Late Stage Functionalisation
A late-stage functionalisation strategy was assessed for its ability to diversify OSM Series 4 compounds, exploring methods of C-H hydroxylation, fluorination and trifluoromethylation.
Methods of C-H substitution that occur through innate reactivity pathways were specifically explored, as it was hoped that such reactivity patterns would exhibit regio-nonselectivity. Such non-selectivity would be advantageous for a drug development strategy, as it would allow the rapid synthesis of derivatives to be assessed for structure-activity relationships.
Further, innate methodologies such as hydrogen-atom transfer are biomimetic in their substitution pattern, reacting similarly to cytochrome P450 enzymes. This was hoped to confer enhanced metabolic stability to the derived molecules by substitution with a metabolically resistant atom (fluorine, deuterium, etc.).
A metric for the comparison of the ability of a reaction to form many derivatives in good yield was developed, the reaction non-selectivity score (RNS). It can be easily computed in three steps, for example, where a reaction has formed three products, n1, n2, n3 and n4 with yields of 8%, 18%, 22% and 33% respectively, a comparison to ideal product formation can be made in three steps as follows:
The use of a metric to define reaction success for multi-product forming reactions allows quantitative analysis to be used to improve total reaction yield, for example, by using it as the response variable in a design of experiments (DOE) protocol.
Images and compound numbering are directly from honours thesis (Conor Graham, University of Sydney) and so may not be in consecutive order.
The late-stage installation of a hydroxyl group (-OH) into a drug-like molecule is a major area of research within the field of LSF. Hydroxylation can improve a lead compound's drug-like properties by increasing hydrophilicity, or by providing hydrogen bond donor and acceptor motifs to the molecule to better affect target-binding.
As a reactive polar protic group, hydroxyl typically requires protection when installed in earlier stages. Despite the obvious advantages of a late-stage method of C-H hydroxylation, there are still only a few methods able to perform such a transformation under mild conditions.
To produce a hydroxylated derivative of an active OSM Series 4 compound, we used two catalyst-based strategies, and two reagent-based strategies.
The two catalysts used have both previously demonstrated hydrogen substitution at unactivated aliphatic positions. The first catalyst, Fe(II)PDP (AKA Chen-White catalyst, 5) was chosen for its ability to hydroxylate unactivated sp3 C-H positions on complex molecules with high regioselectivity [DOI:10.1126/science.1148597], while the second catalyst, Fe(II)TPA (6) was chosen for its ability to functionalise simple substrates with low selectivity [DOI:10.1039/C3CC47830K, DOI: 10.1021/ja010310x, DOI: 10.1038/nchem.648].
The benzylic functionalised derivatives 40 and 41 were hypothesised to be the major products. Using the iterative addition protocol (described in the original Chen-White paper) for both catalysts resulted in an intractable mixture that did not contain derivatised products. Starting material (58%) was reclaimed from the reaction using Fe(II)TPA as catalyst, while none was reclaimed using Fe(II)PDP. Longer reaction times and slower oxidant addition times did not significantly affect the outcome.
With the failure of using catalytic methods to functionalise the substrate, we turned to a reagent-based method. Dimethyldioxirane (DMDO) (42) and methyl(trifluoromethyl)dioxirane (TFDO) (43) are three-membered ring peroxides, used as effective mild oxidising reagents with a wide variety of applications.
TFDO is a similar oxidant to DMDO, but is more electrophilic due to the highly electron-withdrawing CF3 group, making it a useful alternative where DMDO fails. Rather than hydrogen-atom transfer, the mechanism of these reagents occurs through electrophilic oxygen-atom transfer
Due to their sensitivity to heat and light, they are usually synthesised in low yield in a dilute solution with acetone/trifluoroacetone, by oxidation of acetone/trifluoroacetone with oxone (AKA potassium peroxymonosulfate). The volatile product is collected at -78 °C as a distillate via a cold finger condenser. Special care must be taken to ensure glassware is free of any trace metal contaminants, and that the reaction is shielded from light.
Attempts at using DMDO to afford a hydroxylated product of an OSM Series 4 compound were unsuccessful, with no conversion of starting material taking place. However, TFDO produced an oxidised derivative of 20 in low yield (44, 23%), with functionalisation occurring exclusively at the C8 position.
Unfortunately, this compound did not exhibit biological activity. Previous attempts at improving activity with hydroxylation at this position were also unsuccessful (MMV669025 and MMV675960).
The low yield of this reaction was hypothesised to be due to catalytic decomposition of the reagent through N-oxide formation.
A photocatalytic method using tetrabutylammonium decatungstate ((NBu)4W10O32, TBADT) as photocatalyst and N-fluorobenzenesulfonimide as a source of electrophilic fluorine was trialled, using ibuprofen methyl-ester (11) as a model compound. The reaction succeeded with 60% conversion according to an analysis of the crude reaction mixture by 1H NMR spectroscopy.
When the same reaction was attempted on an active substrate, the reaction did not progress, with no new peaks observed in the 19F NMR spectrum of the crude reaction mixture. A small amount of an uncharacterised, intractable mixture of non-fluorine containing products eluted in hexanes, suggesting possible photo-decomposition of the starting material.
