Human Serum Albumin Labelled with Sterically-Hindered Nitroxides as Potential MRI Contrast Agents.

Four albumin-nitroxide conjugates were prepared and tested as metal-free organic radical contrast agents (ORCAs) for magnetic resonance imaging (MRI). Each human serum albumin (HSA) carrier bears multiple nitroxides conjugated via homocysteine thiolactones. These molecular conjugates retain important physical and biological properties of their HSA component, and the resistance of their nitroxide groups to bioreduction was retained or enhanced. The relaxivities are similar for these four conjugates and are much greater than those of their individual components: the HSA or the small nitroxide molecules. This new family of conjugates has excellent prospects for optimization as ORCAs.


Introduction
Magnetic resonance imaging (MRI) is a powerful non-invasive technique for clinical diagnosis. The specificity of MRI is sometimes insufficient and can be improved by enhancing the spin relaxation of water protons by the addition of a contrast agent. The spin relaxation of protons is characterized by two parameters, T 1 and T 2 , called the spin-lattice and spin-spin relaxation times, respectively. Depending on their effect on T 1 and T 2 , contrast agents fall into two classes that afford positive and negative contrast images, respectively [1][2][3][4]. The most efficient of the currently-used T 1 contrast agents are paramagnetic transition metal chelates of gadolinium and manganese. However, the low stability of some chelates is a potential challenge due to the release of potentially toxic metal ions from the chelates [1,[5][6][7][8][9]. Consequently, the European Medicines Agency recently recommended restricting the use of less stable linear Gd(III) complexes and using macrocyclic Gd(III) agents at the lowest possible dose [6][7][8]. Magnetite/maghemite combinations approved for clinical use, as T 2 MRI contrast agents have little or no cytotoxicity but may have harmful cellular effects, including HTL can be chemically N-acylated by a nitroxide group to produce a reagent for making N-Hcyacylated N-homocysteinylated albumin. The wide range of known nitroxide free radicals includes bioreduction-resistant nitroxide groups that can be a basis for combining the attractive properties of HSA, such as penetrating blood-vessel barriers, passing through the blood-brain barrier and high biocompatibility, with metal-free nitroxides that resist bioreduction.
In the present work, we test the validity of this ORCA design strategy by preparing a set of spin labels via N-acylation of HTL with four nitroxide carboxylic acids 1-4 (Chart 1) and testing the suitability of their HSA-nitroxide conjugates (HSA-NIT) as metal-free ORCAs for MRI. The conjugates were characterized by electron paramagnetic resonance spectroscopy (EPR), matrix assisted laser desorption/ionization time of flight mass spectrometry (MALDI ToF MS) and circular dichroism (CD) spectroscopy. Nitroxide reduction rates in these biomaterials were measured; the relaxivities r1 and r2 were estimated; and 1 H-MRI phantom images were obtained. Initial tests of biocompatibility checked for the formation of potentially toxic aggregates using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and for cytotoxicity of the conjugates using standard cellular assays [44,45]. The results from these first four conjugates show great promise for homocysteinylation of HSA in the construction of ORCAs, while the wide variety of known nitroxide compounds provide strong potential for optimizing the performance of this new family of ORCAs.

Synthesis of N-Substituted Homocysteine Thiolactone Derivatives
Homocysteine thiolactone (HTL) is a natural cyclic product from the intramolecular acylation of homocysteine (Hcy) (Chart 1). Some aminoacyl-tRNA synthetases form HTL in their editing and proofreading reactions in order to prevent translational incorporation of Hcy into proteins [46]. HTL easily reacts with nucleophilic and electrophilic reagents at its activated carboxyl and amino groups, respectively. Reaction of HTL with the nucleophilic terminal amine of Lys can be performed in aqueous media, opening the five-membered ring of HTL and producing a stable amide bond ( Figure  1). Electrophilic reactions to attach a nitroxide to the HTL amino group face the difficulty that the hydrochloride of HTL is chemically stable in bulk [46]. Attempts to obtain the free base for reaction results in the self-condensation of two HTL molecules [47,48]. The intrinsic instability of the HTL free HTL can be chemically N-acylated by a nitroxide group to produce a reagent for making N-Hcy-acylated N-homocysteinylated albumin. The wide range of known nitroxide free radicals includes bioreduction-resistant nitroxide groups that can be a basis for combining the attractive properties of HSA, such as penetrating blood-vessel barriers, passing through the blood-brain barrier and high biocompatibility, with metal-free nitroxides that resist bioreduction.
In the present work, we test the validity of this ORCA design strategy by preparing a set of spin labels via N-acylation of HTL with four nitroxide carboxylic acids 1-4 (Chart 1) and testing the suitability of their HSA-nitroxide conjugates (HSA-NIT) as metal-free ORCAs for MRI. The conjugates were characterized by electron paramagnetic resonance spectroscopy (EPR), matrix assisted laser desorption/ionization time of flight mass spectrometry (MALDI ToF MS) and circular dichroism (CD) spectroscopy. Nitroxide reduction rates in these biomaterials were measured; the relaxivities r 1 and r 2 were estimated; and 1 H-MRI phantom images were obtained. Initial tests of biocompatibility checked for the formation of potentially toxic aggregates using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and for cytotoxicity of the conjugates using standard cellular assays [44,45]. The results from these first four conjugates show great promise for homocysteinylation of HSA in the construction of ORCAs, while the wide variety of known nitroxide compounds provide strong potential for optimizing the performance of this new family of ORCAs.

