Towards a Nano-Encapsulated EPR Imaging Agent for Clinical Use

Abstract

Purpose:

Progress toward developing a novel radiocontrast agent for determining pO₂ in tumors in a clinical setting is described. The imaging agent is designed for use with electron-paramagnetic-resonance-imaging (EPRI), in which the collision of a paramagnetic probe molecule with molecular oxygen causes a spectroscopic change which can be calibrated to give the real oxygen concentration in the tumor tissue.

Procedures:

The imaging agent is based on a nano-scaffold of aluminum hydroxide (boehmite) with sizes from 100 – 200 nm, paramagnetic probe molecule, and encapsulation with a gas permeable, thin (10–20 nm) polymer layer to separate the imaging agent and body environment while still allowing O₂ to interact with the paramagnetic probe. A specially designed deuterated Finland trityl (dFT) is covalently attached on the surface of the nanoparticle through 1,3-dipolar addition of the alkyne on the dFT with an azide on the surface of the nano-scaffold. This click-chemistry reaction affords 100% efficiency of the trityl attachment as followed by the complete disappearance of the azide peak in the infrared spectrum. The fully encapsulated, dFT-functionalized nanoparticle is referred to as RADI-Sense.

Results:

Side-by-side in-vivo imaging comparisons made in a mouse model made between RADI-Sense and free paramagnetic probe (OX-071) showed oxygen sensitivity is retained and RADI-Sense can create 3D pO₂ maps of solid tumors

Conclusions:

A novel encapsulated nanoparticle EPR imaging agent has been described which could be used in the future to bring EPR imaging for guidance of radiotherapy into clinical reality.

I. INTRODUCTION:

Low oxygen (hypoxia) is a hallmark of rapidly growing solid tumors. Hypoxia has been indicated in a number of disease states, including anemia, renal failure and cardiovascular disease, in addition to its central role in carcinogenesis. The diverse partial pressure of oxygen (pO₂) in solid tumors has been recognized since 1955^[1]. Hypoxic fractions have been shown to make up to 25% of uterine, cervix, head/neck and breast cancers, whereas no such hypoxia areas exist in normal tissue^[2]. Tumor hypoxic regions are associated with resistance to radiation and a poor patient prognosis^[3].

X-ray irradiation infers a small amount of damage directly to tumor DNA. Indirect damage from the irradiation, caused by oxygen-based radicals (i.e., superoxide or peroxyl) predominates in producing the lethal damage. Strand breaks in tumor DNA are represented by carbon-centered radicals, and by themselves can be repaired by cellular maintenance mechanisms. However, if the carbon-centered radical interacts with oxygen, it produces a peroxyl radical which is more difficult to repair, and therefore more likely to be lethal to the tumor^[4].

A precise knowledge of the 3D distribution of pO₂ and specifically the hypoxic regions which require a higher radiation dose to achieve an acceptable outcome is required. The higher doses required for the hypoxic regions are much more than is required for highly oxygenated regions, and much more likely to do collateral damage to normal tissue surrounding the solid tumor. Thus, having the ability to map the pO₂ in 3-dimensions allows the radiotherapist to deliver the higher radiation doses to hypoxic regions specifically while treating well-oxygenated regions of the tumor with much lighter radiation doses which have a reduced risk of damaging healthy tissue^[5].

Multiple techniques have been developed to try and measure oxygen concentration and pO₂ in tumors. Tumor vasculature is disordered, and oxygen consumption is much higher than in normal tissue. Blood Oxygen Level-Dependent (BOLD) imaging can measure changes in blood oxygenation by detecting changes in T₂-weighted images^{[6, 7]}. BOLD-MRI and dynamic-contrast-enhanced MRI (DCE-MRI) are both clinically available approaches for tumor hypoxia imaging^[8]. A drawback to MR approaches is that these approaches measure the change in blood oxygenation that may be poorly correlated with tumor oxygenation that has leaky and disrupted vasculature ^[9]. Also, BOLD measures blood oxygenation that is poorly correlated with tumor tissue oxygenation. An accurate measurement is needed that can be used as a biomarker in therapy decisions.

NIR-fluorescence probes and phosphorescence lifetime measurements have also been employed to estimate the partial pressure of oxygen in tumors^[10]. The optical approaches are limited by the penetration depth of light into the tissue of living animals. In addition, optical signals are susceptible to interference from biological molecules. Invasive microelectrodes can measure the electric current from reduction of oxygen at the cathode and are the current “gold standard” measurement. However, oxygen is consumed during the measurement, so repeated measurements at the same time and place cannot be taken. Positron emission tomography (PET) probes (i.e. ¹⁸Fluoromisonidazole) are retained in low oxygen environments but cannot provide a quantitative measure of pO₂^[11]. In addition, costly cyclotrons are required to produce the radionuclides. Problems also arise from the high background of non-metabolized ¹⁸F-misonidazole, and PET has low spatial resolution and is not calibrated for measuring O₂ concentration.

In-vivo electron paramagnetic resonance imaging (EPRI) is a quantitative method for measuring oxygen concentration^[12–14]. Exogenously delivered EPR probes interact with oxygen in the tumor or target site. Collision of the paramagnetic probe with O₂ causes a broadening of linewidth and a decrease in the characteristic spin-lattice (T_1e) and spin-spin (T_2e) relaxation times of the probe, thus a change in linewidth or relaxation time in-vivo reports on a change in [O₂]. EPRI is fast enough for successive oxygen images to be acquired between pre-and post- chemo- or radiotherapy treatments. In addition, EPRI facilitates the comparison of quantitative pO₂ measurements between different patients.

The EPRI technique quantitatively maps pO₂^[15]. Higher radiation doses delivered to hypoxic regions indicated by EPRI improved the survivability of test mice compared to control^[5]. A major hurdle has so far prevented translation of this breakthrough technology to helping patients. There are no FDA cleared directly injectable paramagnetic imaging agents for human use, but we have developed one which could be.

Our overall goal is to retain the interaction of molecular oxygen with our EPR probe (deuterated Finland trityl, dFT), while separating the body and probe environments. This is necessary for dFT since it has the longest electron relaxation times (T_1e) compared to other trityl radicals but is also very hydrophobic. The long relaxation times are best for imaging, but the hydrophobic nature is poor from the standpoint of binding to hydrophobic patches of proteins. We accomplish this separation in two ways. First, a covalent triazole linkage between the dFT and alumoxane is achieved through copper-mediated cycloaddition, so called “click-chemistry”, which localizes the dFT to the nanoparticle surface. Second, we encapsulate the nanoparticles with a thin polymer layer which is oxygen permeable. The polymer layer acts as a physical barrier between the dFT and hydrophobic protein patches, while retaining oxygen sensitivity.

The inspiration for encapsulated hydrophobic EPRI agents – separation of probe/body environment while retaining imaging activity - lies in the very successful MRI contrast agents based on Gadolinium (Gd³⁺). Gd³⁺ alone can be toxic to some individuals, especially those with impaired renal function^[16]. In addition, Gd³⁺ has a high affinity for phosphate, citrate and serum albumin (similar to hydrophobic dFT)^[17]. However, the FDA has cleared several Gd³⁺ based agents as long as the chelating ligands (for example, diethylene-triamine-penta-acetic acid, DTPA) are effective at containing the Gd³⁺ and separating the action of Gd³⁺ from the wider body environment^[18].

