Reducing the Lipophilicity of Perfluoroalkyl Groups by CF2-F/CF2-Me or CF3/CH3 Exchange.

Fluorination is commonly employed to optimize bioactivity and pharmaco-kinetic properties of drug candidates. Aliphatic fluorination often reduces the lipophilicity (log P), but polyfluoroalkylation typically increases lipophilicity. Hence, identification of polyfluorinated motifs that nonetheless lead to similar or even reduced lipophilicities is of interest to expand the arsenal of medicinal chemistry tools in tackling properties such as compound metabolic stability or off-target selectivity. We show that changing a CF3-group of a perfluoroalkyl chain to a methyl group leads to a drastic reduction in lipophilicity. We also show that changing a C-F bond of a trifluoromethyl group, including when incorporated as part of a perfluoroalkyl group, to a C-Me group, leads to a reduction in log P, despite the resulting chain elongation. The observed lipophilicity trends were identified in fluorinated alkanol models and reproduced when incorporated in analogues of a drug candidate, and the metabolic stability of these motifs was demonstrated.

The observed lipophilicity trends were identified in perfluorinated alkanol models and reproduced when incorporated in analogues of a drug candidate, and the metabolic stability of these motifs was demonstrated.

Introduction.
The introduction of fluorine is a frequently used medicinal chemistry strategy to optimize the properties of bioactive compounds, including stability against oxidation and acid-mediated degradation processes, pKa, conformations, hydrogen bonding capacity, lipophilicity and off target selectivity. [1][2][3][4] In addition, many synthetic bioisosteric groups have been designed based on the unique electronic and steric properties of fluorine. [5][6][7] The vast majority of cases involve the introduction of fluorine, CF2-and CF3 groups, as well as a range of heteroatom containing fluoroalkyl groups, [8][9] with other types of groups such as pentafluorosulfanyl (SF5) rapidly becoming more popular. 7,10 There are very few polyfluoroalkyl containing drugs on the market (Chart 1). 11 Examples include the herbicide flupoxam 12 and the potent estrogen receptor antagonist fulvestrant. [13][14] Analogues of fulvestrant with a 1,4-perfluorobutylidene group as part of a C11-sulfoxide containing side chain have been reported. 15 Vilaprisan is a highly potent selective progesterone receptor modulator that was found to markedly reduce the growth of human leiomyoma tissue in a preclinical model of uterine fibroids, and is currently in phase II clinical trials for this indication. [16][17] The Vitamin D3 analogue pefcalcitol is currently being investigated as a more potent and more selective anti-psoriatic drug candidate. 18 Chart 1. Perfluoroalkyl containing herbicide and anticancer drug, and drug candidates in clinical trials. context-dependence is also valid for higher polyfluorination: while heptafluorination of 1-butanol leads to a significant lipophilicity increase (cf D1®D4), a logP decrease for a butyl to heptafluorobutyl ester modification has been reported. 33 A detailed structural understanding of how fluorination affects molecular lipophilicity is of great importance. It is now well-recognized that efficient drug development consists of simultaneous potency and physical property optimization, with lipophilicity being one of the important properties, as opposed to a dominating focus on the former. 34 Lipophilicity affects solubility and ADME properties, which determine to a large degree dosing, and thus toxicity effects. New ligand efficiency measures which take both elements into account, such as LLE (lipophilic ligand efficiency), have been proposed as a superior way to guide compounds through a drug development process. [35][36][37] In this context, the development of lipophilicity-reducing motifs is of great relevance.
Our group has an interest in investigating the influence of fluorination on lipophilicity of aliphatic substrates, and we have developed a 19 F NMR based protocol for a convenient and accurate measurement of the octanol-water partition coefficients of such non-UV active compounds. 31  was noticed that alkanols with a terminal -CF2CH3 group (eg B4, E4, Figure 1), have lower logP values compared to the nonfluorinated alcohols B1 and E1. Muller had described the same observation for propyl-and 2,2-difluoropropyl ethers. 28 As shown in Figure 1, -CF2CH3 containing alcohols have lower logP values than their trifluoroethyl (-CH2CF3) equivalents (eg B2, E2), and considerably lower values than the pentafluoroethyl (-CF2CF3) equivalents (eg B3, E3). In the latter case, the opposing C-F dipoles of the pentafluoroethyl CF2 and CF3 moieties, combined with the larger hydrophobic surface, contribute to this difference. It was also noted that these -CF2CH3 alkanols even have a lower logP value than the next lower homologue trifluorinated alcohols (compare B4 with A2, and E4 with D2). A similar observation has been made by O'Hagan within the aryl fluoroalkyl sulfide structure: lipophilicities of ArSCF2CH3 are lower than these of the corresponding ArSCF3 analogues. 25 Using alkanol structures as models for alkyl chains, we wished to explore whether significant attenuation of the lipophilicity of longer perfluoroalkyl chains was possible by CF3/CH3 exchange, and whether that change would also be significant enough for its logP to drop below that of the next lower perfluoroalkanol homologue. Hence, compounds D5 and E6 were targeted ( Figure 2). Their clogP values, obtained by three different calculation methods (see Supporting Information) were all lower than the values of D4/E5, but only one method, which underestimated the value of the perfluoroalkyl derivatives D4/E5, predicted similar or lower values compared to the nonfluorinated alcohols. Hence, experimental determination of the logP values of D5/E6 was justified. In addition, two more difluorinated examples C4 and D6 were synthesized to investigate a wider scope of the lipophilicity change observations as discussed above. Finally, these motifs were also evaluated when incorporated in a drug-like scaffold, via logD7.4 determination of a panel of analogues of evenamide 1a, a schizophrenia drug currently in phase II clinical trials. 38 We report a number of surprising outcomes regarding the magnitude of the measured logP (and logD) changes, which will render these motifs of interest in medicinal chemistry.