Synthesis of N-Substituted Homocysteine Thiolactone Derivatives
Homocysteine thiolactone (HTL) is a natural cyclic product from the intramolecular acylation of homocysteine (Hcy) (Chart 1). Some aminoacyl-tRNA synthetases form HTL in their editing and proofreading reactions in order to prevent translational incorporation of Hcy into proteins [46]. HTL easily reacts with nucleophilic and electrophilic reagents at its activated carboxyl and amino groups, respectively. Reaction of HTL with the nucleophilic terminal amine of Lys can be performed in aqueous media, opening the five-membered ring of HTL and producing a stable amide bond ( Figure 1). HTL can be chemically N-acylated by a nitroxide group to produce a reagent for making N-Hcyacylated N-homocysteinylated albumin. The wide range of known nitroxide free radicals includes bioreduction-resistant nitroxide groups that can be a basis for combining the attractive properties of HSA, such as penetrating blood-vessel barriers, passing through the blood-brain barrier and high biocompatibility, with metal-free nitroxides that resist bioreduction.
In the present work, we test the validity of this ORCA design strategy by preparing a set of spin labels via N-acylation of HTL with four nitroxide carboxylic acids 1-4 (Chart 1) and testing the suitability of their HSA-nitroxide conjugates (HSA-NIT) as metal-free ORCAs for MRI. The conjugates were characterized by electron paramagnetic resonance spectroscopy (EPR), matrix assisted laser desorption/ionization time of flight mass spectrometry (MALDI ToF MS) and circular dichroism (CD) spectroscopy. Nitroxide reduction rates in these biomaterials were measured; the relaxivities r1 and r2 were estimated; and 1 H-MRI phantom images were obtained. Initial tests of biocompatibility checked for the formation of potentially toxic aggregates using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and for cytotoxicity of the conjugates using standard cellular assays [44,45]. The results from these first four conjugates show great promise for homocysteinylation of HSA in the construction of ORCAs, while the wide variety of known nitroxide compounds provide strong potential for optimizing the performance of this new family of ORCAs.

