Volume 2021, Issue 46 p. 6375-6382
Full Paper
Open Access

Some Surprising Transformations of Colchicone and Other Colchicine-Derived Tropolones

Dr. Andreas Stein

Dr. Andreas Stein

Department of Chemistry, University of Cologne, Greinstrasse 4, 50939 Koeln, Germany

Search for more papers by this author
Dr. Persefoni Hilken née Thomopoulou

Dr. Persefoni Hilken née Thomopoulou

Department of Chemistry, University of Cologne, Greinstrasse 4, 50939 Koeln, Germany

Search for more papers by this author
Tim Schulte

Tim Schulte

Department of Chemistry, University of Cologne, Greinstrasse 4, 50939 Koeln, Germany

Search for more papers by this author
Dr. Jörg Neudörfl

Dr. Jörg Neudörfl

Department of Chemistry, University of Cologne, Greinstrasse 4, 50939 Koeln, Germany

Search for more papers by this author
Dr. Martin Breugst

Corresponding Author

Dr. Martin Breugst

Department of Chemistry, University of Cologne, Greinstrasse 4, 50939 Koeln, Germany

Search for more papers by this author
Prof. Dr. Hans-Günther Schmalz

Corresponding Author

Prof. Dr. Hans-Günther Schmalz

Department of Chemistry, University of Cologne, Greinstrasse 4, 50939 Koeln, Germany

Search for more papers by this author
First published: 16 September 2021

In memory of Klaus Hafner

Graphical Abstract

Who would have predicted the reactions outcome? Derivatives of the prominent alkaloid colchicine were found to react in a very surprising fashion with common reagents to afford unexpected products. The unique reactivity of tropolone ethers and their subtle interplay with adjacent functional groups have to be taken into account to solve the mechanistic puzzles.

Abstract

The alkaloid colchicine represents a prominent lead structure for the development of tubulin-binding chemotherapeutics. In the course of explorative research into semi-synthetic colchicinoids we stumbled upon a number of unexpected and mechanistically surprising transformations which reflect the very particular reactivity of the tropolone ether unit and its interplay with adjacent functional groups. Examples are the selective C-8-methylation of colchicone upon reaction with the Wittig reagent triphenylphosphonium methylide, the AlMe3-mediated isomerization of colchicone to a tetracyclic lactone, and the cyanide-induced reaction of the colchicol-derived mesylate to a rearranged and ring B-defunctionalized cyano-allo-colchicinoid. Finally, the fragmenting rearrangement of a tetracyclic colchicinoid with BF3⋅OEt2 in the presence of vinyl-MgBr afforded a tropolonyl difluoroborate under concomitant transfer of an O-methyl group from ring C to ring A.

Introduction

Colchicine (1) belongs to the most prominent organic natural products due to both its fascinating structure and its strong biological activity as a tubulin-binding anti-mitotic agent.1 While it is clinically used as a drug against gout and Mediterranean fever, its high toxicity prevents its application in cancer chemotherapy. However, the search for less toxic derivatives or more selective colchicine analogs represents a highly active field of research.2 Since the isolation of pure colchicine in 18203 it took more than a century until the correct structure was finally suggested by Dewar in 1945 after unraveling the structure and aromaticity of tropones.4 The unusual structure of colchicine (1) with its tricyclic ring system comprising two 7-membered rings, a chiral axis connecting the two aromatic rings A and C, and, in particular, the tropolone ether substructure has challenged synthetic chemists during the past 70 years. As a result, many total syntheses have been elaborated,5 and, additionally, some unusual transformations have been reported, such as the colchicine-allocolchicine rearrangement,6 for which a satisfying mechanistic explanation was published only decades later. [7]

In the course of our research into the synthesis of new bioactive colchicinoids8 we recently exploited the Demjanov rearrangement of desacetylcolchicine (2) to prepare the alcohol 3,9 which in turn was efficiently dehydrated to the exo-methylene B-nor-colchicinoid 4 (Scheme 1). This compound (also named PT100) exhibited outstanding properties as a cytotoxic agent which is able to selectively induce apoptosis in multidrug resistant human cancer cells via the intrinsic pathway.10 This prompted us to further explore the chemistry of colchicinoids with small carbon substituents at C-7. We here report some selected results of this study which has led to the discovery of a number of surprising and mechanistically non-trivial transformations.

