Chemoselective methylene oxidation in aromatic molecules.

Despite significant progress in the development of site-selective aliphatic C-H oxidations over the past decade, the ability to oxidize strong methylene C-H bonds in the presence of more oxidatively labile aromatic functionalities remains a major unsolved problem. Such chemoselective reactivity is highly desirable for enabling late-stage oxidative derivatizations of pharmaceuticals and medicinally important natural products that often contain such functionality. Here, we report a simple manganese small-molecule catalyst Mn(CF3-PDP) system that achieves such chemoselectivity via an unexpected synergy of catalyst design and acid additive. Preparative remote methylene oxidation is obtained in 50 aromatic compounds housing medicinally relevant halogen, oxygen, heterocyclic and biaryl moieties. Late-stage methylene oxidation is demonstrated on four drug scaffolds, including the ethinylestradiol scaffold where other non-directed C-H oxidants that tolerate aromatic groups effect oxidation at only activated tertiary benzylic sites. Rapid generation of a known metabolite (piragliatin) from an advanced intermediate is demonstrated.

The majority of small molecule therapeutics are comprised of functionalized hydrocarbon scaffolds containing a mixture of C(sp)–H bonds and increasingly more C(sp)–H bonds[1-4]. Reactions that effect the direct atomistic exchange of hydrogen with oxygen remote from such aromatic groups would avoid lengthy de novo synthesis for generating analogues and identifying metabolites; however, because of the challenges of hydroxylating an inert C–H bond in the presence of oxidatively more labile π-functionality, such transformations have not been fully realized in the laboratory. Iron enzymes are the only known catalysts capable of hydroxylation of strong aliphatic C–H bonds in the presence of π-functionality by means of restricted substrate access to the oxidant (Fig. 1a, b),[5] however, these enzymes are challenging to use on preparative scales (vide infra) and this supramolecular approach for achieving chemoselectivity has not been successful with small molecule catalysts.

Figure 1 |

Enzymatic and small molecule approaches for C–H oxidation.

A chemoselective small molecule catalyst capable of strong methylene C–H bond oxidations in the presence of the more oxidatively labile π-functionality of aromatic groups was previously unknown. a. Cytochrome P450 enzymes (CYPs) achieve such chemoselective oxidation of methylene C–H bonds by restricting the substrate’s approach to the iron oxidant, however these enzymes are challenging to use preparatively. b. The two major pathways of aromatic oxidation with the iron(oxo) oxidants generated in CYPs are electron transfer (top) or epoxidation followed by hydride shift (e.g. the “NIH shift”, bottom). c. Small molecule catalyst Mn(CF3–PDP) 1 achieves chemoselective hydroxylation of strong methylene C–H bonds in presence of aromatic functionalities. This reactivity is orthogonal to other oxidants that perform weaker bond oxidations (for example benzylic C–H) in the presence of aromatic functionality as demonstrated with an ethinylestradiol derivative (the site of oxidation is highlighted in yellow). d. The crystal structure of the Mn(CF3–PDP)Cl2 precatalyst shows the bulky ligand framework thought to be critical for the high chemoselectivity achieved with Mn(CF3–PDP)(CH3CN)2•(SbF6)2 catalyst [Mn(CF3–PDP)] 1.

Despite recent advances in small molecule catalysts for site-selective aliphatic C–H hydroxylations, none of these catalysts can effect strong aliphatic methylene C–H bond oxidation (BDE = 98 kcal/mol) in the presence of a range of aromatic groups.[6-13] Whereas C–H aminations, alkylations and halogenation reactions tolerate aromatic functionality,[14-16] the only aromatic groups tolerated in higher energy methylene C–H hydroxylations have been electronically deactivated with strong electron withdrawing groups (that is nitro, trifluoromethyl, triflate with >0.5 σp values).[10, 11] Ruthenium[6] and manganese[7, 8] catalysts and stoichiometric oxidant TFDO[17] are tolerant of benzoate (σp = 0.45) and in a few cases phenyl functionality[6] however only in the hydroxylation of weaker benzylic C–H (BDE = 85 kcal/mol, Fig. 1c) or tertiary C(sp)–H bonds (BDE = 96 kcal/mol). The current state-of-the-art suggests that to divert oxidation of labile aromatic functionalities, the catalyst’s capacity for aliphatic hydroxylations must be reduced. We now demonstrate that via the combination of Mn(CF3–PDP) 1 catalyst and chloroacetic acid additive, C(sp)–H oxidation can be achieved with an unprecedented combination of high chemoselectivity, i.e. tolerates medicinally important aromatic functionalities, and high reactivity, i.e. preparatively oxidizes aliphatic methylene C–H bonds in the absence of directing groups or molecular recognition elements (Fig. 1c, d).

Results and discussion

Reaction development.

