¹⁹F NMR Interpretation Complications with 3-Bromo-1,1,1-trifluoroacetone (BTFA) and an Alternative 2-Iodo-N-(2,2,2-trifluoroethyl)acetamide (ITFEA) ¹⁹F NMR Probe

Abstract

3-Bromo-1,1,1-trifluoroacetone (BTFA) is a widely used probe for labeling cysteine (Cys) residues to study protein structure by ¹⁹F NMR. We evaluated the mechanism by which BTFA labels transthyretin (TTR). In aqueous solution, BTFA rapidly undergoes hydration to its corresponding geminal diol (gem-diol), thereby establishing an equilibrium that disfavors the ketone, which is the reactive form of BTFA. The −84 ppm ¹⁹F signal commonly observed in BTFA-substructure labeled biomolecules is attributable to the gem-diol, not the ketone as too often claimed. We demonstrate that the BTFA-substructure-Cys conjugate can undergo further reaction with a nearby nucleophile to produce cyclized products, particularly when the targeted Cys is the most N-terminal residue. If the initially labeled Cys is internal to the protein primary structure, ¹⁹F NMR analysis reveals that subsequent intramolecular cyclization reactions with nearby residues are possible (exacerbated by lyophilization or reduced pressure associated with mass spectrometric analysis), which can perturb the native structure, complicating ¹⁹F NMR structural interpretations. We sought an alternative probe whose conjugates do not undergo hydration and are stable enough not to undergo further reactions post Cys alkylation. We introduce 2-iodo-N-(2,2,2-trifluoroethyl)acetamide (ITFEA) for this purpose, which displays similar labeling efficiency as BTFA without the side product risk.

INTRODUCTION

¹⁹F NMR spectroscopy is a powerful tool for studying protein biology under physiological conditions⁴ by enabling the examination of soluble protein structure and dynamics,^{5, 6} integral membrane protein structures,⁷ protein-protein or protein-ligand interactions,⁸ protein distance measurements,⁹ protein phase separation,¹⁰ protein assemblies,¹¹ protein structural studies in live cells^12-15 or biological fluids,¹⁶ and in scrutinizing aberrant processes, including protein misfolding,^{17, 18} aggregation,¹⁹ and fibrillization.²⁰ Compared to other NMR active nuclei, the ¹⁹F nucleus exhibits certain advantages: with 100% natural abundance, it is a highly NMR-sensitive nucleus, and its large chemical shift dispersion (>400 ppm), paired with its exceptional environmental responsiveness, renders ¹⁹F NMR incredibly powerful in detecting subtle changes in protein structure or in environmental changes.

Since ¹⁹F does not occur naturally in proteins, it must be incorporated into proteins either biosynthetically or through a chemical reaction.²¹ While biosynthetic labeling is well-established, the methodology requires specialization.²² In contrast, chemical labeling is straightforward and provides a low-cost, easily executed option for NMR spectroscopists.²³ Several chemical probes have been designed, synthesized, and scrutinized for biomolecular ¹⁹F NMR applications.^24-27

Cysteine (Cys) is the most targeted residue for site-specific ¹⁹F protein labeling due to the high nucleophilicity of the sulfhydryl (−SH) side chain. The lysine side chain and the N-terminus²⁸ can also be targeted to install a fluorine-based probe, whereas serine and threonine conjugation is rarer.²⁹ Among the Cys-specific protein labeling probes, 3-bromo-1,1,1-trifluoroacetone (BTFA, 1) is one of the most frequently used ¹⁹F-based probes (Figure 1a).^{30, 31} Probe 1 is commercially available, small, reactive, and its ¹⁹F NMR resonance lacks detectable proton coupling. Since its introduction in 1978,³² probe 1 has been applied to study numerous protein systems, including protein-liposome interactions,³³ protein formulations,²² protein dynamics,^{34, 35} membrane proteins,^{36, 37} protein-ligand binding,^38-40 enzymatic reactions,⁴¹ and protein misfolding.^{17, 18} By relying on the single resonance of the trifluoromethyl group of probe 1 in protein conjugates, NMR spectroscopists use this probe to assess homogeneity or to discern structural alternatives associated with protein events.

Figure 1.

Overview of the main findings described herein. (a) Although hydration is well-documented for trifluoromethyl ketones, its presence is infrequently noted within the ¹⁹F NMR literature when BTFA (1) is used.^1-3 (b) A novel reaction pathway of probe 1 has been revealed when the targeted Cys is at the N-terminus that leads to undesired side products, such as dihydro-1,4-thiazine (3). (c) Probe design inspired by iodoacetamide (IAA, 4) led to the demonstration that 2-iodo-N-(2,2,2-trifluoroethyl)acetamide (ITFEA, 5) is a suitable ¹⁹F biomolecular NMR probe.