Details are in the caption following the image

Synthesis of the highly active colchicinoid PT100 (4). Reagents and conditions: a) Boc2O, DMAP, NEt3, CH3CN, reflux, 4 h; b) NaOMe, MeOH, RT, 45 min; c) TFA, RT, 1 h; d) NaNO2, AcOH, RT, 5 h; e) PPh3, DtBAD, CHCl3, 0 °C to RT.

Results and Discussion

Aiming at the synthesis of compound 6 (as the higher homolog of 4) we started our investigation with the preparation of colchicone (5) from desacetylcolchicine (2) through imine formation with 4-formyl-1-methyl-pyridinium benzenesulfonate followed by base-induced isomerization and acidic hydrolysis (Scheme 2).11 Interestingly, however, we completely failed to convert 5 into 6 under a variety of established methylenation conditions.12 Only upon reaction of 5 with the classical Wittig reagent Ph3P=CH2 (generated in situ from Ph3PCH3Br with KOtBu in THF) a new product was formed in 20 % yield, which unexpectedly turned out to be 8-methyl-colchicone (7).

Details are in the caption following the image

Synthesis and attempted methylenation of colchicone (5). Reagents and conditions: a) i. 4-formyl-1-methyl-pyridinium benzenesulfonate, CH2Cl2/ DMF (1 : 1); ii. DBU, iii. saturated aqueous oxalic acid, 48 h, RT, 74 %; b) Ph3PCH3Br, KOtBu, THF, 2 h, RT, 20 %.

The X-ray crystallographic analysis of both colchicone (5) and its 8-methyl derivative 7 (Figure 1) confirmed the constitutional assignments (made initially by NMR) and did not reveal any unusual features except the almost perpendicular orientation of the C-7 carbonyl group with respect to the plane defined by the tropolone ring C (torsion angles of 64.7° and 75.6°, respectively, for compounds 5 and 7). Attempts to increase the yield of 7 by variation of the reaction conditions (base, solvent, temperature, equivalents of ylide, reaction time) were not successful. However, in all cases, about 60 % of the poorly soluble starting material 5 were re-isolated.

Details are in the caption following the image

Structures (ball and stick representation) of 5 (top) and 7 (bottom) in the crystalline state.

To rationalize the surprising formation of compound 7, we assume the Wittig reagent to attack as a nucleophile at C-8 of the tropolone ring of the substrate 5 to form a betaine-type intermediate 8 (Scheme 3). Elimination of PPh3 would then lead to the enol intermediate 9 which could finally tautomerize to the methylated product 7. Noteworthy, no reaction was observed upon treatment of 5 with other Wittig ylides, and no reaction took place upon treatment of 7 with triphenylphosphonium methylide.

Details are in the caption following the image

Proposed mechanism of the formation of compound 7.

A surprising reaction outcome was also observed when colchicone (5) was treated with the Corey-Chaykovsky reagent dimethylsulfoxonium methylide (Scheme 4).13 In this case, the C–8 methylated product 7 was formed again, besides substantial amounts of the allocolchicine-related products 10 a and 10 b (20 % each).

Details are in the caption following the image

Reaction of colchicone 5 with the Corey-Chaykovsky reagent Me2S(O)CH2. Reagents and conditions: Me3S(O)I, NaH, DMSO, RT, 2 h; yields: 20 % 7, 20 % 10 a, and 20 % 10 b

An X-ray crystal structure analysis of 8-methyl-allocolchicone 10 a confirmed the structural assignments also in this case (Figure 2).