We first investigated the oxidation of secondary methylene substrate 2 having a mildly electron withdrawing bromo-substituted aryl moiety (σp = 0.23) with oxidants reported to be capable of oxidizing benzylic and tertiary C(sp)–H bonds in the presence of benzoates. Stoichiometric oxidant TFDO [methyl(trifluoromethyl)dioxirane] was evaluated under low temperature conditions reported to prevent free-radical formation[17] as well as more forcing conditions[18]. Only the later were effective at producing methylene oxidized product albeit in poor yields and chemoselectivity (Table 1, entries 1–2). Ruthenium catalysts purported to be highly chemoselective for benzylic and tertiary C(sp)–H oxidations [Ru(Me3TACN)][8] and cis-[Ru(dtbyp)2Cl2][6] were not effective in oxidizing stronger methylene bonds (entry 3–4). Known manganese catalysts Mn(OTf)2/bipyridine,[8] Mn(PDP)(OTf)2,[7] were evaluated under their previously reported conditions for C(sp)–H bond oxidation and found to furnish 3 in poor yields but encouraging chemoselectivities (entries 5–6).

Table 1 |

Reaction development of Mn(CF3–PDP) 1 catalysis.

Entry	Catalyst	Additive	Oxidant	Temperature	Yield	Sel.
1	—	—	TFDO3 equiv.	−20 °C	trace	N.D.
2	—	—	TFDO3 (6) equiv.	0 °C	trace (11%)	N.D.(18%)
3	Ru(Me3TACN)(2%)	—	CAN2×3 equiv.	RT	trace	N.D.
4	cis-Ru(dtbpy)2CI2(5%)	—	H5IO62 equiv.	RT	0%	N.D.
5	Mn(OTf)2, 0.1%;bipy, 1%	—	AcOOH3 equiv.	RT	26%	43%
6	Mn(PDP)(OTf)2(0.1%)	CH3COOH14 equiv.	H2O22.5 equiv.	0 °C	9%	69%
7	Fe(PDP) 4(3×5%)	CH3COOH3×0.5 equiv.	H2O23×1.2 equiv	RT[a]	1%	2%
8	Mn(PDP) 5(3×5%)	CH3COOH3×0.5 equiv.	H2O23×1.2 equiv	RT[a]	7%	59%
9	5(25%)	CH3COOH5 equiv.	H2O29 equiv.	RT[b]	36%	71%
10	Fe(CF3-PDP) 6(25%)	CH3COOH5 equiv.	H2O29 equiv.	RT[b]	24%	27%
11	Mn(CF3-PDP) 1(25%)	CH3COOH5 equiv.	H2O29 equiv.	RT[b]	21%	95%
12	1(25%)	CICH2CO2H(5 equiv.)	H2O29 equiv.	RT[b]	73%	90%
13	1(25%)	CI2CHCO2H(5 equiv.)	H2O29 equiv.	RT[b]	66%	77%
14	5(25%)	CICH2CO2H(5 equiv.)	H2O29 equiv.	RT[b]	32%	65%
15	1(10%)	CICH2COOH(15 equiv.)	H2O210 equiv.	0 °C[c]	86%	94%
16	5(10%)	CICH2COOH(15 equiv.)	H2O210 equiv.	0 °c[c]	38%	85%
17	6(10%)	CICH2COOH(15 equiv.)	H2O210 equiv.	0 °c[c]	5%	7%
18	1(10%)	CH3COOH(15 equiv.)	H2O210 equiv.	0 °c[c]	20%	75%

Isolated yields are an average of two (TFDO) or three (all catalytic reactions) runs. TFDO, Methyl(trifluoromethyl)dioxirane; Me3TACN, N,N’,N“-trimethyl-1,4,7-triazacyclononane; CAN, ceric ammonium nitrate; dtbpy, 4,4’-di-tert-butyl-2,2’-bipyridine; bipy, 2,2’-bipyridine; PDP, 1,1’-bis(pyridin-2-ylmethyl)-2,2’-bipyrrolidine; CF3–PDP, 1,1’-bis((5-(2,6-bis(trifluoromethyl)phenyl)pyridin-2-yl)methyl)-2,2’-bipyrrolidine; RT, room temperature. SbF6- [hexafluoroantimony(V)] is used as counterion for catalyst 1, 4–6.

Iterative addition protocol; see ref. 9.

Slow addition protocol; see ref. 28.

Single catalyst addition (method A), standard procedure for Mn(CF3–PDP) 1 used unless otherwise noted: substrate (0.3 mmol) with Mn(CF3–PDP) 1 (0.03 mmol) and ClCH2CO2H (4.5 mmol) dissolved in MeCN (0.6 mL) maintained at 0 °C, a solution of H2O2 (50% wt., 3.0 mmol) in MeCN (3.75 mL, 0.8 M) was added via syringe pump over 3 hours.

