**6. Specificity in protein labeling: Thiol groups**

The ideal goal with spin labeling is a universal method to label any tailored site with high specificity. Let's face it; spin labeling of proteins is a protein chemical modification methodology. That aside, it is the chemistry of the functional groups utilized in order to label a protein. If one examines the 20 common amino acids, one finds that the advantageous modification chemistry is both quite limited, ambiguous, and is very much dependent on pKa values where charged sidechains are targeted. This leaves us only with the cysteine thiol as the best candidate for any sort of specific modification. If one looks at the standard array of protein modification functional groups, at least what existed in the 1960s, 1970s and 1980s, we were limited to the maleimde, the alpha halo-acetamide groups and a few disulfide-based reagents, all of which had limitations, particularly the former two. Reagents such as iodoacetamide or N-ethylmaleimide (NEM) will react with thiol groups, amino groups (alpha- amino groups, lysine) and occasionally with hydroxyl groups of a nucleophilic serine or threonine or a tyrosyl side chain. Secondly, the preponderance of these sidechains is usually multifold in proteins, while thiol groups are usually small in number, or occasionally nonexistent. In a quest for highly specific reversible thiol reagents, Berliner and Hideg capitalized on the chemistry pioneered by George Kenyon with methylthiomethane sulfonate, a reagent that undergoes disulfide interchange with a cysteine eliminating the methylsulfonate leaving group [15]. This was definitely advantageous over dithiol reagents, where one loses half of the label in the exercise and it also created other problems involving thiol interchange that could eventually negate of the advantages. Hence the label shown in Figure 10 (bottom), affectionately known as MTSSL or MTSL, was synthesized and was shown to be highly reactive, uniquely specific for cysteine

thiol groups and could be easily released with a small concentration of mercaptoethanol or dithiothreitol, allowing one to recover the protein and also allowing for a second labeling stoichiometry quantitation based on the released label [16]. Berliner and Hideg showed eloquently how this works with the reactive protease papain, which contains a cysteine SH at the active site analogous to the serine OH in chymotrypsin [16]. Initially this label wasn't used much by other research groups, but the advent of molecular biology and the power of site-specific mutation triggered a revolution in this area, pioneered by Wayne Hubbell. The technique, named site directed spin labeling, has really been the method of choice since the 1990s and has created a renaissance in spin labeling [17].

**Figure 9.** Representative spin labeled pyrimidine bases that can be incorporated into nucleic acid structures. Adapted from [13] with permission.

History of the Use of Nitroxides (Aminoxyl Radicals) in Biochemistry: Past, Present and Future of Spin Label and Probe Method 13

12 Nitroxides – Theory, Experiment and Applications

1990s and has created a renaissance in spin labeling [17].

<sup>N</sup> <sup>H</sup> N O

C O

S

C <sup>O</sup> <sup>H</sup> N

C <sup>O</sup> <sup>H</sup> N

C <sup>O</sup> <sup>H</sup> N

structures. Adapted from [13] with permission.

HN N

O

O

**DUAVAP**

NH2

**DCAVAT**

NH2

**DCAVAP**

N N

N N

O

O

HN N

O

O

**DUMBT**

C

C O HN

C O HN

O HN

N O

HN N

HN N

> N N

O

O

O

O

**DUPAT**

O

**DUAT**

NH2

**DCAT**

**Figure 9.** Representative spin labeled pyrimidine bases that can be incorporated into nucleic acid

H N C O

H N C O

H N C O N O

HN N

O

O

N O

HN N

HN N

O

**DIACET**

O

N N

O

O

O

**DUMPT**

S

**DUAP**

NH2

**DCAP**

H N C O

O

C

N O

N O

<sup>N</sup> <sup>H</sup> N O

N O

H N C O

N O

N O

N O

thiol groups and could be easily released with a small concentration of mercaptoethanol or dithiothreitol, allowing one to recover the protein and also allowing for a second labeling stoichiometry quantitation based on the released label [16]. Berliner and Hideg showed eloquently how this works with the reactive protease papain, which contains a cysteine SH at the active site analogous to the serine OH in chymotrypsin [16]. Initially this label wasn't used much by other research groups, but the advent of molecular biology and the power of site-specific mutation triggered a revolution in this area, pioneered by Wayne Hubbell. The technique, named site directed spin labeling, has really been the method of choice since the