Details are in the caption following the image

Structure of 10 a in the crystalline state.

While the formation of ketone 7 can be assumed to follow a similar mechanism as discussed above (Scheme 3, just replacing PPh3 by DMSO) the formation of 10 a and 10 b can be explained by a secondary attack of either methoxide or dimethylsulfoxonium methylide at the tropone carbonyl of 7 to induce the formation of a tetracyclic norcaradiene intermediate (11) through 6π-electrocyclization. Cyclopropane opening under methoxide elimination then leads under rearomatization to 10 a and 10 b, respectively, as in the classical colchicine-allocolchicine rearrangement7 (Scheme 4). In order to demonstrate the simplicity with which this rearrangement proceeds in the colchicone series, we treated 5 with sodium methoxide to obtain allo-colchicone (12) in good yield (Scheme 5). Noteworthy, this compound now smoothly underwent Wittig methylenation to afford 13, i. e. the allo-isomer of 6.

Details are in the caption following the image

Conversion of colchicone (5) into the allo-colchicine derivative 13. Reagents and conditions: a) NaOMe, MeOH, 1 h 75 °C; b) Ph3PCH3Br (1.3 eq.), KOtBu (1.3 eq,), THF, 2 h RT.

Still puzzled why all attempts to attack the C-7 keto function of 5 with a carbon nucleophile had failed, we tested the possibility to convert 5 into the tertiary alcohol rac-14 by treatment with metal-organic (formal methyl anion) reagents (Scheme 6).

Details are in the caption following the image

Formation of lactone rac-15 from colchicone 5. Reaction conditions: a) MeMgBr, THF, 0 °C to RT, 24 h; b) MeLi, THF, −78 °C to RT, 24 h; c) AlMe3, THF, RT, 20 h.

While complex product mixtures resulted from the reaction of 5 with MeMgBr or MeLi, a major new product was formed upon treatment of 5 with an excess of AlMe3 (5.0 equiv) for 20 h. After workup and chromatographic purification, this compound was obtained in 30 % yield and, much to our surprise, turned out to be the tetracyclic lactone rac-15, another isomer of 5, as confirmed by crystal structure analysis (Figure 3).

Details are in the caption following the image

Structure of rac-15 in the crystalline state.

A mechanistic rationalization of the unexpected formation of lactone rac-15 is sketched in Scheme 7. We assume a coordination of AlMe3 to the tropolone carbonyl oxygen, which carries a strong negative partial charge. As in the case of the colchicine allocolchicine rearrangement, a 6π-electrocyclization would then generate a strained cyclopropanone intermediate 17, which could undergo fragmentation to the ketene 18 from which the observed product could easily arise via the isobenzofuranol intermediate 19, which in turn would finally tautomerize to the lactone rac-15.

Details are in the caption following the image

Proposed mechanism of the formation of lactone rac-15 from colchicone 5 upon treatment with AlMe3 according to Scheme 6.

To shed additional light on the AlMe3-induced isomerization of colchicone (5) to the lactone rac-15 we also performed computational studies at the M06-2X-D3/def2-QZVP/IEFPCM(THF)//M06-L-D3/6-31+G(d,p) level of theory. Calculations with other functionals (B2PLYP, B3LYP, M06, M06-L, ωB97X-D) resulted in similar barriers and are shown in the Supporting Information. As illustrated in Figure 4, the sequence starts via coordination of the Lewis acid to the tropone carbonyl group. This interaction is 21 kJ mol−1 more favorable than the coordination to the other carbonyl group (not shown in Figure 4). The overall interaction energy (−54 kJ mol−1) is probably overestimated, as the calculations rely on free AlMe3 for reasons of simplicity (see the Supporting Information for more details). Next, a 6π-electrocyclic reaction takes place via transition state TS1. The barrier for this step, however, is rather large (ΔG=146 kJ mol−1) and may be responsible for the low yields observed in the experimental studies (Scheme 6). The resulting norcaradiene intermediate 17 is rather unstable and no barriers could be located for the cyclopropanone ring opening to yield ketene 18. Subsequent attack of the enol oxygen onto the activated ketene affords after keto-enol tautomerization the lactone rac-15 in an overall exergonic reaction. Although the overall barrier for this transformation is probably overestimated, the formation of cyclopropenone 17 as an intermediate is also supported by the isolation of 16 as a by-product in the experimental investigations (Scheme 6). This side product is obtained via decarbonylation of cyclopropane 17.