The two primary mechanisms for aromatic oxidation with iron oxo enzymes (CYP450s) are single electron pathways at aromatic protein residues to generate radicals[19] and epoxidation of aromatic substrates followed by a rearrangement/hydride shift (“NIH shift’)[20] to generate phenols (Fig. 1b).[5] It is well-precedented that a switch from iron to manganese in catalysts that oxidize C–H bonds via high-valent metal heteroatom species (that is, metal oxos and nitrenes) leads to a reduction in their oxidation potential[21] that may alter their reactivity and chemoselectivity[14]. We evaluated the non-heme FePDP 4 catalyst [Fe(PDP)(SbF6)2], previously demonstrated to hydroxylate strong aliphatic C–H bonds,[9, 10] and its direct manganese analogue MnPDP 5 [Mn(PDP)(SbF6)2], each thought to proceed via metal(oxo) intermediates (Table 1, entries 7–8).[7, 9–13] Under analogous conditions, both catalysts proceeded with poor yields of desired methylene oxidized products; however in the case of the manganese catalyst, the chemoselectivity for 2o oxidized product was significantly higher than that of its iron counterpart (entries 7–8). Although increased reactivity with similar chemoselectivity is observed by switching from the iterative addition protocol to the more forcing slow addition protocol, a preparatively useful yield could not be achieved (36% yield, entry 9).

We hypothesized that the restricted approach trajectory and enhanced electrophilicity provided to the metal(oxo) by appropriate ligand modifications could enhance chemoselectivity. A bulky, electrophilic metal(oxo) should prefer an electron rich methylene C–H bond over a relatively electron deficient and more sterically demanding π-system[15]. The sterically and electronically modified non-heme catalyst Fe(CF3–PDP) 6 [Fe(CF3–PDP)(SbF6)2] was reported to have preferential reactivity for methylene versus tertiary aliphatic C–H bonds.[11] We evaluated Fe(CF3–PDP) 6 and its manganese analogue Mn(CF3–PDP) 1 [Mn(CF3–PDP)(SbF6)2] and found that whereas both catalysts gave lower yields of 3 than Mn(PDP) 5 (entry 9), novel catalyst Mn(CF3–PDP) 1 was unique in affording very high chemoselectivity (Table 1, entries 10–11). Analogous site-selectivity trends to those previously reported with Fe(CF3–PDP) were observed,[9-13] with Mn(CF3–PDP) 1 favoring remote oxidation on the more sterically accessible methylene site (vide infra). Hyperconjugative activation by the alcohol of the methine C–H bond promotes subsequent oxidation to the ketone.

The electrophilicity of the presumed manganese oxo intermediate may additionally be tuned with the inclusion of acid additives, previously shown to modulate reactivity and selectivity with metal(oxo) species.[22-25] Acetic acid in non-heme iron-catalyzed C–H oxidations enhances reactivity, likely by binding and delivery of a proton to the H2O2 at the metal center to promote metal(oxo) carboxylate formation.[9, 23, 26] Carboxylic acid containing substrates are able to direct C–H hydroxylations to electronically and sterically disfavored sites, providing compelling evidence for carboxylate as an ancillary ligand to Fe(PDP) 4 during oxidation.[24] To probe this effect on the electronics of Mn(CF3–PDP) 1, we examined electron deficient carboxylic acids in combination with 1. We found that both chloro- and dichloroacetic acid significantly increased yields for methylene oxidation, however only chloroacetic acid maintained good chemoselectivities (Table 1, entries 12–13). Interestingly, chloroacetic acid did not show the same beneficial effect with the Mn(PDP) 5 catalyst for methylene oxidations (entry 14). Catalyst 1 loadings were decreased (25 mol% → 10 mol%) and the yield was enhanced by increasing acid and H2O2 equivalents, decreasing reaction temperature (room temperature → 0 oC) and modifying the addition protocol (entry 15, see Supplementary Table 1 for details). Under these optimized conditions, neither Mn(PDP) 5 nor Fe(CF3–PDP) 6 showed significant improvements (entries 16–17), indicating that the catalyst is one critical component; however, switching back to acetic acid under these optimized conditions with 1 also led to significant diminishments in the yield and chemoselectivity for this oxidation (entry 18). Collectively, these results suggest a synergy between the acid additive and catalyst to furnish a remarkable amalgamation of reactivity and chemoselectivity.

Reaction scope.