As an alternative to reactions which form C–F bonds, we investigated strategies for the formation of the C-CF3 group instead. Despite a major concern of the OSM Series 4 compounds being their high lipophilicity, trifluoromethylation may still provide important new SAR data and permit the probing of positions on the molecule that are important for metabolic clearance.
We explored a radical trifluoromethylation strategy, following a recent modification of an older reaction, which uses sodium trifluoromethanesulfinate as reagent. We applied this reaction to our model compound, 19, to assess the regioselectivity of the reaction on a compound with more than one heterocycle.
Two additions of Langlois’ salt (7 equivalents) and two additions dropwise of tert-butylhydroperoxide (12 equivalents) yielded a single major product in 29% yield after 48 hours’ reaction time, with a small recovery of starting material (16%). The structure of the major product was elucidated by X-ray crystallography, generating the first structure of a Series 4 compound derived from a late stage functionalisation. Functionalisation was shown to occur at the C8 position
With the X-ray structure of the starting material, and a comparison showed only minor changes to the molecule conformation resulting from the substitution of the aromatic hydrogen with the CF3 group. For example, the dihedral angle between the planes occupied by the triazolopyrazine and pendant aromatic rings.
When 46 was evaluated for biological activity, it was found to have no activity (> 10 μM), which was unsurprising considering the parent compound, 19, also had no activity (> 10 μM).
19F NMR spectroscopy of the crude reaction mixture showed trace amounts of other fluorinated material, with a single main singlet peak at d –66.64 ppm, while baseline spots and streaking visualised on TLC further supported the idea of competitive pathways. A possible source of side-products is from the decomposition of the tert-butylhydroperoxide to isobutene, leading to a fluoroalkyl radical adduct that can add itself to aromatic rings, though such species were not specifically identified in our procedure.
We hypothesised that attack occurred at the C8 position rather than the C6 position for the following reasons:
-
First, the heteroaromatic core is more electron rich, as compared to the pyridine ring: the electronegativity of the pyridine nitrogen makes the adjacent C–H bonds comparatively electron poor, while the two C–H groups of the core benefit from electron donation through the ether side-group. It may be further stabilised through delocalisation of p-electrons from the adjacent triazolo ring. For this reason, the heteroaromatic core may be more prone to attack by an electrophilic radical such as •CF3 than the pyridine side-group.
-
Second, the resonance structures generated from attack at the C8-position may be more stable than the C6-position. Attack at the alpha-ether positon generates a structure for which there are two resonance forms with partially-positively charged radicals (on the ring-N and on the ether-O). The electron deficiency of these atoms destabilises these resonance structures. Attack at the C8 position is comparatively more stabilised, resulting in resonance forms with a partially-positive radical on only the oxygen: the nitrogen-centred radical occurring from attack at the para-position is stabilised by favourable p-system interactions from its remaining lone-pair.
-
Third, approach of the •CF3 radical to the heteroaromatic ring system may also be affected by steric factors. Attack at the alpha-ether position may be less likely due to unfavourable steric issues arising from the attached phenyl ethyl ether group, which blocks attack from one side, while the C8-position has no such hindrance. Further, the high barrier to inversion of the pyramidal •CF3 radical may slow substitutions where steric factors are involved.
The ease and lower costs of this procedure, as compared to other methods, as well as the simplicity of product isolation, demonstrates this procedure to be a useful tool in the LSF toolbox for trifluoromethylation.
Interestingly, when we performed the same procedure on 20, we were surprised to find that reaction did not proceed in the same dichloromethane (DCM)/water (2.5:1) mixture as described for the general procedure of the original report. A dimethyl sulfoxide (DMSO)/water (4:1) mixture was employed instead, following suggestion from the original paper. Despite a more tedious work-up procedure (extensive washing of the organic layer with a water/ice mixture), this solvent mixture allowed conversion to a trifluoromethylated derivative in a yield of 4%.
The very small isolated yield of this reaction is not indicative of conversion, which was near complete; isolation of the product was difficult due to its co-elution with two other products that were not isolated or fully characterised. LCMS analysis was performed on the crude reaction mixture to assess the presence of other compounds. Ions with m/z values 519 and 587 indicate di- or tri-trifluoromethylated products. The ratio of the areas under their peaks, monitored at 254 nm was 8.6:3.5:1, indicating a favourable RNS(3) of 36%. Neither the di- or tri-trifluoromethylated products could be isolated in the time available but this separation is clearly of importance for future work. Analysis showed no other ions with the same mass as the isolated mono-functionalised product, suggesting no other (separable) regioisomers were present.
A number of LSF strategies were assessed, successfully establishing preparative and analytical methods towards the non-selective functionalisation of C–H bonds in drug-like molecules. These methods were used to generate several LSF-derived compounds, providing useful SAR data to a currently-active medicinal chemistry campaign.
Identification and isolation of regioisomers of derivatised products was not successful, although multi-functionalised trifluoromethylated derivatives of 20 were identified by LC-MS.
Photochemical possibilities have been discussed in (Feb, May 2018)