Results.
Chemistry. The heptafluorinated D4 and nonafluorinated E5 were commercially available. The synthesis of the other polyfluorinated targets was achieved using a fluorinated building block strategy.
Hence, commercially available 1,w-diols 2 and 3 (Scheme 1a) were monobenzoylated to give 4 and 5, and the remaining alcohol group was removed via a radical deoxygenation reaction. Deprotection of the resulting 8 and 9 led to the targets D5 and E6. Scheme 1. Synthesis of the model compounds.  Targets D6 and C4 (Scheme 1b,c) were obtained from commercially available 10 and 13 by an alcohol protection, carbonyl deoxofluorination, and deprotection sequence. In the latter case, the deoxofluorination was accompanied by an elimination side reaction, leading to an 85:15 mixture of 15 and 16. Even after deprotection, the alkene and difluoro compound could not be separated. However, this was of no consequence for the lipophilicity determination (see below).
All the synthesized fluorinated alcohols proved to be rather volatile, hampering isolation. Hence, Et2O, CH2Cl2 or n-pentane were used as solvents for the final deprotection reaction, or as part of the eluent mixture used for the final chromatographic purification. Removal of these solvents was typically performed using a controlled pressure pump which was set to 700-750 mbar with a ~30 °C water bath until the volatile alcohols had reached ~80wt% ( 1 H NMR analysis). We found that compound losses were minimized by allowing the final amounts of solvent to be removed by slow evaporation in the open atmosphere.
The evenamide analogues syntheses started from known 18 (Scheme 2). 39 Deprotonation of the BocNH group followed by nucleophilic substitution with N,N-dimethyl chloroacetamide led to 19. Protecting group hydrogenolysis to give 20 then allowed introduction of the alkyl chains a-h. Ether synthesis via conversion of the various alcohol derivatives to the corresponding tosylates 21 followed by reaction of the deprotonated phenolate only worked well when there was no α-CF2 group present. For example, reaction of 20 with base and the tosylate 21d, derived from C8, did not lead to the expected SN2 reaction, but to tosyl exchange, leading to 24 (Scheme 3).

