Mechanisms of chemical protein ¹⁹F-labeling and NMR-based biosensor construction in vitro and in cells using self-assembling ligand-directed tosylate compounds

Abstract

Chemical labeling methods that convert a specific endogenous protein into a semisynthetic biosensor offer numerous new opportunities for biological research and drug discovery. We recently developed a novel protein labeling scheme, termed ligand-directed tosyl (LDT) chemistry, which can site-specifically introduce a synthetic probe to a protein with the concomitant release of the affinity ligand. In previous work, we demonstrated that LDT reagent 1 can be used to modify carbonic anhydrase I (CAI) with a ¹⁹F probe, converting it into a ¹⁹F NMR-based biosensor for CAI inhibitors either in vitro or in red blood cells (RBCs). We herein report the chemical properties of 1, and the mechanisms controlling biosensor construction. It was revealed that the LDT reagent forms self-assembled aggregates in the absence of the target protein. In the aggregated state, nonproductive hydrolysis of the reagent was significantly suppressed, which suggests the potential utility of self-assembly in the design of labeling reagents that have increased stability. In the presence of the target protein, the aggregates were disrupted to form a noncovalent protein–reagent complex, and protein ¹⁹F-labeling proceeded to generate ¹⁹F-labeled CAI. The ligand-binding pocket of the labeled CAI retained the cleaved ligand fragment in vitro, whereas the pocket was vacant in RBC. Further biochemical studies suggested that an anion transporter might play a role in eliminating the cleaved ligand from the interior to the exterior of the cells. The findings provide a fundamental basis for the rational design of reagents applicable to selective protein labeling and biosensor construction in biological contexts.

---

Introduction

Chemical protein labeling (or bioconjugation) is a very attractive strategy for the construction of protein-based biosensors. This semisynthetic approach permits the use of a variety of non-genetically encoded molecules, including organic fluorescent dyes and other biophysical probes such as a NMR-active ¹⁹F nuclide, as reporter units that transduce an analyte-binding event into a detectable signal.¹ The small size of synthetic probes compared to reporter proteins, such as green fluorescent protein and luciferase, is also thought to be advantageous in minimizing loss of protein function. In past decades, much progress has been made in the preparation of semisynthetic biosensors that can specifically detect various biologically relevant species.^2,3 However, the application of these biosensors in intracellular environments has been challenging since semisynthetic biosensors are typically constructed in vitro (in test tubes) first, then transferred into cells, which is technically demanding. Consequently, in recent years, researchers in the chemical biology community have begun developing an alternate (more appealing) approach: the in situ construction of semisynthetic biosensors.⁴ In this approach, a target protein (scaffold) in living cells is specifically modified by a detectable synthetic probe, yielding a labeled protein that allows the monitoring of specific analytes and/or biological events in an in situ manner. The ability to prepare semisynthetic biosensors directly within living systems presents numerous new opportunities in basic biological research, drug discovery and medical diagnosis.⁵

Toward this goal, we recently developed a method, termed ligand-directed tosyl (LDT) chemistry, whereby a target (endogenous) protein can be selectively labeled with a desired probe within its native environment (Fig. 1a).^6,7 Because LDT chemistry is based on the principle of affinity labeling, the labeling site is specific for an amino acid(s) located near the active site of the protein. In addition, in contrast to existing affinity labeling reactions, LDT-based labeling is traceless because the ligand moiety is cleaved off concomitantly with the S_N2-type labeling reaction.⁸ In previous work,⁶ we designed compound 1 as a tool for introducing a ¹⁹F NMR probe into human carbonic anhydrase I (CAI)^9,10 (Fig. 1a). Compound 1 contains a 3,5-bis(trifluoromethyl)benzene derivative (FB) carrying six magnetically equivalent ¹⁹F nuclei connected through a phenylsulfonate (tosyl) ester bond to a protein ligand, benzenesulfonamide (SA),¹¹ which is specific for CA. We demonstrated that CAI could be converted to a ¹⁹F NMR-based biosensor for CA inhibitors not only in a purified form (in vitro), but also in intact red blood cells (RBCs) using 1. However, many details regarding the basic properties of 1, and the mechanisms controlling biosensor construction, remained unexplored.

Fig. 1 (a) Schematic of ligand-directed tosyl (LDT) chemistry for protein labeling and molecular structures of LDT reagents 1 and 2. (b) Schematic of the self-assembling ¹⁹F NMR probe for protein imaging and molecular structure of ¹⁹F-carrying protein imaging probe 3.

The aim of this work was to investigate the chemical properties of 1 and related LDT compound 2 (Fig. 1a), and the mechanisms behind the processes involved in converting native CAI to a ¹⁹F NMR-based biosensor in vitro and in cells. Based on results from our other work (Fig. 1b),¹² it was speculated that 1 would form aggregates in aqueous solution in the absence of CAI. Our earlier study also suggested that the tosyl linkage undergoes both a protein labeling reaction and hydrolysis.^7a We thus first investigated and compared the aggregation properties, labeling efficiencies and hydrolysis rates of the two LDT compounds, and found that 1 self-assembles in aqueous solution and that the formation of aggregates effectively suppresses nonproductive hydrolysis of the tosyl moiety. Next, we investigated the mechanisms controlling in vitro and in cell biosensor construction. The high environmental sensitivity of ¹⁹F NMR chemical shifts allowed the labeling processes to be monitored in real time and the different forms of FB-labeled CAI in its free and complexed state to be characterized. In the in vitro system, CAI labeling with 1 produced FB-modified CAI that remained noncovalently complexed with the cleaved ligand-containing fragment. Removal of the cleaved ligand required an extra purification step, resulting in a CAI-based biosensor with a vacant ligand-binding pocket. In contrast, the in-cell labeling directly generated FB-labeled CAI with an empty ligand-binding pocket, indicating that the ligand-containing fragment is spontaneously removed from labeled CAI in RBCs. Further biochemical and pharmacological studies suggested that an anion transporter in RBCs might play a role in this process by mediating the export of the byproduct. Understanding these aspects will allow us to further modulate the properties and efficacy of LDT reagents and generalize the application of this method to other proteins and other cell types.

Results and discussion

Self-assembly and affinity labeling properties of LDT reagent 1

We previously developed probe 3 for ¹⁹F NMR-based protein detection and imaging (Fig. 1b).^12a Compound 3, designed to be a self-assembling molecule, is comprised of the FB group connected to the CA ligand, benzenesulfonamide, via a stable (unreactive) phenylsulfonamide bond. This probe is NMR-silent in aqueous solution in the absence of CAI due to its self-assembly, but gives a distinct ¹⁹F signal in response to the target protein due to recognition-mediated disassembly of the probe aggregate. Given the high structural similarity between 1 and 3, it was expected that 1 might also form aggregates in aqueous solution.