Details are in the caption following the image

Calculated free energy profile and structures with selected bond lengths (in Å) for the AlMe3-mediated conversion of colchicone (5) to the isomeric lactone rac-15.

Returning to our original goal we envisioned that the conversion of colchicone (5) into the ketone 6 could possibly be initiated by introduction of a cyano substituent at C-7 (to give rac-21) followed by reduction and elimination of the terminal functional group. An initial attempt to achieve the conversion of 5 to rac-21 using TosMIC in a Van Leusen reaction14 was not successful, and we therefore decided to try a SN2 type reaction for the cyanation step. For this purpose, we first converted 5 into the mesylate rac-20,15 an intermediate we already had successfully reacted with NaN3 in DMSO to the corresponding C-7 azide in the course of our total synthesis of colchicine (1).5b

Much to our surprise, however, treatment of rac-20 with sodium cyanide in DMSO did not yield any of the expected nitrile rac-21. Instead, the rearranged and ring B-defunctionalized cyano-allo-colchicinoid 22 was obtained as the only isolated product in 44 % yield (Scheme 8). The structure of 22 was unambiguously secured by X-ray crystallography (Figure 5) and displays a saturated B-ring, a helically twisted biaryl axis and a contracted C ring bearing a nitrile and an ester substituent.

Details are in the caption following the image

Surprising outcome of the reaction of the colchicone-derived mesylate rac-19 with cyanide. Reagents and conditions: a) NaBH4, MeOH/CH2Cl2 (1 : 1), −78 °C to RT, 5 h; b) MsCl, NEt3, CH2Cl2, 0 °C, 30 min; c) NaCN (5 equiv.), DMSO, 50 °C, 3 d.

Details are in the caption following the image

Structure of 22 in the crystalline state.

Mechanistically, the formation of 22 under the given conditions is assumed to be initiated by an attack of cyanide at C-10 of the tropolone ring leading to a nucleophilic aromatic ipso substitution of the methoxy group (Scheme 9). The resulting (less electron-rich) intermediate 23 is then attacked by the methoxide anion to induce a (possibly reversible) 6π-electrocyclization. In contrast to the normal allocolchicine rearrangement, the resulting norcaradiene 25, however, can now eliminate the mesylate anion to form 26 under opening of the cyclopropane ring. Finally, aromatization of 26 to 22 can easily occur via enolization of the ester functionality.

Details are in the caption following the image

Proposed mechanism for the unexpected formation of 22 upon reaction of the mesylate rac-20 with NaCN in DMSO (compare Scheme 8).

In the course of the biological investigations of compound 4 as a particularly active colchicinoid,10 we had repeatedly performed the Demjanov rearrangement of deacetylcolchicine (2) to the B ring-contracted product 3 as shown in Scheme 1.9 Noteworthy, this reaction also gives rise to the tetracyclic compound 27 as a by-product (Scheme 10). And after having collected gram amounts of 27 we asked ourselves whether it is possible to re-convert this compound back into a “normal” colchicinoid, for instance, by opening the four-membered ring with a nucleophile under assistance of a Lewis acid to give a mono-O-demethylated colchicinoid of type 28 under re-aromatization of ring A.