To assess the ability of aromatic groups to tolerate methylene oxidation reactions with 1, we evaluated a cyclopentane substrate with electronically and sterically varied aromatic substituents (Table 2). A cyclopentane substrate with a phenyl substituent afforded poor yields of remote oxidation product 7 due competitive aromatic oxidation. Consistent with the hypothesis that π-oxidation is more sterically demanding than C(sp)–H hydroxylation oxidation, introduction of an electron rich but bulky tert-butyl (t-Bu) substituent blocks aromatic oxidation to afford desired methylene oxidation product 8 in 65% yield. Significantly, electron rich and unsubstituted aromatic groups are also susceptible to CYP450 oxidations and this source of metabolic instability may account the observation that approximately 50% of top leading drugs on the market are halogenated.[4] Mildly electron withdrawing halogen substituted aromatic groups, are all well tolerated in this reaction: bromine (σp = 0.23), chlorine (σp = 0.23) and even fluorine (σp = 0.06) substituted substrates all afford synthetically useful yields of methylene oxidation products 3 (86%, Table 1, entry 15), 9 (81%, Table 2) and 10 (56%). It is worth noting that in gram-scale reactions, catalyst and acid loadings can be reduced to 5 mol% and 7.5 equiv. (respectively) to afford products with comparable isolated yields and chemoselectivities (9,72% yield, 84% chemoselectivity). The phthalimide group diminishes resonance donation of aniline enabling effective remote methylene oxidation to furnish 11 in 63% yield. Compounds with meta-ester (σm = 0.23) and 3,5-difluoro- substituents are oxidized with Mn(CF3–PDP) 1 to give 12 (73%) and 13 (66%) in useful yields. When evaluating aromatic substrates with stronger electron withdrawing substituents such as para-esters and nitriles, somewhat tolerated with Fe(CF3–PDP) 6, we see that 1 affords higher yields of methylene oxidized products 14 and 15 with excellent chemoselectivity. Underscoring the unprecedented level of chemoselectivity with the new Mn(CF3–PDP) 1, direct comparisons made with Fe(CF3–PDP) 6 under previously reported optimal conditions (Table 1, entry 10) show 1 gives significant improvements in yields and selectivities relative to 6 (11% for 12 and 13, 28% for 14, 55% for 15). Substrates with trifluoromethyl and difluoromethyl aromatic groups, increasingly found in pharmaceuticals,[2] afford remote methylene oxidation in 77% (16) and 55% (17) yield, respectively. Cyclopentanes with a nitrophthalimide masked primary amine as well as isobenzofuranone and oxindole cores, seen in fluorescein and sunitinib, afforded remote methylene oxidation products 18, 19, and 20 in good yields and chemoselectivities. Spirocycle substrates are not uniquely effective, methylene oxidations of analogous aromatic compounds containing remotely appended cyclopentane rings and linear aliphatic chains afford products in good yields (64% for 21, 61% for 22, 82% for 23, and vide infra). Following site-selectivity rules established with analogous non-heme iron catalysts[9-13], oxidation by 1 of a 3-butyl 3-acetoxy oxindole substrate gave only one observed methylene oxidized product at the most sterically accessible and electron rich δ site remote from the tertiary acetoxyl moiety (22). When the alkyl chain was extended by one carbon in an N-pentyl benzenesulfonamide substrate, a 3:1 δ:γ mixture of oxidized products (23) was observed still favoring the most remote δ site.

Table 2 |

Substrate scope of Mn(CF–PDP) 1 catalyzed chemoselective C–H hydroxylation

Method A is used unless otherwise noted. Isolated yields are average of two-three runs. Site of oxidation is highlighted in yellow. Chemoselectivity (2° C–H [O]/total conversion) in parenthesis. NPhth, phthalimide; N(4-NO2Phth), 4-nitrophthalimide. aIterative catalyst addition protocol (method B): a total of 15 mol% Mn(CF3–PDP) 1 was added in 3 portions (5 mol%, 3 times). See methods section for more details. b25 mol% Fe(CF3–PDP) 6, slow addition protocol (ref. 28). cStarting material recycled once. dRatios are statistically corrected. e7% chloroacetic ester, 8% recovered starting material also observed.

Small ring carbocycles are increasingly utilized in pharmaceutical compounds to address the physicochemical and pharmacological challenges in classic sp-hybridized planar structures.[27] Cyclohexyl rings are oxidized in good yields, however poor site-selectivities (Table 2, 24, 1:1 γ:δ ratio) due to competing electronic effects favoring the remote δ site and stereoelectronic effects, that is relief of ring strain, favoring the γ site.[9-13] Despite their prevalence in natural products and pharmaceutical molecules, cyclobutanes are rarely demonstrated in intermolecular C–H oxidations due to the partial sp-hybridized character of the C–H bonds. Under Mn(CF3–PDP) 1 catalysis, remote cyclobutyl groups are effectively oxidized in preparative yields with excellent site-selectivity (25). The chemoselectivity and high yield of remote oxidation are maintained in reactions with a series of biaryl bioactive molecule derivatives, further highlighting the generality of this method. A cyclobutyl analogue of the pesticide proclonol is oxidized on the cyclobutyl ring in good yield and site-selectivity (26, 52% yield, 4:1 δ:γ). A derivative of clorindione, a vitamin K antagonist, is oxidized with 1 to afford 60% isolated yield of ketone product 27. A derivative of nonsteroidal anti-inflammatory drug ketoprofen affords remote ketone products 28 in 52% yield and 9:1 site-selectivity.