Despite its widespread use, probe 1 exhibits drawbacks owing to its inherent chemical reactivity. For example, it is well known that trifluoromethyl ketones readily form hydrates in aqueous solution (Figure 1a).^{1, 42} We observed two ¹⁹F NMR signals from probe 1 in organic solvents with trace water due to population of both the ketone and geminal diol (gem-diol) forms, and under certain conditions this equilibrium has the potential to obfuscate ¹⁹F NMR structural interpretations, even though the ketone population is very low in aqueous buffer.^1-3 For example, we report previously undisclosed reactions that can occur after Cys alkylation with this class of labeled proteins (e.g., the formation of product 3 when the Cys is at the N-terminus of a protein labeled with probe 1, Figure 1b). Because of the ketone–geminal diol equilibrium inherent to probe 1 Cys labeling, if the conformation of a given protein positions a suitable nucleophile proximal to the ketone present at a very low population, a vinyl trifluoromethyl ring can form upon dehydration, especially after lyophilization or in the high vacuum environment of a mass spectrometer. With these observations in mind, we were prompted to screen for new alternatives to probe 1 that similarly offered a highly chemoselective reaction toward Cys in a protein, yet were devoid of further reactivity and propensity for side product formation. Through these efforts, we identified 2-iodo-N-(2,2,2-trifluoroethyl) acetamide (ITFEA, 5) as an alternative reagent that possesses these characteristics and whose straightforward ¹⁹F NMR probe design is inspired by the global Cys capping reagent iodoacetamide (4, Figure 1c).

EXPERIMENTAL METHODS

General Methods

Unless otherwise noted, all materials were used as received from commercial sources without further purification. 3-Bromo-1,1,1-trifluoroacetone (BTFA), dipeptide cysteinylglycine (category # C0166), and human serum albumin (HSA) were purchased from Sigma-Aldrich (St. Louis, MO, USA), and 2-bromo-N-(2,2,2-trifluoroethyl)acetamide and 2-iodo-N-(2,2,2-trifluoroethyl)acetamide were purchased from Enamine US (Monmouth Junction, NJ, USA). Recombinant wild-type TTR (WT-TTR) and a mutant form of TTR (C10A/S85C) were expressed and purified from Escherichia coli cells transformed from the pMMHa plasmid as described previously.^43-44 The concentrations of all the protein stock solutions were determined by UV absorption using NanoDrop^™ (Thermo Scientific). Anhydrous solvents (acetonitrile, dichloromethane, N,N-dimethylformamide, tetrahydrofuran, toluene, triethylamine) were purchased from Acros Organics^® (extra dry over molecular sieves, AcroSeal^™) and used as received. Deuterated solvents were purchased from Cambridge Isotope Laboratories, Inc (Tewksbury, MA, USA).

Protein NMR spectra were recorded on a Bruker Avance 700 MHz spectrometer with a ¹H/¹⁹F-¹³C/¹⁵N TCI cryoprobe and shielded z-gradient coil. Each 1D spectrum was acquired with 4k complex points with an acquisition time of 155 ms. The carrier offset was set at −83.3 ppm. The recycle delay was set at 1.0 sec. Unless otherwise noted, 10% D₂O was included as the lock signal in protein NMR experiments, and the temperature was 25 °C (RT). NMR buffer used for protein NMR is 10 mM potassium phosphate and 100 mM KCl at pH 7.0. ¹⁹F NMR spectra were recorded on a Bruker Avance 700 MHz spectrometer (as detailed above) without ¹³C decoupling, using either an external reference (CCl₃F: δ_F = 0 ppm) or an internal reference (CF₃CO₂H: δ_F = −75.51 ppm in D₂O).

BTFA Labeling

C10A/S85C transthyretin labeling by BTFA (1)

100 μL of transthyretin sample (100 μM) in 1× PBS were added with 2 μL of BTFA stock solution in acetonitrile (50 mM) to make a sample solution containing 100 μM protein and 1 mM BTFA reagent. The protein was incubated with BTFA reagent for 1 h at room temperature. The excess BTFA reagent was then removed by Zeba^™ spin desalting column (7 kDa molecular weight cut-off) according to the manufacturer protocol. The labeled protein was analyzed by LC-MS to confirm the BTFA modification. The 1× PBS buffer was prepared by dissolving one PBS tablet in 200 mL water to obtain a 137 mM NaCl, 2.7 mM KCl and 10 mM phosphate buffer solution (pH 7.4 at 25 °C).

Cysteinylglycine (7) labeling by BTFA (1) and ITFEA (5)

Cysteinylglycine (3 mM, 1 eq) was incubated with BTFA (3 mM, 1 eq) in PBS buffer (0.5 mL, 40 mM phosphate buffer, pH=7.4) for 1 h at RT with shaking. The reaction mixture was then immediately purified on an Agilent Auto purification LC with a Phenomenex C18 column (5 μm, 110Å, 150 × 21.2 mm) using a 0.1% aqueous TFA: ACN gradient (20 mL/min, main segment 0–80% ACN + 0.1% TFA over 10 min) at ambient temperature. 9 compounds were separated by peak at 210 nm absorbance. A sample of each (20 μL) was analyzed on an Agilent 1260 Infinity II LC/MS with an Agilent C18 column (1.8 μm, 80Å, 2.1 / 50 mm) using a 0.1% aqueous FA: ACN gradient (0.5 mL/min, main segment 20–95% ACN + 0/1% FA over 9.5 min) at ambient temperature. Both ESI+ and ESI- modes were used to search for m/z values correlating with the reaction components. A sample of each (0.45 mL) was taken and D₂O (50 μL) was added to each sample before ¹⁹F NMR was taken on the Avance-700 (300 scans per sample) as described previously (General Information). The nine compounds were then subjected to lyophilization to dryness before being resuspended in NMR buffer with 10% D₂O (0.5 mL). These samples were then subjected to ¹⁹F NMR spectroscopic analysis. This method was repeated for reagent 5.

