Mapping Local Protein Electrostatics by EPR of pH-Sensitive Thiol-Specific Nitroxide

Abstract

A first thiol-specific pH-sensitive nitroxide spin label of the imidazolidine series -methanethiosulfonic acid S-(1-oxyl-2,2,3,5,5-pentamethylimidazolidin-4-ylmethyl) ester (IMTSL) - has been synthesized and characterized. X- (9 GHz) and W-band (94 GHz) EPR spectral parameters of the new spin label in its free form and covalently attached to an amino acid cysteine and a tripeptide glutathione were studied as a function of pH and solvent polarity. pK_a value of protonatable tertiary amino group of the spin label was found to be unaffected by other ionizable groups present in side chains of unstructured small peptides. The W-band EPR spectra were shown to allow for pK_a determination from precise g-factor measurements. Is has been demonstrated that high accuracy of pK_a determination for pH-sensitive nitroxides could be achieved regardless the frequency of measurements or the regime of spin exchange: fast at X-band and slow at W-band. IMTSL was found to react specifically with a model protein - iso-1-cytochrome c from yeast Saccharomyces cerevisiae - giving EPR spectra very similar to those of the most commonly employed cysteine-specific label MTSL. CD data indicated no perturbations to the overall protein structure upon IMTSL labeling. It was found that for IMTSL, g_iso correlates linearly with A_iso but the slopes are different for the neutral and charged forms of the nitroxide. This finding was attributed to the solvent effects on the spin density at the oxygen atom of the N–O group and on the excitation energy of the oxygen lone-pair orbital.

Electrostatic interactions and hydrogen bonding play fundamental roles in virtually all aspects of protein structure and function. Solvent polarity and presence of localized charges affect protein stability, protein-protein interactions, and modulate binding and insertion of proteins and peptides into lipid bilayers (1,2). While basics of solvation effects for molecules are well understood, applications of the same principles to proteins and lipid-protein complexes are less straightforward because of the large number of interactions involved. Indeed, explicit molecular-level treatment of protein solvation processes requires detailed evaluations and averaging of electrostatic and van der Waals energy terms over a large number of solvent molecules and protein side chain configurations. One very useful approach for predicting electrostatic properties of large molecular assemblies is based on treating the solvent as a dielectric continuum and solving the classical Poisson-Boltzmann equation (PBE)¹ digitally for realistic molecular surfaces (3). This classical electrostatic model is then combined with computationally efficient algorithms for solving PBE for complex shapes and charge distributions on a finite grid. Such an approach has proven to be successful in modeling solvent-and pH-dependent protein phenomena with at least semi-quantitative accuracy (3,4). Recently, some more progress in modeling of electrostatic properties has been achieved by developing methods that include polarizable force fields (5) and/or generalized Born and other techniques for molecular dynamics simulations of proteins and RNA with a continuum solvent (6). One important conclusion of a recent theoretical study of four different proteins under various conditions of pH, temperature, solvation, or ligand binding was that behavior of the charged residues is the primary determinant of the protein effective static dielectric permittivity, ε(0) (7). Furthermore, only environmental changes that altered the properties of charged residues were found to exert a significant effect on ε(0). In contrast, buried water molecules or ligands have little or no effect on proteins (7). Thus, it appears that the charged surface side chains play the primary role in determining proteins’ electrostatic properties. Furthermore, there is accumulating evidence that protein side chains exhibiting perturbed electrostatic behavior tend to occur in enzymes’ active sites with sufficient frequency and, therefore, could serve as useful markers of chemical reactivity (8). Such sites could be identified, for example, by a THEMATICS (Theoretical Microscopic Titration Curves) method that is based on Poisson–Boltzmann calculations of the electrical potential function of a protein structure, followed by a Monte Carlo procedure to compute the mean net charge for each of the ionizable sites as a function of pH (8–10). As the theoretical models evolve to become more detailed and complete, the onus is shifting to the development of experimental methods that are capable to provide quantitative data for a critical comparison (11).

From an experimental point of view, local electrostatic interactions remain to be somewhat elusive parameters because of the scarcity of the experimental methods capable of measuring these effects accurately and unambiguously. One of the most powerful methods of structural biology - high resolution NMR - can be utilized directly or indirectly to access complex electrostatics of biomolecules. Residue-specific pK_a can be determined directly by NMR of protein and RNA samples that are labeled with ¹³C and ¹⁵N isotopes uniformly (12) or selectively (13). Special NMR pulse sequences have been developed to measure the ¹³C chemical shifts even for unfolded protein states where the chemical shift overlap is limiting (12). However, for unambiguous pK_a determination, the dynamics of proton exchange has to be slow on the NMR time scale. NMR can also be used to access electrostatic potential distribution around proteins and other molecules by measuring site-specific proton relaxation enhancement upon collisions of exposed residues with nitroxide spin probes bearing different electrical charges (14). This relaxation enhancement depends upon collision frequency of a particular side chain proton with a charged nitroxide molecule and, therefore, is affected by a local electric field. While this NMR nitroxide accessibility method yields surface potentials that are in a good agreement with theoretically predicted values, it is applicable only to proteins with resolved and assigned NMR spectra. This complicates the adaptation of this NMR method to large membrane proteins.

To address some of the shortcomings of the existing experimental techniques Cohen et al. utilized a labeling approach that is based on an unnatural amino acid Aladan, which can be specifically incorporated into a protein by nonsense suppression and reports on the electrostatic character of a protein by steady-state and time-resolved fluorescence (15).

Another labeling method that could be used to probe local electric fields and hydrogen bonding interactions is based on EPR of stable nitroxide radicals. Magnetic parameters of nitroxides are known to be sensitive to intermolecular interactions and, in particular, to hydrogen bonding and local solvent polarity (16–18). In general, when a spin-labeled molecule is transferred from an aqueous (polar) to a hydrocarbon (non-polar) environment, the isotropic nitrogen hyperfine constant A_iso decreases by up to ~2 G while the isotropic g-factor, g_iso, increases by ~0.0004. At X-band (9 GHz) Larmor frequency this change in g_iso corresponds to ~0.68 G shift of the EPR line that is smaller than the typical nitroxide line width (~1 G). However, this spectral shift can be amplified by carrying out EPR experiments at high magnetic fields (HF EPR). For example, at W-band (ca. 95 GHz, corresponding magnetic field of 34,000 G for g=2) the shift in the positions of the resonance lines due to the g-factor differences is magnified tenfold to ~6.8 G, thus, exceeding the corresponding changes in A_iso.