The site-selectivity of Mn(CF3–PDP) 1 catalysis, analogous to that reported for Fe(PDP) and Fe(CF3–PDP) catalysis, is not solely dependent on bond dissociation energies of the C–H bond: in the 1 catalyzed oxidation of dichlorobenzylketone derivative, C–H oxidation of the remote methylene sites on the cyclopentyl ring are preferred over oxidation of weaker benzylic and tertiary sites that are electronically deactivated by the proximal carbonyl (Table 2, 29). Highlighting the orthogonality of Mn(CF3–PDP) 1 catalysis to alternative methods, 1 preferentially oxidizes at the less electron rich but more sterically accessible methylene sites to afford 30 and 31, whereas electrochemical radical methods[18] and TFDO[17] both afford the tertiary alcohol as the exclusive or major product. Cyclopropanes effectively activate α-methylene sites towards oxidation with 1 to afford 54% combined yield of ketone and alcohol products 32 with no detectable ring-opened products. In contrast, other Mn catalysts, including Mn(PDP)(OTf)2[8] are reported to give no product or trace yields.

Medicinally interesting compounds often contain aromatic rings with varied electronic and steric properties (Table 2). Whereas tert-butoxybenzene moiety in the precursor of 33 is oxidized, addition of a bulky, electronegative chlorine substituent to the aromatic group affords methylene oxidized product 34 in 73% yield. A compound of alternate topology with the same aromatic substitution is also oxidized in preparative yields to afford 35 in 54% yield, albeit in diminished site-selectivity (3.5:1 ε:δ). Replacing Ot-Bu with OMe in the substrate followed by oxidation with 1 provides product 36 in diminished yield and selectivity. Consistent with the lability of the ethereal site, oxidation by 1 of an analogous deuterated anisole compound affords 37 with a slightly higher yield and chemoselectivity. As the aromatic ring is made more electron deficient, ethereal and benzylic C–H bonds are shielded from oxidation with 1: in triflate protected phenol substrates with methyl and methoxyl substituents higher yields and selectivities are obtained for methylene oxidized products 38 and 39, respectively. In a difluorobenzodioxole substrate where the ethereal C–H bonds are replaced with fluorine and the π-donation of the oxygen into the aromatic ring is attenuated, remote oxidized product 40 is furnished in preparative yields.

A cursory glance at pharmaceutical molecules reveals that the combination of halogenated aromatics and nitrogen heterocycles (e.g. piperidines, pyridines, imidazoles) is ubiquitous in their structures.[1,2] We explored Mn(CF3–PDP) 1 catalysis with the known HBF4 protection strategy for enabling remote methylene C–H oxidations in the presence of basic nitrogen heterocycles that additionally contain aromatic moieties (Fig. 2a).[28] 4-Arylpiperidines and benzyl derived piperidines, prevalent motifs in opioid analgesics and cognition enhancing drugs (e.g. ketobemidone and donepezil), were tolerated and afforded remote methylene oxidized products in good yields (41–43). In the absence of HBF4, no desired remote oxidation was observed, indicating the chloroacetic acid alone is not effective in protecting the basic amine functionality. An analogue of haloperidol bearing 4-fluorophenyl butyl ketone and piperidine pharmacophores undergoes remote oxidation with 1 in 84% isolated yield to afford 44. An aliphatic motif with a 1,1-disubstituted pyridine-aryl pharmacophore is effectively oxidized to afford ketone product 45 in 54% yield. Mn(CF3–PDP) 1 oxidation of a chloro-phenylpyridine substrate shows excellent yields (46, 78%) and good site-selectivity (5:1 δ:γ). Additionally, imidazole, benzimidazole and quinoline derivatives are successfully oxidized to afford remote ketone in 50% (47), 55% (48) and 84% (49) yield, respectively. Significantly, without the combination of HBF4 protection and Mn(CF3–PDP) 1 catalysis, no significant remote oxidized products were observed: 1 affords enhanced heterocycle tolerance relative to iron catalyst 6, however, the protonation strategy is still necessary in oxidations with 1 for less basic heterocycles to access 47, 48, and 49.

Figure 2 |

Chemoselective methylene C–H oxidation.

a. Aromatic substrates with heteroaromatic or 3° amine substituentsa. Mn(CF3–PDP) 1 catalyzed remote 2° C–H oxidation in the presence of aromatic functionality with basic nitrogen containing heterocycles or heteroaromatics. b. Aromatic dipeptides. Mn(CF3–PDP) 1 oxidation of peptides containing amino acids with mildly deactivated aromatic groups. This enables for the first time intramolecular functionalizations with arene π-nucleophiles to furnish medicinally relevant tricyclic cores. Isolated yields are average of two-three runs. Site of oxidation is highlighted in yellow. Ns, 4-nitrophenylsulfonyl. HBF4•OEt2 protected as described in ref. 28 followed by slow catalyst addition protocol (method C): both H2O2 (10 equiv.) and Mn(CF3–PDP) 1 (10 mol%) were simultaneously added over 3 hours. See methods section for more details. Method C used at 0 °C. Without HBF4 complexation under optimal condition with 1. 25 mol% Fe(CF3–PDP) 6, slow addition protocol (ref. 28). Method B used. Starting material recycled once. Method A used with AcOH (15 equiv.) and H2O2 (5.0 or 7.5 equiv.) at –36 °C. Crude hemiaminal, TfOH (2.0 equiv.), 90 °C, 2 hours.