Acetylcysteinylglycine (S1) labeling by BTFA (1)

As a control, S1 (3 mM, 1 eq), was incubated with 1 (3 mM, 1 eq) in PBS buffer (40 mM phosphate buffer solution, pH=7.4). After 1 h, the reaction solution was analyzed by LC-ESI-MS, ¹H NMR, and ¹⁹F NMR.

HSA labeling by BTFA (1) and ITFEA (5)

HSA (50 μM) was reduced by TCEP·HCl (5 mM) in PBS buffer (475 μL, pH=7.4) at 37 °C with shaking for 30 min. Excess TCEP·HCl was removed by a PD25 spin desalting column (Cytiva) according to the manufacturer’s protocol. The reduced HSA solution was then incubated with 1 (5 mM, 558 μL final volume) for 1 h at room temperature, followed by the addition of D₂O (58.3 μL) to obtain a final concentration of 10% (v/v). The sample was then directly analyzed by ¹⁹F NMR, with a total of 10,000 scans. The same sample was then subjected to lyophilization overnight to dryness. Post-lyophilization, the sample was resuspended in the same solution (PBS buffer + 10% D₂O) and then directly analyzed by ¹⁹F NMR, with a total of 10,000 scans to obtain sufficient S/N. This method was repeated for reagent 5.

Protein Labeling with Non-BTFA Reagents

C10A/S85C TTR labeling experiments with 3-Bromo-1,1,1-trifluoropropan-2-ol (BTFP, 13), 2-bromo-N-[4-(trifluoromethyl)phenyl]acetamide (BTFMA, 14), N-[4-bromo-3-(trifluoromethyl)phenyl]acetamide (3-BTFMA, 15), BTFA ketal (6), 2-[methyl(trifluoromethyl)amino]-2-oxoethyl 4-methylbenzenesulfonate (16), 2-iodo-N-methyl-N-(trifluoromethyl)acetamide (17), 2-bromo-N-(2,2,2-trifluoroethyl)acetamide (18), 2-iodo-N-(2,2,2-trifluoroethyl)acetamide (5), and 3,3,3-trifluoropropyl 2-bromoacetate (19) were performed as follows: C10A/S85C TTR (50 μM) was incubated with the −CF₃ labeling reagent (500 μM) in PBS (pH 7.4, 50 μL) for 24 h (or 1 h in the case of 5) at room temperature. Zeba^™ Spin Desalting Columns (7k MWCO, Thermo Scientific) were used to remove excess labeling reagent. A sample was then taken and diluted (1 in 10) before injection (1 μL) onto Agilent 6230 TOF LC/MS for analysis, using the described method below. C10A/S85C TTR labeling experiments with 2,2,2-trifluoroethyl-1-thiol (TFET, 12) were performed as follows: C10A/S85C TTR (10 μM) was incubated with DPS (500 μM) in PBS (pH 7.4, 49.5 μL) for 6 h in the dark at RT. Then, TFET was added (final concentration: 100 μM) in PBS (pH 7.4, final volume: 50 μL) for 24 h at room temperature. Zeba^™ Spin Desalting Columns (7k MWCO, Thermo Scientific) were used to remove excess labeling reagent. A sample (1 μL) was then taken and diluted in ddH₂O (10 μL) before injection (1 μL) onto Agilent 6230 TOF LC/MS for analysis, using the described method below. For reagents 5 and 17, the ¹⁹F NMR experiments were performed as follows: C10A/S85C TTR (50 μM) was incubated with the −CF₃ labeling reagent (1 mM) in NMR buffer (0.5 mL) for 24 h at room temperature. PD MiniTrap G-25 was used (spin protocol, Cytiva) to remove excess labeling reagent. Then D₂O (50 μL) was spiked into the desalted sample and subjected to ¹⁹F NMR spectroscopic analysis.

Mass Spectrometric Analysis

LC/MS analysis of wild-type TTR, C10A/S85C TTR, peptides, and their BTFA conjugates were conducted on an Agilent 6230 TOF LC/MS System with a Dual ESI source connected to an Agilent 1260 LC stack with a DAD detector fitted with an Agilent PLRP-S (1000 Å, 2.1 × 50 mm, 5 μm). The various species were separated with 5% buffer B (ACN + 0.1% TFA) isocratic over 2 min, followed by a linear gradient of 5% to 90% buffer B over 13 min, and finally 90% buffer B isocratic for 4 min (0.3 mL/min flow rate). Unmodified C10A/S85C elutes around 11–12 min under these conditions. Mass spectra were processed and deconvoluted using the MassHunter Workstation, and the modification percentage was determined using MassHunter Qualitative Analysis.