We first investigated the self-assembly properties of LDT reagent 1 using several spectroscopic techniques. Atomic force microscopy (AFM) showed spherical or oval aggregates of 1 ranging in diameter from 200 to 500 nm (Fig. 2a). Dynamic light scattering (DLS) measurements confirmed the size of the aggregates in a buffer solution (Fig. S1a, ESI†, mean diameter ca. 250 nm). Furthermore, optical density measurements at 600 nm revealed that a solution containing only 1 is turbid, and that turbidity decreases (by 10-fold) upon the addition of CAI (Fig. S1b, ESI†), indicating that the aggregates efficiently collapse through recognition of and binding by CAI. All these properties of 1 were very similar to those of 3 reported previously.^12a

Fig. 2 Self-assembly and protein labeling properties of reagents 1 and 3. (a) AFM image of self-assembled reagent 1 (25 μM) (scale bar, 500 nm). (b, c) MALDI-TOF mass spectra of a reaction mixture containing CAI (150 μM) and 3 (300 μM) (b) or 1 (300 μM) (c). (○) Native CAI (M_w = 28 781); (●) FB-CAI (M_w = 29 082). (d, e) ¹⁹F NMR spectra of a reaction mixture containing CAI (150 μM) and 3 (300 μM) (d) or 1 (300 μM) (e). (△) Complex of CAI and 3; (▽) complex of CAI and 1; (▼) FB-CAI_SA; (×) FB_OHOH.

The most obvious difference between 1 and 3 is in protein labeling: the former is reactive, but the latter is unreactive. After mixing CAI (150 μM) with 1 or 3 (300 μM) in a buffer solution, the labeling processes were monitored by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS). The labeling reaction was essentially complete within 24 h in the case of 1 (Fig. 2c), whereas no reaction occurred in the case of 3, as expected (Fig. 2b). CAI labeling by 1 was completely abolished in the presence of ethoxzolamide (EZA),¹¹ a strong inhibitor of CA, indicating that the labeling occurs via an affinity-driven proximity effect.⁶ Both the self-assembly/disassembly and labeling processes were next monitored by ¹⁹F NMR spectroscopy. First, when 3 was dissolved in a buffer solution containing trifluoroacetic acid (TFA; internal standard at −75.6 ppm), no ¹⁹F signal was observed because of the formation of aggregates (Fig. 2d).^12a However, a sharp signal appeared at −62.6 ppm (△) upon the addition of CAI, apparently due to disassembly of the aggregates and the formation of a protein–probe complex. This signal remained unchanged over 24 h, consistent with 3 being unreactive. Similarly, a solution containing only 1 showed no signal, but gave a sharp signal at −62.3 ppm (▽) in the presence of CAI (Fig. 2e).¹³ In contrast, the signal corresponding to the complex of CAI and 1 disappeared and a new signal appeared at −62.6 ppm (▼) over a 24 h period as the labeling reaction proceeded (Fig. 2e). A new minor signal also appeared at −62.9 ppm (×) during the incubation, due to a FB-containing fragment (FB_OHOH) generated by the hydrolysis of 1 (Scheme 1).¹⁴ Collectively, these results indicate that (1) LDT reagent 1 self-assembles and thus is NMR-silent in aqueous solution in the absence of CAI, (2) the ¹⁹F signal appears upon recognition-mediated disassembly of the aggregates and (3) after complexation with CAI, 1 reacts with an amino acid residue on the surface of CAI, which further induces a change in the ¹⁹F NMR chemical shift. It was also shown that 1 undergoes an intrinsic hydrolysis process.

Scheme 1 Hydrolytic reaction scheme of reagent 1.

Implications of the self-assembly properties of LDT reagents for protein labeling and hydrolysis efficiencies

The self-assembly properties of labeling reagents may influence on protein labeling processes in LDT chemistry. We thus investigated in detail the labeling efficiency of LDT reagents 1 (FB-type) and 2 (biotin-type) (Fig. 1a). MALDI-TOF MS analysis indicated that the yield of CAI labeling after 24 h incubation varied among the reagents: 90 and 32% for 1 and 2, respectively (in the case of 1 : 1 mixture of CAI (50 μM) with 1 or 2 (50 μM)). In these structures, the distance between the ligand and the reactive site (tosylate) is identical. However, it appears that their hydrophobicities greatly differ. In fact, turbidity measurements of the reagents in buffer solution gave significantly different optical densities at 600 nm: 0.074 and 0.001 for 1 and 2, respectively (Fig. 3a). DLS measurements also showed that 1 form aggregates with a mean diameter of 250 nm in buffer solution (as above described), whereas negligible DLS intensity was apparent in the case of 2. These results indicate that 1 is self-assembling LDT reagents, whereas 2 is homogeneously soluble in aqueous solution.

Fig. 3 Self-assembly properties, hydrolysis rates and labeling efficiencies of LDT reagents 1 and 2. (a) Optical density (O.D.) at 600 nm of an aqueous solution containing each reagent (25 μM). Experiments were performed in triplicate to obtain mean and standard deviation values (shown as error bars). (b) Time profiles of hydrolysis of 1 (○) and 2 (◆) in buffer solution at 37 °C. The hydrolysis product SA_OHOH was detected using HPLC. Fitting curves generated using eqn (1) (see Experimental section) are shown as lines. (c) Time profiles of the generation of the SA_OHOH fragment during labeling experiments. The labeling reaction was carried out as shown in (d), and the SA_OHOH fragment was detected using HPLC. Fitting curves generated using eqn (2), given in the Experimental section, are shown as lines. (d) Time profiles of CAI labeling (50 μM) with 1 (○) and 2 (◆) (50 μM) in buffer solution at 37 °C. The labeling reaction was monitored by MALDI-TOF MS analysis. Fitting curves generated using eqn (3), described in the Experimental section, are shown as lines. All experiments in (b–d) were performed in triplicate to obtain mean and standard deviation values (shown as error bars).

