patz scale, for five NÀFf luorinating reagents.W eu sed a homologouss eries of enolic para-substituted 1,3-diaryl-1,3-dicarbonyl derivatives to report aq uantitative reactivity scale for ten electrophilic NÀFf luorinating reagents ( Figure 2) and deliver Hammett correlationst hat support S N 2f luorine transfer. [21] We recently extended our studies to explore the factors affecting difluorination of our 1,3-dicarbonyl compound series, concluding that solvent effectsd ramatically enhancee nolization rates, thuspromoting difluorination. [22] Quantitative approaches offer the possibility of matching the reactivities of nucleophilic ande lectrophilicpartners, allowing reactionr ates to be estimated andr apid, but controlled, processes to be designed. [23] Given the pharmaceutical importance of 6-fluorosteroidal systems, we soughtt oq uantify the nucleophilicities of enol equivalents of four main classes of steroids, namely;p rogesterone 13 (a progestogen), testosterone 14 (an androgen), cholestenone 15 (a cholesterol precursor) and hydrocortisone 16 (a corticosteroid). Each of theses teroids containsa ne nolisable a,b-unsaturated ketone system that can direct fluorination to their 6-positions via the corresponding enol ester systems 17-20 (Scheme 1), which are readily synthesised in one step. [24,25] We explore ar ange of fluorinating reagents, including NÀF reagents and fluorine gas, and the factors that affect stereo-chemicalo utcome. We also perform kinetic studies upon epimerisation processes that are used to redress the balance between the kinetically favoured b-isomers that arise from fluorination processes and the pharmaceutically desired a-isomers.
Most electrophilic fluorinating reagents are synthesised from fluorineg as (F 2 ), however, there have been no reports of the corresponding fluorinations of steroidal enol ester systemsa t the 6-positionu sing fluorine gas itself, although there are several reports of fluorination of tertiary CÀHp ositions in steroid substrates using F 2 . [26,27] Since selectived irect fluorination of steroidsb yf luorine gas could provide am ore cost-effective, greener route to commercially important 6-fluorosteroid derivatives, we used progesteronee nol acetate 17 to study direct fluorination using fluorine gas. [28] Results and Discussion

Preparation of materials and stereochemical characterisation
In ordert oa ssess kinetics of enol acetate fluorination and subsequent epimerisation of fluorosteroids, we preparedaseries of enol acetate substrates and isolated a-a nd b-fluorosteroid isomers. The facial selectivities of ar ange of reagents were determinedb yN MR spectroscopy.T he absolute configurations of the fluorosteroid products were, in somecases,confirmed crystallographically.
Progesterone enol acetate 17 wass ynthesised in 65 %y ield following am odified literature procedure (Scheme 1). [24] Spectroscopic analyses were in agreement with previousr eports [25] and the structure was further confirmed by X-ray crystallography (see Supporting Information Section 2.1). Fluorination of 17 was conducted using Selectfluor (Scheme1)t oo btain a mixture of a and b isomers of 6-fluoroprogesterone 21. [8] The  fluorination proceeded cleanly,w ith 100 %c onversion as determined by 1 Ha nd 19 FNMR spectroscopy,a nd 96 %y ield of the a/b isomer mixture, where the a and b isomersw erep resent in a3 4:66 ratio. The isomer mixture was resolved using columnc hromatography to afford isolated yields of 19 % aisomer and4 6% b-isomer.T he structures of the isomers were assigned by X-ray crystallography (Scheme 2).
Te stosterone enol diacetate 18,c holestenone enol acetate 19,h ydrocortisone enol tetraacetate 20 and their fluorinated derivatives 22-24 (total yields of 69-78 %f or both isomers) were synthesised and isolated using the same procedures as progesterone. X-ray crystallographic analyses of testosterone derivatives 18 and 22 were also obtained (see Supporting Information Section 2.1).
In the case of fluoroprogesterone 21,t he b-isomer was converted to the more thermodynamically stable a-form by epimerisation of the crude product mixture. We used the procedure reportedb yR ingold, [29] where the crude mixture of isomers of 21 was dissolved in acetic acid, and dry HCl gas was bubbled through the solution for 1.5 h( Scheme2). Following evaporation of solvents and recrystallisation from MeOH, 21-a was obtained in 74 %y ield.