RESULTS AND DISCUSSION

BTFA Hydration States Can Obfuscate ¹⁹F NMR Interpretations

We first investigated gem-diol formation by incubating probe 1 in either 5% or 10% PBS in DMSO at pH 6.0, 7.4, or 8.0, respectively (Supporting Figure S1). The absorbance at 305 nm was used to measure ketone disappearance, revealing fast hydration kinetics, as expected. In 10% PBS (pH 7.4), probe 1 has a hydration t_1/2 of 280 s, whereas at a lower water concentration (i.e., 5% PBS, pH 7.4), hydration is slower with a t_1/2 of 430 s. We also found that probe 1 forms crystals during storage over time, and these were subjected to X-ray crystallographic analysis to unambiguously confirm the gem-diol structure of 2 (Supporting Figure S1b, c). These findings corroborate the proclivity of probe 1 to undergo hydration as previously reported by other research groups employing probe 1.² Since sample preparation for ¹⁹F NMR analysis requires (at minimum) minutes of time (> 60 s) and as protein labeling experiments are mainly performed in > 99% aqueous buffer solution (often 0.1% DMSO) over tens of minutes, we expect probe 1 to largely exist in its hydrated form during the ¹⁹F NMR labeling experiments. Therefore, the ketone population is negligible under standard ¹⁹F NMR analysis. However, the largely undetectable population of the ketone form of probe 1 is the species that is susceptible to nucleophilic attack by a Cys or related nucleophilic side chain in a protein (see results below on the ketal form of BTFA, probe 6).

Given the broad chemical shift range of ¹⁹F, probes 1 and 2 should be well resolved in the ¹⁹F NMR spectrum. To interrogate this, we recorded ¹⁹F NMR spectra of probe 1 in CDCl₃ (Figure 2a) and in 90% NMR buffer / 10% D₂O (NMR buffer = 10 mM aqueous potassium phosphate, 100 mM aqueous KCl, pH 7.0; Figure 2b). In CDCl₃ the major peak is a singlet at −76.3 ppm, corresponding to probe 1 (Figure 2a). Presumably, trace water in the CDCl₃ solvent hydrates a small amount of probe 1, reflected by the minor peak at −84.2 ppm (Figure 2a). Conversely, probe 1 undergoes complete hydration in 90% NMR buffer / 10% D₂O, wherein only probe 2 is observed (Figure 2b).

Figure 2.

¹⁹F NMR analysis of BTFA (1) and BTFA-conjugated transthyretin (TTR) mutant (C10A/S85C) protein compared to that of ITFEA (5). (a) ¹⁹F NMR spectrum of probe 1 in CDCl₃ shows two distinct peaks: one at −84.2 ppm, corresponding to probe 2, and the major peak at −76.3 ppm, corresponding to probe 1. The gem-diol at −84.2 ppm likely results from trace water in CDCl₃ from exposure to air. Inset at the top left is to define the species C10A/S85C TTR–1-sub. This nomenclature (sub) is used to distinguish between original probe structure 1 and the remaining substructure after labeling, 1-sub. (b) ¹⁹F NMR spectrum of probe 1 in 90% NMR buffer/10% D₂O. A single peak is observed at −83.3 ppm, indicating complete hydration of probe 1 in aqueous buffer. (c) ¹⁹F NMR spectrum of BTFA-conjugated C10A/S85C TTR in 90% NMR buffer/10% D₂O. No peak around −76.3 ppm was observed, indicating that probe 1 is fully hydrated in aqueous buffer. When conjugated to a biomolecule, the attached BTFA substructure generates a ¹⁹F NMR peak at −84.2 ppm, corresponding to the gem-diol form (i.e., −83 → −84 ppm).^{1, 2} (d) ¹⁹F NMR spectrum of probe 5 in 90% NMR buffer/10% D₂O exhibits a sharp, single peak at −72.0 ppm. (e) Protein labeling was repeated as in (c) with probe 5, showing the labeled protein peak at −71.5 ppm.

It is well-established that nucleophilic Cys side chains in proteins readily displace bromide from the α-carbon of probe 1 chemoselectively and in high yield. After loss of HBr following protein conjugation, the reaction theoretically yields a product with a 110 Da mass increase. A mutant (C10A/S85C) of transthyretin (TTR) protein was previously reacted with probe 1 and studied by ¹⁹F NMR.^{17, 18} Therefore, we used this TTR mutant as our model protein to explore TTR labeling by probe 1 since we also have experience producing very pure TTR.

The major product resulting from the reaction between probe 1 (the ketone 1 is the reactive probe, vide infra) and either C10A/S85C or wildtype TTR, affords a 128 Da mass increase (see Supporting Figures S2 or S3), corresponding to the hydrate of the remaining BTFA substructure (see inset to Figure 2c for a line drawing). This indicates that the predominant conjugate is C10A/S85C TTR covalently linked through Cys85 to the substructure of probe 2 remaining after conjugation (i.e., C10A/S85C TTR–2-sub, where sub refers to the remaining substructure of probe 2), not C10A/S85C TTR–1-sub (see Figure 2a, c insets for line drawings of potential products). Greater than 90% of the products quantified by mass spectrometry for the reaction between probe 1 and C10A/S85C TTR are the hydrate C10A/S85C TTR–2-sub (see inset to Figure 2c for line drawing and Supporting Figure S2). More C10A/S85C TTR–1-sub is observed by mass spectrometry than by solution ¹⁹F NMR because the high vacuum conditions associated with mass spectrometry support dehydration to the ketone. The reaction between probe 1 and wildtype TTR (harboring one Cys at position 10) also afforded greater than 90% hydrate according to mass spectrometry characterization (Supporting Figure S3). To further scrutinize the form of the substructure of probe 1 that exists after protein conjugation, C10A/S85C TTR (50 μM) was incubated with probe 1 (1 mM) in NMR buffer for 1 h at room temperature. A PD25 spin column was used to remove excess probe 1 prior to D₂O injection (10% v/v), followed by ¹⁹F NMR analysis. Two chemical shifts at −83.3 ppm and −84.2 ppm, corresponding to free probe 2 and the C10A/S85C TTR–2-sub conjugate, respectively, were observed (Figure 2c). This observation favorably aligns with literature reports, which demonstrate that BTFA-protein conjugates typically resonate around −83 / −84 ppm in buffer, but are often misassigned as the ketone (1-sub) and not the corresponding hydrate (2-sub).^{17, 18, 27, 35, 38, 39}