LDT reagents intrinsically undergo nonproductive hydrolysis at the phenylsulfonate ester bond in aqueous medium, along with the S_N2 reaction for protein labeling.^7a Indeed, as mentioned above, the hydrolysis product (FB_OHOH) was detected by ¹⁹F NMR in a solution containing CAI and 1 (Fig. 2e). Interestingly, we found that the amount of cleaved fragment SA_OHOH, which is produced by both hydrolysis and protein labeling, was larger than the amount of labeled protein formed, in the case of 2. This observation also suggests that the hydrolysis reaction concomitantly occurs with the protein labeling reaction. We next conducted a kinetic study to evaluate the labeling rates and hydrolysis rates of the LDT reagents in the presence and absence of CAI. The hydrolysis rates of the reagents (k_H) in buffer solution without CAI were determined by monitoring the time-dependent increase in SA_OHOH by HPLC analysis, giving 0.006 and 0.044 h⁻¹, for 1 and 2, respectively (Fig. 3b and Fig. S3, ESI†). Based on the difference in the aggregation property, it is reasonably considered that the value for 1 corresponds to the hydrolysis rate from aggregates (k_H·aggregate), and the value for 2 corresponds to the hydrolysis rate from monomeric state of the LDT reagent (k_H·monomer). Under the labeling conditions (i.e., with CAI), we determined the apparent rates of the formation of the cleaved SA_OHOH fragment, which correspond to the sum of the rates of the labeling and hydrolysis reactions from the noncovalent CAI-reagent complex, i.e., (k_L + k_H·complex) (Fig. 3c). We also evaluated the time profiles of CAI labeling by MALDI-TOF MS analysis (Fig. 3d). Using nonlinear least-squares curve-fitting, the labeling rates (k_L) were determined to be 0.186 and 0.031 h⁻¹, and the rates of the hydrolysis reaction from the complex (k_H·complex) were determined to be 0.030 and 0.038 h⁻¹, for 1 and 2, respectively (Table 1).¹⁵

Table 1 Kinetics constants of the hydrolysis and labeling reactions of LDT reagents 1 and 2^a

	1	2
a k H·monomer: rate of hydrolysis in the monomeric state in the case of 2; kH·aggregate: rate of hydrolysis in the aggregated state in the case of 1 (in the absence of CAI); kL: rate of labeling; kH·complex: rate of hydrolysis in the complexed state (in the presence of CAI). b Not determined because 1 aggregated in aqueous solution. c Not determined because 2 did not aggregate in aqueous solution.
k H·monomer/h^−1	—^b	0.044 ± 0.005
k H·aggregate/h^−1	0.006 ± 0.0008	—^c
k L/h^−1	0.186 ± 0.009	0.031 ± 0.002
k H·complex/h^−1	0.030 ± 0.003	0.038 ± 0.002

---

Scheme 2 summarizes the mechanism and reaction rates of the processes involved in CAI labeling by 1. As mentioned above, we show the rate of LDT reagent hydrolysis in the monomer state using 2. The most striking finding is that simple hydrolysis of 1 in the absence of the target protein was markedly suppressed by self-assembly: the hydrolysis rate (k_H·aggregate) was more than 7-fold slower than when the reagent was in the monomeric state (k_H·monomer, the value of 2). This is most likely due to sequestering of the reactive tosylate moiety within the hydrophobic interior of the self-assembled aggregates. Generally, in protein labeling using electrophilic reagents, the nucleophilic attack by water is inevitable, leading to decomposition of the reagent. Self-assembly might be a promising strategy for suppressing such undesired side reactions during chemical protein labeling. Upon binding with CAI, self-assembled 1 rapidly disassembled to form a noncovalent protein-reagent (1 : 1) complex. At this stage, the hydrolysis of 1 on the complex occurred at a faster rate that approached the hydrolysis rate of monomeric reagent 2, because the tosylate group was now exposed to the aqueous phase. However, the S_N2 reaction of the reactive tosylate functionality by the nucleophilic amino acid present on the CAI surface (His in this case, see below) was 6-fold faster than the competing hydrolysis reaction, resulting in efficient labeling with greater than 90% yield. The hydrolysis rates from the protein–reagent complex were very similar between the two LDT reagents, whereas the labeling rate was larger for the hydrophobic probe 1, compared to 2. Therefore, the labeling efficiency of 1 was greater than 2.

Scheme 2 Overall scheme of the self-assembly, hydrolysis and labeling processes in 1-mediated CAI labeling.

Functions of ¹⁹F-labeled CAI as a ¹⁹F NMR-based biosensor in vitro and in RBCs

As mentioned above, the ¹⁹F NMR spectrum of the post-labeling solution containing CAI and 1 (after 24 h incubation) showed a signal at −62.6 ppm (▼, Fig. 2e and Fig. 4a). The labeled CAI was then purified by gel filtration to remove small molecules including FB_OHOH and SA_OHOH (and unreacted 1, if any). The original signal at −62.6 ppm disappeared and a distinct new signal appeared at −62.0 ppm (□, Fig. 4a). Based on these data, the ¹⁹F signals observed at −62.6 and −62.0 ppm could be assigned to a noncovalent complex of FB-labeled CAI and SA_OHOH (FB-CAI_SA) and the ligand-free, labeled CAI (FB-CAI), respectively. We previously confirmed that a single FB group attaches specifically at His67 of CAI, which is located in the proximity of the active site.⁶

Fig. 4 Conversion of CAI and endogenously expressed CAI (eCAI) to a NMR-based semisynthetic biosensor. (a) ¹⁹F NMR spectra of the post-in vitro-labeling solution containing CAI (150 μM) and 1 (300 μM) before (top) and after (middle) purification by gel filtration. To the purified FB-CAI solution was added AAZ (bottom). (b) ¹⁹F NMR spectra of a RBC suspension containing 1 with (top spectrum) or without (the middle three spectra) EZA. After 48 h incubation, AAZ was added to the post-in-cell-labeling suspension (bottom spectrum). (c) Quantitative NMR detection of AAZ by FB-(e)CAI in vitro (◇) and in RBCs (◆). The response shown on the y axis is defined as the relative area of peak ■ (FB-CAI_AAZ) divided by the sum of the relative areas of peak □ (FB-CAI) and ■ and was plotted against the concentration of AAZ. (d) Molecular structures of CAI inhibitors used in this study. The abbreviations are as follows: EZA, 6-ethoxy-2-benzothiazolesulfonamide; AAZ, acetazolamide; NBS, 4-nitrobenzenesulfonamide; SBA, 4-sulfamoylbenozic acid; BS, benezenesulfonamide; SFA, sulfamic acid; TA, 4-toluenesufonic acid.