The a:b isomer ratios of each fluorinated steroidc rude reaction mixture,p repared using NÀFr eagents, are summarised in Ta ble 1. When Selectfluor was used as the fluorinating reagent,t estosterone enol diacetate 18 resulted in the highest levels of the desired a isomer compared with the other steroidal nucleophiles. The crude a:b isomer ratios for fluorinationso fp rogesterone enol acetate 17 by seven NÀFr eagentsw ere determined, andt he reaction with Selectfluor gave the highest proportion of the a isomer.T he stereoselectivity does not follow the trend in reactivities of the NÀFr eagents.A sw eh ave previously reported, [21] diCl-NFPy TfO À /BF 4 À 11 a/b are 4-fold lessr eactive than Selectfluor and show the lowests electivities for the a isomer.N FSI 8,N FPy TfO À 9 and triMe-NFPy TfO À 10 are 4-6 orders of magnitude less reactive than Selectfluor,a nd pentaCl-NFPyT fO À 12 is 1o rder of magnitude more reactive, yet, all lead to similars tereoselectivities.
These results are surprisingc onsidering the vast range of reactivities and differing steric requirements of the fluorinating reagents.T he near-identical a:b isomer ratios suggested that epimerisation at the newly-formed CÀFc entre could be in operation, with the isomer ratio being determined by solventproduct interactions. Closer inspection of 1 HNMR kinetic data for the fluorinations of progesterone enol acetate 17 with NFPy TfO À 9 and NFSI 8 showedc onstant a:b isomer ratios over the courses of the reactions. In order to furtheri nvestigate the potential for in situ epimerisation we explored whether 'spent' Selectfluor (ClCH 2 -DABCO + ·BF 4 À )c ould play an acid/base catalysis role in this process.W hen 21-b was incubated with ClCH 2 -DABCO + ·BF 4 À in MeCN-d 3 ,n of ormationo f21-a was observed by 19 FNMR spectroscopy over the course of 1 week.T op robe the possibility of protonated 'spent' Selectfluor catalysing in situ epimerisation,w ea ttempted to prepare as ample of this protonated species, although our efforts were unsuccessful (see Supporting Information Section 2.1.12). Thus, at present, we are unable to confirm that in situ epimerisation is in operation.
To explore the effects of other enol derivatives upon fluorination kinetics and stereoselectivity,t he ethylated form of the progesterone enol, 3-ethoxy-pregna-3,5-dien-20-one 25,w as prepared using ap reviously described method. [30] This compound is the starting material for the synthesis of birth control drug quingestrone and severalrelatedc ompounds. [31] The electrophilicchlorination of 25 with N-chlorosuccinimide (NCS) was Scheme2.Epimerisation of 6-fluoroprogesterone 21-b to 21-a in HCl/AcOH and X-ray crystallographic structures for each isomer. conducted by Ringold et al. [32] although, to the best of our knowledge,t he fluorination of 25 has not been reported. On mixing 25 and Selectfluor (0.95 equiv) in CDCl 3 ,a ni mmediate colour change from yellow to red was observed (Scheme 3). CDCl 3 was used duet ot he lower solubility of 25 in MeCN and acetone. Analysis of the reactionm ixture by 19 FNMR spectroscopy showed the disappearance of the NÀF signal, however,n ew signals weren ot observed. Analysisb y high resolution mass spectrometry (HR-MS) showedt he formation of ap roduct consistentw ith oxidation of 25 by introduction of as ingle oxygen atom. The fragment ions showed loss of Ac-and AcO-groups, which indicate that the oxidised product is progesterone enol acetate 17.S electfluor is ak nown oxidant, for example, the copper-mediated oxidation of amides to imides by Selectfluor has been reported. [33] The oxidation of an ethoxyg roup to an acetyl group was previously reportedf or the conversion of 6-ethoxybenzothiazole-2-sulfonamide to the corresponding acetate with trichloroisocyanuric acid, [34] ar eagent that is known to carry out both chlorination ando xidation. [35] Oxidation may be specifica tt his position due to the non-bondinge lectrons on the ethereal oxygen atom that could activate the neighbouring CÀHb onds by hyperconjugation effects. [36] We believe oxidation processes are likely to result in the formation of HF,w hich will be lost from the reaction mixture, thus leading to the loss of detectable 19 FNMR signals over the course of the reaction. When the reactionw as conducted in MeCN-d 3 with 1equiv of Selectfluor (with sonication to fully solubilise 25), small amountso f21-a and 21-b were detected by 1 Ha nd 19 FNMR spectroscopy,asw ell as traces of unidentified fluorinated products (NMR spectra are included in the Supporting Information Section2 .1.11). Progesterone enol acetate 17 was not detected;h owever,t he major product of the reaction was progesterone 13,which could have formed by oxidation of 25 to 17,followed by hydrolysis of the ester group to form 13.T he formation of 13 was confirmed by comparison of the spectra with an authentic sample of 13.