In the past, magnetic parameters (both g-matrices and A-tensors) determined with high accuracy from the rigid limit EPR spectra at high magnetic fields have been utilized to characterize nitroxide-solvent interactions and solvent polarity (18). More recent applications are exemplified by HF EPR studies of a polarity profile along the lipid binding channel of Sec14p protein using a series of doxyl-labeled stearic acids (19,20) and an examination of spin-labeled side chains of a membrane protein bacteriorhodopsin (21,22). The latter studies employed a site-directed spin labeling (SDSL) method. Briefly, in application to proteins, SDSL involves site-directed cysteine substitution mutagenesis at desirable sites of the protein sequence (23). Then thiol-specific reagents, such as methanethiosulfonate (MTS), maleimide derivatives, or others, are utilized to covalently attach a nitroxide label to a cysteine SH group. The MTS attachment group has a clear advantage over the maleimide because of the high specificity to the sulfhydryl group of proteins (24). Although, in principle, several different nitroxide labels can be linked to the protein sulfhydryl groups using MTS chemistry, almost all SDSL EPR studies up to date are carried out with methanethiosulfonate spin label MTSL, (1-oxyl-2,2,5,5-tetramethyl-3-pyrroline) methanethiosulfonate. This label has a rather small size and, when attached to a cysteine, causes little or negligible perturbations to the protein structure/function (25).

While MTSL has been proven to be a very useful nitroxide for SDSL EPR protein studies, the arsenal of EPR experiments could be further expanded through developing new free-radical bearing molecular probes with additional reporting capabilities. Particularly, pH-sensitive nitroxide labels could be very useful for probing the complex landscape of protein electrostatics. Among nitroxides, imidazoline and imidazolidine derivatives containing protonatable functionalities within the structure of heterocycle are considered to be the most promising for biophysical applications because of i) high sensitivity of the EPR spectrum to pH changes, ii) tunable pK_a range and, iii) reversibility of the pH effect (26). These properties have been utilized by Khramtsov and co-workers for evaluating surface potentials and polarity of phospholipid bilayers (27) and human serum albumin (28) although the nitroxide covalent attachment was not fully specific.

Here we describe a general approach for mapping local pK_a of protein side chains with spin-labeling EPR. We show that accurate g-factor determination achieved at high magnetic fields provides some new information on electric field effects and allows for differentiating between polarity and pH-effects on EPR spectra of the imidazolidine nitroxide in a single experiment. We characterize a new thiol-specific pH-sensitive nitroxide - methanethiosulfonate derivative of an imidazolidine nitroxide (methanethiosulfonic acid S-(1-oxyl-2,2,3,5,5-pentamethylimidazolidin-4-ylmethyl) ester; IMTSL) [see ref. 29]. We demonstrate that IMTSL reacts specifically with a unique cysteine of a model protein – iso-1-cytochrome c from yeast Saccharomyces cerevisiae – giving rise to EPR spectra that are very similar to those of MTSL. We also report on EPR spectral parameters of the new nitroxide label in its free form and upon attachment to free amino acid cysteine and a tripeptide glutathione as a function of pH and polarity of solvents. Finally, we report on the use of this new pH-sensitive spin label to study local pK_a in a peptide P11, which constitutes the residues 925–933 of the B1 chain of the basement membrane-specific glycoprotein laminin.

Materials and Methods

Materials

All chemicals were purchased from Sigma-Aldrich (St. Louis, MS) or Acros Organics (Morris Planes, NJ) unless otherwise indicated. All solvents were reagent grade and used as received. Synthesis of the thiol-specific pH-sensitive spin label IMTSL was described in our preliminary communication (29).

Two other nitroxides utilized in this study include perdeuterated Tempone (4-oxo-2,2,6,6-tetramethylpiperidine-d₁₆ 1-oxyl, or PDT, purchased from Cambridge Isotope Laboratories, Andover, MA) and 3-carboxyl-2,2,5,5,-tetramethylpyrrolidin-1-oxyl (CP, purchased from Sigma-Aldrich). These nitroxides were used as received for preparation of 10 mM stock solutions in deionized water.

EPR titration experiments were carried out with the following set of standard buffer solutions purchased from VWR International (West Chester, PA): a) potassium tetraoxalate at pH=1.68, b) hydrochloric acid/glycine at pH=2.0, c) potassium hydrogen phthalate/hydrochloric acid at pH=3.0, d) potassium hydrogen phthalate at pH 4.0, e) acetic acid/sodium acetate at pH=4.63, f) potassium hydrogen phthalate/sodium hydroxide at pH=5.0, g) potassium hydrogen phthalate/sodium hydroxide at pH=6.0, h) sodium phosphate/potassium phosphate at pH=6.86, 7.0, 7.38, and 8.0, i) boric acid/potassium chloride/sodium hydroxide at pH=9.0, j) sodium borate at pH=9.18, and k) sodium bicarbonate/ sodium carbonate at pH=10. All buffer solutions contained 0.5% of a biocide Dowicide A^® (sodium o-phenylphenate tetrahydrate), and were at 50 mM concentration, except the phthalate buffer (pH=6.86) which was 25 mM concentration. These reference buffer solutions are specified to be accurate to at least ±0.02 pH at 25°C and deviate from the specified values by not more than ±0.05 pH at 40°C. Some measurements were carried out with freshly prepared potassium hydrogen phthalate/hydrochloric acid buffers. In addition, 0.1 N solution of HCl with pH=1.1 (30) was used.

A fragment of the laminin B1 chain Cys-Asp-Pro-Gly-Tyr-Ile-Gly-Ser-Arg (P11 peptide) was synthesized by SynPep (Dublin, CA) and specified to be 95.3% pure by HPLC.

Labeling of proteins and peptides

Typically, 80 mM IMTSL stock solution was prepared in 0.5 ml of dry DMSO. The solution was kept under an airtight seal in a −20 °C freezer. Fresh 100 mM solutions of glutathione or L-cysteine were prepared in distilled water. Labeling of the thiols was carried out by adding IMTSL stock to at least twofold excess of glutathione or L-cysteine solutions. These reactions were allowed to proceed at room temperature for 40 min or longer. The resulting mixtures were used without further purification. To verify the absence of unreacted IMTSL, labeling experiments with a five- and a tenfold excess of amino acids were also carried out. No changes in EPR spectra were observed under these conditions.

Labeling of the P11 peptide

For labeling of the fragment of the laminin B1 chain, a 50 mM peptide solution in distilled water was freshly prepared and mixed with a threefold excess of IMTSL from DMSO stock. The reaction was allowed to proceed for 1.5 hr at room temperature. The products were separated by a reverse-phase HPLC (Varian C₁₈ column, Varian, Walnut Creek, CA) using acetonitrile in 0.1% aqueous trifluoroacetic acid (linear gradient from 15% to 40%). After the separation, the samples were immediately lyophilized and stored in a −20 °C freezer.