We next evaluated the function of FB-CAI as a ¹⁹F NMR-based biosensor. When a strong inhibitor, acetazolamide (AAZ, Fig. 4d),¹¹ was added to the FB-CAI solution, the peak shifted from −62.0 to −63.0 ppm due to the complexation of FB-CAI with AAZ (FB-CAI_AAZ, Fig. 4a). This result clearly indicates that FB-CAI can sensitively readout the presence or absence of a CAI ligand by a chemical shift change in ¹⁹F NMR spectroscopy. Titration with AAZ resulted in a decrease in the ¹⁹F signal of FB-CAI (−62.0 ppm, □) which was synchronized with the increase in the ¹⁹F signal of FB-CAI_AAZ (−63.0 ppm, ■) (Fig. S4a, ESI†). In the case of EZA, a ¹⁹F signal from the complex of FB-CAI and EZA (FB-CAI_EZA) was observed at −62.5 ppm,⁶ whereas no change was observed upon the addition of a non-aromatic-ligand sulfamic acid (SFA, Fig. S4c, ESI†).¹¹ These data strongly indicate that the observed changes in chemical shift were caused by a change in the microenvironment of the labeled FB group, which varies depending on the chemical structure of the inhibitor. The titration curve of the chemical shift change with AAZ showed saturation behavior (Fig. 4c). Similar titration curves were obtained by the addition of other sulfonamide class inhibitors (Fig. S4f, ESI†). Table 2 summarizes the chemical shifts and the association constants of FB-CAI determined from those titration curves for various benzenesulfonamide derivatives. The relative order of the affinity of the tested compounds for FB-CAI was very similar to that for native CAI.¹⁶

Table 2 Chemical shifts and association constants of FB-eCAI and native-CAI for various inhibitors

Inhibitor	Chemical shift/ppm				K app/M^−1
	in vitro		In RBCs		in vitro ^a	In RBCs^a	Native-CAI^b
	Before	After	Before	After
a Determined by nonlinear curve-fitting analysis. b Reported values determined by in vitro assays.^11 c Not determined because no signal change was observed. d Not performed. e No literature value is available.
EZA	−62.0	−62.5	−62.0	−62.5	>10^6	>10^6	1.3 × 10^8
AAZ	−62.0	−63.0	−62.0	−63.0	>10^6	5.9 × 10^4	4.0 × 10^7
NBS	−62.0	−63.2	−62.0	−63.2	>10^6	1.2 × 10^4	1.6 × 10^7
SBA	−62.0	−62.3	−62.0	−62.3	6.4 × 10^5	1.2 × 10^3	2.9 × 10^6
BS	−62.0	−62.6	−62.0	−62.6	5.5 × 10^4	1.0 × 10^4	3.0 × 10^6
SFA	−62.0	—^c	−62.0	—^d	—^c	—^d	4.7 × 10^3
TA	−62.0	—^c	−62.0	—^d	—^c	—^d	—^e

---

In previous work we demonstrated that LDT reagent 1 is cell-permeable and thus applicable to the specific modification of endogenously expressed CAI (eCAI) in human RBCs (note that CAI is a cytosolic protein).^6,17 eCAI labeling in RBCs was monitored by in-cell¹⁹F NMR. At first, when 1 (200 μM) and EZA (1 mM) were added to a suspension of RBCs, no ¹⁹F signal was observed (Fig. 4b). In a separate experiment, we confirmed that 1 did indeed enter the RBCs (data not shown). Since no hemolysis occurred in any of the RBC-based experiments, these results indicate that 1 is NMR-silent even inside RBCs when it is not recognized by eCAI.¹⁸ This phenomenon can be explained by the self-assembly properties of 1 discussed above, and is similar to the result previously obtained using 3.^12a Without EZA, a ¹⁹F signal was clearly observed at −62.3 ppm (▽) in a suspension containing 1 and RBCs, which is identical to the chemical shift obtained in a simple mixture of (purified) CAI and 1 (also see Fig. 2e). This signal, corresponding to the complex of 1 and eCAI, gradually disappeared, and a new signal appeared at −62.0 ppm over a period of 48 h. Interestingly, the chemical shift observed in the cell-based post-labeling solution was different from that obtained in the aforementioned in vitro post-labeling solution, i.e., −62.6 ppm (FB-CAI_SA). Instead, it was consistent with the chemical shift of FB-CAI (−62.0 ppm, □ in Fig. 4a) in the ligand-free state, indicating that RBC-based labeling directly led to the formation of ¹⁹F-labeled eCAI with a free active site pocket inside the cells (more details regarding this phenomenon are discussed in the next section). After intracellular labeling, the cells were washed and incubated with several CA inhibitors. By increasing the extracellular AAZ concentration, a new peak appeared at −63.0 ppm (■, FB-eCAI_AAZ) in a saturating manner (Fig. 4c). Table 2 summarizes the chemical shifts and the association constants of FB-eCAI determined from titration curves generated using various benzenesulfonamide derivatives. Comparison of the in vitro data and cell-based data indicates that the chemical shifts of the inhibitor-bound FB-CAI and FB-eCAI are essentially the same in both settings. On the other hand, the affinity values in the cell-based system were lower than compared with those of native CAI determined by in vitro experiments. This can be explained by lower effective concentrations of the inhibitors inside RBCs, which is due to the limited cell permeability of the molecules as well as their potential nonspecific interactions with other components in the cells such as cell membranes or non-target proteins.

Mechanisms of ¹⁹F NMR biosensor construction in red blood cells

The in vitrolabeling of CAI with 1 generated ¹⁹F-labeled CAI which remained complexed with the cleaved SA_OHOH, i.e., FB-CAI_SA, so that subsequent purification was required for the removal of the SA_OHOH fragment to afford FB-CAI with a free active pocket. These observations initially led us to consider that a purification step is essential for the construction of a NMR biosensor in cells, as is the case in vitro. However, in the RBC experiments, it was found that simple incubation of a RBC suspension with 1 gave rise to direct (in situ) formation of FB-eCAI whose ligand-binding pocket is vacant. Consequently, we were able to convert endogenous CAI (eCAI) into the FB-eCAI biosensor inside the cells without special purification. However, the molecular mechanism of this intriguing phenomenon remained unclear. Thus, we first focused our attention on the relative intracellular concentrations of 1 and eCAI. If eCAI is present in excess compared to 1 inside the cells under the labeling conditions, the SA_OHOH ligand would be transferred to non-labeled, native eCAI via equilibrium, allowing the ligand-binding pocket of the labeled FB-eCAI to be reopened. We investigated this scenario by determining the quantities of eCA and 1 present inside the cells in the labeling suspension (at 0 h). Using densitometric analysis following SDS-PAGE (CBB staining), the intracellular concentration of eCA was determined to be approximately 520 μM in 1 mL bed volume of RBCs (Fig. S5, ESI†). With this, the mole number of total eCA in the labeling sample (1 mL bed volume of RBCs, total 2 mL suspension volume) was calculated to be ca. 520 nmol. On the other hand, it was estimated that ca. 190 ± 14 nmol of 1 was present (uptaken) inside the total RBCs used, as determined by HPLC analysis of the supernatant (after centrifugation) of the labeling suspension (Fig. 5a).¹⁹ Thus, approximately 48% of total reagent 1 used for the labeling experiment (400 nmol) was incorporated into the cells. Overall, comparison of the intracellular mole numbers of eCA (520 nmol) and 1 (190 nmol) revealed that eCA was present in 2.7-fold excess over 1 in the RBC labeling setting. This moderate excess ratio is probably one factor responsible for the observation that the ligand-binding pocket of FB-eCAI reopens in RBC-based labeling.