**Figure 10.** Comparison of thiol labeling reagents. The top reaction utilizes the NEM analog, which reacts irreversibly with the protein thiol group, but can also partially react with other nucleophiles. On the other hand, MTSSL reacts specifically with SH groups and can be reversed in the presence of another thiol reagent, such as mercaptoethanol or DTT. Adapted from [18] with permission

Wayne Hubbell's important contribution was to realize that one could incorporate thiol groups into protein sequences with ease, almost at choice. If there was an example where the disulfide bridge or a few free thiol group caused a major perturbation in the structure or the folding of the protein, it was usually pretty obvious by some functional or conformational (e.g. CD, ORD) analysis. Hubbell attacked the most pressing problems in protein science, which is membrane proteins, which are neither soluble nor amenable to xray crystallography or NMR. He started first in collaboration with Nobel Laureate H. G. Khorana on bacteriorhodopsin, a protein whose structure and function had eluded us up to this point, particularly with respect to the light induced conformational changes that occur [19]. The technique, in concert with the well known molecular oxygen Heisenberg exchange relaxation (broadening) of the N-O, allowed assessment of secondary structure characteristics, particularly that of bundled helical structures, which are typical of membrane spanning proteins. If one mutates every residue in a helical protein to a Cys, in

each case labels the protein, and then assesses accessibility under increased oxygen, a periodicity of about 3 -4 would be expected in the oscillation of oxygen exposure (since O2 has a higher solubility in the interior vs the solvent environment). For β-sheet structures, the periodicity would be 2 since every other residue is exposed to the solvent and vice-versa. The two figures below depict the theoretical behavior for a β-sheet and α-helical domain, respectively. Further confirmation occurs when using aqueous paramagnetic reagents such as chromium oxalate or potassium ferricyanide, which selectively relax (broaden) spin labels on the exterior of the protein pointing into the solvent [17]. Hence these accessibility parameters could be quantified and used as sensitive probes of secondary and supersecondary structure. Theoretical plots are shown below in Figure 11. In a study on lac permease, an SDSL 'scan' was done and the results are shown in Figure 12 below [21].

**Figure 11.** Idealized accessibility data plots indicating β-strand and α-helical secondary structure. From [20] with permission

[20] with permission

each case labels the protein, and then assesses accessibility under increased oxygen, a periodicity of about 3 -4 would be expected in the oscillation of oxygen exposure (since O2 has a higher solubility in the interior vs the solvent environment). For β-sheet structures, the periodicity would be 2 since every other residue is exposed to the solvent and vice-versa. The two figures below depict the theoretical behavior for a β-sheet and α-helical domain, respectively. Further confirmation occurs when using aqueous paramagnetic reagents such as chromium oxalate or potassium ferricyanide, which selectively relax (broaden) spin labels on the exterior of the protein pointing into the solvent [17]. Hence these accessibility parameters could be quantified and used as sensitive probes of secondary and supersecondary structure. Theoretical plots are shown below in Figure 11. In a study on lac permease, an SDSL 'scan' was done and the results are shown in Figure 12 below [21].

**Figure 11.** Idealized accessibility data plots indicating β-strand and α-helical secondary structure. From

**Figure 12.** Π(O2) (solid line) and 1/ΔH (broken line) versus sequence position for the nitroxide-labeled single-Cys residues at positions 387-402 in lac permease. The dotted curve is that for a function of period 3.6, and comparison with the Π(O2) and 1/ΔH functions confirms that the data are consistent with an α-helical structure. Ada[ted from [21] with permission.

The site directed spin labeling (SDSL) method blossomed by the early to mid-1990s, with dedicated sessions at meetings on the use of SDSL and EPR in protein structure. Of course, the MTSL label had some disadvantages: it still had some conformational flexibility, it could perturb the protein structure and lastly, in order to obtain an unambiguous assessment of protein structure and function, the use of additional spin labels would be desirable. Hence the Hideg lab propagated several more labels and analogues [22]. Recent work has involved distance measurements within proteins, i.e., the incorporation of two cysteines at selected positions in protein with the idea of mapping the structure by distance triangulation. This is a major effort since one obtains only the distance between the electrons on the two labels, respectively, and each spin label must be correlated back to the protein backbone with inferences from amino acid side chain structure and the aspects of motion of the label in multiple orientations. Consequently, one pair of incorporated cysteines yields just one distance. Figure 13 shows the dilemma of attaining very accurate distance measurements from a double labeling experiment. Nonetheless, the technique has still been valuable and people have developed sophisticated motional simulations in order to localize the label in the protein structure. One looks at motion around a cone and workers have attempted to come up with distances within 15 to 20% accuracy. Needless to say, the x-ray crystallographers have a major advantage (assuming that the protein can be crystallized) but the NMR spectroscopists have a bigger advantage because they can incorporate one single,