Synthesis of N-Substituted Homocysteine Thiolactone Derivatives
Homocysteine thiolactone (HTL) is a natural cyclic product from the intramolecular acylation of homocysteine (Hcy) (Chart 1). Some aminoacyl-tRNA synthetases form HTL in their editing and proofreading reactions in order to prevent translational incorporation of Hcy into proteins [46]. HTL easily reacts with nucleophilic and electrophilic reagents at its activated carboxyl and amino groups, respectively. Reaction of HTL with the nucleophilic terminal amine of Lys can be performed in aqueous media, opening the five-membered ring of HTL and producing a stable amide bond ( Figure  1). Electrophilic reactions to attach a nitroxide to the HTL amino group face the difficulty that the hydrochloride of HTL is chemically stable in bulk [46]. Attempts to obtain the free base for reaction results in the self-condensation of two HTL molecules [47,48]. The intrinsic instability of the HTL free  Electrophilic reactions to attach a nitroxide to the HTL amino group face the difficulty that the hydrochloride of HTL is chemically stable in bulk [46]. Attempts to obtain the free base for reaction results in the self-condensation of two HTL molecules [47,48]. The intrinsic instability of the HTL free base requires an electrophile with enhanced reactivity for efficient reaction with the amino group; examples include acid halides, activated carboxylic acids and anhydrides [39,49,50]. Therefore, commercially-available nitroxide carboxylic acids 1 and 2, as well as the sterically-shielded, reduction-resistant nitroxide 3 [51], were first converted in situ into chloroanhydrides with thionyl chloride and then used to construct ORCAs. A racemic mixture of D and L isomers of HTL hydrochloride was then added as a powder and the solution was carefully basified again (Scheme 1). The reaction with 1 afforded the racemic spin-labeled HTL 5 as a yellow crystalline solid. The IR spectrum showed strong bands of thiolactone and amide carbonyl vibrations at 1706 and 1664 cm −1 , respectively; pyrroline C=C band at 1620 cm −1 ; and multiple broad bands of hydrogen-bonded NH vibrations at 3361, 3321 and 3054 cm −1 (See Supplementary Materials for full synthetic and analytical details). High-resolution mass spectral and element analysis data support the assigned structure. For further confirmation, the nitroxide was converted into its corresponding methoxyamine [52] and the 1 H NMR spectrum was recorded. Acylation of HTL with 2 and 3 gave mixtures of diastereomers, 6a, b and 7a, b. The isomers 6a and 6b were partly separated using chromatography and the structure of 6a was confirmed (Figure S11.1) by single-crystal X-ray analysis (all calculations were performed using SHELXTL-2018/3 [53]). The isomers 7a, b could not be separated, so their mixture was characterized.
spectrum showed strong bands of thiolactone and amide carbonyl vibrations at 1706 and 1664 cm −1 , respectively; pyrroline C=C band at 1620 cm −1 ; and multiple broad bands of hydrogen-bonded NH vibrations at 3361, 3321 and 3054 cm −1 (See Supplementary Materials for full synthetic and analytical details). High-resolution mass spectral and element analysis data support the assigned structure. For further confirmation, the nitroxide was converted into its corresponding methoxyamine [52] and the 1 H NMR spectrum was recorded. Acylation of HTL with 2 and 3 gave mixtures of diastereomers, 6a, b and 7a, b. The isomers 6a and 6b were partly separated using chromatography and the structure of 6a was confirmed (Figure S11.1) by single-crystal X-ray analysis (all calculations were performed using SHELXTL-2018/3 [53]). The isomers 7a, b could not be separated, so their mixture was characterized.
The sterically-shielded nitroxide 7 (as a diastereomeric mixture) is hydrophobic and poorly soluble in water, which complicates its use for labeling proteins in water solutions, so the more hydrophilic spin label 8 was prepared from the recently-synthesized nitroxide 9 [54]. Conversion of 9 into 4 (Scheme 1) followed the strategy for its 2,2,5,5-tetramethyl analog [55,56]. Mesylation afforded 10 in nearly quantitative yield. The reaction of 10 with LiBr appeared much smoother than the similar reaction with NaI and the resulting dibromide 11 required somewhat longer time for dehydrobromination than did the dehydrohalogenation of 3,4-bis-iodomethyl-2,2,5,5-tetraethylpyrrolidine-1-oxyl [56]. Reduction of the resulting diene 12 with the 2-propanol-hydrogen chloride system afforded 13, which was brominated as described for its 2,2,5,5-tetramethyl analog [56]. Nucleophilic substitution with NaOAc, aminolysis and two-step oxidation afforded 4. The spectral parameters of the new compounds are similar to those of their tetramethyl analogs [55,56]. The structures of 16 and 13 were confirmed by X-ray analysis data ( Figure S11.1). The acid 4 was heated with acetic anhydride and sodium acetate to form the cyclic anhydride 17 (cf. [55]), which was immediately used for HTL acylation without further purification. The spectral parameters of 8 were similar to those of 5; the structure was confirmed with elemental analysis and high-resolution mass spectra. The sterically-shielded nitroxide 7 (as a diastereomeric mixture) is hydrophobic and poorly soluble in water, which complicates its use for labeling proteins in water solutions, so the more hydrophilic spin label 8 was prepared from the recently-synthesized nitroxide 9 [54]. Conversion of 9 into 4 (Scheme 1) followed the strategy for its 2,2,5,5-tetramethyl analog [55,56]. Mesylation afforded 10 in nearly quantitative yield. The reaction of 10 with LiBr appeared much smoother than the similar reaction with NaI and the resulting dibromide 11 required somewhat longer time for dehydrobromination than did the dehydrohalogenation of 3,4-bis-iodomethyl-2,2,5,5-tetraethyl-pyrrolidine-1-oxyl [56]. Reduction of the resulting diene 12 with the 2-propanol-hydrogen chloride system afforded 13, which was brominated as described for its 2,2,5,5-tetramethyl analog [56]. Nucleophilic substitution with NaOAc, aminolysis and two-step oxidation afforded 4. The spectral parameters of the new compounds are similar to those of their tetramethyl analogs [55,56]. The structures of 16 and 13 were confirmed by X-ray analysis data ( Figure S11.1). The acid 4 was heated with acetic anhydride and sodium acetate to form the cyclic anhydride 17 (cf. [55]), which was immediately used for HTL acylation without further purification. The spectral parameters of 8 were similar to those of 5; the structure was confirmed with elemental analysis and high-resolution mass spectra.

Synthesis and Characterization of HSA-Nitroxide Conjugates
N-Homocysteinylation of HSA by HTL-nitroxide derivatives 5-8 was done under physiological-like conditions (PBS, 37 • C, pH 7.4) (Figure 1 and Supplementary Materials). Unreacted reagent and other low MW compounds were removed by centrifugal filtration using Centricon concentrators. The yields of HSA-NIT conjugates were 95%. For comparison, N-homocysteinylated HSA was synthesized. The reaction conditions were optimized with respect to DMSO content, HSA concentration and the excess of HTL or HTL derivatives. The incorporation of Hcy-nitroxide residues into HSA was proven by MALDI-ToF, EPR and Ellman's assay for free SH-groups [57], with unmodified HSA as control.
A new SH-group is introduced in the resulting HSA by each N-homocysteinylation reaction. The albumin SH-group content and measured extent of reaction increased with increasing molar ratios of acylating reagent (HTL or HTL derivative) to protein ( Figure S3.1). The extent of HSA modification does not increase significantly with greater than a 30-fold excess of HTL derivatives, which can be explained as the acylation of the main albumin N-homocysteinylation sites.
The extent of modification derived from Ellman's test was much lower than from MALDI ToF or EPR (Tables S1 and S2), an artifact of the use of DMSO. Despite its wide use as a safe solvent for pharmacological substances and for biopolymer modification reactions, DMSO can cause protein conformational changes and oligomerization [58,59]. DMSO is sometimes used as a mild oxidizing agent for SH groups [60,61]. Therefore, the lowest possible dose of DMSO was used to minimize its adverse effects.
N-homocysteinylation, like any protein modification, might cause protein damage, loss of function, conformational changes and precipitation; or might change susceptibility to oxidation, proteolysis and oligomerization [46,48,62,63]. Gel electrophoresis (SDS-PAGE) was used to check the HSA-NIT for any changes in mobility from that of HSA (Table 1) that could indicate aggregation or protein damage. Bands corresponding to monomeric protein (MW = 66.5 kDa) and its dimer (MW~130 kDa) were observed in all samples, including the starting HSA, but the HSA-NIT conjugates showed no significant changes from the starting, natural HSA ( Figure S3.3). Table 1.
HSA conjugate secondary structure from CD and aggregates from quantitative SDS-PAGE analysis. Changes in the secondary structure of the protein were examined by deconvolution of CD spectra to determine the α-helix and β-sheet content (Table 1) based on CCA+ software [64,65]. The secondary structure of the HSA remained intact: α-helical content of the HSA-NIT conjugates decreased slightly from 55% to 46%-48%, while the β-sheet content was almost unchanged. N-homocysteinylation of HSA by HTL produced a slight increase in β-sheets (Hcy-HSA, Table 1) [47,66]. The positive charge of the Hcy amino group and its interaction with a protein side chain cause these slight conformational adjustments [47], but CD and SDS-PAGE show that the structural integrity of the HSA was preserved in all four conjugates.