Fig. 5 Investigation of the involvement of an anion transporter in the RBC-based labeling system. (a) Quantification of the amount of reagent 1 incorporated in RBCs before incubation (left) and of the SA_OHOH fragment released into the supernatant of the post-labeling solution after incubation (right). (b) Dependence of the extracellular concentration of the SA_OHOH ligand on DNDS concentration. All experiments in this figure were performed in triplicate to obtain mean and standard deviation values (shown as error bars).

Human red blood cells possess a membrane anion transporter protein, termed anion exchanger 1 (AE1) or band 3.²⁰AE1 catalyzes transport of various anionic substrates such as halogen ions, carbonate, phosphate and sulfate. Interestingly, even larger synthetic molecules, including NBD-taurine, a fluorescent sulfonate derivative, can act as a substrate for AE1-mediated anion excretion.^20a Thus we hypothesized that AE1 or other transporters might be involved in the release of the anionic SA_OHOH fragment from the interior to the extracellular region of the RBCs during the labeling process. To explore this possibility, we investigated whether the SA_OHOH fragment could be detected in the supernatant of the post-labeling suspension after 48 h incubation.²¹ Surprisingly, approximately 167 ± 21 nmol of SA_OHOH was present in the supernatant, which corresponds to ca. 88% of the total amount of 1 observed inside the cells before incubation (Fig. 5a). This is in agreement with the observation that the formation of ca. 160 ± 20 nmol of FB-eCAI was detected by ¹⁹F NMR measurements of the post-labeling suspension (Fig. 4b). Interestingly, the addition of 4,4′-dinitro-2,2′-stillbenedisulfonic acid (DNDS), a specific inhibitor of AE1-mediated anion transport,^20a to the labeling suspension considerably lowered the extracellular concentration of SA_OHOH in a DNDS-concentration-dependent manner (Fig. 5b). These results strongly suggest that the SA_OHOH fragment was efficiently eliminated from the RBCs with the assistance of a natural anion transporter, AE1.

Conclusions

It is now known that many organic molecules form (colloid-like) aggregates at micromolar concentrations in aqueous media.²² The relevance and application of such aggregation to biological systems has recently begun to be addressed in the areas of drug discovery²² and drug delivery.²³ In this work, we demonstrated that the LDT-based ¹⁹F-labeling reagent 1 forms self-assembled aggregates in the absence of the target protein in buffer solution. Through a comprehensive study involving the microscopic observation of aggregation morphology, light scattering measurements, product analysis and ¹⁹F NMR spectroscopy, it was shown that 1 has the following characteristics: (1) the reagent alone is NMR-silent because of its aggregation, (2) the aggregates are disrupted in the presence of the target protein (CAI) to form a protein–reagent complex that displays a clear ¹⁹F signal in a turn-on manner, and (3) after the complexation with the target protein, the reagent reacts with an amino acid residue (His67) on the protein surface, resulting in ¹⁹F-labeling (Scheme 2). Additionally, detailed kinetic investigations revealed that the self-assembly properties of the LDT reagent effectively suppresses its nonproductive hydrolytic decomposition. As well as the CAI LDT reagent, we found the suppression of hydrolysis by self-assembly in the case of LDT reagents for FKBP12 labeling (data not shown).⁶ Utilization of the self-assembly properties of reagents might serve as a new promising chemical strategy for selective protein labeling under biological contexts, including cells, tissues and whole organisms, using chemical electrophiles. Since the reactivity can be significantly lowered (masked) in the aggregated state until the reagent encounters the target protein, i.e., its reactivity is restored only when the reagent is recognized by the target protein, both hydrolysis and nonspecific (intermolecular) labeling of non-target proteins would be effectively suppressed. Aggregation (self-assembly) is therefore an important but largely neglected factor for determining the efficiency and selectivity of chemical labeling reagents.

Detailed investigation of the processes involved in the construction of the biosensor indicated that the ligand-binding pocket of the ¹⁹F-labeled endogenous CAI (FB-eCAI) is empty, even without any special treatment after labeling in RBCs. This is in sharp contrast to in vitrolabeling, which generates FB-CAI whose active site is occupied by the cleaved ligand-containing fragment (SA_OHOH). Based on the results from ¹⁹F NMR measurements, product analysis and biochemical inhibitor-based assays, this unexpected removal of the cleaved SA_OHOH is likely due to: (1) re-distribution of the cleaved ligand from the FB-eCAI pocket to the native, unmodified eCAI (because the latter is present in excess) via equilibrium, and (2) the effective elimination of the anionic ligand from the interior to the exterior of the RBCs with the help of a cellular anion transporter system. Although not established, cellular transporter systems could be useful in removing labeling byproducts (in this case, a ligand-tethered phenylsulfonic acid derivative), from the cells, which might otherwise interfere with subsequent applications. Consequently, we demonstrated that it is feasible to convert an endogenous natural protein to a semisynthetic biosensor within its native habitat (in cells) in a ship-in-a-bottle manner by exogenously adding a chemical labeling reagent. We believe that these new findings reported in this work will provide a fundamental basis for the design of more general and versatile chemical methods for selective protein labeling and biosensor generation in crude biological environments.

Experimental procedures

General materials and methods

All proteins and chemicals were obtained from commercial suppliers (Sigma, Aldrich, Tokyo Chemical Industry (TCI), or Wako Pure Chemical Industries) and used without further purification. UV-visible spectra were recorded on a Shimadzu UV-2550 spectrometer. AFM measurements were performed with a SEIKO SPA-400 microscope in tapping mode. DLS measurements were performed using a NICOMP 380zls. The scattering angle was 108° and the wavelength of the laser was 520 nm. MALDI-TOF MS spectra were recorded on a Bruker Autoflex III using sinapinic acid as the matrix. ¹⁹F NMR spectra were recorded on a JEOL ECX-400P (376.5 MHz) spectrometer and calibrated with TFA (−75.6 ppm). Reversed-phase HPLC (RP-HPLC) was carried out on a Hitachi LaChrom L-7100 system equipped with a LaChrom L-7400 UV detector. UV-visible detection was performed at 254 nm, with a flow rate of 1.0 mL min⁻¹. All runs used a mixture of acetonitrile containing 0.1% TFA (solvent A) and 0.1% aqueous TFA (solvent B).