Investigation into BTFA’s Labeling Mechanism using a Transthyretin Mutant

Given the rapid hydration of probe 1 in aqueous solution, it is possible that probe 2 is the form reacting with protein sulfhydryl groups. To test this hypothesis, we synthesized a ketal form of BTFA (6) and incubated it with C10A/S85C TTR in 10-fold excess for 24 h before analysis by LC-MS (Figure 1).

We do not observe any TTR labeling, i.e., no C10A/S85C TTR–6-sub is detected by LC-MS (Figure 1a). Furthermore, when a 100-fold excess of ketal 6 is employed, the major species eluted (Figure 1b) is still identified as unlabeled TTR dimer (Figure 1c), although a minor population of singly labeled TTR dimer (8%) is also observed, i.e. C10A/S85C TTR–6-sub (Figure 1a, c). This result supports the hypothesis that ketone 1 is the form of BTFA that is susceptible to nucleophilic attack by Cys, resulting in high labeling efficiencies. While probe 1 is present at a negligible concentration in aqueous conditions, undetectable by ¹⁹F NMR (Figure 2b), it appears to be the reactive form in equilibrium with probe 2. Therefore, although the ketone form of probe 1 is largely undetectable, it is the species that is susceptible to nucleophilic attack by a Cys. It is important to note, however, that the ketal in probe 6 and the gem-diol in probe 2 are different chemical species, with the latter being more electron-deficient at the α-carbon.

BTFA is Capable of Reactions Beyond Cys Alkylation, Leading to Cyclized Products

By virtue of its strongly electron-withdrawing trifluoromethyl group directly attached to its ketone substructure, probe 1 is very reactive and thus predisposed to nucleophilic attack by even weak nucleophiles such as water, exemplified above. Therefore, we hypothesized that in cases where there is a nucleophilic atom 5– or 6–atoms away from the carbonyl carbon (e.g., an amino group (−NH₂)) following Cys alkylation by probe 1 within a peptide / protein conjugate, nucleophilic attack followed by elimination of water to afford a cyclized peptide / protein could reasonably ensue.⁴⁵ We tested this hypothesis by reacting the dipeptide cysteinylglycine (7) that comprises two nucleophiles, a sulfhydryl group that displaces the bromide of probe 1 and a free −NH₂ at the N-terminus that could readily undergo intramolecular imine formation, with the ketone form of the nascent substructure of 1–7-sub conjugate (9) to generate a six-membered ring (Figure 4).

Figure 4.

Proposed mechanism for the cyclization reaction of dipeptide 7 following treatment with probe 1. The reactive ketone product 9, in equilibrium with gem-diol 8, undergoes a subsequent cyclization reaction with the proximal amino group to form a six-membered hemiaminal intermediate 10, followed by a dehydration step to form cyclic imine 11 en route to the thermodynamically stable dihydro-1,4-thiazine 3 during lyophilization. Molecular weights of the species are shown in red, and the ¹⁹F NMR resonances of the observed species 8 and 3 are shown in blue. HMBC confirmation of cyclic product 3 is highlighted in green.

We reacted dipeptide 7 stoichiometrically with probe 1 in PBS (pH 7.4) for 1 h at room temperature (~23 °C). The reaction mixture was then purified by preparative HPLC and 9 peaks were collected by 210 nm absorbance (Supporting Figure S4a). Each peak was analyzed by LC-MS (ESI+) and ¹⁹F NMR, pre- and post-lyophilization (Supporting Figure S4b). The peak that eluted around 4 min contained two major species (m/z = 271, 307), corresponding to product 3 and intermediate 8, respectively (Figure 4). When analyzing these species by ¹⁹F NMR, we observed a major peak around −84 ppm and a minor peak around −69 ppm (Supporting Figures S4c c.f. S5a). We assigned the −84 ppm peak to intermediate 8, corresponding well with the previously observed shifts of probe 2 (Figure 2). After lyophilization, the ¹⁹F NMR spectrum revealed a shift of the major peak to −69 ppm ((Figure 4 product 3), a result observed multiple times by different scientists), demonstrating that a small amount of product 3 can form in solution, but reduced pressure that facilitates water removal (as in a lyophilizer) will favor the formation of product 3 through a Le Chatelier effect (Supporting Figure S5a). We suspect the dehydration step to form product 3 is irreversible, as after 1 week in aqueous solution, the −69 ppm shift was still present (Supporting Figure S5b). The structure of product 3 was unambiguously confirmed via HMBC (Supporting Figure S6). It is also important to note that cyclization to afford product 3 after reaction by probe 1 is only possible with a free N-terminus (N-terminal acetylation blocks this intramolecular cyclization reaction, Supporting Figure S7).