Previous reports demonstrated an impressive ability of iron catalysts to afford remote aliphatic C–H hydroxylations on peptides;[29] however, the aromatic side chains present were restricted to strongly electron withdrawing triflate (σ = 0.53) protected tyrosine. Using Mn(CF3–PDP) 1 catalysis, dipeptide containing chlorophenylglycin residues can be oxidized at aliphatic norleucine and cyclopentyl residues to afford single oxidation product in 55% (50) and 59% (51) yield respectively (Fig. 2b). An isoleucine residue is selectively oxidized at the more sterically accessible methylene site to furnish 52 with 9:1 secondary:tertiary ratio. Mn(CF3–PDP) 1 selectively installs the diversifiable C5-hydroxyl group onto proline in dipeptide settings containing halogenated phenylalanine and now readily removable pivalate protected tyrosine amino acid residues in good yields (53–56). Arene π-nucleophiles can now be linked directly to an amide nitrogen (57) prior to C–H hydroxylation to enable N-acyl iminium intramolecular cyclization to furnish tricyclic isoquinolone derivative 58. Importantly, in all cases examined Fe(CF3–PDP) 6 oxidations run under their reported optimal conditions do not tolerate these aromatic functionalities and affords only trace remote oxidized products (e.g. 50, 55 and 56).

Late-stage oxidation of pharmaceuticals.

The unprecedented chemoselectivity and reactivity of Mn(CF3–PDP) 1 for remote methylene oxidations in the presence of pharmaceutically relevant aromatics and heteroaromatic moieties, provides an opportunity to effect late-stage diversification of drug leads (Fig. 3a). The atomistic change of C–H to C–O is a validated way to impact physiological responses of small molecules due to altered interactions with proteins and/or changes to physical properties. Evaluation of the hexanoyl derivative 59 of efavirenz, a WHO essential medicine for HIV-1, demonstrates that the singular chemoselectivity of 1-catalyzed aliphatic C–H oxidation persists in the presence of the π-system of an aromatic and an alkyne moiety to furnish 60 in 58% yield with no observed products from oxidation of the aromatic or alkyne functionality. Compound 61, a γ-secretase modulator analogue,[30] was effectively oxidized by 1/HBF4 strategy to give remote ketone product 62 in 55% yield. The activated tertiary benzylic sites, generally susceptible to TFDO and radical-mediated C–H abstraction, did not undergo oxidation or epimerization with 1. Antidepressant citalopram 63 housing two aromatic rings and a tertiary amine, underwent selective C–H hydroxylation alpha to the ethereal oxygen. Treatment of the hemiacetal with BF3•Et2O furnished a reactive oxocarbenium ion intermediate that underwent diastereoselective nucleophilic attack by an electron rich arene to furnish the 2-naphthol adduct 64 in 42% overall yield (2 steps, 3:1 d.r.). Significantly, generation of hemiacetals has not been described under iron catalysis, likely because the highly reactive hemiacetal rapidly undergoes further oxidation to a carbonyl.

Figure 3 |

Late-stage methylene hydroxylation of synthetic and natural product, aromatic drugs derivatives.

a. Derivatives of the HIV-1 drug efavirenz and a γ-secretase modulator analogue that house oxidatively sensitive π-functionality such as aryl halides, acetylenes, piperidine, and tertiary amines are oxidized with Mn(CF3–PDP) 1 at remote methylene sites in preparative yields. Antidepressant citalopram is oxidized to a hemiacetal and arylated with an arene π-nucleophiles via an intermediate oxocarbenium. b. Drug candidate piragliatin, identified as a metabolite of drug lead 65, required de novo synthesis to furnish quantities for further evaluation. Mn(CF3–PDP) 1 catalysis enables an advanced lead intermediate to be rapidly transformed to piragliatin. c. Sequential Mn(CF3–PDP) 1 catalyzed 2o benzylic/2o aliphatic C–H oxidation of an ethinylestradiol derivative. 2° benzylic oxidation can be tuned with oxidant to give ketone 70 or diastereomerically pure alcohol 71 that undergo further diastereoselective C12 β methylene hydroxylation to alcohols 73 and 74 (crystal structures confirm the 12β configuration). In contrast, TFDO furnishes oxidation at the doubly activated 3° benzylic site. All reactions run with limiting substrate and isolated yields reported as an average (2–3 reactions, using 1). See the Supplementary Information Section VIII and IX for more details. Ac, acetyl; Piv, pivaloyl. Method A used at 0 °C. HBF4•OEt2 protected, ref. 28 followed by method C at 0 °C. Method A using 2 equiv. H2O2; arylation protocol, ref. 29. LiOH•H2O (5 equiv.); oxalyl chloride (1.1 equiv.); 2-aminopyrazine (2.2 equiv.), pyridine (2.2 equiv.). Method A with modifications: 1 (5 mol%; 2 mol% for gram scale), ClCH2CO2H (7.5 equiv.), H2O2 (10 equiv. for 70 or 2 equiv. for 71), 4:1 MeCN:CH2Cl2 at –36 °C. Ac2O (2.4 equiv.), NEt3 (2.4 equiv.), CH2Cl2, 86%. Method A with 4:1 MeCN:CH2Cl2 at 0 °C. TFDO: 72 (0.05 mmol) in CH2Cl2 (0.5 mL) at –20 °C. TFDO (0.4 M solution, 0.25 mL, 2 equiv.) added at –20 °C and stirred for 40 minutes in dark.

