Can FtmOx1 catalyze diverse C(sp3)-H oxyfunctionalization for late-stage diversification?

Repurposing a native endoperoxidase for non-native substrates

The paper asks whether an enzyme evolved for one reaction can be redirected to oxidize C(sp3)-H bonds in drug-like scaffolds. FtmOx1 normally abstracts a C21 hydrogen from fumitremorgin B and partitions the resulting radical toward an endoperoxide or a decomposition product. The authors instead incubated FtmOx1 with phenethylamine and tetrahydroisoquinoline derivatives, compounds that share the phenethylamine skeleton of many bioactive alkaloids but are not its natural substrate.

The reaction setup was simple enough to be a screening platform: 0.1 mM wild-type or variant enzyme, 1 mM Fe2+, 2 mM substrate, 2 mM sodium ascorbate and 4 mM 2-oxoglutarate (2OG), shaken in air at room temperature for 4 h, with products quantified by LC-MS at 254 nm and stereochemistry checked by chiral HPLC. Ascorbate was needed to keep the iron reduced and sustain multiple turnovers; without it, the β-hydroxylated product still formed at 20% yield, and only a catalytic amount restored multiple turnovers. Product formation tracked 2OG consumption, consistent with the canonical Fe/2OG mechanism.

Four oxidation outcomes from the same catalyst

Across the substrate set, FtmOx1 produced four product families. β-Hydroxylation gave 1,2-amino alcohols (2), β-ketone formation gave 1,2-amino ketones (3), α,β-desaturation gave enamines (4), and oxidation at the α' or α position gave dealkylated ureas or amides (5, 7, 8). These are different oxidation depths and positions, not variants of one measurement, and the paper treats them separately.

The representative substrate 1a, an N-morpholine-carbonyl phenethylamine bearing a trimethoxybenzyl group, converted to the β-hydroxylated product 2a in about 95% yield with roughly 2% ketone 3a and only trace dealkylation detectable by extracted ion chromatography. The β-selectivity metric used throughout combines β-hydroxylation and β-ketone yield over the total yield of all oxidation species, reaching about 98% for 1a. A preparative-scale reaction gave 2a in 85% yield (50 mg), and NMR confirmed that the β-CH2 signal at δ 2.65 ppm became new resonances at δ 4.6 and 5.5 ppm, with chiral HPLC assigning the R configuration.

A kinetic isotope effect of about 2 points to hydrogen abstraction

To test which step limits β-hydroxylation, the authors compared 1a with a β-deuterated analog, 1a-d2. In the first 2 minutes the reaction was linear and the deuterated substrate showed a kinetic isotope effect of approximately 2, which the authors read as hydrogen atom transfer from the β-C(sp3)-H being rate-determining. The deuterated substrate still gave the β-hydroxylated product as the major product at 73% yield, with an M+1 mass shift indicating loss of a single β-deuterium.

Deuteration also changed the product distribution: β-selectivity fell from 98% with protium to 82% with deuterium, and new α'-debenzylation (7% yield) and α'- or α-ketone species (8% yield) appeared. The authors interpret this as evidence that β-, α- and α'-C-H bonds all sit near the iron center in the flexible substrate, so even a subtle isotopic change can shift regioselectivity. That is an author explanation of a measured shift, not a demonstration that the same shift would occur in every substrate.

Ring substituents decide whether hydroxylation or over-oxidation wins

Substrate scope was mapped by varying three regions of 1a: the amide R1 group, ring A, and ring A'. When R1 was a nitrogen-containing 5- to 7-membered heterocycle such as piperidine, pyrrolidine or azepane, β-hydroxylation gave 70-86% yields with regioselectivity above 98% and enantiomeric excess above 97%. Replacing that group with N,N-dimethylamine or 4-pyridyl dropped yields to 19-37% and β-selectivity to 59-80%, showing that the cyclic urea-like moiety is not decorative.