Kinetics of fluorinationofsteroidenol acetates
The kinetics of fluorination of progesteronee nol acetate 17 by NFSI 8,N FPy TfO À 9 and triMe-NFPy TfO À 10 were monitored by quantitative 1 HNMR spectroscopy in MeCN-d 3 (Figure 3). These kinetics experimentsw ere carried out with excessN ÀF reagent to achieve pseudo-first order conditions. Due to the wide range of reactivity of the NÀFr eagents, reactions involving Selectfluor 7,d iCl-NFPy TfO À 11 a,d iCl-NFPy BF 4 À 11 b and pentaCl-NFPy TfO À 12 were too rapid to be monitored by NMR spectroscopy.H ence, we used UV/Vis spectrophotometry, where the use of lower concentrations of the reaction partners was expected to afford lower observed rates. Although the fluorinations of steroide nol acetates discussed in the previous section werec arried out in MeCN-acetone mixtures to maximise the solubilities of the reactionp artners, solubility was not an issue at the lower concentrationsu sed in UV/Vis spectrophotometry and NMR spectroscopy studies. Hence, kinetics studies were conducted in MeCN only.E xtinction coefficients were determined for 17, 21-a and 21-b (see Supporting Infor-mationS ection2 .4.4) with the aim of enabling us to determine 21-a:21-b ratios in our reactionm ixtures spectrophotometrically.A lthough there was ad ifferenceb etween the extinction coefficients for 21-a and 21-b,w ew ere unable to reliably extract ratios.S ince Selectfluor 7 is not chromophoric,w ew ere able to monitor the decaysi na bsorbance of progesterone enol acetate 17 at 236 nm with an excesso fS electfluor 7. However,t he UV/Vis spectra of 11 a, 11 b and 12 contain absorbance bands between 200-350 nm and it was not possible to selectively monitort he steroidb and at 236 nm. Therefore, the kinetics experimentsi nvolving reagents 11 a, 11 b and 12 were conducted with excess 17 by monitoring the NÀFr eagent absorbance at l max = 288 nm (for 11 a and 11 b)a nd l max = 320 nm (for 12).
Ar epresentative example is shown in Figure4 for the fluorination of progesterone enol acetate 17 by Selectfluor 7.  Clean exponential decays of absorbance of the nucleophile were observed in all runs, and the first-order rate constants k obs were obtained from the fitting of plots of absorbance versus time (Figure 4a). The k obs values were plotted against NÀFreagent concentration and alinear (i.e.,first order) correla-tion was observed (Figure 4b). The directd ependence upon NÀFr eagent concentration demonstrates rate-limiting fluorination and thus the slope of this graph gave the second-order rate constant, k 2 [M À1 s À1 ], according to the second-order rate Equation (1): Kinetics studies were conducted on the fluorination of steroid enol acetates 18, 19 and 20 using similarp rocedures and the k 2 values are reported in Ta ble 2, as well as those of 17.A ll spectrar elatingt ok inetics studies on fluorination of 17-20 are included in the Supporting Information Section 2.4. Relative rate constants (k rel )w ere determined using Equation (2), with Selectfluor as the reference electrophile, thus enabling a comparison of reactivities ( Table 2). The reactivity trendso ft he NÀFf luorinating agentsm atch those we previously reported for fluorinations of enolic 1,3-dicarbonyl systems 26, [21] which reinforces the predictiven ature of our reactivity scale. More detailedc omparisons of reactivities between the two nucleophile systemsa re included in the Supporting Information Section 2.5.