Identification of N-Homocysteinylation Sites in HSA-NIT Conjugates
The HSA sites modified by the HTL derivatives were identified using trypsin digestion with MALDI-ToF peptide mapping. Each HSA-NIT conjugate is a heterogeneous mixture having Hcy-nitroxide residues attached at different lysine residues (Table S4). The major modified sites coincide with the N-homocysteinylation sites of HSA in vitro and in vivo [41][42][43] (Table 2). We were able to map sites over a large coverage of the HSA. Surprisingly, many additional Lys residues were modified by the N-substituted HTL derivatives:  (Table S4). N-homocysteinylation by HTL-nitroxide derivatives can occur at many previously unreported sites and offers the potential for ORCAs with very high loadings, or with additional diagnostic or therapeutic groups. Various factors, including protein conformation, nucleophilicity, local charges and pK a , determine the reactivity of potential acylation sites, and can help explain the sites of modification. The pK a value and the fractional accessible surface area (ASA) were estimated for each Lys side chain [67][68][69]. Lys residues with low pK a values should be unprotonated, and hence more active to nucleophilic substitution. Side chains with high ASA values are more accessible to modifying reagents. For example, Lys-525 has the lowest calculated pK a among the residues usually N-homocysteinylated, reflecting this trend (Table S5). Consequently, the pattern of Lys modification in HSA-NIT broadly resembles that of natural HSA post-translational modification which helps preserve biocompatibility.
Attachment of the nitroxide groups to HSA was verified by room-temperature continuous-wave (CW) EPR spectra ( Figure 2). Binding of a nitroxide to any protein significantly increases its rotation correlation time to >1.0 ns, producing pronounced changes in its CW EPR spectrum [70]. Spectral modeling indicates that 99% of the nitroxides are attached to the HSA and exhibit slowed rotation with τ c >1.0 ns (Table S3). The remaining <1% are trace amounts of incompletely washed-out free 5-8, freely tumbling in solution and unaffected by the protein (τ c~0 .6 ns).
Attachment of the nitroxide groups to HSA was verified by room-temperature continuous-wave (CW) EPR spectra ( Figure 2). Binding of a nitroxide to any protein significantly increases its rotation correlation time to >1.0 ns, producing pronounced changes in its CW EPR spectrum [70]. Spectral modeling indicates that 99% of the nitroxides are attached to the HSA and exhibit slowed rotation with τc >1.0 ns (Table S3). The remaining <1% are trace amounts of incompletely washed-out free 5-8, freely tumbling in solution and unaffected by the protein (τc~0.6 ns). X-band CW EPR spectrum of spin-labelled conjugates at 300 K. Black and blue lines with noise are experimental spectra. The experimental spectra were simulated using EasySpin software [71,72]. Red lines are simulated spectra with parameters listed in Table S3. Red arrow-«fast», black-«slow» motion component. Figure 2. X-band CW EPR spectrum of spin-labelled conjugates at 300 K. Black and blue lines with noise are experimental spectra. The experimental spectra were simulated using EasySpin software [71,72]. Red lines are simulated spectra with parameters listed in Table S3. Red arrow-«fast», black-«slow» motion component.
The EPR spectra from nitroxides at each modification site overlap, preventing characterization of individual modification sites by EPR. However, the partially-immobilized labels can be classified by motional correlation time (τ c )-2-3 ns («fast») or 10-14 ns («slow»)-and attributed to labels located outside and inside pockets of the protein, respectively (Table 3 and Table S3). Most labels are in the «slow» class whose long τ c reveals a rather confining environment [15,70]. This characterization is supported by their nitrogen hyperfine constants (HFC): the «slow» labels have a smaller HFC resulting from their less polar environment inside the protein, while the «fast» labels are less hindered in the polar surroundings at the exposed HSA surface [65]. The strongly immobilized «slow» labels comprise the majority in all samples, especially those with tetraethyl substituted radicals. The EPR spectra from nitroxides at each modification site overlap, preventing characterization of individual modification sites by EPR. However, the partially-immobilized labels can be classified by motional correlation time (τc)-2-3 ns («fast») or 10-14 ns («slow»)-and attributed to labels located outside and inside pockets of the protein, respectively (Table 3 and Table S3). Most labels are in the «slow» class whose long τc reveals a rather confining environment [15,70]. This characterization is supported by their nitrogen hyperfine constants (HFC): the «slow» labels have a smaller HFC resulting from their less polar environment inside the protein, while the «fast» labels are less hindered in the polar surroundings at the exposed HSA surface [65]. The strongly immobilized «slow» labels comprise the majority in all samples, especially those with tetraethyl substituted radicals.