Details are in the caption following the image

Possible recycling of 27 formed as a by-product in the Demjanov rearrangement of 2 ? Reagents and conditions: a) NaNO2, AcOH, RT, 5. H.

After testing various reaction conditions (combinations of Lewis acids and nucleophiles) we found that simultaneous treatment of 27 with an excess of boron trifluoride etherate and vinyl magnesium bromide resulted in the formation of the two new products 29 and 30 (Scheme 11). Based on the NMR data and an X-ray crystal structure analysis of 30 (Figure 6) both products could be unambiguously assigned as (zwitterionic) tropolonyl difluoroborates.16 Most surprisingly, compound 30 (obviously resulting from fragmentative elimination) carried three methoxy substituents at ring A instead of the expected two (see above).

Details are in the caption following the image

Formation of the tropolonyl difluoroborates 29 and 30 from 27. Reagents and conditions: a) BF3⋅OEt2 then vinyl-MgBr, THF, 0 °C, 12 h.

Details are in the caption following the image

Structure of 30 in the crystalline state.

As sketched in Scheme 11, we assume the initial formation of 31 as an intermediate from which the main product 29 arises through simple O-demethylation by reaction with a nucleophile. However, in the presence of a Grignard reagent (acting as a base) deprotonation can induce fragmentation of the 4-membered ring to generate the phenolate 32 which itself acts as a (reactive) nucleophile to take over the methyl group from 31 to afford the observed trimethylated product 30. Noteworthy is the fact that the formation of 29 and 30 only occurred when 27 was reacted with the specific combination of boron trifluoride and either vinyl or allyl magnesium bromide. The use of other Grignard reagents (MeMgX, iso-PrMgBr/LiCl), lithium organyls (MeLi, BuLi) or Me2CuLi (also in combination with other Lewis acids such as TiCl4, FeCl3, AlCl3) only led to decomposition or gave no conversion.

Inspired by the structure of 30 we also prepared the corresponding tropolonyl difluoroborates from colchicine (1) and PT100 (4). To obtain higher yields, the starting compounds were first hydrolyzed to the corresponding tropolones 33 (colchiceine) and 35, respectively, prior to the borylation step (Scheme 12). Again, the structure of the colchicine-derived difluoroborate 34 was secured by single crystal X-ray analysis (Figure 7). Like 30, this compound displays a virtually undistorted tropylium ring C fused to the 2,3-dioxaborol. Also noteworthy is the strongly twisted ring B with the NHAc substituent in a Ψ-equatorial position.

Details are in the caption following the image

Synthesis of the tropolonyl difluoroborates 34 and 36. Reagents and conditions: a) 1 M HCl, AcOH, reflux, 3 h; b) BF3⋅Et2O, THF, 0 °C to RT.

Details are in the caption following the image

Structures of 34 (top) and 36 (bottom) in the crystalline state.

Conclusion

In summary, we have documented and discussed a whole series of rather unexpected reactions of colchicone (5), its reduced and mesylated derivative rac-20, and the tetracyclic colchicinoid 27. The discovered transformations, for which we suggest plausible mechanisms, reflect the unique reactivity of tropolone ethers and their subtle interplay with adjacent functional groups. More specifically, the presented results emphasize the pronounced tendency of tropolone ethers to react with nucleophiles and to undergo rearrangements involving norcaradiene intermediates.7, 17 We are convinced that the results reported herein will be of value also for other researchers interested in the chemistry of colchicinoids or other natural or non-natural compounds containing a tropone or a tropolone substructure.17