Oxidative metabolism of pharmaceuticals via CYP450 enzyme mediated C–H hydroxylation is a major pathway for drug clearance in vivo and their identification is crucial for the evaluation of their safety and effectiveness.[31] Piragliatin, a drug candidate advanced in the clinic for type 2 diabetes, was discovered via in vivo metabolite studies (MetID) of lead compound 65 (Fig. 3b).[32] Since MetID studies only generate analytical amounts of the metabolite, piragliatin, its diastereomer as well as several other potential cyclopentyl ring metabolites were produced in quantities needed for identification and in vivo safety and efficacy profiling via independent total syntheses that each proceeded with a higher step counts than that of the lead 65 (8 steps). Advanced synthetic intermediate 66 whose acid was used in the synthesis of 65 can be intercepted and directly oxidized with 1 to afford the C3 ketone 67 (1:1 d.r.) in 54% yield. Hydrolysis of methyl ester followed by amide coupling with 2-aminopyrazine rapidly affords piragliatin and its diastereomer (54% yield, 1:1 d.r.).[32] Mn(CF3–PDP) 1 catalysis may provide a powerful means for rapidly accessing aliphatic metabolites of aromatic drugs in preparative quantities.

Sequential C–H oxidation of complex natural products.

Natural products (e.g. polyketides, steroids, flavonoids) containing aromatic moieties showcase a wide range of biological activities and serve as impacting starting points for generating new pharmaceuticals.[3, 33] Chemical modifications to these complex structures is generally limited to manipulations of native functionality. We envisioned that Mn(CF3–PDP) 1 may be used on such medicinally important aromatic steroids like estrogen supplement ethinylestradiol, containing a sensitive aromatic ring and an alkyne, to enable the rapid, sequential buildup of remote hydroxylation on a steroid core (Fig. 3c). In contrast to TFDO which hydroxylates the doubly activated tertiary benzylic site on estrone derivatives (e.g. 68),[34] performing oxidation with 1 on ethinylestradiol derivative 69 effectively oxidizes the sterically more accessible secondary benzylic C–H site in high yields. The oxidation state of the newly formed C–O bond is controlled by the amount of hydrogen peroxide oxidant: 10 equivalents affords the benzylic ketone product 70 (72% yield; gram-scale synthesis with 2% of 1, 74% yield) whereas 2 equivalents affords a mixture of ketone 70 and alcohol 71 formed as a single diastereomer. Further hydroxylation at the non-activated methylene C12 position can be effected with 1 in to afford diastereomerically pure alcohol products: oxidation of ketone 70 and the protected benzylic alcohol 72 afford single diastereomers of C12 β-alcohol in 47% yield of 73 and 32% yield of 74. The observed diastereoselectivity and absence of over-oxidized C12 ketone products is likely due to the inability of bulky catalyst 1 to access the sterically hindered C12 hydrogen. In sharp contrast, TFDO oxidation of 72 led to hydroxylation at the tertiary benzylic site to afford 75 in 84% yield. The ability to access hydroxylation of methylene sites on steroids has been primarily limited to enzymes and engineered directing group approaches.[35, 36] Despite the fact that C12 hydroxylation occurs in numerous classes of natural steroids (e.g. polyoxypregnanes, digoxigenin, cholic acid), it is significant to note that the only non-directed C–H hydroxylations at this position have been effected with (PDP)-inspired small molecule catalysts.[28, 37, 38]

Conclusion

Recognition of the benefits of moving away from flat architectures in drug design is stimulating the incorporation of increasing C(sp)–H bonds into aromatic medicinal compounds. The development of the chemoselective and reactive Mn(CF3–PDP) 1 catalyst system described herein enables for the first time the strategic advantages of late-stage aliphatic C–H oxidation to be leveraged in these settings. Following the same predictable selectivity rules first established with Fe(PDP) and Fe(CF3–PDP) catalysts, the Mn(CF3–PDP) 1 catalyst system site-selectively oxidizes methylene sites distinguished by subtle differences in electronic, steric, and stereoelectronic environments in a range of carbocyclic and linear alkane structures. In sharp contrast to all previous small molecule C–H oxidation systems, Mn(CF3–PDP) 1 affords preparative methylene oxidations in medicinally relevant aromatic compounds substituted with halogen, oxygen, nitrogen, heterocyclic, and biaryl moieties. We anticipate that this discovery will benefit future catalyst design in developing chemoselective and reactive aliphatic C–H oxidation catalysts. Additionally, small-molecule therapeutics will be empowered with 1 to rapidly diversify aromatic drugs and natural products and quickly identify their metabolites. Future studies will probe the mechanism that biases the relatively simple Mn(CF3–PDP) 1 catalyst system towards methylene C–H oxidation and impedes non-productive aromatic oxidation.