The use of Nanoparticles in Cancer Diagnostics and EPRI

Nanoparticle based approaches are widely used in cancer therapy and diagnostics. Both tissue and cellular distribution profiles of anticancer drugs can be controlled by entrapment or association with submicronic colloidal systems (i.e. nanoparticles)^[19]. Nanoparticles are also decorated with ligands for active targeting of cancerous cells, as well as used as imaging agents. Nanoparticles for cancer therapy can also be vesicles, designed to protect the body from the paramagnetic agent, vice-versa, or both.

Redox nanoparticles (RNPO) with sizes around 40 nm have been used for oral chemotherapy in the treatment of colon cancer^[20]. In the RNPO, nitroxide radicals are contained within a core nanostructure and scavenge reactive oxygen species (ROS). Orally administered RNPO accumulate in cancer tissues, but less so in normal tissue. No noticeable toxicities were observed in long-term oral administration studies done on mice. The combination of RNPO with conventional therapeutic drug irinotecan improved therapeutic efficacy and suppressed the adverse effects of irinotecan on the GI tract.

Mesoporous silica nanoparticles of approximately 70 nm in size loaded with trityl radical (FMSN-trityl) have been reported for measuring intracellular oxygen distributions in real-time^[21]. Ionic forces between positively charged amine groups inside the channels of the mesoporous silica and the negatively charged carboxylic acids of the trityl radical encourage the radical to stay localized to the nanostructure. Oxygen sensitivity is retained, and the FMSN-trityl performs as well as free trityl in cellular oxygen uptake assays and in-vivo imaging of solid tumors in mice.

A large effort has gone into creation and use of paramagnetic molecules associated with liposomal nanoparticles. Liposome encapsulation of Finland trityls has also been used for in-vivo EPR oximetry. Liposomes prepared with a mean size of 167.5 nm were used to encapsulate Finland trityl^[22]. EPR was used to measure the oxygen sensitivity of the Finland trityl entrapped in the liposome, and the liposome-encapsulated radical was delivered intracellularly in HepG2 and HUVECs cell lines^[22]. The negatively charged nature of free Finland trityl prevents the crossing of the cell membrane and limits oximetry uses to the extracellular region.

Lipid nanoparticles with sizes of about 60 nm loaded with either the nitroxide tempo-benzoate or the trityl were used to protect the radicals from reduction in-vivo. Compared to free tri-aryl-methyl radical (TAM), the half-life of the liposome-encapsulated TAM exceeded one hour^[23]. Calibration of the EPR linewidth showed a two-fold increase in sensitivity to pO₂ compared to free hydrophilic trityl radical in aqueous media. Use of liposomal nanoparticles has also been used to demonstrate radical “self-quenching”^[24], wherein the quenching interaction refers to an extreme of collisional exchange broadening for locally high concentrations of radical result in a near zero-amplitude in the derivative EPR spectrum. When combined with exterior ligands that target the liposomal carrier to specific cell types, the high concentration nitroxide or trityl probes are released into the environment and the maximum amplitude signal results as the local concentration decreases upon delivery^[25].

Carboxylic-acid Derivatized Alumoxane Nanoparticles

TDA pioneered the development of carboxylic acid derivatized alumoxane nanoparticles in the late 1990’s along with researchers at Rice University. The process has been patented and several publications describe characterization^[26], doping with main group and transition metal ions^[27] and use in the formation of ternary aluminum oxides^[28]. Alumoxane nanoparticles can be synthesized from aluminum alkoxide starting materials. The solubility and size of the particles are determined by the functional groups attached through the carboxylic acid linkage. The attachment proceeds through one of two paths. IR data suggests a bridging group between two aluminum atoms, while magic angle ¹⁷Al NMR analysis indicates a single covalent attachment with hydrogen bonding to a nearby hydroxide group. This is also supported by the amount of water displaced upon refluxing the alumoxane particles in a suitable polar solvent. The alumoxane nanoparticles are stable for long periods of time, and their solubility in different solvents is tailored by differential grafting with hydrophilic or hydrophobic substituents on the other end of the carboxylic acid head group.

Our imaging agent, deemed RADI-Sense (Figure 1), is based on a nano-scaffold of aluminum hydroxide (boehmite) with sizes from 100 – 200 nm, an alkyne containing deuterated Finland trityl probe molecule (1). The probe molecule is covalently attached on the outside of the nanoparticle through 1,3-dipolar cycloaddition of the alkyne on the paramagnetic trityl molecule with an azide attached to the surface of the nano-scaffold (2). The final step is encapsulation with a gas permeable, thin (10–20 nm) polymer layer to separate the imaging agent and body environment while still allowing O₂ to interact with the paramagnetic probe (3). For simplicity, the fully constructed imaging agent has been termed “RADI-Sense”. Side by side in-vivo comparison in a mouse model made between RADI-Sense and free hydrophilic paramagnetic probe OX-071 showed oxygen sensitivity is retained and that RADI-Sense can be used to create 3D pO₂ maps of solid tumors. These initial results are a promising first step in evaluating this approach in existing pre-clinical imaging experiments and for potential use in future clinical EPR imaging work.

Figure 1.

Illustration of the imaging agent described in this paper. (1) A specially designed deuterated Finland trityl with an alkyne group. (2) The alkyne group allows covalent attachment via click-chemistry to the surface of an alumoxane nanoparticle functionalized with an azide group. The dFT-NP is encapsulated with a thin polymer film. The fully constructed imaging agent is referred to as RADI-Sense (3).

II. METHODS

a. Synthesis of Click-Capable Deuterated Trityl Radicals (dFT)

Synthesis of deuterated Finland trityl (dFT) was carried out as described in ^{[29, 30]}. Deuterated propargyl tosylate was synthesized as described in ^{[31, 32]}. The creation of “clickable”-dFT was previously described in ^[33]; Briefly, dFT (350 mg) was dissolved in 10 mL DMF, then sodium carbonate (74 mg, 2 eq.) and the deuterated propargyl tosylate (60 mg, 0.8 eq.) were added. The mixture was stirred for 2h at 50°C. 50 mL of HCl 1M were added and the products were extracted with dichloromethane. The organic phase was dried with magnesium sulfate, filtered and evaporated under reduced pressure (10–200 mBar) and dried under vacuum. The crude mixture was dissolved in 2 mL acetonitrile and was purified on C18 column using a Combiflash RF+ System; gradient from 30% methanol and 70% ammonium acetate 0.1M to 85% methanol and 15% ammonium acetate 0.1M. 77 mg of monoester were recovered after evaporation of methanol under reduced pressure follow by freeze-drying (21% yield).

