Species-specific lipophilicities of fluorinated diketones in complex equilibria systems and their potential as multifaceted reversible covalent warheads

The study addresses the design of reversible covalent inhibitors that target non-catalytic residues, particularly cysteine, to improve specificity and reduce off-target toxicity associated with irreversible covalent drugs. Fluorinated ketone warheads (di- and trifluoromethyl ketones) can form reversible covalent adducts (e.g., hemiketals) and often exist as hydrates in aqueous environments. Although such warheads have been mostly exploited at enzyme catalytic sites, targeting non-catalytic cysteines is rare yet promising. Drug design must balance target engagement with physicochemical properties, especially lipophilicity (log P), which governs ADME. For warheads existing as equilibrating species (keto, hydrate, hemiketal, enol), species-specific lipophilicity is critical but challenging to determine. Building on a 19F-NMR method enabling direct, species-specific log P measurements for α-fluorinated ketones, this work proposes that geminal fluorinated diketones CF2(CO)2 can serve as multifaceted reversible covalent warheads whose equilibrating species span a wide lipophilicity range and display flexible regio-reactivity toward nucleophiles such as cysteine.

Prior work established reversible inhibition by α-fluorinated ketones (di- and trifluoromethyl ketones) forming hemiketals at catalytic sites in enzymes including acetylcholinesterase, phospholipases, and viral serine proteases. α-Difluorinated carbonyl functions have served as protease inhibitors, with emerging examples against SARS-CoV-2 main protease. A rare example of targeting a non-catalytic cysteine with an aromatic trifluoromethyl ketone warhead in a reversible kinase inhibitor highlighted this strategy’s potential. Medicinal chemistry has systematically studied how fluorination patterns modulate lipophilicity, polarity, and H-bonding, yet such analysis is largely absent for targeted covalent inhibitors. A 19F-NMR method enabling species-specific log P determination for α-fluoro ketones and their hydrates provides a foundation to extend such measurements to more complex, equilibrating warhead systems.

- Compounds: Model FDKs 1–6 were studied to probe effects of fluorination patterns (symmetric/asymmetric CF2(CO)2, difluorostatone, mono- and terminally fluorinated diketones). Compounds 1–4 were synthesized using Selectfluor; 5 and 6 were commercial. - Species identification: Equilibrium compositions were characterized in water and in octanol saturated with water using 19F- and 13C-NMR. Assignments leveraged characteristic chemical shift trends (hydrated forms upfield vs ketones) and, in octanol, AB quartet patterns for hemiketals due to diastereotopic fluorines. 13C-NMR supported differentiation of regioisomeric gem-diols/hemiketals where 19F-NMR overlap occurred. - Species-specific log P: A direct 19F-NMR-based “stir-flask” method was used. After equilibrating biphasic octanol/water mixtures, species concentrations in each phase were quantified by integrating species-specific 19F signals relative to a reference (trifluoroethanol, known log P). Species-specific log P values were calculated from integration ratios. Detection is limited for species with log P outside roughly −4 to 4 or for species present in minute amounts. - 19F-DOSY-NMR: Employed for complex systems (notably FDK 2 and 1) to corroborate signal assignments and probe species–solvent interactions (inter- and intramolecular H-bonding) via diffusion coefficients in water and octanol. - DFT calculations: Performed at the M06-2X/6-311++G(d,p) level (gas phase) to examine conformational preferences, dipole moments, NBO charges (H-bond basicity/acidity), and intramolecular H-bonds for diketones and hydrated/enol tautomers (FDKs 1, 2, 5, 6 and analog 7). XYZ coordinates provided in Supplementary Data. - Membrane partitioning: Solid-state 19F-MAS-NMR monitored partitioning of 1 and its hydrates between water and artificial membranes or POPC liposomes to assess equilibrium shifts and species accumulation within membranes. - Reactivity with nucleophiles: Chemoselectivity and regioselectivity of asymmetric FDK 2 toward nucleophilic amino acid models were studied in DMSO-d6 and in neat 2-mercaptoethanol by 19F-NMR, including reactions with protected cysteine (P-Cys) and serine (P-Ser) to form hemi-thioketals/hemiketals.