relatively rigid, spin label and obtain more than 100 distances from electron-proton paramagnetic relaxation enhancements over distance ranges of 5 to 15 Å. In fact, the EPR double label method was quite limited on distances, about that of the electron-proton distance limits. Remembering that we only obtain one distance for each labelled pair, it wasn't until Jack Freed and coworkers incorporated DHQC methodology enabling one to assess distances upwards of 50 to 60 Å [23].

**Figure 13.** Uncertainty about spin label motion and orientation(s). The spin label ring can orient in several directions, depending on the flexibility of the tether. Ideally, one would like a rigid unique orientation, but the likelihood is not high.

Some possible approaches are to anchor the label in two points on the protein. The label shown below in Figure 14 would require two mutated cysteines at proper spacing in order to meet that requirement.

**Figure 14.** Rigid two site attachment spin label. bis-MTSSL [pyrrolinyl] 2,5-Dihydro-3,4 bis(methanesulfonylthiomethyl)-2,2,5,5-tetramethyl-1H-pyrrol-1-yloxyl radical

### **7. Nitrones and spin traps: The adducts form nitroxides**

16 Nitroxides – Theory, Experiment and Applications

assess distances upwards of 50 to 60 Å [23].

orientation, but the likelihood is not high.

to meet that requirement.

relatively rigid, spin label and obtain more than 100 distances from electron-proton paramagnetic relaxation enhancements over distance ranges of 5 to 15 Å. In fact, the EPR double label method was quite limited on distances, about that of the electron-proton distance limits. Remembering that we only obtain one distance for each labelled pair, it wasn't until Jack Freed and coworkers incorporated DHQC methodology enabling one to

6-8Å 6-8Å 6-8Å 6-8Å +1416Å

**Figure 13.** Uncertainty about spin label motion and orientation(s). The spin label ring can orient in several directions, depending on the flexibility of the tether. Ideally, one would like a rigid unique

**Figure 14.** Rigid two site attachment spin label. bis-MTSSL [pyrrolinyl] 2,5-Dihydro-3,4-

bis(methanesulfonylthiomethyl)-2,2,5,5-tetramethyl-1H-pyrrol-1-yloxyl radical

Some possible approaches are to anchor the label in two points on the protein. The label shown below in Figure 14 would require two mutated cysteines at proper spacing in order


These compounds are actually a class of chemical functional groups that had been known quite early since one of the synthetic methods of producing of nitroxides is by a controlled, specific oxidation of a nitrone compound. However, these have found tremendous use in the characterization of free radicals in solution, particularly in the biological field where a plethora of potential radicals are possible. In fact, the reaction of a nitrone with carbon or oxygen-based radicals yields a nitroxide adduct with the spectrum that is characteristic of the chemistry of that particular initial radical (with some caveats that are discussed below). The use of radical-addition reactions to detect short-lived radicals was first proposed by E. G. Janzen in 1965 [24]. The early pioneers in this field studied two classes/types of spin traps which were commercially available at the time and are still on the market today: DMPO, 5,5 dimethyl-1-pyrroline-N-oxide, and PBN, alpha-phenyl N-tertiary-butyl nitrone. These and other second generation spin traps are shown in Figure 15 below.