b. Synthesis of Encapsulated Nanoimaging Agent

The full synthetic method has been described in ^[34]. In general, a small molecular weight carboxylic acid and deionized water were added to a 100 mL round-bottom flask, equipped with stir bar and reflux condenser. Aluminum tri-sec-butoxide was added and allowed to hydrolyze at ambient temperature and the resulting slurry was heated to 80 °C. The reaction mixture was aged (~2 hrs) and then heated to 92 °C. A bubbler was added and sec-butanol was removed by a stream of argon. The bubbler was removed and the reaction mixture cooled to 80 °C. The resulting colorless, viscous solution was cooled to ambient temperature. Carboxylic acids used included propionic acid, 3-carboxy-proxl, 2-azidoacetic acid and 6-azidoacetic acid.

dFT was attached to the azido-functionalized nanoparticle surface through 1,3 dipolar Huisgen cycloaddition catalyzed by copper (I). Azide-functionalized aqueous nanoparticles, sodium carbonate and deionized water were combined in a 25 mL round-bottom flask. The alkyne-functionalized dFT, dissolved in dimethyl sulfoxide, was added and the resulting brown solution was sparged for 1 hr with nitrogen. Copper bromide was added and the reaction mixture was purged with argon for 5 minutes. The reaction mixture was aged for 18 hours, then exposed to air. Copper salts were precipitated with the addition of sodium carbonate dissolved in deionized water. The resulting slurry was filtered, and solvents were removed from the filtrate in vacuo.

The solution containing trityl-functionalized nanoparticles was acidified, rendering them hydrophobic enough to be dissolved into chloroform. Pluronic F-127 was dissolved in deionized water in a vial with a stir bar. The aqueous Pluronic F-127 mixture was combined with the trityl-functionalized nanoparticles in chloroform. The resulting biphasic mixture was heated to 80°C for ½ hour with stirring. The chloroform layer evaporated, leaving behind a turbid solution that eventually became a clear golden solution. For fully encapsulated nanoparticle imaging agent aliquots to be used for mouse imaging, the solution salinity was adjusted to 0.9% w/v and pH adjusted to between 7.2 and 7.4. The fully encapsulated dFT-NP used for imaging is referred to as “RADI-Sense”.

c. Nanoparticle size measurements

Nanoparticle size was assessed through three different approaches. Single angle dynamic light scattering was done at TDA using a microtrac nanosizer in Golden, CO. The single angle DLS measurement is based on scatter of incident laser light based on Brownian motion of the particles in the sample solution. For samples with a non-uniform distribution of particle sizes, the influence of a small group of larger particles can skew the median nanoparticle size of the distribution (the D₅₀). Because of this, some particles were also sent out for analysis at Horiba Scientific (Irvine, CA) using their Viewsizer 3000 particle tracking analysis (PTA) instrument. The Viewsizer uses three different laser sources and images each particle individually. This largely removes the bias introduced by a small population of larger nanoparticles. SEM was done on fully encapsulated samples at 25% dFT functionalization at the Colorado School of Mines using a FEI HELIOS NANOLAB 600I SEM prior to sending the fully constructed imaging agent (RADI-Sense) to O2M for mouse imaging.

d. Characterization of Oximetric Properties with EPR

EPR was used extensively to monitor reaction conditions and to demonstrate oxygen sensitivity of the dFT anchored to the nanoparticle surface, and of the fully encapsulated imaging agent. Both continuous wave (CW) and pulse EPR experiments were carried out as needed, and all measurements were done at room temperature (c. 19–21°C) at the University of Denver. Many CW measurements were performed on a Bruker EMXnano benchtop spectrometer. A few CW measurements and all pulse-EPR measurements were performed on the Bruker Elexsys X-band EPR spectrometer. For CW studies, multiple spectra were collected at different powers, modulation amplitude and modulation frequency to prevent experiment-based line-broadening of the dFT lineshape.

For experiments where the O₂ sensitivity was measured, samples were drawn up with a syringe into a 1 mm i.d. Teflon tube and placed into a 4 mm o.d. quartz EPR tube. Purging with N₂ was accomplished by insertion of a thin Teflon tube in parallel with the sample. The N₂ can pass through the wall of the Teflon and deoxygenate the solution. The linewidth decreased to its most narrow value within 30 minutes of N₂ purging.

For pulse-EPR, T_m and T_1e were measured with standard 2-pulse echo or 3-pulse inversion recovery, respectively. The pulse response of narrow trityl spectra is often dominated by free-induction-decay (FID). To partially suppress the FID and better observe a spin-echo, a purposeful magnetic field distortion was created by placing a metal object (bent Allen wrench) between the resonator and magnet pole face. Relaxation time constants were analyzed with locally-written program based on single or double exponential fits, or when a distribution of relaxation time constants was present, modeling with the uniform penalty (UPEN) method^[35–37].

e. Mouse Imaging Studies

A donor mouse was established by grafting fibrosarcoma tissues (a gift from Dr. Howard Halpern, University of Chicago) to the left and right back of 8-week-old female C3H mice. When the grafted tumor grew to 1 g at each side of the back in the donor mouse, the fibrosarcoma tissues were harvested. A single cell suspension was generated by digesting the minced tumor tissues using 0.033% trypsin in RPMI 1640. Fibrosarcoma was induced in mice intended for imaging by injecting 1 × 10⁶ cells in 50 μL RPMI1640 to the right calf of the C3H female mouse.

Eight mice were selected for imaging experiments. Of those eight, six were appropriate for imaging and one died before MRI could be completed. A total of five mice were successfully imaged to compare naked trityl radical delivered by intravenous (IV), and the prototype encapsulated agent delivered by intratumoral (IT) injection. Within these six mice, two different injection schemes were experimented (Figure S1) using injected volumes into the tumor of between 200 and 500 μL (approx. 200 mm³ to 500 mm³). Imaging for mice #1 - #3 were done with a more dilute RADI-Sense formulation, where the concentration of paramagnetic imaging agent was estimated to be ~ 0.6 mM. Image for mice #4 and #5 was done with a higher concentration formulation, achieved by decreasing the volume of the solution by a factor of 2. All imaging solutions used in mice were pH balanced (7.2–7.4) and made saline (0.9% NaCl) prior to injection.

Mouse Imaging Sequence

The tumor (Fibrosarcoma) grew intramuscularly in the leg to no greater than 400 μL (400 mm³), as assessed by ellipsoidal approximation. The mouse was then mounted in the mouse bed, and an MRI measurement was performed. The MRI parameters for each mouse were: FSE (RARE), TE 36 ms, TR 3000 ms, 128×128 matrix size, FOV = 36 mm, slice thickness 1mm and total experiment time of 3 minutes. Immediately after MRI, EPRI measurement was performed using the inversion-recovery-electron-spin-echo (IRESE) sequence using a 25 mT EPRI instrument, JIVA-25^™ (O2M Technologies, LLC). Imaging parameters were $\frac{\pi }{2}=$ 60 ns, tau = 400 ns, delays (410 ns, 612 ns, 912 ns, 1.361 μs, 2.03 μs, 3.029 μs, 4.518 μs, 6.74 μs, 10.055 μs, 15 μs), 16 step phase cycling, 90 shots, TR = 35 μs, acquisition time: 10 min. Reconstruction parameters: Matrix size: 64×64×64, field of view: 4.246 cm.