Selectfluor 7 shows excellent solubility and good stability in water; [22] additionally,t he use of benign solvents such as water is attractive due to the potential for reducing the environmental impact of the process. Furthermore, we recently showed that the addition of water significantly increased fluorination rate constants of the enol forms of 1,3-dicarbonyl species. [22] We also wondered whether the trapping of the cationic intermediate generated upon fluorinationo f17,w ith aw eak nucleophilic species, could improve product yield by sequestering this reactive intermediate (Scheme 4, PathwayB ). Thus, taking these factorst ogether,w ep erformed kinetics studies using water and MeOH as co-solvents for fluorination of 17.  Thus, water is not as uitable cosolventf or increasing the rate of fluorination of 17,a nd we tentatively attributei ts inhibitory effectst ot he differing solvation requirementso ff luoroenol species in our earlier study [22] and enol ester 17 along their fluorination reaction coordinates. Product analyses of crude reaction mixturesb yL C-MS were also unchanged by the addition of MeOH or water,a nd thus we concludet hat the use of these co-solvents offers no advantage to the fluorination process.
We also attempted to employ fluorine gas to fluorinate progesterone enol acetate 17.E xploratory experiments were carried out in formic acid solution,which is apreferred solventfor the direct fluorination of enolic systems. [37] The crude product mixture contained progesterone 13,6 a-fluoroprogesterone (21-a)a nd 6b-fluoroprogesterone( 21-b)a long with some minor impurities. However,u pon analysis by HPLC, integration of the chromatogram revealed that only half of the crude product mass could be accounted for by these three compounds. Direct fluorination in MeCN yieldedm ixtures of 21-a, 21-b and unreacted progesterone enol acetate 17 as well as other fluorinated side-products, but progesterone 13 was not detected. The selectivity of the direct fluorinations were a:b, 38:62, whichi sasimilar ratio to that obtained using Selectfluor 7.U ltimately, we found batch-based direct fluorinations to be ineffective, however,f low-baseds ystems [38][39][40] may offer improved performance. Further details of direct fluorination experiments are contained in SupportingI nformation Section 2.3.

Comparison of nucleophilicities
The relative nucleophilicities, k 0 rel ,o fe nol acetates 17-20,e xpressed, defined by Equation (3) (see Ta ble 4) were determined using the second-order rate constants, k 2 (from Table 2). Unsurprisingly,t he reactivity differencesa re small across the four compounds. Progesterone enol acetate 17 and testosterone enol diacetate 18 have, on average, very similarr eactivities. Cholestenone enol acetate 19 is, on average, 1.4-fold more reactive than 17,a nd hydrocortisone enol tetraacetate 20 is 2.3-   The nucleophilic reactivity of progesteronee nol acetate 17 was compared with those of the 1,3-dicarbonyl derivatives 26 a-f using the second-order rate constants, k 2 ,f or fluorination of these substrates by Selectfluor andN FSI (from ref. [21]). These two NÀFr eagents were selected for this comparisons ince they show markedly different reactivities, as well as havingt he most extensive datasetsf or fluorination kinetics. Equation (4) defines k 00 rel .T he reactivities of the nucleophiles span 5o rders of magnitude ( Figure 6) and enol ester 17 is one order of magnitude more reactive than enol 26 a.M ore detailed comparisons of reactivitiesa re included in the Supporting Information Section2.5.

Kinetics of epimerisation
The a-isomers of 6-fluorosteroids are generally desired because they display higherl evels of biological activity. [16,17] However, the low stereoselectivities of the fluorination reactions dis-cussede arlier result in the formation of the 6b-isomer as the major product (Table1). We carriedo ut studies on the rates of epimerisation of 6b-fluoroprogesterone( 21-b)t o6 a-fluoroprogesterone (21-a)b yH Cl in AcOH, using appropriately diluted solutionso facommerciallya vailabler eagent. Reactions were monitored directly by quantitative time-arrayed" in-magnet" 19 FNMR spectroscopy with four differentc oncentrationso fH Cl in acetic acid (0.25-1.00 m). Ar epresentative example is shown in Figure 7f or an epimerisationr eactionc onducted with 0.50 m HCl in acetic acid. The triplet of doubletsa td = À165.60 to À165.90 ppm, corresponding to 21-b,d ecreased in intensity over time, whereas the doublet of doublet of doubletsa td = À183.00t oÀ183.16 ppm associated with 21-a increased in intensity.A dditional peaks were present at d = À165.56 ppm, which overlapped with parto ft he adjacent 21-b signals.S imilarly,s mallp eaks appeared over time adjacent to the 21-a signals.