**Figure 15.** Structures of various spin trap types including second generation nitrones. Structure abbreviations: PBN α-phenyl N-tert-butyl nitrone, 4PyOBN; DMPO 5,5-dimethylpyrroline N-oxide, EMPO 5-(ethoxycarbonyl)-5-methyl-1-pyrroline N-oxide, DEPMPO 5-(diethoxyphosphoryl)-5-methyl-1 pyrroline-N-oxide, DIPPMPO 5-diisopropoxy-phosphoryl-5-methyl-1-pyrroline-N-oxide, AMPO 5 carbamoyl-5-methyl-1-pyrroline N-oxide, MNP 2-methyl-2-nitroso propane, DBNBS 3,5-dibromo-4 nitrosobenzenesulfonate, CPCOMPO 7-oxa-1-azaspiro[4.4]non-1-en-6-one 1-oxide

DMPO could cross the cell membrane and was in fact sensitive to radicals that were present in both the aqueous media as well as the lipid medium. Some derivatives of DMPO with selected substitutions at the 1- or 4- position allow one to affect its partitioning properties between aqueous and lipid environments. Janzen successfully showed that DMPO could

trap the important reactive oxygen species: superoxide and hydroxyl radical [25]. He clearly demonstrated the differences in their EPR spectra, allowing characterization of these radicals in vivo. However it wasn't until a few years later that Rosen and colleagues found that the superoxide radical adduct of DMPO could decompose to the hydroxyl adduct by a mechanism which, to date, is still not totally clear. Hence one has to take special care, e.g., including SOD in an experiment in order to include or exclude superoxide [26].

**Figure 16.** Reaction of DMPO with an oxyradical.

The PBN spin trap was, by virtue of its non-polar lipophilic behavior, quite valuable for trapping lipid radicals and those in a membrane milieu. If PBN reacted with oxygen radicals, such as hydroxyl radical, it decomposed without any radical adduct remaining. On the other hand, the PBN-lipid adducts, if partitioned into the membrane, were stable for long periods of time and one could e.g., isolate lipid-radical adducts of PBN in erythrocytes, then extract and concentrating them later. In fact, some lipid biologically produced lipid radicals are stable in lipid media and erythrocytes could be 'post labeled' with PBN after some oxidative stress event [26].

**Figure 17.** Reaction of PBN with a carbon based radical.

These two spin traps remained with us for almost 20 years and, along with the particular annoying side reaction noted with DMPO and superoxide, the DMPO adducts tended to be fairly unstable. In the late 1980s, Tordo's group prepared a DMPO analog that contained a phosphoester-type group in one of the positions of the flanking methyl groups [28]. One of these compounds, DEPMPO, was quite successful in that the half life of the radical adduct was much longer than that known for DMPO. It is important to point out at this juncture that the reaction kinetics for all of these radical traps were quite poor, involving the necessity of having 50 – 100 mM concentrations of spin trap in the solutions. Further development of other analogs by Tordo's group and also some quality control by the commercial laboratory that were selling DEPMPO was quite an advance for the community.

18 Nitroxides – Theory, Experiment and Applications

**Figure 16.** Reaction of DMPO with an oxyradical.

**Figure 17.** Reaction of PBN with a carbon based radical.

some oxidative stress event [26].

trap the important reactive oxygen species: superoxide and hydroxyl radical [25]. He clearly demonstrated the differences in their EPR spectra, allowing characterization of these radicals in vivo. However it wasn't until a few years later that Rosen and colleagues found that the superoxide radical adduct of DMPO could decompose to the hydroxyl adduct by a mechanism which, to date, is still not totally clear. Hence one has to take special care, e.g.,

The PBN spin trap was, by virtue of its non-polar lipophilic behavior, quite valuable for trapping lipid radicals and those in a membrane milieu. If PBN reacted with oxygen radicals, such as hydroxyl radical, it decomposed without any radical adduct remaining. On the other hand, the PBN-lipid adducts, if partitioned into the membrane, were stable for long periods of time and one could e.g., isolate lipid-radical adducts of PBN in erythrocytes, then extract and concentrating them later. In fact, some lipid biologically produced lipid radicals are stable in lipid media and erythrocytes could be 'post labeled' with PBN after

These two spin traps remained with us for almost 20 years and, along with the particular annoying side reaction noted with DMPO and superoxide, the DMPO adducts tended to be fairly unstable. In the late 1980s, Tordo's group prepared a DMPO analog that contained a phosphoester-type group in one of the positions of the flanking methyl groups [28]. One of these compounds, DEPMPO, was quite successful in that the half life of the radical adduct was much longer than that known for DMPO. It is important to point out at this juncture

including SOD in an experiment in order to include or exclude superoxide [26].