In the first step of the imaging sequence for mice, EPRI measurements were performed using the hydrophilic free-trityl OX-071 (120 μL of 72 mM) delivered intravenously through the tail vein. After the imaging was performed, a time delay of 60–90 minutes allowed the free-OX-071 to clear the tumor and mouse body through the kidneys. RADI-Sense probe was then delivered to the same tumor by intratumoral injection, using one of two schemes. For each injection site, 60–100 μL of RADI-Sense could be injected. The needle was inserted fully along the length of the tumor. As the needle was withdrawn back along the injection path, RADI-Sense was injected to load the tumor. We observed that only minimal amounts of blood or RADI-Sense agent came back out of the injection site as the needle was being withdrawn. Therefore, we estimate that ~80% of the delivered RADI-Sense agent stayed in the tumor tissue.

For the images acquired with the RADI-Sense agent, the signal-to-noise-ratio (SNR) of the novel encapsulated imaging agent was lower than observed with free trityl OX-071 delivered in the same way. In some cases, the tumor outline provided by the MRI was used to help guide the fitting of EPR amplitudes to generate the pO₂ maps.

III. RESULTS

Characterization of Nanoparticle Size

PTA and SEM allow counting of each particle individually but only sample a very small portion of the entire nanoparticle population. Single angle DLS can address the entire population of nanoparticles but is easily skewed by a smaller number of larger particles. The light scattered by a 50 nm particle will be 10⁶ (one million times) larger than that scattered by a 5 nm particle^[38]. Hence, a danger of all dynamic-light-scattering (DLS) techniques is that light scattering from a much smaller concentration of larger particles will dominate the light scattering of a much larger concentration of small particles that represent the real nature of the nanoparticles in solution.

In general, the single angle DLS values for nanoparticles made with ¹⁴N-3-carboxy-proxyl (3-CP), 2-azido-acetic acid (2-AA), or six-azido-acetic acid (6-AA), functionalized with dFT and encapsulated fell into the range D₅₀ = 90 – 150 nm. A slight, but statistically significant increase was observed comparing 25% labeled dFT nanoparticles before (D₅₀ = 88 nm) and after encapsulation (RADI-Sense@25%, D₅₀ = 114 nm). Analysis with SEM of the 25%-dFT labeled nanoparticles showed two clear populations of nanoparticles centered around 90 nm and 30 nm (Figure S2). We can infer the larger particle in the SEM is an outlier, by nature of the similarity in the particle size results between DLS and PTA. Single-angle DLS measurements are heavily influenced by small percentages of much larger particles as these larger particles scatter the incident light much more efficiently. Thus, if the single larger particle in the SEM were part of a much larger population, we would expect to see a much wider variation in the particle size results between DLS and PTA.

Overall, there is good agreement between the three techniques on the general size of the dFT functionalized nanoparticles. The differences in the nanoparticle size measured by each technique agree within the known limitations of each measurement approach.

3-CP as a Surrogate to Optimize Nanoparticle Synthetic Conditions

The primary structure of the boehmite nanoparticles consists of a double layer of aluminum atoms surrounded by six oxygen atoms. Two layers of the primary boehmite structure are held together through hydrogen bonding. Full (100%) coverage is 8 aluminums per 1 carboxylic acid, and we often achieve this ratio and small nanoparticle sizes with small carboxylic acids such as propionic acid.

Any small molecule with a carboxylic acid and size similar to that of propionic acid can also be incorporated directly onto the surface at approximately the same rate and therefore yield nanoparticles of similar size. The nitroxide 3-carboxy-proxyl (3-CP) is one such molecule. The changes in lineshape due to increased anisotropy in a bound vs. free state have been well characterized in biochemical research through the use of similar nitroxides to 3-CP and the spin-labeling technique for proteins^[39–41]. In a similar way to biomolecular spin-labeling, we can use 3-CP to give us a first look at the motional environment of a radical attached to the nanoparticle surface.

Control of the nanoparticle size in the alumoxane synthesis is governed by the kinetics of the carboxylic acid addition as the alumoxane particle is forming. Incorporation of the carboxylic acid prevents additional growth of the alumoxane structure. For this reason, the synthesis of nano-sized alumoxane structures relies on a great excess concentration, on the order of 40 mM of the carboxylic acid. Since the linewidth of 3-CP is sensitive to changes in concentration via Heisenberg exchange collisional interaction, the EPR spectrum could also provide information on how much free carboxylic acid has been incorporated onto the surface, and how much is left in the reaction mixture.

We created a concentration vs. linewidth calibration for 3-CP in air from 0.25 mM to 40 mM (Figure S3). The upper end was chosen because the 3-CP, and both aziodacetic acid linkers are added at 37 mM in the synthesis reaction. We found that ~40 mM was about the limit of solubility of 3-CP in aqueous solution. The dependence of linewidth on concentration for 3-CP is linear from 10 – 40 mM, with a slope of 76 mG/mM over that range. The linewidth in air is much less dependent on concentration below 10 mM, with measurements between 1.2 and 1.4 G ΔBpp. In a prior study of several nitroxides at room temperature and X-band, the linewidth of 3-CP was insensitive to the presence of oxygen, and hyperfine coupling constants could only be simulated for the radical at low concentrations in toluene. This was attributed to asymmetry in the pyrrole ring, and slow ring dynamics which result in four inequivalent methyl groups around the N-O fragment and a broad, unresolved spectrum^[42]. In that study, the T_2e determined spin packet linewidths indicated a concentration dependence of 80 mG/mM for deoxygenated samples. However, the concentration dependence of the CW-lineshape will not be observed until the concentration is high enough to clearly exceed the linewidth (~1.2 ΔBpp) determined by the hyperfine pattern of the four inequivalent methyl groups.

The EPR of the crude reaction mixture after nanoparticle synthesis is very broad (Figure 2, spectrum was recorded in air). We expect the most dominant, narrow lines to belong to remaining free 3-CP. From our calibration curve (Figure S3), the starting concentration of ~ 37 mM of free 3-CP has a linewidth of ~3.5 G ΔBpp in air. Once the reaction has been run to completion, the linewidth of the remaining free 3-CP in the reaction mixture is about ΔBpp = 2.5 G, indicating a remaining concentration of ~23 mM free 3-CP. This implies ~14 mM of 3-CP has been incorporated onto the nanoparticle surface. At the time of this work, we did not have access to a routinely available particle sizing technique which could also give particle concentration, but in future studies the change in linewidth of 3-CP before and after synthesis could be combined with particle concentration to estimate number of 3-CP molecules per nanoparticle.

Figure 2.