Labeling of the iso-1-cytochrome c

Labeling of the iso-1-cytochrome c cys 102 residue was carried out for 1.5 hr at room temperature by mixing 100 mM protein solution in a 10 mM MOPS buffer (pH=6.6) with at least twofold excess of IMTSL. The mixture was vortexed occasionally. Unreacted spin label was removed using Sephadex G 25M disposable columns (Amersham Biosciences, Piscataway, NJ). Purified protein was further concentrated using a Microcon YM-30 centrifugal filter device (Millipore Corporation, Billerica, MA). CD spectra of spin-labeled iso-1-cytochrome c were measured at room temperature with a Jasco J-720 spectropolarimeter installed at the Laboratory for Fluorescence Dynamics (University of Illinois, Urbana, IL) and compared with those of unlabeled protein. pH-induced denaturation of IMTSL-labeled iso-1-cytochrome c was monitored by far-UV CD signal at 222 nm. This wavelength reflects pH-dependent changes in the regular secondary structure of the protein N- and C-terminal helices (31). CD spectra were measured at room temperature with a PiStar-180 spectrometer (Applied Photophysics, Leatherhead, Surrey, United Kingdom) installed at the Macromolecular Interactions Facility (University of North Carolina, Chapel Hill, NC). 50 mM stock solution of freshly IMTSL-labeled protein was prepared and kept on ice between the measurements. In a typical experiment, an aliquot from the stock solution was diluted by freshly prepared buffer solutions of required pH to the final protein concentration of ca. 6–8 µM.

EPR measurements

X-band (9.5 GHz) EPR measurements were carried out at room temperature with a Varian (Palo Alto, CA) Century Series E-109 spectrometer. Typically, aqueous solutions were drawn into polytetrafluoroethylene (PTFE) capillaries (0.81×1.12 mm, NewAge Industries, Inc., Southampton, PA), the capillaries were folded twice and inserted into standard 3 by 4 mm quartz EPR tubes. Rigid-limit EPR spectra were recorded in the aqueous buffer − 20% glycerol mixture with a finger Dewar (Wilmad-Labglass, Buena, NJ) filled with liquid nitrogen (77 K).

The W-band (94 GHz) EPR spectrometer is described elsewhere (32,33). The magnetic field sweep up to ±550 G from the main field was provided by a water-cooled coaxial sweep coil. The feedback circuit was based on a Danfysik Ultrastab 860R current transducer (GMW Associates, Redwood City, CA) that provided linearity better than 5 ppm and a resolution of 0.05 ppm. Because the transducer operates on a zero-flux principle, it allowed us to avoid a problem of temperature stabilization at high currents. Measured resettability was better than 100 µA (out of ±15.4 A maximum) and the stability was better than 10 ppm of the maximum scan width. This scanning coil provided magnetic field scanning in steps as fine as 30 mG.

The scan and the center of magnetic field were calibrated with a Metrolab precision NMR teslameter PT 2025 (GMW Associates, Redwood City, CA). The microwave frequency was measured with a source-locking microwave counter (Model 578, EIP Microwave Inc., San Jose, CA). The cavity was a cylindrical type TE_01n (n=2, 3 depending upon tuning) with an unloaded quality factor of ca. 4000.

Typically, for W-band EPR spectroscopy, liquid samples were drawn into thin-wall, clear-fused quartz capillaries (purchased from VitroCom, Inc., Mountain Lakes, NJ) by capillary action. Diameter of the capillaries was varied depending on the solvent. The smallest capillaries, 0.20×0.33 mm in diameter, were used for polar aqueous samples, while for non-polar solvents such as toluene, the size of the capillaries was 0.70×0.87 mm. After filling the capillaries with solutions, both ends of the capillaries were sealed with a Critoseal^® (Fisher Scientific, Pittsburgh, PA) in a way that the seals were always outside the EPR resonator. Typical volume of aqueous samples inside the W-band EPR cavity was less than 150–200 nanoliters.

Special care was taken to ensure accurate and repeatable g-factor measurements. A few days before the experiments, the main superconductive coil was parked within 10 G of the target magnetic field. Typically, the magnetic field drift was stabilized within the first six hours. After this stabilization, the center field of the magnet and magnetic field scan range of the water-cooled coil were calibrated with a Metrolab NMR teslameter. Magnetic field scans of the same width were used for both calibrations and experiments. To simplify the calibration procedure in experiments with nitroxide radicals, we have used a 100 µM aqueous solution of PDT for cross-calibration of the magnetic field. The measurements of A_iso from independently calibrated scans at X- and W-bands were found to be consistent and equal to 16.01±0.01 G while g_iso=2.00550(6) was determined from W-band EPR measurements. By acquiring Tempone spectra before and after the experiments we have found that the magnetic field was stable and that g-factors were repeatable to 1 ppm. The drift of the main superconductive coil was determined to be less than 0.1 ppm/hr.

Results and Discussion

EPR calibration of IMTSL and IMTSL-labeled thiols

Initial characterization of IMTSL was carried out using X-band EPR. For aqueous solutions of IMTSL, single-component fast-motion EPR spectra were observed within pH range from 0.1 to 6.5 units. Upon increasing pH from 0.1 to 3.0 the isotropic nitrogen hyperfine splitting increased from A_iso=14.34±0.05 G to 15.75±0.03 G. The change in A_iso was found to be reversible and is known to be associated with the N3 atom protonation 34.

The pH effect on the nitroxide EPR spectra is associated with proton exchange between a radical R^• and its conjugated acid R^•H⁺ :

$$R^{•}+B{H}^{+}\rightleftarrows R^{•}{H}^{+}+B$$ [1]

Previously, Khramtsov and co-workers considered conditions for fast and slow proton exchange between the R^• and R^•H⁺ forms of pH-sensitive nitroxides (28). Assuming that for IMTSL this reaction is diffusion-controlled with the rate constant of k₁≈10¹⁰ M⁻¹s⁻¹ and that the difference in the resonant frequencies of the high field nitroxide components at X-band does not exceed Δv=4×10⁶ s⁻¹ (i.e., about 1.4 G), then the EPR spectra are in a fast exchange if (28):

$$p{K}_a<\text{log}(k_1/\mathrm{\Delta }\mathrm{\nu })\approx 3.4$$ [2]

This estimate is in agreement with fast-exchange EPR spectra observed for IMTSL at X-band for all the pH values studied.

Figure 1A shows an example of an experimental IMTSL X-band EPR spectrum taken at pH=2.06. At this intermediate pH, the radical is expected to exist as a mixture of protonated and nonprotonated forms. The spectrum was least-squares simulated using a fast exchange model and software we described earlier (Fig. 1B) (35,36). Residual of the fit - a difference between the experimental and the best-fit spectrum (Fig. 1C) - shows that the model of a single nitroxide component fits the data exceptionally well. Thus, at this pH chemical exchange between the R^• and R^•H⁺ species is fast on the X-band EPR time scale.

Figure 1.