- Complex equilibria resolved: Up to six species for symmetric diketone 1 (1, 1-D, 1-T, 1-HK, 1-HKD, 1-DHK) and nine species for asymmetric diketone 2 (keto, two gem-diols 2-Da/2-Db, tetrol 2-T, two hemiketals 2-HKa/2-HKb, two hemiketal–diols 2-HKaD/2-HKbD, and a di-hemiketal 2-DHK) were identified by NMR. - Species distributions: For 1 in octanol saturated with water: 1 (7.1%), 1-HK (74.5%), 1-D (17.6%), 1-T (0.06%), 1-HKD (0.42%), 1-DHK (0.26%). For 2: 2 (22.7%), 2-HKa (20.9%), 2-HKb (21.2%), 2-DHK (0.38%), 2-HKaD (2.9%), 2-HKbD (2.0%), 2-Da (4.8%), 2-Db (24.5%), 2-T (0.57%). - Species-specific log P (RT): 1 (1.73), 1-D (0.44), 1-T (−1.84). For 2: 2 (2.62), 2-Da (0.96), 2-Db (1.33), 2-T (0.69). For 3 (difluorostatone): 3 (1.75). For terminally fluorinated systems: 5-enol (3.60), 5-D (1.64); 6-enol (3.01), 6-D (0.99). At 37 °C, 2 retained similar values: 2 (2.70), 2-Da (1.28), 2-Db (0.88), 2-T (0.65). - Lipophilicity shifts: First hydration typically reduced log P by about 1.3 units (e.g., 1→1-D; 2→2-Da/2-Db). Full hydration to tetrol caused a large decrease for rigid 1 (Δlog P 1-D→1-T = 2.28) but only modest decreases for flexible 2 (Δlog P 2-Da→2-T = 0.27; 2-Db→2-T = 0.64), attributed to strong intramolecular H-bonding in 2-T. - DOSY evidence: In water, D values decreased with hydration and molecular size (2 highest D; 2-T lowest), with additional modulation consistent with increased H-bonding to solvent by gem-diols. In octanol, hydrated species showed markedly reduced D values vs keto, consistent with strong H-bonding to octanol. Hemiketals had D values similar to gem-diols. Representative D (×10−10 m2/s): water—2: 7.53±1.05; 2-Da: 5.90±0.19; 2-Db: 5.84±0.19; 2-T: 5.40±0.59. Octanol—2: 1.92±0.05; 2-Da: 0.75±0.04; 2-Db: 0.75±0.02; 2-T: 0.58±0.10. - DFT insights: For 1, hydration increases H-bond basicity (gem-diol O charges ca. −0.74 vs carbonyl O ca. −0.49), explaining decreased log P; 1-T adopts an envelope conformation with a single intramolecular H-bond but overall remains hydrophilic due to four OH groups. For 2, hydration increases H-bond basicity substantially; 2-T favors a bicyclo[3.3.1]nonane-like conformation with two strong intramolecular H-bonds, reducing solvent interactions and mitigating log P decrease. For 5 and 6, enol tautomers with intramolecular H-bonds are more stable (by 4.1 and 3.2 kcal/mol, respectively) and less polar, consistent with higher log P. - Membrane partitioning: 19F-MAS-NMR showed equilibrium shifts for 1 in the presence of membranes/liposomes, with accumulation of more lipophilic species 1 and 1-D in membranes and reduced 1-T in aqueous phase, indicating environment-dependent adaptation. - Reactivity with nucleophiles: FDK 2 reacts preferentially with thiols. With 2-mercaptoethanol neat: hemi-thioketals formed predominantly (51% and 41% for two regioisomers), with ~4% hemiketals and ~4% unreacted 2. With protected cysteine in dry DMSO-d6: hemi-thioketals 10 (48%) and 11 (24%) formed, with 26% unreacted 2; addition of P-Ser and water maintained predominant thiol addition. This demonstrates flexible regio-reactivity of CF2(CO)2 toward cysteine.

The findings support the hypothesis that CF2(CO)2-based fluorinated diketones operate as multifaceted reversible covalent warheads. Their equilibria encompass multiple species spanning a broad lipophilicity range, enabling adaptation to diverse biological environments. Species-specific log P data show that hydration state dramatically modulates lipophilicity, allowing hydrophilic forms to predominate in aqueous compartments while lipophilic keto/enol forms can partition into membranes and hydrophobic protein pockets. DOSY and DFT analyses provide a mechanistic basis: hydration increases H-bond basicity (enhancing solvent interactions and lowering log P), whereas intramolecular H-bonding and conformational locking (e.g., in 2-T and enols of 5/6) reduce polarity and solvent binding, maintaining lipophilicity. Solid-state MAS-NMR demonstrates environment-driven redistribution among species, aligning lipophilicity with membrane partitioning. Chemically, the CF2(CO)2 warhead reacts reversibly and regioflexibly with thiols over alcohols, forming hemi-thioketals at either carbonyl, which is desirable for targeting non-catalytic cysteine residues with tunable engagement depending on local microenvironment.

This work introduces CF2(CO)2 fluorinated diketones as multifaceted reversible covalent warheads whose complex equilibria can be deconvoluted to yield species-specific lipophilicities via direct 19F-NMR. Key contributions include: (i) simultaneous identification and quantification of up to nine interconverting species; (ii) robust species-specific log P values demonstrating large, hydration-dependent lipophilicity modulation; (iii) mechanistic insight from 19F-DOSY and DFT into solvent interactions and intramolecular H-bonding; (iv) demonstration of environment-driven membrane partitioning; and (v) proof of flexible, thiol-selective reactivity toward cysteine. These features position CF2(CO)2 as a promising warhead for reversible covalent inhibitors of non-catalytic cysteines. Future research should extend measurements to peptide/drug conjugates, evaluate kinetics and selectivity in protein contexts, quantify membrane partitioning and permeability in cellular models, and explore structure–lipophilicity–reactivity relationships across broader FDK scaffolds.

- Measurement constraints: Species with log P values outside approximately −4 to 4 or present at very low abundance may fall below 19F-NMR detection in one phase, precluding log P estimation (e.g., highly lipophilic hemiketals not observed in water; some systems like 4 lacked common species in both phases). - Spectral complexity: Overlapping signals and low-level species necessitated deconvolution and auxiliary methods (13C-NMR, DOSY), which may introduce uncertainty in minor components. - Modeling scope: DFT was conducted in the gas phase and may not fully capture solvent and biomolecular environments; membrane partitioning experiments were preliminary and performed with model systems. - Generalizability: Studies focused on small-molecule FDK models; translation to larger drug-like entities and in vitro/in vivo efficacy, selectivity, and safety were not assessed.