Reduction Rates of HSA-NIT Conjugates
Bioreduction of nitroxides to diamagnetic hydroxylamines, which is typical under physiological conditions [26][27][28]74], negatively affects ORCA relaxivity and complicates their use as MRI contrast agents. Bio-oxidation is rare and not a problem [28,75]. We evaluated the reduction rates of nitroxide groups in the conjugates and in 1-4 by ascorbate, the reducing agent standardly used to estimate the redox properties of nitroxides [21,29,54]. Reduction rates were measured under pseudo-first-order conditions using a 400-fold excess of ascorbate (≈100-150 per nitroxide unit) and a 200-fold excess of The EPR spectra from nitroxides at each modification site overlap, preventing characterization of individual modification sites by EPR. However, the partially-immobilized labels can be classified by motional correlation time (τc)-2-3 ns («fast») or 10-14 ns («slow»)-and attributed to labels located outside and inside pockets of the protein, respectively (Table 3 and Table S3). Most labels are in the «slow» class whose long τc reveals a rather confining environment [15,70]. This characterization is supported by their nitrogen hyperfine constants (HFC): the «slow» labels have a smaller HFC resulting from their less polar environment inside the protein, while the «fast» labels are less hindered in the polar surroundings at the exposed HSA surface [65]. The strongly immobilized «slow» labels comprise the majority in all samples, especially those with tetraethyl substituted radicals.

Reduction Rates of HSA-NIT Conjugates
Bioreduction of nitroxides to diamagnetic hydroxylamines, which is typical under physiological conditions [26][27][28]74], negatively affects ORCA relaxivity and complicates their use as MRI contrast agents. Bio-oxidation is rare and not a problem [28,75]. We evaluated the reduction rates of nitroxide groups in the conjugates and in 1-4 by ascorbate, the reducing agent standardly used to estimate the redox properties of nitroxides [21,29,54]. Reduction rates were measured under pseudo-first-order conditions using a 400-fold excess of ascorbate (≈100-150 per nitroxide unit) and a 200-fold excess of The EPR spectra from nitroxides at each modification site overlap, preventing characterization of individual modification sites by EPR. However, the partially-immobilized labels can be classified by motional correlation time (τc)-2-3 ns («fast») or 10-14 ns («slow»)-and attributed to labels located outside and inside pockets of the protein, respectively (Table 3 and Table S3). Most labels are in the «slow» class whose long τc reveals a rather confining environment [15,70]. This characterization is supported by their nitrogen hyperfine constants (HFC): the «slow» labels have a smaller HFC resulting from their less polar environment inside the protein, while the «fast» labels are less hindered in the polar surroundings at the exposed HSA surface [65]. The strongly immobilized «slow» labels comprise the majority in all samples, especially those with tetraethyl substituted radicals.

Reduction Rates of HSA-NIT Conjugates
Bioreduction of nitroxides to diamagnetic hydroxylamines, which is typical under physiological conditions [26][27][28]74], negatively affects ORCA relaxivity and complicates their use as MRI contrast agents. Bio-oxidation is rare and not a problem [28,75]. We evaluated the reduction rates of nitroxide groups in the conjugates and in 1-4 by ascorbate, the reducing agent standardly used to estimate the redox properties of nitroxides [21,29,54]. Reduction rates were measured under pseudo-first-order conditions using a 400-fold excess of ascorbate (≈100-150 per nitroxide unit) and a 200-fold excess of The EPR spectra from nitroxides at each modification site overlap, preventing characterization of individual modification sites by EPR. However, the partially-immobilized labels can be classified by motional correlation time (τc)-2-3 ns («fast») or 10-14 ns («slow»)-and attributed to labels located outside and inside pockets of the protein, respectively (Table 3 and Table S3). Most labels are in the «slow» class whose long τc reveals a rather confining environment [15,70]. This characterization is supported by their nitrogen hyperfine constants (HFC): the «slow» labels have a smaller HFC resulting from their less polar environment inside the protein, while the «fast» labels are less hindered in the polar surroundings at the exposed HSA surface [65]. The strongly immobilized «slow» labels comprise the majority in all samples, especially those with tetraethyl substituted radicals.

Reduction Rates of HSA-NIT Conjugates
Bioreduction of nitroxides to diamagnetic hydroxylamines, which is typical under physiological conditions [26][27][28]74], negatively affects ORCA relaxivity and complicates their use as MRI contrast agents. Bio-oxidation is rare and not a problem [28,75]. We evaluated the reduction rates of nitroxide groups in the conjugates and in 1-4 by ascorbate, the reducing agent standardly used to estimate the redox properties of nitroxides [21,29,54]. Reduction rates were measured under pseudo-first-order conditions using a 400-fold excess of ascorbate (≈100-150 per nitroxide unit) and a 200-fold excess of The EPR spectra from nitroxides at each modification site overlap, preventing characterization of individual modification sites by EPR. However, the partially-immobilized labels can be classified by motional correlation time (τc)-2-3 ns («fast») or 10-14 ns («slow»)-and attributed to labels located outside and inside pockets of the protein, respectively (Table 3 and Table S3). Most labels are in the «slow» class whose long τc reveals a rather confining environment [15,70]. This characterization is supported by their nitrogen hyperfine constants (HFC): the «slow» labels have a smaller HFC resulting from their less polar environment inside the protein, while the «fast» labels are less hindered in the polar surroundings at the exposed HSA surface [65]. The strongly immobilized «slow» labels comprise the majority in all samples, especially those with tetraethyl substituted radicals.