Left: A) Experimental room temperature X-band (9.5 GHz) EPR spectrum of IMTSL taken at pH=2.06; B) least-squares simulated spectrum of the nitroxide at pH=2.06; C) residual of the fit – the difference between experimental and the best-fit spectrum. Right: D) Experimental room temperature W-band (95 GHz) EPR spectrum of IMTSL taken at pH=2.1; E and F – the best-fit components for protonated and nonprotonated forms of IMTSL, respectively; G) residual of the fit.

Fig. 2 (filled circles) shows A_iso of IMTSL as a function of pH as determined from least-squares simulations of fast-exchange X-band EPR spectra. The overall A_iso titration curve fits well to the Henderson-Hasselbalch equation:

$$A_{iso}=\frac{A_a·{10}^{(pH-p{k}_a)}+A_b}{1+{10}^{(pH-p{K}_a)}}$$ [3]

where A_a and A_b are the isotropic nitrogen hyperfine coupling constants for the acidic and the basic form of IMTSL respectively. The best-fit results are shown as a solid line in Fig. 2 and are also summarized in the Table 1. The difference between A_a and A_b for IMTSL was found to be Δ_iso = 1.39±0.06 G, which is similar to that of the “parent” radical 2,2,3,4,5,5-hexamethylimidazolidine 1-oxyl, RH3 (ΔA_iso≈1.3 G (26,37)) and is one of the largest ΔA_iso among the nitroxides of the imidazolidine series (26).

Figure 2.

Weighted average A_iso determined from least-squares simulations of fast-exchange X-band EPR spectra as a function of pH. Filled circles - IMTSL, open circles - IMTSL-cys, filled squares - IMTSL-glu; corresponding least-squares Henderson-Hasselbalch titration curves are shown as solid (IMTSL and IMTSL-glu) and dashed (IMTSL-cys) lines.

Table 1.

Titration parameters of free IMTSL and with thiols attached as determined from least squares fitting of room temperature X-band (9.5 GHz) EPR data to the Henderson-Hasselbalch equation. Isotropic g-factors were not measured because of insufficient accuracy of X-band data.

	Aiso(acid)	Aiso(base)	pKa
IMTSL	14.34±0.05	15.75±0.03	1.58±0.03
IMTSL –cys	14.56±0.02	15.82±0.02	3.21±0.04
IMTSL –glu	14.57±0.02	15.86±0.03	3.15±0.03
IMTSL- iso-1-cytochrome c	14.46±0.03	15.81±0.05	1.78±0.07

It was observed that above neutral pH IMTSL would rapidly form biradicals as became evident from characteristic five line EPR spectra (not shown). These species were not characterized for pH dependence. Biradicals have also been observed for MTSL under similar conditions. One could expect that at basic pH, some of the labels’ methanethiosulfonate groups are hydrolyzed to thiols, which would then react with the remaining methanethiosulfonates forming disulfide biradicals.

It is well established that upon proton exchange reactions both the nitrogen hyperfine coupling constant A and g-factor of nitroxides are affected; however, at X-band the line shifts, ΔB, due to Δg_iso are smaller than those originating from ΔA_iso, or the line width and, therefore, are typically not resolved (28,34,37). Specifically, for IMTSL the difference in g_iso between R^• and R^•H⁺ species, Δg_iso, is ≈2.8×10⁻⁴ that translates into ΔB≈0.47 G at X-band (9.5 GHz) – smaller than the line width of about 1–1.2 G. With a tenfold increase in resonance field/frequency achieved at W-band (95 GHz), ΔB raises to ≈4.7 G exceeding both the field-independent change in ΔA_iso≈1.4 G and the line width of about 2 G. This leads to a significantly improved spectral resolution. Figure 1D illustrates this on an example of IMTSL W-band EPR spectrum that shows a partially resolved protonated and nonprotonated nitroxide species at intermediate pH=2.1 while the X-band spectrum measured at essentially the same pH=2.06 (Fig. 1A) does not. The two-component IMTSL W-band EPR spectra are indicative of a slow chemical exchange regime on the W-band EPR time scale.

At intermediate pH values the W-band EPR spectra of IMTSL, such as shown in Fig. 1D, were modeled as a superposition of two nitroxide species which parameters were adjusted during the Levenberg-Marquardt optimization. Figures 1E and 1F show the best-fit components for protonated and nonprotonated forms of IMTSL. Residual of the fit (Fig. 1G) demonstrates that this approximate model fits the experimental spectrum (Fig. 1D) rather well. From such a fitting, giso, A_iso, line widths, and double integrals of the individual components, f₁ and f₂, were determined. In order to compare the results of one- and two-component fitting (such as W- and X-band titration curves) it is more convenient to plot a weighted sum of, for example, A_iso=(f₁ A_1,iso+f₂ A_2,iso)/(f₁ + f₂) as a function of pH rather than just the fraction of the individual component. Fig. 3 shows titration data for both g_iso and A_iso obtained from fitting of 19 °C W-band EPR spectra (filled circles). Corresponding least-squares Henderson-Hasselbalch titration curves are shown as dashed lines and the best-fit titration parameters are listed in Table 2 along with the standard deviations predicted from such a fitting.

Figure 3.

Titration data for A_iso (A, top) and g_iso (B, bottom) obtained from fitting of W-band EPR spectra of IMTSL recordered at 19 °C (filled circles) and 37 °C (open circles). Corresponding least-squares fits to the Henderson-Hasselbalch equation are shown as dashed and solid lines respectively.

Table 2.

Titration parameters of free IMTSL (at 19 and 37 °C) and with thiols attached (at 19 °C) as determined from least-squares fitting of W-band (94 GHz) EPR data to the Henderson-Hasselbalch equation.

	Aiso(acid)	Aiso(base)	pKa (Aiso)	giso(base)	giso(acid)	pKa (giso)
IMTSL, 19 °C	14.34±0.04	15.75±0.03	1.58±0.03	2.005363±3.10−6	2.005638±5.10−6	1.59±0.03
37 °C	14.36±0.06	15.75±0.04	1.54±0.08	2.005363±6.10−6	2.005643±13.10−6	1.42±0.07
IMTSL-cys	14.52±0.04	15.79±0.04	3.29±0.08	2.005356±5.10−6	2.005613±5.10−6	3.29±0.08
IMTSL-glu	14.51±0.01	15.82±0.02	3.17±0.03	2.005353±2.10−6	2.005620±4.10−6	3.12±0.04
CP	16.01±0.04	16.20±0.04	4.03±0.08	2.005315±2.10−6	2.005349±2.10−6	4.18±0.07

Comparison of A_iso(acid), A_iso(base), and pK_a for IMTSL obtained from X- (Table 1) and W-band (Table 2) experiments at 19 °C demonstrates essentially the same results being obtained in these two experiments. Thus, very similar accuracy of pK_a determination for free tumbling pH-sensitive nitroxides could be achieved regardless the microwave frequency of EPR experiments or regime of the spin exchange: fast at X-band and slow at W-band. Notably, W-band measurements allow one to monitor pK_a from g-factor changes: pK_a(g_iso)=1.59±0.03 is identical to that derived from A_iso (pK_a(A_iso)=1.58±0.03). The importance of g-factor measurements for local pK_a determination of larger macromolecules will be discussed in the following sections of this report.