Alumoxane nanoparticles functionalized with 3-CP characterized by EPR. Spectrum was recorded in air. The crude reaction mixture linewidth post-synthesis is dominated by an excess free 3-CP. () Derivative data, () Single integration, () Double integration

In order to extract motional information for the 3-CP associated with the nanoparticle surface, we need to remove some of the free 3-CP to see the contribution more clearly from the slower tumbling, surface bound 3-CP. This was accomplished by passing the crude reaction mixture through an Amberlite ion-exchange resin. The result EPR spectrum is shown in (Figure 3). One very broad feature in the spectrum comes from 3-CP attached through the COO⁻ to the alumoxane, while the second feature is three narrower lines which come from free CP in solution which has not been attached to the alumoxane. In the case of free 3-CP, the rapid tumbling averages anisotropic components of the spectrum and produces three narrow lines (nearly isotropic spectrum). The large broad feature is a produced when the 3-CP is rigidly bound to alumoxane scaffold. In the bound state, motional rotation is expected to be much slower (on the scale of nanoseconds), leading to less averaging of the anisotropic hyperfine component, and resulting in a very broad spectrum.

Figure 3.

Alumoxane nanoparticles functionalized with 3-CP characterized by EPR, after passing the reaction mixture through an anion exchange media. () Derivative data, () Single integration, () Double integration

Using MATLAB, the EasySpin package^[43], and a graphical user interface designed for spin-label simulation (SimLabel^[44]), we simulated the spectrum with contributions from the free and bound species of 3-CP and compared to the experimental spectrum (Figure 4). The simulation indicated 84% of the EPR signal in the cleaned reaction mixture is coming from slow tumbling 3-CP. The remaining 16% is due to residual free 3-CP in solution. The tumbling correlation time for the free 3-CP in solution is ~80 ps, while that for the slower tumbling species, which we have assigned to bound 3-CP, is ~3.6 ns.

Figure 4.

Simulation of the spectrum from Figure 3 - 3-CP functionalized nanoparticle after running the crude reaction mixture through an anion exchange column to remove remaining free 3-CP. Simulation was done with EasySpin add on for Matlab. () Derivative data, () Bound 3-CP, () Free 3-CP, () fit with both species present.

Attachment of dFT through Click-Chemistry

For the covalent attachment of alkyne-dFT to the surface of the alumoxane nanoparticle, the reaction is run with an azidoacetic acid instead of propionic acid or 3-CP. The presence of the azido group on the surface of the nanoparticle can be monitored directly via the appearance of peak in the infrared (IR) spectrum near 2100 cm⁻¹(Figure 5). Based on the observations using 3-CP as a reporter molecule, crude reaction mixtures were filtered through an anion exchange column to remove excess unbound azidoacetic acid after the reaction was completed prior to attachment of the dFT to the nanoparticle surface.

Figure 5.

The infrared spectrum (IR) is useful for following the successful attachment of the azido group to the surface of the alumoxane nanoparticle. Nanoparticle prepared with only propionic acid () does not show the presence of the azide stretch near 2100 cm⁻¹, while the nanoparticle synthesized with 2-azidoacetic acid does (). For comparison, the neat spectra of 2-azidoacetic acid is also shown ().

The azido peak near 2100 cm⁻¹ is also used to monitor completion of the click chemistry reaction after the azido group has been combined with the alkyne group to create the triazole bond. This approach was demonstrated previously when the dFT was covalently linked to azido-functionalized PEG^[33]. IR spectra were collected for dFT-alkyne alone, the nanoparticle functionalized with an azide group, and the nanoparticle functionalized with dFT after the click-chemistry reaction was complete (Figure 6). The peak at 2100 cm⁻¹ is eliminated after the click-chemistry reaction with dFT has been completed. For the azido-NP, the region from 1600 cm⁻¹ to 550 cm⁻¹ has only very few absorbances, while for free dFT this region has multiple peaks due to the aromatic rings in the molecule. After the click reaction is completed, the dFT-NP shows many more peaks in the region of 1600 cm⁻¹ to 550 cm⁻¹ than the nanoparticle without any dFT attached.

Figure 6.

Following the click-chemistry reaction between azido-functionalized nanoparticle surface and dFT-alkyne through changes in the IR spectrum.

To further confirm the attachment of the dFT to the alumoxane scaffold, electron relaxation times were measured for the samples in both air and in the absence of oxygen (N₂ purge). For a dFT anchored to the nanoparticle surface, we expect slowing rotational motion compared to free dFT in solution. The dominant relaxation mechanisms (which determine the electron relaxation times) for trityl radicals are spin rotation, and modulation of the dipolar interaction between the radical and solvent molecules^[45–47]. Attaching the dFT to the alumoxane both slows down the radical and decreases the interaction with solvent molecules compared to a freely tumbling molecule in solution. Thus, we would expect longer values of T_1e for a bound dFT molecule compared to un-bound. Measurements of T_1e with pulse EPR show this to be the case, with the T_1e of the presumed bound dFT (15 μs) longer than both the free dFT in PBS (10.2 μs) and the T_1e of dFT in the nanoparticle reaction mixture but unbound (T_1e = 7 μs). Thus, the analysis with EPR and IR both indicate we have successfully attached the dFT to the alumoxane scaffold.

Encapsulation of dFT-functionalized nanoparticles

The encapsulation procedure requires transformation of the dFT-functionalized nanoparticle from an aqueous phase during attachment to the nanoparticle to an organic phase in preparation for the encapsulation. This is because the encapsulation is driven by hydrophobic forces bringing a hydrophobic core molecule together with the hydrophobic end of a polymer with both hydrophobic and hydrophilic segments. This is facilitated by acidification of the click-chemistry reaction mixture to ensure the carboxylic acid groups on the dFT are in a protonated state. Electron relaxation times were measured for aliquots of the chloroform fraction, and the encapsulation product (Table 1).

Table 1.

Electron Relaxation Times of dFT throughout the synthesis of the encapsulated imaging agent. All times are in microseconds. Recorded at X-band (9.5 GHz).

	Tm		T1
	N2	Air	N2	Air
Click-Chemistry (Cleaned)	3.74	1.05	15.8	1.39
Organic Phase	1.3	0.07	6.1	0.06
Organic Phase (10x dilution)	6.5		9.7
Encapsulated	7.9	0.83	15.1	0.99
Free alkyne-dFT in PBS	2.94		10.2
Free alkyne-dFT in reaction mixture	3.35		6.96

The extremely low values of T_1e in the chloroform fraction indicate a high concentration of dFT in close proximity undergoing a high frequency of collisional exchange. Upon extraction of the labeled nanoparticle into the aqueous phase, the T_1e returns to that of the dFT attached to the nanoparticle. To further test the effect of a locally high concentration of dFT-NP in the organic phase, a 10x dilution in chloroform was made, with the T_1e and T_2e increasing substantially. The oxygen sensitivity of the encapsulated, dFT-functionalized nanoparticle is very good. T_1e decreases from15 μs in N₂ atmosphere to 7 μs in air-saturated solution (atmospheric pressure in Denver is ca. 12 psi or 0.85 kg/cm²).