To further scrutinize the potential for cyclization after Cys conjugation with probe 1 in the context of proteins with multiple proximal nucleophiles, we speculated whether subsequent cyclization could afford larger rings than observed in product 3, or if the imine or enamine could stably form in the context of a protein, as certain protein tertiary structures could facilitate such reactions.⁴⁶ We probed this possibility in human serum albumin (HSA), which comprises 35 cysteines. Given that the Cys-Gly dipeptide-based enamine product 3 possesses a distinct ¹⁹F chemical shift (−69 ppm) relative to its hydrated intermediate 8 (−84 ppm) (Figure 4 and Supporting Figures S4, S5), we probed cyclization while monitoring for enamine formation using ¹⁹F NMR to observe the presence or absence of a chemical shift corresponding to the vinyl trifluoromethyl group (−69 ppm) (Figure 5).

Figure 5.

¹⁹F NMR spectra comparing HSA labeled by probe 1 post lyophilization (a) and HSA labeled by probe 5 post lyophilization (b). We observe a peak at −69 ppm in the case of probe 1, corresponding to the chemical shift observed in Figure 4, suggesting the formation of a vinyl trifluoromethyl group. In the case of probe 5, we observed the expected shifts around −72 ppm for Cys labeling (Figure 2e), and no −69 ppm shift after lyophilization, suggesting the inability of probe 5 to cyclize. X = heteroatom.

HSA was first reduced by tris(2-carboxyethyl)phosphine (TCEP), prior to labeling by probe 1 for 1 h in PBS buffer. Residual label and reducing agent were removed by desalting column, then D₂O (10% v/v) was spiked in before ¹⁹F NMR was taken. We first recorded the ¹⁹F NMR spectrum of probe 1 + HSA in solution, without lyophilization (Supporting Figure S8a). No ¹⁹F signals were detected around −69 ppm, indicating that the vinyl trifluoromethyl group did not form under these standard aqueous conditions. However, we observed a −69 ppm chemical shift from this sample following lyophilization (Figure 5a), correlating with the shift previously observed for product 3 (Figure 4). Proteins other than HSA were not evaluated; however, we suspect that some proteins with Cys residues located proximal to nucleophilic residues (in sequence or 3D space) would be susceptible to this further reactivity under comparable lyophilization conditions, leading to cyclized products. Therefore, this finding is especially important for peptide chemists and mass spectroscopists who rely on lyophilization and/or reduced pressure conditions for mass spectrometry analysis of conjugates of probe 1.

Systematic Evaluation of Known and Newly Designed ¹⁹F NMR Probes

Given that BTFA can undergo further reactions beyond Cys alkylation, especially with mass spectrometry analysis or using solution NMR analysis after lyophilization, we were eager to explore other known or newly designed ¹⁹F NMR probes with a trifluoromethyl group as a fluorine tag and that are designed to avoid further reactivity after labeling. To this end, we evaluated a range of thiol-reactive ¹⁹F NMR probes⁴⁷ (Figure 6) using the C10A/S85C TTR mutant as a model system, containing a single Cys residue per TTR subunit.

Figure 6.

Summary of ¹⁹F NMR probes incubated with C10A/S85C TTR, then analyzed by LC-MS (ESI+). The reactive center is highlighted in orange and the trifluoromethyl tag in blue. The reagents are differentiated according to whether they are literature-known ¹⁹F NMR reagents (1, 12-15) or newly designed (5, 6, 16-19). Within the designed probes, some are synthetic only (6, 16, 17), whereas others are commercially available (5, 18).

All probes, except 2,2,2-trifluoroethyl-1-thiol (TFET, 12, Supporting Figure S9),⁴⁸ were incubated under identical conditions (10 equivalents in PBS buffer, pH 7.4, 24 h) with the C10A/S85C TTR mutant (50 μM). 3-bromo-1,1,1-trifluoropropan-2-ol (BTFP, 13),⁴⁹ a structural analogue of probe 1, exhibited significantly reduced reactivity, with only 11% modification observed (Table 1, Supporting Figure S10). This is likely due to its weaker electron-withdrawing ability relative to the evaluated reactivity of α-haloketones and α-haloacetamides, consistent with the low reactivity of ketal 6.

Table 1.

¹⁹F NMR labeling reagents evaluated using C10A/S85C TTR

19F NMR Reagent	Cys Selectivity (Y/N)	TTR Mutant Modification (%)a
1	Y	90
5	Y	86
6	Y	6
12	N/A	N/A
13	Y	11
14	N	N/A
15	Not reactive	Not reactive
16	Y	2
17	Y	71
18	Y	50
19	N	42

^aPercent labeling determined via LC-MS analysis after a 24 h reaction period. See Supporting Information for details.

Both probes 1 and 13 demonstrated high Cys specificity, yielding exclusively mono-modified TTR mutant species. In contrast, 2-bromo-N-(4-(trifluoromethyl)phenyl)acetamide (BTFMA, 14)^50-51 labeled other residues beyond the anticipated Cys side chain within the TTR mutant subunit (Table 1, Supporting Figure S11). At 10 equivalents, probe 14 resulted in multiple-modified species, despite each TTR mutant subunit containing only a single Cys residue. Thiol reagent 12 was reactive but led to the formation of several species that could not be accurately assigned (Table 1, Supporting Figure S9). Surprisingly, N-(4-bromo-3-(trifluoromethyl)phenyl)acetamide (3-BTFMA, 15)⁵⁰ was completely unreactive toward this TTR mutant under the tested conditions (Table 1, Supporting Figure S12). This is presumably due to its large steric bulk hindering nucleophilic attack and subsequent labeling.