To examine whether the pK_a of IMTSL is affected by moderate changes in temperature during EPR experiments, g_iso and A_iso titration curves were also measured at 37 °C using W-band EPR (Fig. 3, open circles; corresponding least-squares fits to the Henderson-Hasselbalch equation are shown as solid lines). Titration parameters are given in the Table 2. It is clear that the W-band data at 37 °C follow closely the results at 19 °C. This indicates a relatively small enthalpy of the IMTSL proton exchange reaction and is in accord with literature data for other compounds (38). However, the 37 °C data show somewhat larger scatter as evident from larger standard deviations predicted from the fit (Table 2). The likely reason for the scatter is that with increasing temperature, the rate of proton exchange increases and the spectra start to approach an intermediate regime for which neither one- nor two-site simple models are fully applicable.

It is well established that chemical modifications of nitroxide side chains close to N3 would affect the pK_a of the imidazolidine nitroxides (26). Thus, in order to utilize IMTSL in SDSL EPR studies the label has to be calibrated when it is attached to cysteine-containing molecules, such as amino acids and/or small peptides. To assess these changes two additional W-band EPR titration calibration experiments, one with the label attached to a cysteine (IMTSL-cys) and the other with IMTSL-labeled glutathione (IMTSL-glu) were carried out. Similar to the free IMTSL, the EPR spectra of these compounds taken at 19 °C and intermediate pH values were indicative of a slow exchange at W-band and fast exchange at X-band. Both IMTSL-cys and IMTSL-glu EPR titration experiments were carried out up to pH=9.0. The disulfide bond formed in these adducts is sufficiently stable under these alkaline conditions, as was ascertained from the reversible character of the titration curves. All spectra were least-squares simulated as described above. For clarity and to avoid overlapping data points, the W-band titration curves for IMTSL-cys and IMTSL-glu (Fig. 4) are plotted separately from the free IMTSL data (Fig. 3) while X-band titration data are shown in the same graph (Fig. 2). The least squares parameters of the Henderson-Hasselbalch fits are given in Tables 1 and 2.

Figure 4.

Experimental titration data for g_iso (lower panel) and A_iso (top panel) obtained from fitting of W-band EPR spectra of IMTSL-cys (open circles) and IMTSL-glu (filled squares); the corresponding least-squares Henderson-Hasselbalch titration curves are shown as solid and dashed line respectively.

Several observations could be made from the EPR titration data given in Tables 1 and 2. Firstly, for each of the two IMTSL thiol adducts studied, EPR titration parameters derived from either X- and W-band spectra were found to be the same, demonstrating that the same accuracy could be achieved regardless of the spin exchange regime. Secondly and most notably, the pK_a’s of these two adducts were shifted to more basic pK_a=3.21±0.04 for IMTSL-cys and pK_a=3.15±0.03 for IMTSL-glu (X-band measurements) from the pK_a of the free label (1.58±0.03 units). Also, A_iso(acid) for IMTSL-thiol adducts was somewhat larger than that of free IMTSL while the increase in A_iso(base) was comparable with the experimental error.

EPR titration parameters in Tables 1 and 2 indicate that the main effect of replacing MTS with the disulfide group, as a result of the reaction with thiols, is the change in the pK_a of the tertiary amino group rather than in isotropic A_iso and g_iso for the both basic and acidic forms of the nitroxide. This is expected because the N3 protonation site is only three chemical bonds away from the site of modification while the EPR-reporting nitroxide moiety is distanced by five chemical bonds. In nitroxide radicals the electronic spin density is chiefly localized on the π orbital of the NO fragment and the modification of the side chain away from that fragment is not expected to cause any significant electron redistribution. Therefore, A_iso and g_iso for both basic and acidic nitroxide forms are not affected by the side chain modification. However, the pK_a of the protonatable tertiary amino group that is positioned closer to the side chain becomes affected: reaction with a thiol converts the methanethiosulfonate group into a less electronegative disulfide side chain, thus, shifting the pK_a of the formed adduct to more basic values.

In principle, the presence and the ionization state of other groups in the side chain could also affect the nitroxide EPR spectra. Such groups could be found in both IMTSL adducts. For example, the carboxylic and amino functionalities of IMTSL-cys are expected to exhibit the pK_a’s that are close to those of cysteine (respective pK_a values are 1.71 and 8.33) and cystine (pK_a values are 1.0 and 2.1 for two carboxylic and 8.02 and 8.71 for two amino groups, respectively) (39). However, for IMTSL-cys no measurable changes in either A_iso or g_iso were observed at these pH values: both EPR parameters followed the single-transition Henderson-Hasselbalch titration curves. The same result was observed for IMTSL-glu adduct which side chain contains groups with pK_a’s that could be close to those of a glutamine (pK_a1=2.2, pK_a2=9.1) and to the carboxylic group of the glycine moiety of the tripeptide itself (pK_a=3.59) (40). The absence of any measurable effects of the ionizable side chain groups on EPR titration parameters is in line with the fast decay of inductive electronic effects along the chain of σ-bonds; any through-space electric field effects on either nitroxide moiety or pK_a of tertiary amino group are expected to be small. Thus, when attached to thiols, IMTSL reports on the local pH experienced by the N3 amino group. The pK_a of this group is found to be unaffected by other ionizable groups present in the side chain, at least when this chain is short and interchain interactions/tertiary structure are absent as in these small unstructured peptides. However, one could expect that for more complex macromolecules, the presence of tertiary interactions and local folds would affect the electrostatic environment of the probe resulting in measurable changes in pK_a reported by the label.

Labeling of model proteins with IMTSL

Comparison of the chemical structure of IMTSL in its basic form to that of MTSL indicates many similarities that raise expectations for essentially the same EPR parameters. Moreover, similar physical properties, including solubility and the same structure of the attachment group, make MTSL labeling protocols applicable for site-directed protein chain modification with IMTSL. Uncharged IMTSL is expected to affect the protein structure essentially the same way as MTSL does: the latter label has shown to have small, if any, effects on the protein structure and function as was demonstrated by scanning 30 different cysteine mutants of T4 lysozyme (25). Moreover, based on similarity of molecular sizes and geometry of nitroxide rings of MTSL and the basic form of IMTSL and the same length and the structure of the attachment tether one would expect very similar rotational dynamics for the two spin-labels and, therefore, similar EPR spectra.

To illustrate these similarities we have closely followed the MTSL protocol to label the single native cysteine 102 (C102) of iso-1-cytochrome c with IMTSL (41). Figure S1 (Supporting Information) shows that the shapes of far-UV CD spectra of the native protein (C102) and the one labeled with IMTSL (C102-IMTSL) are essentially identical confirming absence of nitroxide perturbation effects on the content of the α-helical structure (31).