The T_2e is longest for the encapsulated, dFT-functionalized sample. The spin-spin relaxation time, T_2e, is most sensitive to collisional interactions which result in a shorter electron relaxation time. Collisions can be decreased either by a barrier (i.e. encapsulation), or by a decrease in concentration of dFT. While the dFT is hydrophobic, the surface of the alumoxane nanoparticle is very hydrophilic, so much so that concentrations of dFT-NP can be prepared up to ~7 mM (100% labeling of all available sites) in aqueous solution, while alkyne-dFT alone can only be prepared up to ~200 μM in aqueous solution. To facilitate transfer of dFT-NP into the organic phase in preparation for encapsulation, the solution is made extremely acidic (pH~2) in order to protonate all hydroxide groups on the surface of the alumoxane. Unfortunately, the low pH partially degrades the dFT signal. Thus, it is plausible one part of the increase in the T_2e from 3.74 μs to 7.9 μs is a lower overall dFT concentration, comparing the dFT covalently attached to the alumoxane nanoparticle before and after encapsulation. Optimization of the encapsulation procedure is an ongoing endeavor.

The amount of 25% dFT-NP which was encapsulated was further probed by measuring the effect of bovine-serum-albumin (BSA) on the linewidth of dFT. Free dFT interacts strongly with bovine-serum-albumin (BSA)^[48]. The interaction stems from hydrophobic interactions between dFT and the sub-domain IIA (Sudlow’s site I) of BSA, and the interaction broadens the dFT linewidth. The encapsulation should inhibit this interaction. In the first test, a sample of free dFT and BSA was made up with corresponding concentrations of 100 μM and 50 μM, respectively. A sample of free dFT alone with no BSA, also at 100 μM was made up as a reference. EPR spectra for both samples were recorded in air and the linewidths were compared. For free dFT in PBS, the in-air linewidth is 140 mG, while in the presence of BSA the linewidth is severely broadened and is ca. 350 mG. This corresponds to 150% line broadening due to the interaction with the BSA (Figure, 7A).

Figure 7.

Encapsulation of dFT + nanoparticle partially blocks the line-broadening interaction between free dFT and BSA. (A) The comparison of 100 μM dFT alone (black) or 100 μM dFT in the presence of 50 μM BSA (blue). (B) The comparison of 77 μM encapsulated dFT + nanoparticle (RADIsense) in solution alone (black) or in the presence of 50 μM BSA (blue). The line broadening interaction between dFT and BSA is reduced in the presence of encapsulation.

The same test was done for dFT on the alumoxane nanoparticle and after encapsulation. The concentration of encapsulated dFT + nanoparticle was ~77 μM in the presence or absence of 50 μM BSA. In the case of anchored, encapsulated dFT the linewidth is 200 mG in the absence of BSA, and broadens to 265 mG when BSA is present. This corresponds to only a ~33% line broadening in the presence of BSA (Figure 7B). The encapsulation is able to block the interaction between dFT and BSA, at least partially. Thus, some of the increase in T_2e may also be due to a majority of the nanoparticles being encapsulated, and that encapsulation barrier preventing the collisional interactions between dFT which would lengthen T_2e relative to the un-encapsulated state.

Finally, in preparation for mouse-imaging studies, the encapsulated dFT-NP imaging agent was brought up to saline with a pH = 7.2. Measurements of electron relaxation times were made at 23°C (room temperature) and 37°C, and before and after autoclaving (20 minutes at 120°C at 17 psi). These measurements are shown in (Table 2). Both T_2e and T_1e increase slightly after autoclave. The T_1e is still indicative of dFT attached to the nanoparticle, and CW-EPR spectra post-autoclave did not show any change that would indicate the presence of free dFT which had come off the nanoparticle. Both T_1e and T_2e decrease slightly from 23°C to 37°C, which is in line with slightly faster relaxation rates at faster tumbling correlation times at the higher temperature.

Table 2.

Relaxation Times of dFT in RADI-Sense at X-band (9.5 GHz)

Condition	T2e (μs)	T1e (μs)
RADI-Sense, saline, pH 7.2, 23 C	5.5	14.5
RADI-Sense, saline, pH 7.2, 23 C, post-autoclave	6.4	16.4
RADI-Sense, saline, pH 7.2, 37 C	4.4	12.0

Oxygen Imaging with Encapsulated dFT-Nanoparticle in Mouse FSa tumors

Electron paramagnetic relaxation times were recorded against oxygen concentrations of 0%, 6%, 9%, 12% and 21% to create the calibration for converting RADI-Sense EPR amplitude into pO₂. The RADI-Sense slopes for both T_1e and T_2e are shown in (Table 3), as are the slopes for free molecular OX-071. For both cases the calibration was done at 37°C in phosphate-buffered-saline (PBS). The slope and intercept data for free OX-071 is an average of pO₂ calibrations at OX-071 concentrations of 0.5 and 1 mM. For RADI-Sense, the effect of concentration is decreased because Heisenberg exchange would have to take place between dFT on different nanoparticles. Three different pO₂ calibrations were done for RADI-Sense to generate the slope and intercept values in (Table 3).

Table 3.

Calibration Statistics for oxygen response for RADI-Sense encapsulated imaging agent compared to control free molecule OX-071.

	T1e Slope (torr O2/MHz)	Intercept (MHz)	T2e Slope (torr O2/MHz)	Intercept (MHz)
OX-071	111.4 ± 2.5	0.13 ± 0.003	89.7 ± 5.4	0.18 ± 0.03
RADI-Sense	118.4 ± 14.3	0.20 ± 0.02	108.3 ± 1.5	0.18 ± 0.01

Within the uncertainty in the measurement, the response of T_1e to oxygen is the same for free molecular OX-071 and the trityl anchored to the nanoparticle and encapsulated as the RADI-Sense agent. The response of T_2e to oxygen is about 20% higher for the RADI-sense agent compared to free molecular OX-071. This is an improvement over the current state of the art (OX-071). Improved O₂ response could be derived from a decreased contribution of inter-molecular broadening in the RADI-Sense agent, driven either by a lower concentration of the trityl in general, or better separation between individual trityl molecules anchored onto the nanoparticle.

For the images acquired with the RADI-Sense agent, the signal-to-noise-ratio (SNR) of the novel encapsulated imaging agent was lower than what is observed with free trityl OX-071 delivered in the same way at 25 mT on JIVA-25^™ (Figure S4). The imaging was done on the same mouse tumor, first by intravenous injection with OX-071, followed by intra-tumoral (IT) injection of RADI-Sense. A delay time between imaging with OX-071 and RADI-Sense of between 60–90 minutes was allowed, during which the tumor was imaged until all OX-071 had been cleared. Since the imaging was done on the same tumor volume, one measure of the SNR is the number of voxels with sufficient signal amplitude to yield information on pO₂ (Table 4 and Table 5). There is better agreement between images acquired with OX-071 vs. RADI-Sense for Mouse #4 and #5 because the RADI-Sense solution was concentrated to increase the signal amplitude before injection.