On ring A, methoxy, bromo, methyl or trifluoromethyl substituents, and even a thiophene replacement, kept β-hydroxylation as the major pathway but lowered yields to 24-57% and raised ketone products to 4-28%. Stereoselectivity followed a steric pattern for methoxy, bromo and methyl analogs, with para substitution giving the highest enantiomeric excess and ortho the lowest; the para-trifluoromethyl case broke the pattern and is attributed by the authors to its electron-withdrawing character. On ring A', a single para-methoxy group was enough to retain 64% yield, 99% enantiomeric excess and 96% β-selectivity, whereas unsubstituted or para-chloro analogs gave only 9-13% hydroxylated product and ketone as the major outcome. The authors conclude that a para substituent able to accept a hydrogen bond is required for efficient β-hydroxylation, with the trimethoxy pattern performing best.

Tetrahydroisoquinoline substrates shift the outcome toward ketones

The tetrahydroisoquinoline series behaved differently from the phenethylamines. For substrates 1w-1ac, the four-electron β-ketone products 3w-3ac were the major products at 27-75% yield regardless of the configuration at the α'-carbon, with smaller amounts of β-hydroxylated and α,β-desaturated products. This is a meaningful contrast for anyone planning synthesis: the same enzyme that gives mainly the alcohol on a flexible phenethylamine can push the related rigid scaffold further to the ketone.

The paper also reports that ketone formation is not a single route. FtmOx1-catalyzed 1,2-dehydrogenation of the alcohol 2 is the major but not exclusive path to 3, so ketone yield cannot be read simply as a second hydroxylation step. The authors further show that a substrate with two distinct CH2 positions, 1v, gave the β-hydroxylated product 2v at 47% yield and 88% enantiomeric excess, while the β'-hydroxylated isomer was not detected, indicating that the pocket discriminates between the two chains rather than oxidizing whichever is nearest.

Crystal structures link chain flexibility to product branching

Four ternary crystal structures (FtmOx1 with 2OG and 1t, with 2OG and 1w, with succinate and 2t, and the L222F variant with 2OG and 1t) support the mechanistic picture. The benzyl chain of ring A' sits in a narrow pocket formed by F233, Y224, S119 and T134, a region the native substrate cannot reach; Y224 and T134 flip by about 73° and 119° relative to their orientation in the 1t complex to accommodate different ring A' substituents. The phenethyl chain of ring A is held near F72, F115 and L222, and the authors argue that para substitution, especially interaction with F72, matters more for regio- and enantioselectivity than meta or ortho substitution.

Mutagenesis tested parts of this model. F115A and L222F increased the ratio of β-hydroxylation to ketone in some cases, and the L222F structure showed the substrate positioned with its β-C-H farther from the iron, which the authors use to explain reduced over-oxidation. By contrast, all tyrosine-to-phenylalanine or alanine variants at positions 224 and 68 retained activity and still produced hydroxylated products 2a and 2t, albeit at slightly reduced yields, which the authors attribute to weaker substrate binding. They therefore conclude that this non-native β-hydroxylation follows the canonical Fe/2OG mechanism without requiring Y224 or Y68, unlike proposals for the native endoperoxidation pathway.

Oxyfunctionalized products gain RXFP4 activity, with scope still to define

The pharmacological test is small but pointed. Because the substrates resemble the previously reported RXFP4 activator DC591053, the authors assayed 1a, 2a, 3a, 1c, 2c, 1s and 2s in their established RXFP4 activation assay. They report that the oxyfunctionalized products 2 or 3 consistently showed enhanced activity relative to their parent compounds 1 across three representative examples. This is a structure-activity observation from a limited compound set, not a demonstration that the platform yields a clinical candidate, and the paper does not report potency values in the extracted text.

The scope conditions are what a reader should carry forward. Yields and selectivity were measured on analytical-scale reactions with LC-MS peak integration at 254 nm, with preparative isolation reported for selected products; the 48-substrate count describes compounds that proved suitable, not a random or exhaustive library. The strongest β-selectivity came from substrates combining a trimethoxy or otherwise hydrogen-bond-accepting ring A' with a cyclic urea-like R1 and an appropriate ring A substituent, while tetrahydroisoquinolines favored ketones. The practical priority is therefore to choose the scaffold for the product you want: phenethylamines with the right para substituent for amino alcohols, tetrahydroisoquinolines when the ketone is the target, and the α'-oxidation pathway when no hydrogen-bond acceptor is present.

Sources used in this answer