Experimental Section

Synthesis of 7: To a suspension of 65.0 mg (0.18 mmol, 1.30 eq.) of Ph3PCH3Br in 2.00 mL of anhydrous THF were added 20.0 mg (0.18 mmol, 1.30 eq.) of KOtBu. The mixture was allowed to stir at RT for 45 min before 50.0 mg (0.14 mmol, 1.00 eq.) of ketone 5 were added. After stirring for 2 h at RT n-pentane was added followed by filtration through a short pad of Celite® and rinsing with n-pentane. The filtrate was concentrated under reduced pressure and the crude product purified by column chromatography (CyHex/EtOAc/MeOH 4 : 4 : 1) to afford 10.5 mg (28.3 μmol, 20 %) of 7 as a yellow foam besides 30.0 mg (84.2 μmol, 60 %) of re-isolated starting material 5. The product 7 was recrystallized from CH2Cl2/CyHex to afford yellow crystals suitable for X-ray structure analysis. Mp.: 138 °C. 1H NMR (500 MHz, CDCl3): δ [ppm]=2.29 (s, 3H), 2.66–2.72 (m, 1H), 2.78–2.90 (m, 1H), 3.05–3.11 (m, 1H), 3.12–3.19 (m, 1H), 3.55 (s, 3H), 3.89 (s, 6H), 3.99 (s, 3H), 6.54 (s, 1H), 6.82 (d, 3JH,H=10.5 Hz, 1H), 7.19 (d, 3JH,H=10.5 Hz, 1H). 13C NMR (125 MHz, CDCl3): δ [ppm]=19.3, 30.2, 48.7, 56.0, 56.5, 61.0, 61.2, 107.1, 111.5, 125.8, 131.4, 134.4, 135.2, 138.9, 141.8, 148.0, 151.9, 153.5, 162.7, 179.9, 207.1. HRMS (ESI): calcd. for [M+Na]+ (C21H22NaO6): 393.13086; found: 393.13078. FT-IR (ATR): 1706 (m), 1586 (s), 1254 (s), 1103 (s) [cm−1].

Synthesis of rac-15: To a suspension of 100 mg (0.28 mmol, 1.00 eq.) of ketone 5 in 2.80 mL of anhydrous THF were added at 0 °C dropwise 0.70 mL (101 mg, 1.40 mmol, 5.00 eq.) of a 2.0 M solution of AlMe3 in THF. The mixture was allowed to stir at 0 °C for 30 min before the ice bath was removed and stirring was continued at RT for 24 h while the initial yellow suspension cleared to a bright brownish solution. The mixture was then cooled to 0 °C before aqueous HCl (1.0 M) was slowly and carefully added to destroy the excess of AlMe3. The aqueous layer was extracted with EtOAc (3x), and the combined organic layers dried over MgSO4 and concentrated under reduced pressure. Column chromatography (CyHex/ EtOAc/MeOH, 8/1/1) afforded 30.0 mg (0.084 mmol, 30 %) of lactone rac-15 (RF=0.34) as a white solid besides 3.0 mg (0.009 mmol, 3 %) of ketone 16 (RF=0.42) as a yellow solid. A sample of rac-15 was recrystallized from CH2Cl2/CyHex to afford white crystals suitable for X-ray structure analysis. Mp.: 66 °C. 1H NMR (600 MHz, CDCl3): δ [ppm]=2.17–2.21 (m, 1H), 2.50–2.60 (m, 2H), 2.89–2.95 (m, 1H), 3.71 (s, 3H), 3.92 (s, 3H), 3.92 (s, 3H), 4.06 (s, 3H), 5.21 (dd, 3JH,H=9.6 Hz, 7.6 Hz, 1H), 6.65 (s, 1H), 7.04 (d, 3JH,H=8.5 Hz, 1H), 7.89 (d, 3JH,H=8.5 Hz, 1H). 13C NMR (150 MHz, CDCl3): δ [ppm]=31.2, 38.4, 56.0, 56.1, 60.9, 61.0, 77.2, 108.7, 111.0, 112.2, 122.0, 122.8, 134.8, 137.0, 141.3, 150.2, 151.0, 152.9, 157.3, 168.7. HRMS (ESI): calcd. for [M+Na]+ (C20H20NaO6): 379.11521; found: 379.11542. FT-IR (ATR): 1760 (s), 1287 (s), 1106 (s) [cm−1].