Table 4.

pO₂ Statistics Generated from EPRI Images using OX-071

Animal #	Mean (torr)	Std. Err. Of Mean (Torr/(Voxel)1/2)	Med. (torr)	Tumor pO2 Hetero. (torr)	Std (torr)	# of Voxels
1	30.19	0.53	30.45	22.96	11.48	472
2	9.89	0.32	6.63	23.3	11.65	1340
3	12.61	0.28	9.21	24.82	12.41	1991
4	17.92	0.38	17.17	21.1	10.55	776
5	15.98	1.03	11.78	31.74	15.87	236

Table 5.

pO₂ Statistics Generated from EPRI Images using RADI-Sense

Animal #	Mean (torr)	Std. Err. Of Mean (Torr/(Voxel)1/2)	Median (torr)	Tumor pO2 Hetero. (torr)	Std (torr)	# of Voxels
1	15.01	1.02	10.57	32.36	16.18	250
2	6.98	0.38	3.43	23.62	11.81	947
3	6.00	0.27	2.97	21.38	10.69	1591
4	15.63	0.61	11.15	32.24	16.12	692
5	13.30	0.94	8.24	29.76	14.88	253

In some cases, the tumor outline provided by the MRI was used to help guide the fitting of EPR amplitudes to generate the pO₂ maps (Figure 8). Despite the SNR of the RADI-Sense agent being lower than that of the free molecular OX-071 delivered intravenously, similar maps of pO₂ are produced. Statistics for the five mice which completed the three-imaging-set are shown for OX-071 (Table 4) and RADI-Sense (Table 5). The number of voxels which are used in the calculation of median pO₂ (in torr) are much lower in Mouse #1 and #2, because the SNR of the imaging agent as sent was lower. For Mouse #3, #4 and #5, a higher volume of the low concentration RADI-Sense agent (Mouse #3 at 500 uL), or the increased concentration (by decreasing solvent volume) RADI-Sense agent (Mouse #4 and #5) were used.

Figure 8.

Comparison of pO₂ images acquired with OX-071 delivered intravenously, or RADI-Sense delivered intratumorally for Mouse #5. (A) Registration with MRI. The tumor area is outlined in yellow (B) pO₂ map of tumor using OX-071 (C) pO₂ map of tumor using RADI Sense.

No quantitative clearance studies were undertaken in this initial work. Two qualitative observations related to retention were made. For OX-071 the excreted urine of the mouse takes on a slight rust color attributed to an oxidation product of OX-071 cleared from the tumor through the kidneys. No such color was observed in the urine of the mice after imaging with RADI-Sense. This requires further study, as it could be due to protection of the dFT from oxidation by the encapsulation, or indicate the RADI-Sense is not renally cleared.

A second observation regarded diffusion through the tumor and surrounding muscle after injection. When free trityl OX-071 is injected intratumorally, the free trityl spreads into tumor and muscle tissue, and quickly washes out of the muscle tissue, remaining only in the tumor. The signal from the RADI-Sense agent decayed gradually within the tumor, rather than a spread into the surrounding muscle tissue. The signal strength of the RADI-Sense agent which in the tumor remained stable for 45–60 minutes. This is a significant advantage for RADI-Sense relative to other trityl or nitroxide based imaging, because it could better facilitate time-dependent studies.

IV. DISCUSSION

3-CP as a Reporter for Motional Time Scales on the Nanoparticle Surface

In aqueous solution at low concentration (0.25 mM), and in the absence of oxygen, the 3-CP has a τ_C = 19 ps^[49]. This was the longest τ_C in the set of nitroxides studied, which included ¹⁴N/¹⁵N pairs of 3-CP, deuterated-Tempone (τ_C = 9 ps), and deuterated mHCTPO (τ_C = 13 ps). The difference in tumbling correlation time of these three small molecule nitroxides was ascribed to increasing solute-solvent interaction ketone<amine<acid^[42]. Specifically, the hydroxide group on 3-CP interacts through hydrogen bonding with the surrounding water molecules, and this “sticking” interaction slows tumbling of the radical slightly compared to nitroxides without the hydroxide group.

The simulated value of 3-CP in the nanoparticle reaction mixture is τ_C = 80 ps, 4x that reported at low concentration in aqueous solution. The linewidth associated with the free 3-CP in the simulated spectrum is ~1.4 G ΔBpp (in air), putting the concentration of free 3-CP in the range of 5–10 mM based on our calibration curve. This concentration is high enough for collisional broadening between 3-CP molecules to increase the linewidth beyond the spin-packet derived linewidth, which could account for a fitted value of τ_C = 80 ps. In addition, the sample was collected in air, rather than purging with N₂ to remove the line broadening effects of O₂. The solvent environment from the 3-CP point of view contains not only water, but a large concentration of alumoxane nanoparticles with surfaces covered in hydroxide groups which could be another interaction to slow down free tumbling 3-CP on the scale of tens of picoseconds. Thus, given the uncertainty in the measurement from a high concentration, in-air measurement, and the very different solvent environment of the reaction mixture compared to aqueous, the fit value of a τ_C = 80 ps seems reasonable.

Natural abundance dFT does not have enough hyperfine anisotropy to calculate a tumbling correlation time based on lineshape change. Very recently, the first ¹³C-isotope enriched trityl has been reported with large enough anisotropic hyperfine that tumbling correlation times can be calculated, as is done for nitroxide radicals^[50]. In the future, if an alkyne containing version of the ¹³C-enriched dFT can be synthesized, a direct comparison can be made between the tumbling correlation time of 3-CP and the tumbling correlation time of dFT on the surface of the alumoxane nanoparticle. In the interim, we will use the tumbling correlation time of the 3-CP on the nanoparticle surface of ~3.6 ns to guide interpretation of changes in relaxation time for dFT after it has been covalently linked to the nanoparticle.

Electron Relaxation Times of dFT in Solution or Attached to Nanoparticles

Study of T_1e for protonated-Finland Trityl in water (FT-CH₃), and the deuterated analog (FT-CD₃) and water:glycerol solutions showed little dependence on viscosity up to 80% glycerol content. Inspection of Figure 2 in^[46] show the viscosity for glycerol solutions up to 80% corresponds to a region where FT-CH₃ and FT-CD₃ are less sensitive to changes in the tumbling correlation time. However, once this tumbling correlation time region is exceeded, the value of T_1e increases with tumbling correlation time in the range of 1 – 6 ns.

The value of ${\tau }_C$ = 3.6 ns obtained from fitting the CW spectrum of attached 3-CP is a lower limit on the ${\tau }_C$ which would be observed for dFT. The dFT is a bulkier molecule than the proxyl, however it has greater motional freedom as it is held to the surface of the nanoparticle with a longer carbon chain that is six carbons in the case of 6-AA or two carbons in the case of 2-AA instead of directly attached in the case of proxyl. To what extent increased size and greater motional flexibility cancel each other out is unknown, but it seems plausible that the dFT is in the 1–6 ns range where T_1e increases with tumbling correlation time.

One paper measured the relaxation times of immobilized FT-CH₃ on either Nucleosil or Trehalose^[51]. Measured values of T_1e of 22 and 25 μs, respectively, were 1.3–1.5x longer than for FT-CH₃ in solution. This is in line with lengthening T_1e for an immobilized trityl vs. one in solution, and also supports the idea that while FT-CH₃ and FT-CD₃ have a range of tumbling correlation times over which their T_1e is not dependent on tumbling, once this is exceeded the T_1e becomes linearly dependent on increased tumbling. This makes sense from the vantage point of the combination of relaxation mechanisms as a function of tumbling correlation times. At sufficiently long tumbling correlation times, the dominant relaxation mechanism for trityls is spin rotation. In spin rotation, modulation of the g-anisotropy by tumbling means the effect of this mechanism decreases as ${\tau }_C$ increases (i.e. T_1e gets longer at longer ${\tau }_C$). Trityl radicals have much less g-anisotropy, compared to nitroxide radicals, so it seems plausible that much longer tumbling correlation times are required to see the effect of modulation of much smaller amplitude g-anisotropy in the spin rotation relaxation mechanism.