Room-temperature X-band EPR spectra of iso-1-cytochrome c labeled with MTSL and IMTSL are shown in Fig. 5. The spectrum of IMTSL- iso-1-cytochrome c, taken at pH=6.6 (Fig. 5 B) is very similar to that of MTSL- iso-1-cytochrome c (Fig. 5 A) demonstrating that for this site the nanosecond-scale rotational dynamics of the nitroxide and, thus, the spin label interactions with the protein backbone, are unaffected by the presence of an uncharged amine group in the imidazolidine ring. The isotropic nitrogen hyperfine coupling constant for MTSL-iso-1-cytochrome c is A_iso=16.13±0.05 G which is within the experimental error of A_iso observed for this spin label in water (16.09±0.03 G). For IMTSL- iso-1-cytochrome c A_iso=15.81±0.05 G is the same as A_iso=15.82±0.02 G observed for IMTSL-cys in the basic form (Table 1). This comparison of isotropic nitrogen hyperfine coupling constants indicates that the nitroxide moieties of both spin-labeled side chains are fully exposed to aqueous phase.

Figure 5.

Representative room-temperature X-band EPR spectra of iso-1-cytochrome c labeled at the unique Cys102 with MTSL (A) taken in 10 mM MOPS buffer at pH=6.6 and IMTSL (B–E) measured in 50 mM buffers of various pH: (B) – 6.6; (C) – 5.0, (D) – 3.0, and (E) – 1.68. Vertical dashed lines correspond to approximate positions of the zero cross points of the nitrogen hyperfine coupling components of the IMTSL- iso-1-cytochrome c spectrum measured at pH=1.68.

To demonstrate utility of IMTSL for monitoring local protein electrostatics we have carried out EPR titration of spin-labeled iso-1-cytochrome c. Fig. 5 shows a series of X-band EPR spectra of this protein in the pH range from 6.6 to 1.68 units. For all pH values, the label tethered at C102 - the penultimate C-terminal amino acid residue located in the loop region – experienced only little rotational restrictions as indicated by EPR spectra characteristic for fast motion regime. This is typical for surface-exposed protein side chains (42). One could also expect that such a progressive lowering of pH would trigger denaturing of this protein. Indeed, far-UV CD of IMTSL- iso-1-cytochrome c in the pH range from 7.03 to 1.68 units measured at 208 and 222 nm, indicated a gradually decreased content of the α-helical fold (Fig. S2, Supporting Information). Specifically, CD spectra taken in the range from 7.03 to 3.99 pH units revealed the presence of a native-like but reduced α-helical signature. The α-helical content drastically decreased when pH dropped from 3.99 to 3.54, and the protein was completely denatured below pH=2.99. These CD data are in a good agreement with spin-labeling EPR data that show progressive disappearance of the residual slow/intermediate motion spectral components at pH=3.0 and below (Fig. 5). Thus, the local environment of this spin label attached to the protein loop becomes more disordered as the protein unfolds.

Lowering the pH value also resulted in a decrease in the apparent isotropic nitrogen hyperfine coupling constant A_iso of IMTSL- iso-1-cytochrome c measured as a distance between the low field and the central nitrogen hyperfine EPR components. The A_iso vs. bulk pH titration plot (open circles, Fig. 6) yielded pK_a=1.78±0.07 that was largely different from that of unstructured IMTSL-cys (pK_a=3.21±0.04). This difference is attributed to the local electrostatic environment of the label that remains present even after the protein unfolds. It should be noted here that protonation of the spin probe itself is not the reason for the protein denaturing: according to the CD data, it occurs between pH≈4.0 and 3.0 that is well above pK_a=1.78±0.07 determined for IMTSL- iso-1-cytochrome c by EPR.

Figure 6.

Experimental X-band EPR titration data for IMTSL- iso-1-cytochrome c measured at room temperature from changes of isotropic nitrogen hyperfine coupling constant, A_iso (open circles, left axis), and from an anisotropic hyperfine component A_z (filled circles, right axis) defined as the splitting between the outer peaks of rigid limit EPR spectra measured at T=77 K. Solid lines show the least-squares Henderson-Hasselbalch titration curves that yielded nearly identical pK_a: 1.78±0.07 for the A_iso plot and 1.76±0.10 for A_z. Dashed line defines average pK_a=1.77.

IMTSL- iso-1-cytochrome c X-band titration experiments performed under the rigid-limit EPR conditions (77 K) produced a pK_a essentially the same as for room temperature (Fig. 6, filled circles). This result indicates that the local electrostatic environment of this spin-labeled protein side chain is unaffected by freezing the protein solution at 77 K. Representative rigid-limit EPR spectra of the IMTSL- iso-1-cytochrome c are shown in the Figure S3, Supporting information.

Differentiating effects of local polarity and proton exchange: g_iso-A_iso correlation plots

In applications of pH-sensitive nitroxides to studies of local electrostatics of proteins and membranes from changes in magnetic parameters of the radical, one should also consider factors other than proton exchange reactions for potential contributions. Some of the most important of such factors are the solvent dielectric constant and hydrogen bonding. In order to investigate these solvent effects we have carried out a calibration of g_iso and A_iso of IMTSL for a series of protic and aprotic solvents including water, water/ethanol mixtures, several alcohols, acetone, acetonitrile, and toluene. Figure 7 (open squares) shows a correlation plot for g_iso versus A_iso obtained from fitting room-temperature W-band EPR spectra. The estimated errors are within the size of the symbols. The data shown in Fig. 7 strongly suggest the existence of a linear correlation. Previously, Kawamura and co-workers reported that for another nitroxide, di-tert-butyl nitroxide (DTBN), g_iso versus A_iso plots have two different linear correlations for protic and aprotic solvents (16). Recent W-band measurements of isotropic magnetic parameters of another nitroxide, Tempo, strongly suggested that g_iso is proportional to A_iso regardless whether the solvent is protic or not (43).

Figure 7.

Isotropic magnetic parameters g_iso vs. A_iso obtained from solution W-band EPR spectra. Open squares: IMTSL in a series of protic and aprotic solvents and their mixtures: 1 - toluene, 2 – acetonitrile, 3 – acetone, 4 – iso-propanol, 5 – ethanol, 6 – water/ethanol solution (3:7, v/v), 7 – water/ethanol solution (7:3, v/v), 8 – water (buffered to pH 6.0). Titration parameters for: IMTSL (open circles) and IMTSL-P11 (filled squares). Linear regressions are shown as solid lines and discussed in the text.