Scheme 3. Side reaction with perfluoroalkylmethyl tosylates.
It is well-known that leaving groups adjacent to perfluoroalkyl groups are very unreactive towards SN2 reactions, even when present on a primary CH2-group, due to an electronic deactivation effect and, certainly with anionic nucleophiles, a destabilizing repulsion with the fluorine atoms. [40][41][42] This results in slow reactions and, as observed here, S-O instead of C-O cleavage. 43 This can be mitigated by the use of a more reactive triflate leaving group. [44][45] Interestingly, an electrophile with a heptafluoropropyl substituent has a significantly reduced reaction rate compared to when a trifluoromethyl substituent is present. 44 Hence, the process was optimized starting from alcohol D4, and the best conditions were found when the reaction mixture obtained after addition of Tf2O and 2,6-lutidine in dichloromethane had Cs2CO3 added after 1 h, directly followed by a solution of 20 in DMF. Following this protocol, D5, E5, and E6 were also introduced leading to the collection of evenamide 1a and seven analogues 1b-h in good yields.  Lipophilicity.
The direct determination of logP values by measuring the octanol/water partition coefficients P is typically efficiently accomplished by concentration measurements through UV-spectroscopy, which is a standard procedure and often high-throughput in most companies. However, for compounds that do not contain a chromophore, this is unsuitable, and a number of NMR-based methodologies can be used for such cases. [46][47][48] Our group has developed a 19 F NMR based method for shake-flask logP measurement of fluorinated derivatives. 31 The use of an internal standard as a mixture with the compound of interest conveniently obviates the need for accurate weight/partition volume/sample volume measurements. The NMR sensitivity limitation translates into a logP window of ~ ±2-2.5 before NMR-experiment times become impractically long, although the use of a cryoprobe extends the logP window to ~ ±3-3.5. A very useful practical advantage is that the logP of impure compounds (eg C4 and 17, see Scheme 1) can be easily measured provided the fluorine chemical shift values of the desired compound and that of any impurities are different. This method was used to measure the logP of all compounds below, unless indicated.
The results for the polyfluoroalkyl CH3/CF3 exchange are shown in Figure 3. As expected the measured logP of E5 was significantly higher than that of 1-pentanol E1. For both the nonafluorobutyl and heptafluoropropyl group containing compounds E5 and D4, changing the CF3-group into a CH3 group resulted in a dramatic decrease in logP. Both perfluoroalkyl groups have the same distance to the alcohol group, hence the alcohol functional group is not expected to play a part in this logP modulation. Very unexpectedly, the logP of the internally tetrafluorinated butanol D5 is almost identical to that of the parent butanol. Strikingly, this is also the case for the internally hexafluorinated pentanol E6 vs pentanol E1. It is also interesting to contrast these observations with a CF3®CH3 exchange applied to the trifluorinated D2 and E2: this has either no effect (eg D2®D1) or even shows a logP increase (eg E2®E1). Hence, these results suggest that polyfluorination of alkyl groups need not lead to a punishing increase in lipophilicity, if the terminus remains nonfluorinated. Next, the lipophilicity of the two novel difluorinated alkanols C4 and D6 was determined ( Figure 4). For comparison purposes the known data for B4 and E4 are also included. Clearly the same large lipophilicity decrease is observed upon changing a pentafluoroethyl group for a CH3CF2-group.  Finally, the lipophilicities of Me-CF2-R with the lower homologues CF3-R were compared ( Figure 5).
As can be seen from the logP values of the nonfluorinated alcohols, extending the alkyl chain leads to an increase in logP. However, in all cases investigated, regardless of the length of the perfluoroalkyl chain, extending a CF3-terminated alkyl chain by one carbon through replacing a single fluorine by a methyl group, leading to a terminal Me-CF2-group, results in a lipophilicity reduction. Hence, these results are consistent with the work of O'Hagan regarding the lower lipophilicities of ArSCF2CH3 compared to the corresponding ArSCF3 analogues. 25 This type of motif change was further investigated by the synthesis of analogues of evenamide, an antipsychotic drug candidate now in Phase IIa clinical trials, not only to establish whether the logP differences will be maintained when incorporated in a different structural context, but also to evaluate metabolic stabilities of these fluorinated moieties. In the event, the linear butyl chain of evenamide was replaced by fluorinated butyl chains b-d and pentyl chains e-g (cf Scheme 2), leading to the evenamide analogues 1a-g (Table 1). These analogues were subjected to logD7.4, solubility determination, metabolic stability studies and hERG activity. hERG (human ether-a-go-go-related gene) potassium channels are essential for normal electrical activity and function of the heart. Arrhythmia can be induced by a blockage of hERG channels by a diverse set of drugs, with this side effect being a common reason for drug failure in preclinical safety trials.
Therefore minimization of hERG channel blocking activity is an important aspect of drug discovery.  The logD values of the nonfluorinated evenamide 1a is +1.8. As expected, its homologue 1e shows an increased value +2.3 owing to the longer alkyl chain. Difluorination at the penultimate position led to a significant logD decrease (0.6-0.7 logD units, 1b, 1f), and perfluorination to a significant logD increase (1d, 1h). Pleasingly, changing the CF3 group of the perfluoroalkylmethyl groups in 1d and 1h to the corresponding methyl groups (1c and 1g respectively) leads to a reduction in logD to a value that is similar and even slightly lower than that of the nonfluorinated chains (compare 1c with 1a, and 1g with 1e). In addition, the evenamide analogue with the hexafluoropentyl chain has a lower logD compared to that with a heptafluorinated butyl chain (compare 1g with 1d). Hence, the logP trends that were obtained upon introducing these fluorinated motifs in the butanol and pentanol models are fully replicated in the logD7.4 trends of a pharmaceutically relevant drug candidate when they are introduced as part of an aromatic butoxy/pentoxy chain.
Importantly, the introduction of the fluorine atoms did not lead to a decrease in aqueous solubility; on the contrary, in all cases the solubility increased, even for the analogues with increased logD values. Human plasma protein binding correlates with the lipophilicity. For example, 1e and its hexafluorinated analogue 1g are isolipophilic and have the same levels of human plasma binding. Metabolic stability studies (human Cl microsomes and rat hepatocytes) showed the increased stability of the fluorinated derivatives towards oxidative degradation. Hence, a terminal methyl group adjacent to a perfluoroalkylidene moiety is not metabolically labile, which was presumed to be due to the electronegativity of the adjacent fluorine atoms.
With respect to percentage inhibition of the hERG receptor at a concentration of 10 µM, this correlated broadly with lipophilicity in that only the most lipophilic heptafluorinated 1d and nonafluorinated 1h analogues showed increased inhibition relative to the parent compounds (compare 1d with 1a, and 1h with 1e). All of the internally fluorinated compounds (1b, 1c, 1f, 1g) showed similar or lower levels of inhibition relative to the parent compounds.