Dynamic light scattering and optical density measurements

The optical density was measured at 25 °C in 50 mM HEPES buffer (pH 7.2, 0.2 mM TFA) using a quartz cell (1 cm). A DMSO stock solution of each reagent (1 or 2) was slowly added to the buffer solution to give a final concentration of 25 μM (0.25% DMSO (v/v)). The DLS measurements were performed under the same conditions with a tubular-type cell. All measurements were carried out in triplicate.

Evaluation of in vitro CAI labeling using 1 and 3

The concentration of human CAI (Sigma) was determined by measuring the absorbance at 280 nm using a molar absorption coefficient of 49 000 M⁻¹ cm⁻¹.²⁴CAI (150 μM) was incubated with 1 or 3 (300 μM) in 50 mM HEPES buffer (pH 7.2, 0.2 mM TFA, 10% D₂O (v/v), 1.2% DMSO (v/v)) at 37 °C. Control reactions with EZA (750 μM) were also conducted. The reactions were monitored by MALDI-TOF MS and ¹⁹F NMR spectroscopy.

Kinetic studies of hydrolysis and labeling reactions

To determine the hydrolysis rates of the LDT reagents, each reagent (1 or 2: 50 μM) was dissolved in 50 mM HEPES buffer (pH 7.2, 0.2 mM TFA, 0.5% DMSO (v/v), glass tube) and incubated at 37 °C. Aliquots were taken at various times, mixed with EZA (100 μM, as an internal standard), and subjected to RP-HPLC equipped with a YMC-Pack PROTEIN column (5 μm, 250 × 4.6 mm). The elution conditions (solvent A/solvent B (v/v)) were as follows: from 0 to 10 min, 5/95; from 10 to 50 min, a linear gradient from 5/95 to 85/15. In the case of 1, the aliquots were diluted with acetonitrile (to 20%) in order to homogenize the aggregates prior to HPLC analysis. The concentrations of the hydrolysis product, SA_OHOH, were estimated by determining the relative peak area of SA_OHOH to the internal standard. The first-order rate constants of the hydrolysis reaction (k_H·monomer or k_H·aggregate) were obtained by fitting a plot of the SA_OH concentration ([SA_OH]) as a function of incubation time (t) to eqn (1) using KaleidaGraph (Synergy Software).

[SA_OHOH] = [reagent]₀(1 − exp(−k_Ht)) (1)

where [reagent]₀ is the initial concentration of the LDT reagent (50 μM in this case) and k_H refers to k_H·monomer or k_H·aggregate.

To determine the reaction rates of protein labeling and hydrolysis from the protein-reagent complex (k_L and k_H·complex, respectively), each reagent (1 or 2: 50 μM) and CAI (50 μM) was dissolved in 50 mM HEPES buffer (pH 7.2, 0.2 mM TFA, 0.5% DMSO (v/v)), and incubated at 37 °C. Aliquots were taken at various times, mixed with EZA (100 μM), and subjected to RP-HPLC as described above. First, the apparent first-order rate constants of SA_OHOH formation, which correspond to (k_H·complex + k_L), were obtained by fitting a plot of the SA_OHOH concentration ([SA_OHOH]) as a function of incubation time (t) to eqn (2).

[SA_OHOH] = [reagent]₀(1 − exp(−(k_H·complex + k_L)t)) (2)

In separate experiments based on MALDI-TOF MS analysis, the concentrations of labeled CAI were estimated by determining the relative peak area of labeled CAI to parental CAI. Finally, the first-order rate constants of protein labeling (k_L) and hydrolysis from the protein-reagent complex (k_H·complex) were obtained by fitting a plot of the concentration of labeled CAI ([labeled CAI]) as a function of incubation time to eqn (3).

[labeled CAI] = (k_L/(k_H·complex + k_L))[reagent]₀(1 − exp(−(k_H·complex + k_L)t)) (3)

Titration of ¹⁹F NMR biosensor in vitro

¹⁹F-labeling of CAI was performed as described in the Evaluation of in vitroCAI Labeling using 1 and 3 section using CAI (150 μM) and 1 (300 μM). After 24 h incubation, the nearly quantitative labeling was confirmed by MALDI-TOF MS. The ¹⁹F-labeled CAI (FB-CAI_SA) was purified by size-exclusion chromatography using a TOYOPEARL HW-40F column (Tosoh Corp.). The purified FB-CAI was dissolved in 50 mM HEPES buffer (pH 7.2, 0.2 mM TFA, 10% D₂O (v/v)). The in vitrotitration experiments with various sulfonamide derivatives were carried out using a solution containing FB-CAI (20 μM) in 50 mM HEPES buffer (pH 7.2, 0.2 mM TFA, 10% D₂O (v/v)) in a NMR tube (550 μL) at 25 °C. The titration curves were analyzed with the nonlinear least-squares curve-fitting method to evaluate the apparent binding constants (K_app, M⁻¹).

eCAI labeling and titration in RBCs

Blood was taken from one of the authors by Dr Eishi Ashihara (Department of Molecular Cell Physiology, Graduate School of Medical Science, Kyoto Prefectural University of Medicine). After anti-coagulation treatment with heparin, RBCs were separated from plasma by centrifugation. The separated RBC suspensions were washed three times with HEPES-buffered saline (20 mM HEPES, 107 mM NaCl, 6 mM KCl, 1.2 mM MgSO₄, 2 mM CaCl₂, 11.5 mM glucose, pH 7.4, HBS). A 1 mL solution of reagent 1 (133 μM) in HBS was mixed with a 1-mL (bed volume) suspension of RBCs, incubated at room temperature for a few minutes, and centrifuged (1500 rpm, 10 min). After removing the supernatant, the same procedure was repeated two more times to treat the cells with a total of 400 nmol of 1. The 1-treated RBCs were resuspended in HBS containing 20% D₂O (v/v) and 200 μM TFA (HBS^DT, 1 mL, yielding a final TFA concentration of 100 μM), and subjected at several time points to ¹⁹F NMR measurements. After 48 h of incubation at 25 °C, the labeled RBCs were collected by centrifugation, and resuspended in fresh HBS^DT. The in-cell titration experiments with various sulfonamide derivatives were carried out using a portion of the RBC suspension (125 μL bed volume) in a NMR tube (500 μL) at 25 °C. The titration curves were analyzed with the nonlinear least-squares curve-fitting method to evaluate the apparent binding constants (K_app, M⁻¹).