Next, we sought to synthesize novel ¹⁹F NMR probes inspired by probe 1 that retain high reactivity and chemoselectivity toward Cys, but with the added benefit of improved stability in aqueous solution to prevent further hydration, and to form conjugates that are not prone to further side reactions. Taking inspiration from a recently disclosed synthesis of N-trifluoromethyl tertiary amides,⁵² 2-tosyl-N-methyl-N-(trifluoromethyl)acetamide (16) and 2-iodo-N-methyl-N-(trifluoromethyl)acetamide (17) were synthesized and evaluated. Unlike probe 1, where the carbonyl group is directly bonded to the strongly electron-withdrawing −CF₃ group, probes 16 and 17 incorporate an amide linkage directly attached to the −CF₃ group. The difference between these two probes lies with the leaving group: probe 16 has a tosylate, whereas probe 17 has an iodide. The two probes were tested separately by incubating them with the TTR mutant under the same conditions as described above. Unfortunately, probe 16 demonstrated only minimal reactivity towards the C10A/S85C TTR mutant (2% modification, Table 1, Supporting Figure S13) to generate the expected conjugate, presumably due to the steric bulk of the tosylate leaving group. Although probe 17 appeared promising based on mass spectrometry results, displaying 71% modification at the Cys (Table 1, Supporting Figure S14), its ¹⁹F NMR signal was unexpectedly broad (FWHH = 699.71 Hz), perhaps due to an intermediate rate of cis-trans amide bond isomerization (Supporting Figure S15). We hypothesize that the tertiary amide probes (probe 16 and probe 17) show reduced labeling efficiency compared to the ketone and secondary amide probes due to the rotational restriction around the N atom. Furthermore, the synthetic preparation of these compounds was tedious, and probe 17 was a non-solid volatile compound that rendered its handling inconvenient. Thus, these probes were rejected as potential candidates to be considered along with probe 1 as a suitable ¹⁹F NMR probe.

ITFEA is an Alternative ¹⁹F NMR Probe to BTFA

Inspired by the success of probe 17 in modifying the C10A/S85C TTR mutant, we sought other amides possessing a −CF₃ group. We identified ITFEA (5) as a potential ¹⁹F labeling agent, despite the lack of literature precedence for its use in this way. Trifluoromethylphenyl-based iodoacetamides, such as N-(4-trifluoromethylphenyl)iodoacetamide and 4-perfluoro-tert-butyl-phenyliodoacetamide, have historically been applied to ¹⁹F NMR, but their lack of commercial availability, and therefore the synthetic requirements, have limited their utility.^53-56

Interestingly, the corresponding brominated derivative of probe 5, 2-bromo-N-(2,2,2-trifluoroethyl)acetamide (BTFEA, 18), showed reduced labeling efficiency, with around 50% of TTR mutant labeled (Supporting Figure S16); we attribute this to the less stable bromide leaving group. Furthermore, when a structurally similar ester, 3,3,3-trifluoropropyl 2-bromoacetate (TFPBA, 19) was employed, the labeling efficiency further decreased to 42% (Supporting Figure S17). Within the 42% of TTR mutant labeled, there was a small amount of TTR mutant labeled beyond its Cys side chain, suggesting that probe 19 is not Cys-specific. When we screened probe 5 under the same conditions as the other tested ¹⁹F NMR labeling agents, probe 5 showed the highest labeling efficiency of the alternative tested probes (Table 1, Supporting Figure S18). After 24 h, probe 5 labeled 86% of the C10A/S85C TTR mutant, and when repeated at 1 h, probe 5 labeled at least 74% of the C10A/S85C TTR mutant. Therefore, probe 5 has very high labeling efficiency, comparable to that of probe 1. It is important to note that neither probe 1 nor probe 5 went to completion (100% labeling). We hypothesize that this could be due to potential oxidation of the model protein system (C10A/S85C TTR) due to freeze-thaw cycles during its storage (−80 °C) and use (25 °C). Therefore, the small population of oxidized C10A/S85C TTR would not be accessible for labeling, leading to these reduced labeling efficiencies. It is imperative to directly compare probe 1 and probe 5 under the same conditions. Therefore, we then subjected probe 5 to the same set of experiments as probe 1 to investigate whether we observed similar challenges with probe 5 as we do with probe 1.

When we directly compared the ¹⁹F NMR spectra of probes 1 / 2 (Figure 2a, b) with that of probe 5 (Figure 2d), we observed that probe 5 does not detectably undergo subsequent transformations, such as hydration. There is only a single species observed for free 5 (Figure 2d) or for C10A/S85C TTR–5-sub after conjugation (Figure 2e), compared to the two possible hydration states observed by ¹⁹F NMR for 1-sub / 2-sub (Figure 2c). Additionally, the ¹⁹F NMR peaks for free 5 and C10A/S85C TTR–5-sub are resolved (0.5 ppm) similarly to free 2 and C10A/S85C TTR–2-sub (0.9 ppm).