Discussion
Having observed a significant logP decrease for a perfluoroalkyl CF3/CH3 exchange, leading to internally polyfluorinated butoxy and pentoxy groups having similar lipophilicities as their nonfluorinated counterparts, we were keen to understand the underlying factors in terms of conformation and polarity contributing to this.
Polarity effects can be gauged by estimation of the overall molecular dipole moments, which require  Table 2). These were always lower in the octanol phase. This trend is expected as it illustrates that a polar medium stabilizes polar conformations. While this difference was typically <0.2 D, the exception was D5, for which the dipole moment was much larger (about 1.0 D) in the water phase compared to that in the octanol phase. This was also the compound with the most biased conformational profile, its two most abundant conformers in octanol having a much lower dipole moment than those in water ( Figure 6).
Hence, D5 adopts a very different conformational profile in water compared to octanol, with significantly populated but close in energy conformers possessing very different dipole moments, a behavior which has been referred to by Muller as a "lipophilicity chameleon". 28 Octanol phase (24.6%, 2.14 D) (19.6%, 2.16 D) Water phase : A qualitative comparison of the calculated dipole moments is instructive ( Table 2). All fluorinated derivatives have a higher dipole moment compared to the parent nonfluorinated alkanols. The hexafluorinated E6 has a larger weighted dipole moment than the nonfluorinated E5, both in the water and octanol phases. When comparing tetrafluorinated D5 with heptafluorinated D4, this is only true in the water phase, but not in the octanol phase, which originates, as explained before, by its ability to adopt very apolar conformers ( Figure 6). Indeed, while the C-F dipoles in these conformers are opposed and hence cancel each other out, this is not possible for the hexafluorinated derivative, explaining its higher averaged dipole moment. Nevertheless, the low-polarity conformers of E6 in the octanol phase do not lead to a high logP, as may be expected. Hence, the stabilization of the polar conformations in water must dominate the partitioning equilibrium.
In addition, the C-H bonds of the resulting CH3 group are also strongly polarized by the electron withdrawing effect of the perfluoroalkylidene group. This was apparent from their experimental chemical shift, showing a deshielding, as well as by the partial atomic charges calculated at the SMD/MN15/augcc-pVTZ//MN15/cc-pVTZ level of theory using the Natural Population Analysis methodology (Table 3).
Owing to the fluorine inductive effect, an increase of the hydrogen positive charge is found as more internal fluorines are introduced. The values, which are the mean values over the three hydrogen atoms weighted upon the conformation relative populations, are just slightly lower in octanol than in water. They are very similar between conformations. Hence, a CF3/CH3 exchange introduces a more polar moiety than perhaps expected, further leading to a downwards effect on lipophilicity. Interestingly, The Polar Surface Area was calculated to be very similar for all compounds, both in the water and in the octanol phase. Table 2. Calculated dipole moments a and polar surface areas b in the octanol and water phase and measured c lipophilicities.  The other main contributing factor involves the reduction in hydrophobic surface which results from replacing a CF3-group with a CH3-group. Comparing E5 with E6 (Table 2), the observed logP change of ~1 unit is consistent with Muller's predictive model, which assigns a 0.3 logP unit per H/F exchange, together with a contribution for polarity change (in this case ~0.1-0.2 µC-F).
The similarity in lipophilicity between the nonfluorinated and internally polyfluorinated alkanols could therefore be rationalized by a compensation of the increased dipole of the latter with the increase in hydrophobic surface due to fluorine introduction.