Quantification of reagent 1 and SA_OHOH following RBC-based labeling

RBC labeling with 1 was performed as described in the previous section. To determine the mole number of 1 uptaken into the cells at 0 h, the supernatants obtained during the administration of 1 were combined, mixed with EZA (100 μM, as an internal standard), and subjected to RP-HPLC on a YMC-Pack PROTEIN column (5 μm, 250 × 4.6 mm). The mole number of reagent 1 that was not uptaken into the cells was first estimated by determining the ratio between the sum of the peak areas of 1 and SA_OHOH (which was produced by hydrolysis during the experiment) and the area of the internal standard. By subtracting the obtained value from 400 nmol, the mole number of 1 present in the cells was determined. After 48 h of incubation, the supernatant was collected from the labeled RBC suspension. The supernatant was mixed with EZA (100 μM) and subjected to RP-HPLC. The mole number of SA_OHOH released from the cells was estimated by determining the relative peak area of SA_OH to the internal standard. In the AE1 inhibition experiments, following the treatment of RBCs with 1, RBCs were resuspended in HBS^DT containing DNDS (0, 20, 50 or 100 μM). After 48 h of incubation, the supernatant was obtained and analyzed as described above.

Acknowledgements

We thank Drs Masahiro Shirakawa and Hidehito Tochio (Kyoto University) for assistance with ¹⁹F NMR measurements during the early stages of this project. We thank Dr Eishi Ashihara (Kyoto Prefectural University of Medicine) for his help in taking blood samples. Y. T. and Y. S. acknowledge the JSPS Research Fellowships for Young Scientists.