However the real, major effort in our understanding and designing nitrone spin traps, based on the DMPO skeleton, was a result of an intense effort at Ohio State University by Villamena, who did synthetic, kinetic and computational design studies of these traps as well as their aminoxyl radical adducts. Hence Villamena studied both their reactivity and the stability of the adducts [29]. Overall, spin traps are powerful reagents, albeit more limited for in vivo studies due to their low sensitivity and kinetics and the concentration limits of reactive radicals in vivo. The hope was to accumulate radicals in vivo up to levels where the trapped adducts exceeded the normal in vivo level. Suffice it to say we are 'part way' there. But we still suffer from the breakdown of the radical adducts and have not yet attained optimal kinetics. Some future concepts for applications of these type of compounds would be to prepare 'spin trap labels' that could be incorporated at specifically targeted organ sites in vivo which would then would convert to the radical adduct at the time and place of radical generation. There have also been some efforts to attach fluorescent moieties onto these labels, whereby the fluorescence is quenched upon formation of the nitroxide radical adduct [30-31]. This area has great promise and some examples are diagrammed below under future developments (Section 8).

### **8. In-vivo EPR using aminoxyl radicals: History and fate**

The last frontier of applications of aminoxyl (a.k.a. nitroxide) radicals is in their applications to in-vivo studies. Since the aminoxyl (nitroxide) radicals were the first, and for a long time the only radicals that were stable and detectable in aqueous solution, such as cell and other components, it was straightforward and logical to try to examine the fate and behavior of these radicals in living systems. Early on, attempts were made to mix development of spin labeling that one attempted to mix aminoxyl radicals and living systems. In fact, in an undocumented experiment in the McConnell laboratory the toxicity of a nitroxide was tested on a goldfish. A beaker-full of the radical, t-butylnitroxide, was emptied into a Bell jar containing a goldfish. While the concentration was not accurately estimated, it was certainly in the tens to hundreds of millimolar; needless to say the fish lived, and as we learned later these labels are actually life-sustaining compounds). But someone accidentally left the hot water slowly dripping into the Bell jar and eventually the fish expired. One postdoc in the lab actually monitored his urine for the ingestion of these compounds as several of them are volatile and there was an extensive amount of synthesis and gas chromatography ongoing them lab with these volatile compounds. Needless to say, no adverse effects were found.

So what happens if you mix a nitroxide with e.g., a cell or tissue suspension? If it is the sixmembered ring species, i.e., the piperidinyl nitroxides, and easily those that can cross the cell membrane into the cytoplasm, they are rapidly reduced, i.e., 'neutralized,' within a few minutes, since a plethora of intracellular biological reducing agents ready to take on their antioxidant role and convert the nitroxide to its corresponding hydroxylamine. For example,

TEMPONE, or for that matter TEMPOL, are rapidly converted to the hydroxylamine with an immediate loss of the paramagnetism. This occurs within a few minutes. The fivemembered ring species, however, have much longer half-life, i.e. of the order of 15 to 30 min., allowing one to study some aspects of the metabolism and perhaps the ability to image this paramagnetic material in a living species. The first experiments were done by the Brasch group where they were evaluating nitroxides as MRI contrast agents[33-34]. This was followed by a plethora of studies on animals, tissue samples, blood samples, etc. where we obtained a wealth of pharmacokinetic data (although no totally clear understanding of the mechanism and detailed rate constants) [35-36]

Suffice it to say, imaging by EPR methods is challenging, if not hopelessly low resolution, since most nitroxide labels have linewidths of, at best, 0.3-0.5G for a compound that is deuterium and N-15 enriched. It has only been with the trityl radicals mentioned early where any hope of imaging was possible. However, if one takes advantage of the power and high resolution of magnetic resonance imaging (MRI) and the fact that the contrast agents in this methodology are paramagnet, then organic radicals have a place. Therefore nitroxide spin label/spin probe analogs have been tested as MRI contras agents and have met with some success. One must overcome the problem of biological reduction, also a problem with the radical adducts of nitrone spin traps since the cellular/tissue milieu contain many reducing agents such as NADH, ascorbic acid, and mitochondrial reduction sources [33]. The real quest here is to produce a well protected, aminoxyl radical that is highly resistant to biological reduction yet can be incorporate into the tissue system of choice. A few examples have been reported to date, particularly where the tetramethyl groups that flank the N-O group are replaced by long aliphatic chains such as lipids or tertiary butyl chains or cages.