Reduction Rates of HSA-NIT Conjugates
Bioreduction of nitroxides to diamagnetic hydroxylamines, which is typical under physiological conditions [26][27][28]74], negatively affects ORCA relaxivity and complicates their use as MRI contrast agents. Bio-oxidation is rare and not a problem [28,75]. We evaluated the reduction rates of nitroxide groups in the conjugates and in 1-4 by ascorbate, the reducing agent standardly used to estimate the redox properties of nitroxides [21,29,54]. Reduction rates were measured under pseudo-first-order The EPR spectra from nitroxides at each modification site overlap, preventing characterization of individual modification sites by EPR. However, the partially-immobilized labels can be classified by motional correlation time (τc)-2-3 ns («fast») or 10-14 ns («slow»)-and attributed to labels located outside and inside pockets of the protein, respectively (Table 3 and Table S3). Most labels are in the «slow» class whose long τc reveals a rather confining environment [15,70]. This characterization is supported by their nitrogen hyperfine constants (HFC): the «slow» labels have a smaller HFC resulting from their less polar environment inside the protein, while the «fast» labels are less hindered in the polar surroundings at the exposed HSA surface [65]. The strongly immobilized «slow» labels comprise the majority in all samples, especially those with tetraethyl substituted radicals.

Reduction Rates of HSA-NIT Conjugates
Bioreduction of nitroxides to diamagnetic hydroxylamines, which is typical under physiological conditions [26][27][28]74], negatively affects ORCA relaxivity and complicates their use as MRI contrast agents. Bio-oxidation is rare and not a problem [28,75]. We evaluated the reduction rates of nitroxide groups in the conjugates and in 1-4 by ascorbate, the reducing agent standardly used to estimate the redox properties of nitroxides [21,29,54]. Reduction rates were measured under pseudo-first-order R=OH 0.063 [26,29] 0.097 [ The EPR spectra from nitroxides at each modification site overlap, preventing characterization of individual modification sites by EPR. However, the partially-immobilized labels can be classified by motional correlation time (τc)-2-3 ns («fast») or 10-14 ns («slow»)-and attributed to labels located outside and inside pockets of the protein, respectively (Table 3 and Table S3). Most labels are in the «slow» class whose long τc reveals a rather confining environment [15,70]. This characterization is supported by their nitrogen hyperfine constants (HFC): the «slow» labels have a smaller HFC resulting from their less polar environment inside the protein, while the «fast» labels are less hindered in the polar surroundings at the exposed HSA surface [65]. The strongly immobilized «slow» labels comprise the majority in all samples, especially those with tetraethyl substituted radicals.

Reduction Rates of HSA-NIT Conjugates
Bioreduction of nitroxides to diamagnetic hydroxylamines, which is typical under physiological conditions [26][27][28]74], negatively affects ORCA relaxivity and complicates their use as MRI contrast agents. Bio-oxidation is rare and not a problem [28,75]. We evaluated the reduction rates of nitroxide groups in the conjugates and in 1-4 by ascorbate, the reducing agent standardly used to estimate the redox properties of nitroxides [21,29,54]. Reduction rates were measured under pseudo-first-order The EPR spectra from nitroxides at each modification site overlap, preventing characterization of individual modification sites by EPR. However, the partially-immobilized labels can be classified by motional correlation time (τc)-2-3 ns («fast») or 10-14 ns («slow»)-and attributed to labels located outside and inside pockets of the protein, respectively (Table 3 and Table S3). Most labels are in the «slow» class whose long τc reveals a rather confining environment [15,70]. This characterization is supported by their nitrogen hyperfine constants (HFC): the «slow» labels have a smaller HFC resulting from their less polar environment inside the protein, while the «fast» labels are less hindered in the polar surroundings at the exposed HSA surface [65]. The strongly immobilized «slow» labels comprise the majority in all samples, especially those with tetraethyl substituted radicals.

Reduction Rates of HSA-NIT Conjugates
Bioreduction of nitroxides to diamagnetic hydroxylamines, which is typical under physiological conditions [26][27][28]74], negatively affects ORCA relaxivity and complicates their use as MRI contrast agents. Bio-oxidation is rare and not a problem [28,75]. We evaluated the reduction rates of nitroxide groups in the conjugates and in 1-4 by ascorbate, the reducing agent standardly used to estimate the redox properties of nitroxides [21,29,54]. Reduction rates were measured under pseudo-first-order