While this paper is not intended for developing a detailed theoretical understanding of the observed linear correlation, a qualitative explanation could be built based on the existing theories. Indeed, it is well established that both g_iso and A_iso are approximately proportional to the spin density at the nitrogen atom (44–47) while the electronic g-factor has some additional dependence on the excitation energy of the oxygen lone-pair orbital, which is affected by the solvent dielectric constant and hydrogen bonding (48). If a hydrogen bond donor molecule would be approaching the oxygen atom only at certain defined angles that are asymmetric with respect to the N-O bond, then one would expect anisotropic effects on g-factor and, likely, some different correlations between g_iso and A_iso for hydrogen-bonding vs. non-hydrogen-bonding solvents. However, for random rapid collisions that are typical for solutions studied here at room temperature, one cannot expect any asymmetry in formation of transient hydrogen bonds with respect to the nitroxide N-O axis. Thus, the main effect of the collisions with the protic molecules would be on the spin density at the oxygen atom - the same as for aprotic molecules. It is worthwhile to note here that fast spin exchange between nitroxides in different local environments would not affect this linear correlation, and the same empirical plot could be used to assign effective “polarity” in the vicinity of the nitroxide, as was demonstrated for a nitroxide partitioned in two compartments of a phospholipid bilayer interdigitated at high ethanol concentrations (43).

One interesting question is whether the same correlation between g_iso and A_iso would hold for the protonated form of IMTSL. Figure 7 demonstrates that the correlations are different: open circles show that with decrease in pH, weighted parameters A_iso and g_iso clearly fall below the linear correlation plot for the IMTSL in the nonprotonated form. This deviation is related to the effect of an asymmetric charge, located chiefly on N3, on the spin density at the oxygen atom. As one would expect, the protonated form of IMTSL would have somewhat different electronic configuration of the nitroxide moiety that cannot be mimicked by symmetric rearrangement of the solvent molecules with respect to the axis of the N-O bond. Then, g_iso(base) and A_iso(base) for the non-protonated form of the IMTSL in various solvents should fall onto one straight line while g_iso(acid) and A_iso(acid) would belong to another one. It is also easy to see that as long as both g_iso and A_iso report on the same proton exchange reaction (and follow the same Henderson-Hasselbalch equation), the plot of weighted g_iso vs. A_iso upon protonation (open circles, Fig. 7) would also follow a straight line but with a different slope. Such linear correlations (solid lines in Fig.7) could be used for differentiating “polarity” and proton exchange effects on IMTSL isotropic magnetic parameters. It is worthwhile to note here that while A_iso=14.34±0.05 G for the acidic form of IMTSL is very close to A_iso=14.47±0.05 G observed for the basic form of this nitroxide in acetonitrile, the difference in isotropic g-factors between these two species is Δg_iso=(5.1±0.7)×10⁻⁵. While at X-band this Δg_iso translates into only ca. 85 mG line shift, which is hard to measure, at W-band this difference amplifies to measurable 850 mG. Thus, the high g-factor resolution achieved by W-band EPR is essential for assigning whether changes in magnetic parameters of IMTSL are caused by proton exchange or just by local polarity effects.

The deviation of g_iso from the g_iso/A_iso polarity correlation plot as shown in Fig. 7 is still rather small when compared with the overall change in g_iso in the course of proton exchange reaction: Δg_iso=(5.1±0.7)×10⁻⁵ vs. (27.5±0.7)×10⁻⁵ respectively. This is because the primary effect of the electrical charge localized at N3 is on the spin density at the oxygen atom of the nitroxide moiety and the effect on the oxygen lone-pair orbital contributes as a smaller correction. This also means that for reliable differentiation of polarity and pH effects it is desirable to have a nitroxide with large changes in g_iso and A_iso upon protonation.

To illustrate the latter we have carried out W-band EPR titration and solvent polarity calibration experiments for another nitroxide - 3-carboxy-2,2,5,5-tetramethylpyrrolidin 1-oxyl (CP). Results of these titrations are summarized in Fig. 8 A and B and listed in the Table 2. The data are in general agreement with the previously published 220 GHz EPR results of Gulla and Budil, who utilized this nitroxide for studies of electric field effects on the nitroxide g-matrix (37). For this spin probe the changes in g_iso and A_iso upon proton exchange reaction are small when compared with those reported here for IMTSL or with another imidazolidine nitroxide RH3 (37). The main reasons for smaller Δg_iso and ΔA_iso observed for CP are (i) a longer distance between the charged carboxylic group and nitroxide moiety and therefore smaller effective electric field when compared with either IMTSL or RH3 and (ii) a compensating effect of polar solvent molecules surrounding the carboxylic group (37). In addition, for CP the electric charge is expected to be delocalized in a greater degree than for IMTSL. Based on these two factors -small Δg_iso and ΔA_iso and a larger charge delocalization - one could expect that for CP, g_iso should be approximately proportional to A_iso regardless of the ionization state of the carboxyl group. Indeed, the Fig. 8 C demonstrates that for this nitroxide, all the isotropic magnetic parameters measured in solvents of different polarities and also in aqueous buffers with pH from 1.0 to 10.0 (Fig. 8, open circles) follow the same linear dependence regardless the charge of the carboxylic group or whether the solvent is protic or aprotic. Thus, for this nitroxide, polarity and titration effects cannot be differentiated as was illustrated for IMTSL (at least at magnetic fields of W-band EPR). It is also worthwhile to note here, that CP is yet another example of a nitroxide with the same linear correlation between g_iso and A_iso for protic and aprotic solvents.

Figure 8.

Room-temperature W-band EPR g_iso (A) and A_iso (B) titration experiments and solvent polarity calibration experiments (C) for 3-carboxy-2,2,5,5-tetramethylpyrrolidin 1-oxyl (CP). Top: Experimental titration data for g_iso (A) and A_iso (B) with the corresponding least-squares Henderson-Hasselbalch titration curves shown as solid lines with the best-fit parameters listed in the Table 2. Bottom (C): a combined g_iso vs. A_iso correlation plot. Filled circles: 1 – toluene, 2 – acetonitrile, 3– acetone, 4 – iso-propanol, 5 – ethanol, 6 – water/ethanol solution (3:7, v/v), 7 – water/ethanol solution (7:3, v/v), 8 – water, buffered to pH 6.0. Titration data from (A) and (B) plots are shown as open circles in front of the filled circles corresponding to water and water-ethanol mixture data.

Measurements of local pK_a of a P11 peptide fragment of glycoprotein laminin B1 chain

To illustrate utility of IMTSL in site-directed pK_a measurements and the use of our HF EPR method for differentiating pH and polarity effects, we have labeled a synthetic P11 peptide – a fragment of the laminin B1 chain (Cys-Asp-Pro-Gly-Tyr-Ile-Gly-Ser-Arg) - with IMTSL at the single cysteine. This peptide was found to reduce colonization of lung cells with malignant melanoma by >90%, when administered together with the melanoma cells by tail vein injection into mice (49). It was speculated that this peptide could inhibit lung tumor colony formation by blocking tumor cell invasion through basement membranes (49).