Conclusions
This work concerns an investigation on lipophilicity modulation upon changing the CF3-group of a perfluoroalkyl moiety into a Me-group. We have confirmed an observation that changing a CF3CF2-group for a MeCF2-group leads to a drastic reduction in lipophilicity. A key finding was that this is also the case when a CF3(CF2)n-group (n>1) is converted to a Me(CF2)n-group. Remarkably, tetrafluorinated MeCF2CF2CH2OH and hexafluorinated MeCF2CF2CF2CH2OH have the same or very similar lipophilicities compared with the nonfluorinated BuOH and PentOH. In addition, we have shown that extending a CF3-group into a MeCF2-group, or a CF3(CF2)n-group into a MeCF2(CF2)n-group, despite resulting in a longer alkyl chain, still leads to a reduction in lipophilicity (in contrast to nonfluorinated nalkanols). These lipophilicity changes were also investigated when various fluorinated butyl and pentyl ethers were incorporated into evenamide, a schizophrenia drug candidate currently in Phase II clinical trials. Pleasingly, all the lipophilicity trends observed for the alkanol model compounds were confirmed.
In addition, these evenamide analogues have slightly improved aqueous solubilities, and enhanced metabolic stabilities compared to the nonfluorinated analogues, showing that a MeCF2(CF2)n-group is not metabolically labile. Conformational analysis shows that in general, the conformational profile in the water and octanol phases is different, with a higher weighted dipole moment in the more polar water phase. This was especially pronounced for the internally tetrafluorinated butanol, making this a new example of a 'lipophilicity chameleon'. The internally fluorinated compounds have an increased dipole moment compared to the perfluorinated alkanols, which in turn was much increased compared to the nonfluorinated species.
With lipophilicity control typically being a standard concern throughout the drug optimization process, the results described herein will be of great interest to medicinal chemists as it demonstrates the ability to introduce polyfluoroalkylation without logP penalization.