Reduction Rates of HSA-NIT Conjugates
Bioreduction of nitroxides to diamagnetic hydroxylamines, which is typical under physiological conditions [26][27][28]74], negatively affects ORCA relaxivity and complicates their use as MRI contrast agents. Bio-oxidation is rare and not a problem [28,75]. We evaluated the reduction rates of nitroxide groups in the conjugates and in 1-4 by ascorbate, the reducing agent standardly used to estimate the redox properties of nitroxides [21,29,54]. Reduction rates were measured under pseudo-first-order conditions using a 400-fold excess of ascorbate (≈100-150 per nitroxide unit) and a 200-fold excess of glutathione per albumin molecule in pH 7.4 PBS. Second-order rate constants for reduction were obtained: (1) from the initial decay of the low-field EPR peak amplitude (labelled init.) with calibration by the double integral of the EPR spectrum; and (2) from the change of the 1 H NMR r 1 -relaxation rate with time. As seen in Table 3, the r 1 -relaxation measurement is suitable only for the more stable nitroxides because it requires more time than is available with rapidly-decaying nitroxides.
The nitroxide environment in the HSA-NIT conjugates can alter the rate of nitroxide reduction [15]. For example, reduction of 1 was significantly slower in its HSA-NIT-5 conjugate (k = 0.096 M −1 s −1 ) than in a free solution of 1 (k = 0.22-0.58 M −1 s −1 ). In contrast, the reduction of radical 3 seems unaffected by environment whether it is conjugated in HSA-NIT-7 or free in solution ( Table 3). The reduction-resistance of nitroxides 3 and 4 can be preserved in HSA-conjugate ORCAs to enable MRI measurements over longer time scales.
We found above ( Table 2, Table S4, and Section 5 of Supplemental Materials) that chemical reaction of HSA with HTL derivatives depends on the accessibility of the site. The same effect is seen in the reduction of nitroxides on the outside and inside of virus particle-based ORCAs, and can be expected in the HSA-NIT. Spectral simulations, made throughout the reduction, determined the amount of the «slow» and «fast» motional components. Those values were used to determine reduction rates of the nitroxides belonging to each motional component ( Table 3). The reduction rate constants for the «slow» component are 2-5 fold slower than for the «fast» component. This difference is presumably because the protein provides reducing agents less access to the «slow» component. This effect provides a possible strategy to produce even more stable conjugates.
The decay rates obtained from the initial decay of all the nitroxides (init.) agree well with the rates from the 1 H r 1 measurements for the sterically-hindered HSA-NIT-7 and HSA-NIT-8. This suggests that nitroxides in both motional components contribute similarly to the relaxivity. There is little correlation between the EPR-derived reduction rates and those from the r 1 measurements for the rapidly reduced conjugates, which is likely an artifact from making a lengthy r 1 measurement on a rapidly-changing sample.
A slow reduction rate is vital for a nitroxide-based ORCA to provide consistent relaxivity for the MRI measurement. Thus, it is significant that under similar conditions, HSA-NIT-7 and HSA-NIT-8, with k = 0.0018 ± 0.0002 M −1 s −1 and 0.0022 ± 0.0002 M −1 s −1 , respectively, are reduced more than an order of magnitude slower than the nitroxides on a dendrimer-based carrier proposed as a metal-free ORCA for MRI with k = 0.0376 M −1 s −1 [15].

Cytotoxicity of HSA-NIT Conjugates
The viability of cells in the presence of the HSA-NIT conjugates was investigated by the widely-applied 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) test using breast cancer MCF-7 and human glioblastoma T98G cells [44,45]. During exponential growth phase, cell cultures were treated for 72 h with amounts of the HSA conjugates in the range expected for ORCA MRI applications (see discussion below). The HSA-NIT conjugates show no significant difference from native HSA (Figure 3).
The viability of cells in the presence of the HSA-NIT conjugates was investigated by the widelyapplied 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) test using breast cancer MCF-7 and human glioblastoma T98G cells [44,45]. During exponential growth phase, cell cultures were treated for 72 h with amounts of the HSA conjugates in the range expected for ORCA MRI applications (see discussion below). The HSA-NIT conjugates show no significant difference from native HSA (Figure 3).

Characterization of HSA-NIT Magnetic properties
The 1 H water relaxivities r1 and r2 of the HSA-NIT conjugates in PBS were measured using a 7 T Bruker Avance III 300 MHz spectrometer at 25 °C (Table 4). The plots of the 1 H relaxation rate of water (1/T1) versus the nitroxide concentration were linear for all agents, indicating the absence of aggregation ( Figure S7.1). The per-nitroxide r1 values ranged from 0.33 to 0.51 mM −1 s −1 (Table 4), making them comparable to those reported in various nitroxide-bound polymers and dendrimers [15,29] and higher than for potential low-molecular-weight, nitroxide-based MRI probes at 7 T [20,21]. Moreover, per-nitroxide r2 values ranged from 4.7 to 7.2 mM −1 s −1 , much greater than what is typical for individual nitroxides in solution; e.g., 2 has r1 and r2 of 0.15 and 0.17 mM −1 s −1 , respectively [76].