As a first step in understanding the role of electrostatics in interaction of this peptide with membranes, we have carried out room temperature EPR titration experiments. Figure 9 shows representative X- and W-band EPR spectra from ca. 0.3 mM of IMTSL-P11 in a 50 mM phosphate buffer at pH 6.0. At both EPR frequencies the spectra fall into fast motional regime as demonstrated by least-squares simulation of the W-band spectrum (Fig. 9 B–D). For this spectrum, isotropic nitrogen hyperfine coupling constant A_iso=15.75±0.01 was essentially the same as A_iso(base) for IMTSL-cys and IMTSL-glu (Table 2), indicating that the nitroxide is fully exposed to aqueous phase.

Figure 9.

Representative room-temperature fast-motion X- (A) and W-band (B) EPR spectra from ca. 0.3 mM of IMTSL-P11 in a 50 mM phosphate buffer at pH=6.0; (C) is the least-squares simulation of the W-band spectrum (B); (D) is the fit residual – a difference between the experimental and the simulated spectrum.

With decrease in pH the W-band EPR spectrum of IMTSL-P11 was split into two components (not shown) similar to that of IMTSL-cys and IMTSL-glu. While at pH=4.0 and 3.0 the effective A_iso was decreased only slightly, at pH=2.0 A_iso(acid) was essentially the same as for other IMTSL-modified thiols (see Fig. 10 and Table 1). Since one could expect some conformational changes to occur for a peptide at an acidic pH, only the data from pH=3.0 to pH=6.0 were fitted to the Henderson-Hasselbalch equation. Moreover, because all the data presented here for IMTSL and its adducts with thiol-containing molecules indicate that the A_iso(base) and A_iso(acid) values are approximately unaffected by the side-chain substitutions, the latter values were fixed during the optimization and only pK_a was varied. The results of the fit are shown in Fig. 10 as a solid line together with the least-squares Henderson-Hasselbalch curves for IMTSL-glu and IMTSL-cys (short- and long-dashed lines respectively). Fig. 10 demonstrates that pK_a of IMTSL-P11 (pK_a=2.5±0.1) is lower than that of either IMTSL-glu or IMTSL-cys (pK_a≈3.18). Unfortunately, the titration curve for IMTSL-P11 cannot be completed at low pH because EPR signal from IMTSL-P11 rapidly decayed even at pH=2 and no EPR signals were observed at pH<2.

Figure 10.

Experimental titration data for A_iso obtained from simulation of W-band EPR spectra of IMTSL-P11 (filled circles). The least-squares Henderson-Hasselbalch titration curves for IMTSL-glu and IMTSL-cys (short- and long-dashed lines respectively) are shown for comparison.

One could argue that the observed changes in A_iso could be related to a decrease in an effective local polarity arising from peptide conformational changes rather than the reversible protonation of the spin label. For example, the tyrosine residue could approach the nitroxide ring, thus, decreasing local polarity and, as a result, lowering A_iso. To discriminate between these two scenarios we have compared g_iso and A_iso at each pH values with the g_iso/A_iso correlation plot. Figure 7 shows that isotropic magnetic parameters for IMTSL-P11 (filled squares) closely follow the IMTSL pH titration correlation plot but not the polarity plot. Thus, the observed changes in g_iso and A_iso are related to proton exchange reaction of the nitroxide label but not the local polarity changes.

As discussed above we have experimentally shown that the pK_a’s of IMTSL-glu and IMTSL-cys are not affected by the ionization state of the functional groups in the side chain. However, larger peptides, such as P11, could adopt a folded or a partially folded state through tertiary interactions. Specifically for P11, the energy-optimized structure (Fig. S4, Supporting Information) shows that the nitroxide ring attached to a cysteine could be placed in close proximity to the carboxylic group of the aspartic acid. Previously, it has been shown that in proteins the pK_a of the carboxylic group in an aspartic acid residue could vary from 2.0 to 4.93 (50,51). However, for IMTSL-P11, the shift of the titration curve to more acidic region as compared with that for IMTSL-cys and IMTSL-glu could not be explained by the effect of the aspartic acid. The negatively charged carboxylate should stabilize the protonated form of the proximal imidazolidine moiety, shifting its apparent pK_a to more basic region of pH. Another possible reason for the shift of the pK_a to the acidic region could be destabilization of the protonated form of the nitroxide by an effective positive charge, which was acquired by a globular peptide structure as the result of protonation of other basic groups (e.g., guanidine moiety of Arg).

To summarize, we have synthesized a new pH-sensitive thiol-specific imidazolidine nitroxide label methanethiosulfonic acid S-(1-oxyl-2,2,3,5,5-pentamethylimidazolidin-4-ylmethyl) ester (IMTSL) and characterized this nitroxide in its free form and covalently tethered to a cysteine, short peptides (glutathione and P11 peptide), and the native cysteine 102 of iso-1-cytochrome c from Saccharomyces cerevisiae. It was shown that EPR titration experiments could be carried out at W-band by measuring the effect of protonation on either A_iso or g_iso, as both magnetic parameters respond to an asymmetric charge acquired upon protonation and located chiefly on N3. In order to differentiate reversible protonation and solvent effects on magnetic parameters of IMTSL, g_iso and A_iso were also measured for a series of protic and aprotic solvents. It was found that for IMTSL, g_iso correlates linearly with A_iso but the correlations are different for the neutral and charged forms of the nitroxide. This finding was attributed to solvent effects on the spin density at the oxygen atom of the N–O group and on the excitation energy of the oxygen lone-pair orbital. It is proposed to employ g_iso versus A_iso correlation plots for differentiating polarity and protonation effects in spin-labeling experiments. The utility of such an approach has been demonstrated on an example of a synthetic P11 peptide. IMTSL was also employed for labeling of the unique cysteine of iso-1-cytochrome c. EPR spectra of this protein labeled with IMTSL were found to be very similar to that labeled with MTSL and CD data revealed no perturbations to the overall protein structure. Titration curves deduced from either room-temperature or rigid-limit IMTSL EPR spectra yielded nearly identical pK_a’s demonstrating that both approaches are equally applicable. Notably, the protonation of the spin probe itself is not the reason for the protein denaturing as the latter, according to the CD data, occurred at pH higher than the pK_a of the spin-labeled side-chain.

The present study shows that pH-sensitive nitroxides in combination with enhanced spectral resolution of high-field EPR spectroscopy could considerably extend the range of biophysical experiments carried out with site-directed spin labeling techniques. While this manuscript was being prepared, a report by Möbius and co-workers on probing local pH within the proton channel of bacteriorhodopsin light-driven proton pump from Halobacterium Salinarium by means of multi-frequency EPR of IMTSL has been published (52). One of the conclusions in this paper was that high field EPR in combination with pH-sensitive spin labels opens a new avenue for investigating protein systems. The latter finding further reaffirmed conclusions of our study regarding IMTSL being an essential tool for mapping the complex landscape of proteins’ electrostatics.