Synthesis of 22: 35 mg (0.08 mmol, 1.0 equiv.) of mesylated colchicol (rac-20) and 20 mg (0.4 mmol, 5.0 equiv.) of NaCN were dissolved in 5 ml of DMSO. The mixture was heated to 50 °C for 3 d. After dilution with sat. aq. NaHCO3, the mixture was extracted with CH2Cl2. The organic phase was washed with 1 M HCl and sat. aq. NaHCO3, dried over MgSO4 and concentrated under reduced pressure. The crude product was purified by column chromatography (CyHex/EtOAc/EtOH 4 : 1 : 1) to give 13 mg (0.035 mmol, 44 %) of the nitrile 22 besides 11 mg of re-isolated starting material (rac-20). Mp.: 148 °C. 1H NMR (500 MHz, CDCl3): δ [ppm]=2.08–2.16 (m, 1H), 2.19–2.32 (m), 2.47–2.53 (m, 1H), 2.74–2.80 (m, 1H), 3.60 (s, 3H), 3.91 (s, 3H), 3.92 (s, 3H), 4.03 (s, 3H), 6.60 (s, 1H), 7.63 (d, 3JH,H=8.0 Hz, 1H), 7.65 (d, 3JH,H=8.1 Hz, 1H). 13C NMR (125 MHz, CDCl3): δ [ppm]=28.8, 31.3, 32.7, 52.9, 56.0, 61.1, 61.1, 107.8, 109.3, 117.6, 123.6, 130.2, 132.5, 135.4, 135.9, 139.4, 140.9, 142.8, 150.9, 153.8, 167.1. HRMS (ESI): calcd. for [M+Na]+ (C21H21NaNO5): 390.13119; found: 390.13132. FT-IR (ATR): 1729 (s), 1598 (m), 1135 (s), 1096 (s), 1000 (s).

Computational details: The conformational space for each structure was explored using the OPLS-2005 force field18 and a modified Monte Carlo search algorithm implemented in MacroModel.19 An energy cut-off of 84 kJ mol−1 was employed for the conformational analysis, and structures with heavy-atom root- mean-square deviations (RMSD) less than 1.5 Å after the initial force field optimizations were considered to be the same conformer. The remaining structures were subsequently optimized with the dispersion-corrected M06-L functional20] with Grimme's dispersion-correction D321 and the double-ζ basis set 6–31+G(d,p). Vibrational analysis verified that each structure was a minimum or a transition state. Following the intrinsic reaction coordinates (IRC) confirmed that all transition states connected the corresponding reactants and products on the potential energy surface. Thermal corrections were obtained from unscaled harmonic vibrational frequencies at the same level of theory for a standard state of 1 mol L−1 and 298.15 K. Entropic contributions to free energies were obtained from partition functions evaluated with Grimme's quasi-harmonic approximation.22 Electronic energies were subsequently obtained from single-point calculations of the M06-L−D3 geometries employing the meta-hybrid M06-2X functional,23 Grimme's dispersion-correction D3 (zero-damping),21 the large quadruple-ζ basis set def2-QZVP,24 and IEFPCM for THF.25 An ultrafine grid was used throughout this study for numerical integration of the density. All density functional theory cal- culations were performed with Gaussian09.26

Acknowledgements

This work was supported by the Deutsche Forschungsgemeinschaft (DFG; Project SCHM 857/18-1), the Jürgen Manchot Foundation (doctorate Stipend to P.T.), and the Fonds der Chemischen Industrie. We thank the Regional Computing Center of the University of Cologne for providing computing time of the DFG- funded (Funding number: INST 216/512/1FUGG) High Performance Computing (HPC) System CHEOPS as well as for their support. Open Access funding enabled and organized by Projekt DEAL.

    Conflict of interest

    The authors declare no conflict of interest.