Notes and references

1. Selected reviews of semisynthetic biosensors: (a) K. A. Giuliano, P. L. Post, K. M. Hahn and D. L. Taylor, Annu. Rev. Biophys. Biomol. Struct., 1995, 24, 405–434; (b) H. W. Hellinga and J. S. Marvin, Trends Biotechnol., 1998, 16, 183–189; (c) R. Jelinek and S. Kolusheva, Chem. Rev., 2004, 104, 5987–6015; (d) H. Wang, E. Nakata and I. Hamachi, ChemBioChem, 2009, 10, 2560–2577.
2. (a) M. Brune, J. L. Hunter, J. E. T. Corrie and M. R. Webb, Biochemistry, 1994, 33, 8262–8271; (b) G. Gilardi, L. Q. Zhou, L. Hibbert and A. E. Cass, Anal. Chem., 1994, 66, 3840–3847; (c) J. S. Marvin, E. E. Corcoran, N. A. Hattangadi, J. V. Zhang, S. A. Gere and H. W. Hellinga, Proc. Natl. Acad. Sci. U. S. A., 1997, 94, 4366–4371; (d) J. S. Marvin and H. W. Hellinga, J. Am. Chem. Soc., 1998, 120, 7–11; (e) K. Hirose, H. Takeshima and M. Iino, Anal. Commun., 1999, 36, 175–177; (f) M. Renard, L. Belkadi, N. Hugo, P. England, D. Altschuh and H. Bedouelle, J. Mol. Biol., 2002, 318, 429–442; (g) R. M. de Lorimier, J. J. Smith, M. A. Dwyer, L. L. Looger, K. M. Sali, C. D. Paavola, S. S. Rizk, S. Sadigov, D. W. Conrad, L. Loew and H. W. Hellinga, Protein Sci., 2009, 11, 2655–2675; (h) T. Morii, K. Sugimoto, K. Makino, M. Otsuka, K. Imoto and Y. Mori, J. Am. Chem. Soc., 2002, 124, 1138–1139; (i) A. Toutchkine, V. Kraynov and K. M. Hahn, J. Am. Chem. Soc., 2003, 125, 4132–4145; (j) P. H. Chan, H. B. Liu, Y. W. Chen, K. C. Chan, C. W. Tsang, Y. C. Leung and K. Y. Wong, J. Am. Chem. Soc., 2004, 126, 4074–4075; (k) P. Nalbant, L. Hodgson, V. Kraynov, A. Toutchkine and K. M. Hahn, Science, 2004, 305, 1615–1619; (l) B. E. Cohen, A. Pralle, X. Yao, G. Swaminath, C. S. Gandhi, Y. N. Jan, B. K. Kobilka, E. Y. Isacoff and L. Y. Jan, Proc. Natl. Acad. Sci. U. S. A., 2005, 102, 965–970; (m) S. Namiki, H. Sakamoto, S. Iinuma, M. Iino and K. Hirose, Eur. J. Neurosci., 2007, 25, 2249–2259; (n) J. Zhu and D. Pei, ACS Chem. Biol., 2008, 3, 110–119.
3. (a) I. Hamachi, T. Nagase and S. Shinkai, J. Am. Chem. Soc., 2000, 122, 12065–12066; (b) T. Nagase, S. Shinkai and I. Hamachi, Chem. Commun., 2001, 229–230; (c) T. Nagase, E. Nakata, S. Shinkai and I. Hamachi, Chem.–Eur. J., 2003, 9, 3660–3669; (d) E. Nakata, T. Nagase, S. Shinkai and I. Hamachi, J. Am. Chem. Soc., 2004, 126, 490–495; (e) Y. Koshi, E. Nakata and I. Hamachi, ChemBioChem, 2005, 6, 1349–1352; (f) E. Nakata, Y. Koshi, E. Koga, Y. Katayama and I. Hamachi, J. Am. Chem. Soc., 2005, 127, 13253–13261; (g) Y. Takaoka, H. Tsutsumi, N. Kasagi, E. Nakata and I. Hamachi, J. Am. Chem. Soc., 2006, 128, 3273–3280; (h) Y. Koshi, E. Nakata, H. Yamane and I. Hamachi, J. Am. Chem. Soc., 2006, 128, 10413–10422; (i) T. Anai, E. Nakata, Y. Koshi, A. Ojida and I. Hamachi, J. Am. Chem. Soc., 2007, 129, 6232–6239; (j) E. Nakata, H. Wang and I. Hamachi, ChemBioChem, 2008, 9, 25–28; (k) Y. Koshi, E. Nakata, M. Miyagawa, S. Tsukiji, T. Ogawa and I. Hamachi, J. Am. Chem. Soc., 2008, 130, 245–251.
4. (a) C. Hoffmann, G. Gaietta, M. Bünemann, S. R. Adams, S. Oberdorff-Maass, B. Behr, J.-P. Vilardaga, R. Y. Tsien and M. H. Lohse, Nat. Methods, 2005, 2, 171–176; (b) D. Maurel, L. Comps-Agrar, C. Brock, M.-L. Rives, E. Bourrier, M. A. Ayoub, H. Bazin, N. Tinel, T. Durroux, L. Prézeau, E. Trinquet and J.-P. Pin, Nat. Methods, 2008, 5, 561–567; (c) C. A. Jost, G. Reither, C. Hoffman and C. Schultz, ChemBioChem, 2008, 9, 1379–1384; (d) A recent example in which the basic strategy is likely to be applicable to cellular systems: M. A. Brun, K.-T. Tan, E. Nakata, M. J. Hinner and K. Johnsson, J. Am. Chem. Soc., 2009, 131, 5873–5884.
5. (a) J. A. Prescher and C. R. Bertozzi, Nat. Chem. Biol., 2005, 1, 13–21; (b) K. Johnsson, Nat. Chem. Biol., 2009, 5, 63–65.
6. S. Tsukiji, M. Miyagawa, Y. Takaoka, T. Tamura and I. Hamachi, Nat. Chem. Biol., 2009, 5, 341–343.
7. (a) S. Tsukiji, H. Wang, M. Miyagawa, T. Tamura, Y. Takaoka and I. Hamachi, J. Am. Chem. Soc., 2009, 131, 9046–9054; (b) S. Uchinomiya, H. Nonaka, S. Fujishima, S. Tsukiji, A. Ojida and I. Hamachi, Chem. Commun., 2009, 5880–5882.
8. Other recent examples of traceless-type affinity-based labeling: (a) Y. Hwang, P. R. Thompson, L. Wang, L. Jiang, N. L. Kelleher and P. A. Cole, Angew. Chem., Int. Ed., 2007, 46, 7621–7624; (b) C. C. Hughes, Y.-L. Yang, Y.-T. Liu, P. C. Dorrestein, J. J. La Clair and W. Fenical, J. Am. Chem. Soc., 2009, 131, 12094–12096; (c) H. Nonaka, S. Fujishima, S. Uchinomiya, A. Ojida and I. Hamachi, Bioorg. Med. Chem. Lett., 2009, 19, 6696–6699.
9. (a) C. T. Supuran, Nat. Rev. Drug Discovery, 2008, 7, 168–181; (b) V. M. Krishnamurthy, G. M. Kaufman, A. R. Urbach, I. Gitlin, K. L. Gudiksen, D. B. Weibel and G. M. Whitesides, Chem. Rev., 2008, 108, 946–1051.
10. Other examples of ¹⁹F NMR probe-incorporated proteins: (a) M. A. Danielson and J. J. Falke, Annu. Rev. Biophys. Biomol. Struct., 1996, 25, 163–195; (b) J. C. Jackson, J. T. Hammill and R. A. Mehl, J. Am. Chem. Soc., 2007, 129, 1160–1166; (c) C. Li, G.-F. Wang, Y. Wang, R. C. Allen, E. A. Lutz, H. Scronce, K. M. Slade, R. A. S. Ruf, R. A. Mehl and G. J. Pielak, J. Am. Chem. Soc., 2010, 132, 321–327.
11. (a) P. W. Taylor, R. W. King and A. S. V. Burgen, Biochemistry, 1970, 9, 2638–2645; (b) A. Innocenti, S. Zimmerman, J. G. Ferry, A. Scozzafavaa and C. T. Supuran, Bioorg. Med. Chem. Lett., 2004, 14, 4563–4567; (c) J.-Y. Winum, J.-M. Dogné, A. Casini, X. de Leval, J.-L. Montero, A. Scozzafava, D. Vullo, A. Innocenti and C. T. Supuran, J. Med. Chem., 2005, 48, 2121–2125.
12. (a) Y. Takaoka, T. Sakamoto, S. Tsukiji, M. Narazaki, T. Matsuda, T. Tochio, M. Shirakawa and I. Hamachi, Nat. Chem., 2009, 1, 557–561; (b) K. Mizusawa, Y. Ishida, Y. Takaoka, M. Miyagawa, S. Tsukiji and I. Hamachi, J. Am. Chem. Soc., 2010, 132, 7291.
13. The ¹⁹F NMR signal of 1 was silent with CAI and EZA over 24 h in vitro.
14. Chemically synthesized FB_OHOH in buffer solution also showed a signal at −62.9 ppm (Fig. S2, ESI†).
15. Since we set the experimental conditions where all of the labeling reagents are complexed with CAI, the determined values of k_L and k_H·complex are considered as the rates from the complexed state between CAI and the reagents.
16. It should be noted that because of the need for a relatively high concentration of FB-CAI (>20 μM) for ¹⁹F NMR measurements, we were unable to determine the precise affinity values of FB-CAI towards the strong inhibitors, EZA, AAZ and NBS.
17. The labeling site of FB-eCAI was shown to be at His67, which is identical to the site identified for FB-CAI prepared in vitro.
18. Under the conditions using less than 400 μM of 1, we did not observe any toxicity for RBCs over the time period of the assay (over 48 h).
19. The validity of this mole number was also supported by in-cell¹⁹F NMR measurements, in which the intracellular concentration of 1 can be determined from an integral of the signal at −62.3 ppm corresponding to the 1-eCAI complex (Fig. 4b). The mole number estimated in this manner was 194 ± 2 nmol.
20. (a) O. Eidelman, M. Zangvill, M. Razin, H. Ginsburg and Z. I. Cabantchik, Biochem. J., 1981, 195, 503–513; (b) M. L. Jennings, Annu. Rev. Biophys. Biophys. Chem., 1989, 18, 397–430.
21. Note that following mixing of the RBC suspension with 1, the supernatant is replaced with fresh buffer before starting incubation (also see the Methods section).
22. (a) B. Y. Feng, A. Shelat, T. N. Doman, R. K. Guy and B. K. Shoichet, Nat. Chem. Biol., 2005, 1, 146–148; (b) K. E. D. Coan and B. K. Shoichet, Mol. BioSyst., 2007, 3, 208–213; (c) B. Y. Feng, B. H. Toyama, H. Wille, D. Y. Colby, S. R. Collins, B. C. H. May, S. B. Prusiner, J. Weissman and B. K. Shoichet, Nat. Chem. Biol., 2008, 4, 197–199; (d) K. E. D. Coan and B. K. Shoichet, J. Am. Chem. Soc., 2008, 130, 9606–9612.
23. Y. V. Frenkel, A. D. J. Clark, K. Das, Y.-H. Wang, P. J. Lewi, P. A. J. Janssen and E. Arnold, J. Med. Chem., 2005, 48, 1974–1983.
24. C. Chazalette, B. Masereel, S. Rolin, A. Thiry, A. Scozzafava, A. Innocenti and C. T. Supuran, Bioorg. Med. Chem. Lett., 2004, 14, 5781–5786.

---

Footnotes

† Electronic supplementary information (ESI) available: DLS analyses, optical densities, representative HPLC chromatograms, ¹⁹F NMR spectra and SDS-PAGE analyses. Figs. S1–S5. See DOI: 10.1039/c0sc00513d

‡ Present address: Top Runner Incubation Center for Academia-Industry Fusion, Nagaoka University of Technology, 1603-1 Kamitomioka, Nagaoka, Niigata 940-2188, Japan.

---