Characterization of HSA-NIT Magnetic Properties
The 1 H water relaxivities r 1 and r 2 of the HSA-NIT conjugates in PBS were measured using a 7 T Bruker Avance III 300 MHz spectrometer at 25 • C ( Table 4). The plots of the 1 H relaxation rate of water (1/T 1 ) versus the nitroxide concentration were linear for all agents, indicating the absence of aggregation ( Figure S7.1). The per-nitroxide r 1 values ranged from 0.33 to 0.51 mM −1 s −1 (Table 4), making them comparable to those reported in various nitroxide-bound polymers and dendrimers [15,29] and higher than for potential low-molecular-weight, nitroxide-based MRI probes at 7 T [20,21]. Moreover, per-nitroxide r 2 values ranged from 4.7 to 7.2 mM −1 s −1 , much greater than what is typical for individual nitroxides in solution; e.g., 2 has r 1 and r 2 of 0.15 and 0.17 mM −1 s −1 , respectively [76]. HSA-NIT-6 and HSA-NIT-7 contain nitroxides with similar ring structures and have good relaxivities, but have reduction rates differing by about half an order of magnitude; so we compared their relaxivities at the physiological temperature of 37 • C ( Table 4). The r 1 and r 2 values decreased similarly as expected, with the r 2 /r 1 ratios virtually unchanged. The molecular structure features of the nitroxide group that control reduction rates do not seem to significantly affect relaxivity or its temperature dependence.
In fact, the r 2 per nitroxide group (4.3-5.3 mM −1 s −1 ) for both conjugates, at 37 • C and 7 T, match or exceed those reported for other ORCAs (0.17-4.7 mM −1 s −1 ) [23]. This was achieved at a relatively light loading of 3.9 nitroxides per HSA (Table 4). However, HTL spin labels can be possibly attached to at least 23 Lys residues (Table S4), providing scope for increasing, as much as 5-fold, the loading and the relaxivity per HSA.  Figure 4). There is no significant difference in contrast efficiency between the different types of nitroxides. The contrast differences among samples correlate principally with their level of loading with nitroxide.
In fact, the r2 per nitroxide group (4.3-5.3 mM −1 s −1 ) for both conjugates, at 37 °C and 7 T, match or exceed those reported for other ORCAs (0.17-4.7 mM −1 s −1 ) [23]. This was achieved at a relatively light loading of 3.9 nitroxides per HSA (Table 4). However, HTL spin labels can be possibly attached to at least 23 Lys residues (Table S4), providing scope for increasing, as much as 5-fold, the loading and the relaxivity per HSA.
T1-and T2-weighted MRIs on tube phantoms of the HSA-NIT biomaterials in PBS solutions show excellent contrast between samples with 0.0, 0.2 and 0.5 mM concentrations at 7 T and RT (Figure 4). There is no significant difference in contrast efficiency between the different types of nitroxides. The contrast differences among samples correlate principally with their level of loading with nitroxide. Positive and negative contrast enhancements are clearly observed in T1 and T2 images, respectively. The r2/r1 values of 10.4-14.3 obtained in a 7 T magnetic field suggest that at high field strengths the conjugates may work most effectively as T2 contrast agents [15,19]. The r2/r1 values per nitroxide exceed those of most other nitroxide-based ORCAs [15,20,23] and the r1 relaxivities are similar to values at 7 T for other ORCAs.
The MRI performance of HSA-NITs is quite encouraging because the four conjugates characterized here already show good relaxivity and indicate a scope for substantial performance increases for this new class of ORCAs. The large number of known nitroxide molecules provides a range of chemical and physical properties for optimizing an HSA-based ORCA. The high solubility T1-weighted MRI T2-weighted MRI Tube Positioning Positive and negative contrast enhancements are clearly observed in T 1 and T 2 images, respectively. The r 2 /r 1 values of 10.4-14.3 obtained in a 7 T magnetic field suggest that at high field strengths the conjugates may work most effectively as T 2 contrast agents [15,19]. The r 2 /r 1 values per nitroxide exceed those of most other nitroxide-based ORCAs [15,20,23] and the r 1 relaxivities are similar to values at 7 T for other ORCAs.
The MRI performance of HSA-NITs is quite encouraging because the four conjugates characterized here already show good relaxivity and indicate a scope for substantial performance increases for this new class of ORCAs. The large number of known nitroxide molecules provides a range of chemical and physical properties for optimizing an HSA-based ORCA. The high solubility of HSA and biocompatibility of HSA-NIT ORCAs can also bring advantages. HSA readily crosses blood-vessel and blood-brain barriers and can accumulate in tumors; hence HSA conjugates can be injected remotely from sensitive organs/tissues and be delivered by the circulation. The properties of HSA are preserved when conjugated with HTL or its derivatives [40,47,77] (and see above), and could be exploited with ORCAs to specifically enhance contrast for tumors. HSA-based constructs often bind to albondin on the endothelium and SPARC in the tumor interstitium, thereby accumulating in and highlighting tumors, including brain tumors (glioma) [32,40,47,77,78]. Finally, the lifetime of HSA in the body is about 25 days with a normal turnover of more than 10 g per day [79], readily accommodating the HSA added by a conjugate optimized for use as an ORCA.

Conclusions
A new family of nitroxide-bearing conjugates (HSA-NIT) was constructed by the attachment of four different HTL-based spin labels to human serum albumin. These conjugates retain important physical and biological properties of their HSA component, and the chemical resistance of their nitroxide groups was retained or enhanced. The relaxivities of these four conjugates are similar and are much greater than those of their components: the HSA or the small nitroxide molecules.
It is possible to homocysteinylate HSA at previously-undocumented Lys locations, providing the potential of much heavier loading of nitroxides for HSA-NIT conjugates with even greater relaxivities, potentially surpassing those of the best current ORCAs. The combination of the inherent biocompatibility and long blood half-life of albumin, in addition to the possibility for these albumin conjugates to accumulate in tumors and the excellent stability of their nitroxide radicals under reducing conditions, can provide promising contrast media for MRI diagnostics. Last but not least, these HSA-NIT conjugates are also an extensible basis for the development of theranostics that combine ORCAs with therapeutic payloads to achieve simultaneous drug delivery and tumor imaging.