Determination of logP
Lipophilicities of the fluorinated alkanols were determined using a previously published protocol: 31  to obtain the logP value of the compound. The logP measurement of each compound was run in triplicate.
LogP values of non-fluorinated compounds were taken from the literature.

Determination of logD7.4
LogD7.4 measurements were made using a shake-flask method where the extent of partitioning between pH 7.4 buffer and octanol was measured. Compounds were dissolved in a known volume buffer, and following the addition of a known amount of octanol, the solutions were shaken for 30 min.

Determination of solubility
Assessments of aqueous solubility were made using a shake-flask approach that uses 10 mM DMSO solutions which are supplied from AstraZeneca Compound Managements Liquid Store and is a high throughput method. The dried compounds were equilibrated in an aqueous phosphate buffer (pH 7.4) for 24 hours at 25 °C, the portion with the dissolved compound was then separated from the remains. The solutions were analyzed and quantified using UPLC/MS/MS, QC-samples were incorporated in each assay-run to ensure the quality of the assay.

Determination of human plasma protein binding
The automated equilibrium dialysis assay in human plasma used a RED (Rapid Equilibrium Dialysis) Device and sample handling. After dialysis for 18 h, plasma and buffer samples were prepared for analysis by liquid chromatography and mass spectrometry. Samples were generally tested in singlicates and quantified by LC/MSMS by using a 7-point calibration curve in plasma. The compounds were pooled together in plasma pools up to 10 compounds. Three reference compounds (propranolol, metoprolol and warfarin) were used in each run. Warfarin was used as a control in each pool and propranolol and metoprolol were placed randomly in each run.

Determination of human Microsome Metabolism Clint
Test compounds were incubated with human liver microsomes and NADPH in up to three 96-well plates

The hERG Binding Assay
The hERG single shot assay was run on an IonWorks Quattro device. Each compound was tested at 33

Calculations
The theoretical calculations were carried out with the Gaussian16 program. 49 The conformational analysis of the various compounds investigated was performed with the MN15 functional 50  were taken into account using the SMD solvation continuum model. 54 The vibrational spectrum of each optimized conformer was computed to confirm its nature of true minimum and to obtain free energies.
The relative populations, pi, of the various conformers were evaluated at 298K from the computed free energies through a Boltzmann distribution (Eq. (1)). (1) The various theoretical descriptors (molecular dipole moments, partial atomic charges) computed for each conformer were weighted according to these populations.
The To a solution of the alcohol (1 equiv) in CH2Cl2 was added triflic anhydride (1.05 equiv) and 2,6-lutidine (1.1 equiv). Once complete consumption of the alcohol was shown by 19 F NMR analysis (roughly 1 h), Cs2CO3 (3 equiv) was added to the reaction mixture followed by a solution of 20 (1.2 equiv) in DMF.
After 16 h the solvent was removed in vacuo. The residue was dissolved in water (similar volume as DMF) and extracted with EtOAc. The combined organic layers were dried over Na2SO4 and concentrated.
The crude mixture was purified by column chromatography (1:1 EtOAc/heptane) to afford the ether 23.

General procedure D for tosylate formation
To a solution of alcohol (1 equiv) in CH2Cl2 was added Et3N (1.1 equiv), DMAP (0.05-0.1 equiv) and tosyl chloride (1.1 equiv). After 1 h the reaction was quenched with 2M aq. HCl and the layers were separated. The organic layer was washed with aq. sat. NaHCO3, dried over Na2SO4 and concentrated to afford the tosylate 21, which was used immediately without further purification.
The reaction mixture was allowed to warm to rt and then heated at 40 °C. After 48 h the reaction mixture was cooled to 0 °C and quenched with sat. aq. NaHCO3 till pH 7. The layers were then separated and the aqueous phase was extracted with CH2Cl2 (2´150 mL), dried over MgSO4 and concentrated. The crude was purified by column chromatography  To a suspension of sodium hydride 60% in mineral oil (0.57 g, 2 equiv) in DMF (35 mL)