The reaction profiles for each species in the epimerisation mixture are shown in Figure 8. Due to the overlapb etween the peaks, partial signal integration was employed. The integrals corresponding to the growth of 21-a (black data points) were fitted to an exponential rise function for all concentrations of HCl. Ap lot of k obs versus HCl concentration showed a linear relationship (Figure 9). The integrals of the small signals   The small signals at d = À165.56 and À183.20 ppm in Figure 7a re likelyt ob ep roduct-related species, due to their similarityi nc hemical shift and coupling patterns.W hen an authentic sample of 21-a was incubated in as olutiono f0 .50 m HCl in AcOH for 45 min, the proton-coupled 19 FNMR spectrum of the solution showedt he presenceo fs ignals corresponding to 21-a (at d = À183.03 to À183.15 ppm) as well as smaller adjacent signals due to the intermediate species( at d = À183.10 to À183.20 ppm). The proton-decoupled 19 FNMR spectrum confirmed that two differents pecies were present (see Sup-porting Information Section2 .6). However,w hen 21-a was incubated in AcOH alone, signals corresponding to only one species, 21-a,w ere observed. These results suggesta cid-catalysed formation of intermediates such as hemiacylals 29-b and 29-a (Scheme5,b lue pathways) or enol esters, however,w eh ave no direct evidence of their structures. The change in epimer preference under the reactionc onditions from 21-b to 21-a probably derives from differing solvent-product interactions. These interactions are likelyt ob em arkedly different in AcOH in comparison to the MeCN solvent that was applied for fluorinations.

Conclusions
The kinetics of fluorination of progesterone enol acetate 17 using seven NÀFr eagents were studied. The methodo fa nalysis was tuned to the reactivity of the system:l ess powerful electrophiles weres tudied by 1 HNMR spectroscopy while more reactive reagents were studied using UV/Vis spectrophotometry. Relative rate constantsw ere determined from absolute rate constants, and they correlate well with our recently reported reactivity scale. [21] These resultsh ighlightt he successful predictive powero fo ur scale towards ad ifferent class of carbon nucleophiles. Activation parameters were determined for the fluorination of progesteronee nol acetate 17 by Selectfluor 7 and diCl-NFPy TfO À 11 a.T he moderately negative valueso fDS°are consistentw ith those from our previous studies on fluorination of enolic 1,3-dicarbonyl systems [21,22] and recent studies by Nelson et al. [19] on closely-related tetralone systems. Kineticss tudies on the fluorination of testosterone enol diacetate 18,c holestenonee nol acetate 19 and hydrocortisone enol tetraacetate 20 were conducted.T he substituenta tt he C-17 positionh as as mall but measurable effect upon the rate of fluorination.
The epimerisation of 6b-fluoroprogesterone 21-b to 6a-fluoroprogesterone 21-a with increasingc oncentrationso fH Cl in acetic acid proved to be more rapid. Additional signals in the 19 FNMR spectra of the product mixtures also gave evidence for the formation of intermediates in an acid catalyst-dependent manner.
Overall, we have delivered ac learer understanding of the kinetics of fluorination of steroidal systemsa nd the subsequent epimerisation of fluorosteroids. These resultsh ighlightt he op- Figure 8. Reactionprofiles for the species present in the epimerisation of 21-b (60 mm)t o21-a with 0.50 m HCl in AcOH. Red: 21-b,blue: b-isomer intermediate 29-b,green: a-isomer intermediate 29-a,b lack: 21-a.D ue to significant overlapo fpeaks present, partial signal integrationwas used for each species, therefore, the integral intensity of 21-a (black) at the end of the reaction is higher than that of 21-b (red) at the start. Scheme5.Apotential mechanism for epimerisation of 21-b to 21-a.Another pathway,hemiacylal formation via AcOH (in blue), is proposed.