Methods

Method A: Single Catalyst Addition Protocol

A 40 mL vial was charged with substrate (0.3 mmol, 1.0 equiv.), Mn(CF3–PDP) 1 (0.03 mmol, 10 mol%), ClCH2CO2H (425 mg, 4.5 mmol, 15.0 equiv.) and a stir bar. Acetonitrile (MeCN, 0.6 mL, 0.50 M) was added along the wall to ensure all compounds were washed beneath the solvent level and the vial was sealed with a screw cap fitted with a PTFE/Silicone septum. The vial was cooled to 0 °C with an ice/water bath. A separate solution of H2O2 [(204 mg, 3.0 mmol, 10.0 equiv.), 50% wt. in H2O, purchased from Sigma-Aldrich] in MeCN (3.75 mL) was loaded into a 10 mL syringe fitted with a 25G needle and was added dropwise to the stirring reaction over 3 hours via a syringe pump (1.25 mL/h addition rate) while maintaining the reaction vial at 0 °C. Upon completion, the reaction mixture was concentrated to a minimum amount of solvent. The residue was dissolved in ~20 mL dichloromethane (DCM) and washed with 9 mL sat. NaHCO3 solution (CAUTION: CO2 released) to remove ClCH2CO2H. The aqueous layer was extracted with ~15 mL DCM two times and the combined organic layer was dried with Na2SO4. The filtrate was concentrated and purified by flash chromatography on silica gel.

Method B: Iterative Catalyst Addition Protocol

A 40 mL vial was charged with substrate (0.3 mmol, 1.0 equiv.), Mn(CF3–PDP) 1 (0.015 mmol, 5 mol%), ClCH2CO2H (425 mg, 4.5 mmol, 15.0 equiv.) and a stir bar. MeCN (0.6 mL, 0.50 M) was added along the wall to ensure all compounds were washed beneath the solvent level and the vial was sealed with a screw cap fitted with a PTFE/Silicone septum. The vial was cooled to –36 °C with a 1,2-dichloroethane/dry ice bath or to 0 °C with ice/water bath. A separate solution of H2O2 [(204 mg, 3.0 mmol, 10.0 equiv.), 50% wt. in H2O, purchased from Sigma-Aldrich] in MeCN (3.75 mL) was loaded into a 10 mL syringe fitted with a 25G needle and was added dropwise to the stirring reaction over 3 hours via a syringe pump (1.25 mL/h addition rate) while maintain at the corresponding temperature. The initial time is recorded as the time the first drop of H2O2 solution was added into the reaction. One hour after the initial time, another batch of catalyst (0.015 mmol, 5 mol%) was dissolved with 0.1 mL MeCN in a 0.5-dram vial and added dropwise into the reaction via syringe followed directly by another 0.1 mL MeCN that was used to rinse the vial. The addition of 5 mol% catalyst was repeated at two hours after the initial time using the same procedure. A total of 15 mol% of catalyst was used in this protocol. Upon completion, the reaction was worked up and purified as described in Method A.

Method C: Slow Catalyst Addition Protocol

A 40 mL vial was charged with substrate (0.3 mmol, 1.0 equiv.), ClCH2CO2H (425 mg, 4.5 mmol, 15.0 equiv.) and a stir bar. MeCN (0.6 mL, 0.50 M) was added along the wall to ensure all compounds were washed beneath the solvent level and the vial was sealed with a screw cap fitted with a PTFE/Silicone septum. The vial was cooled to –36 °C with 1,2-dichloroethane/dry ice bath or to 0 °C with ice/water bath. A 1.0 mL syringe was filled with a solution of Mn(CF3–PDP) 1 (0.03 mmol, 10 mol%) in MeCN (0.375 mL, 0.083 M). A few drops of this solution were added to the reaction. A 10 mL syringe was filled with a solution of H2O2 (204 mg, 3.0 mmol, 10.0 equiv., 50% wt. in H2O, purchased from Sigma-Aldrich) in MeCN (3.75 mL, 0.8 M). Both syringes were fitted with 25G needles and loaded to a syringe pump resulting a slow simultaneous addition of catalyst and oxidant solutions over 3 hours while maintain at the corresponding temperature (1.25 mL/h addition rate for the H2O2 syringe; 0.125 mL/h for the catalyst syringe). Upon completion, the reaction was worked up and purified as described in Method A.

Synthetic procedures for Mn(CF3–PDP) 1 and for all substrates in the manuscript are available in the Supplementary Information Section II to IX.