We then investigated the ability of probe 5 to intramolecularly cyclize post-Cys alkylation using both dipeptide 7 and HSA. When dipeptide 7 is reacted stoichiometrically with probe 5 in PBS (pH 7.4) for 1 h at room temperature (~23 °C), we observed the expected singlet peak around −72 ppm (Supporting Figure S19c, cf. Figure 2e). We obtained virtually the same spectrum pre- and post-lyophilization (Supporting Figure S19c-d), demonstrating the inability of probe 5 to cyclize, as expected. When we labeled HSA with probe 5 according to the same method previously described for probe 1, we observed the same expected chemical shifts around −72 ppm, consistent with previous findings (Figure 2e). Furthermore, we observed the same chemical shifts either before (Supporting Figure 20a) or after (Figure 5b) lyophilization. Therefore, probe 5 does not further undergo cyclization reactions, in contrast to our observations with probe 1.

These experiments demonstrate that probe 5 reacts chemoselectively with the Cys85 residue of the C10A/S85C mutant of TTR and is a reliable labeling reagent for ¹⁹F NMR experiments. Moreover, probe 5 exhibits a sharp ¹⁹F resonance in its conjugate (FWHH = 19.63 Hz, Figure 2e), lacks observable ¹H coupling, displays conjugate stability, and is unable to undergo further reactions under similar conditions as probe 1 after conjugation (Figures 4-5). In addition, probe 5 is commercially available and displays convenient physicochemical properties. In this way, probe 5 maintains the benefits of probe 1: small, commercially available, and reactive; while also offering certain advantages: no hydration, no cyclization, and being a non-volatile, easily handled solid. Thus, probe 5 is a very attractive alternative for general use as a ¹⁹F NMR labeling probe.

CONCLUSIONS

In summary, we systematically evaluated the BTFA probe (1), one of the most frequently used ¹⁹F NMR probes for protein structural characterization. BTFA exhibits a highly chemoselective Cys reaction and its conjugates exhibit sharp ¹⁹F resonances. Once the Cys conjugate is formed, we established that the species observed by ¹⁹F NMR is the corresponding hydrate or gem-diol (−84 ppm) of probe 1, not the ketone as too often claimed (−76 ppm). Post-conjugation cyclization is particularly problematic when the Cys is located at the N-terminus with a proximal free amino group, which eventually forms a stable dihydro-1,4-thiazine. While the post-conjugation risk is lower when probe 1 modifies internal Cys residues, the risk increases when reduced pressure is applied to the protein–BTFA conjugate (e.g. lyophilization, mass spectrometric analysis). Users of probe 1 must remain vigilant for the possibility of intrachain cyclization, which will likely become more prominent when favorable protein tertiary / quaternary structures position a nucleophile adjacent to the carbonyl of the conjugate of probe 1 that exists in equilibrium with the hydrate form, especially when the sample is lyophilized or analyzed by mass spectrometry.

These findings prompted the search for new fluorine-containing reagents for peptide or protein conjugation that do not possess the chemical liabilities of probe 1 described herein.^57-64 Our survey of five known ¹⁹F NMR labeling agents along with six previously uninvestigated trifluoromethyl-based small molecules for ¹⁹F NMR labeling provided new insights for ¹⁹F NMR probe design. Encouraged by the stability of our initially synthesized −CF₃ based tertiary amides, we explored structurally related amide derivatives. Gratifyingly, we identified 2-iodo-N-(2,2,2-trifluoroethyl)acetamide (ITFEA, 5) as a competent and worthy candidate that exhibits comparable labeling efficiency but is not capable of undergoing subsequent carbonyl reactions or hydration post-conjugation. Furthermore, probe 5 is a small, commercially available, stable solid. However, because the substructure remaining after probe 1 protein labeling is very rapidly and effectively completely hydrated, as long as the labeled cysteine is not at the N-terminus or in proximity to a nucleophile, BTFA (Probe 1) can still be a good, if not better, ¹⁹F conformational protein probe than probe 5 because it is less bulky (less perturbing) and has fewer rotatable bonds (good for structural measurements). Therefore, probe 1 would be more suitable for certain structural studies compared to probe 5. Vinyl trifluoromethyl ring formation is particularly problematic if the protein is lyophilized before ¹⁹F NMR analysis or if mass spectrometry is used for analysis (vide supra). Collectively, we envision that our findings will encourage practitioners to exercise caution when employing probe 1 as a ¹⁹F NMR labeling reagent and to consider ITFEA (5) as an alternative reagent that also reliably reacts with proteins.

Supplementary Material

Additional experimental details, materials, and methods are available (DOC)

Figure 3.

The labeling of C10A/S85C TTR by ketal 6. (a) Reaction scheme showing the expected mass difference when C10A/S85C TTR is labeled by ketal 6. (b) The total ion count (TIC) from the LC-MS run showing a single peak around 12 min. (c) The deconvoluted mass spectrum of the peak at 12 min, showing a major peak at 27752 Da that corresponds to the mass of the TTR dimer. There is a minor peak at 27906 Da, corresponding to the mass of the TTR dimer being singly labeled by ketal 6. This was quantified as 8% labeling following 24 h incubation with a 100-fold excess of ketal 6.