For the deuterated Finland trityl (FT-CD₃), modulation of a dipolar interaction with solvent molecules has also been shown to effect T_1e. Natural isotope abundance (¹²C) FT-CD₃ showed a similar independence as FT-CH₃ over the range of glycerol water solutions up to 80% glycerol^[46]. An isotopically enriched ¹³C at the central carbon, ¹³C₁-FT-CD₃ was studied in PBS and PBS: glycerol mixtures. Tumbling correlation times were calculated from simulation of the spectra, which was facilitated by the very large ¹³C hyperfine couplings in the enriched sample. For tumbling correlation times from 290– 3400 ns, the measured T₁ increased from 5.9 to 12 μs^[50].

The variation in T_2e is much wider. The T_2e and T_m are more sensitive to changes in radical concentration, the solvent environment than T_1e, which is one reason why T_1e-driven EPRI has been championed by the Halpern group at the University of Chicago^[15]. The solvent environment changes drastically between each step, even to the encapsulation stage where, hypothetically, the trityl is being shielded from the solvent by the encapsulating Pluronic polymers. More investigation is required into changes in T_2e for the dFT attached to the alumoxane nanoparticle as a function of synthetic step. Once fully developed, T_2e measurements should be able to yield insight specifically into the success of the encapsulation step.

Comparison of pO₂ Statistics of OX-071 and RADI-Sense

Mouse #4 and Mouse #5 show that SNR was improved to the point where the number of voxels used to compute pO₂ for free molecular OX-071 and RADI-Sense agent were nearly the same. The median pO₂ values of the RADI-Sense agent for Mouse #4 and #5 are less than the median pO₂ values for OX-071 by about 30%. Some of this difference is reflective of the difference between administering EPRI imaging agent via IV or IT. Even when using free-molecular trityl OX-071 in both cases, median pO₂ values are slightly lower for IT delivery compared to IV. Our current hypothesis is that this reflects IV-OX-071 reporting on pO₂ in tumor vasculature plus tumor tissue, while IT-OX-071 reflects primarily the pO₂ within the tumor tissue. Another explanation could be that the free OX-071 is exposed to different oxygen concentrations due to different solubility in tissues, while RADI-Sense are always present inside the encapsulation envelope and may not be influenced by the heterogeneity of oxygen solubility in tissues as readily since O₂ must pass through the encapsulation layer. This suggests head-to-head tests between free OX-071 and RADI-Sense agents where both are delivered via IT injection would be a valuable part of future comparisons.

The RADI-Sense agent used for initial imaging studies had dFT covalently attached to the alumoxane nanoparticle at only 25% of the total available labeling sites. While initial imaging studies were ongoing at O2M, efforts to improve the labeling efficiency continued at TDA. This 25% labeling limitation was imposed by the solubility of 6-hexanoic acid in solutions of pH<6.5 which is only 249 mg/mL. 6-hexanoic acid was initially selected under the hypothesis that the azide on a short “tether” extending from the surface of the nanoparticle would be more amenable to successful click-chemistry with the bulky alkyne-dFT. We revisited 2-azidoacetic acid which has an aqueous solubility of 6,570 mg/mL at pH<6.5. We found that 2-AA was easily dissolved at high enough concentration to permit 100% labeling with dFT during the nanoparticle synthesis.

Experimentally determined relaxation widths and linewidths are shown in (Table 6) for 25% and 100% labeling of alumoxane nanoparticles with dFT using 2-AA. These measurements were made for unencapsulated dFT-functionalized nanoparticles. Decreases in in T_2e from 25% to 100% labeling reflect a higher probability for collisional exchange when the maximum number of sites on each nanoparticle is occupied with dFT. Due to the hydrophilic nature of the alumoxane nanoparticle and the covalent attachment, much higher concentrations dFT can be dragged into aqueous solution than could be dissolved when using dFT alone. In our studies, the maximum aqueous concentration which could be achieved for alkyne-dFT alone was ~ 200 μm.

Table 6.

EPR Characteristics of Alumoxane Nanoparticles with dFT covalently attached through click-chemistry. CW-EPR linewidths recorded at X-band (9.5 GHz).

	[dFT]	LW (air)	LW (N2)	T1e (air)	T1e (N2)	T2e (air)	T2e (N2)
25% Label	0.77 mM	77 mG	28 mG	1.33 μs	16.1 μs	1.04 μs	3.90 μs
100% Label	3.08 mM	116 mG	77 mG		15.9 μs		1.55 μs

Towards Clinical EPR Imaging

Currently the paramagnetic probe most far along in the journey to clinical use is the OxyChip for making subcutaneous oxygen measurements^[52–54]. OxyChip is a particulate EPR probe (a Lithium Naphthalocyanine derivative, LiNC-BuO) crystal embedded in polydimethylsiloxane (PDMS). Free LiNC-BuO has limited biocompatibility, problems with biodegradation and migration of individual crystals within tissue^[53]. The PDMS encapsulation offers an oxygen permeable, biocompatible barrier which separates the body and imaging agent environment, while still allowing the EPR probe to report on oxygen concentration. OxyChip has finished Phase I clinical trials and is a proof of concept that separating the paramagnetic agent from the body environment with an oxygen-permeable membrane is a viable route toward clinical use.

We have shown that a similar body/probe separation can be accomplished at the nano-scale. The success of the mm-sized Oxychip in Phase I clinical trials, and several MRI-based gadolinium contrast agents which are FDA approved and in use are encouraging. While other nanoparticle structures have been FDA approved, the alumoxane architecture offers fine control over nanoparticle size down to about 20 nm and a variety of morphologies which can accessed through careful control of pH and temperature. In addition, through judicious selection of ligands incorporated on the surface during synthesis, there are a wide variety of radicals of both nitroxide and trityl types which can be covalently attached. Future efforts must also include toxicological characterization of paramagnetic probe functionalized alumoxane and polymer encapsulated alumoxane nanoparticles.

V. SUMMARY

Using the probes with the longest possible relaxation times is a major enabler of higher SNR in imaging, and faster imaging times, possibly even generation of “real-time” 3D pO₂ maps. The dFT trityl has the longest relaxation times of any trityls, as the functional added groups which make a trityl more water soluble (i.e., OX-063, OX-071) also decrease its relaxation time T_1e. Because the naked dFT radical binds hydrophobic regions in proteins, it has never before been considered for use in clinical imaging. However, with RADI-Sense, we secured the dFT molecule into a biocompatible form. There are additional benefits from covalently attaching the dFT to the nanoparticle. First, the tumbling correlation time of the radical is slowed, further increasing the T_1e which makes the radical more sensitive to changes in T_1e such as collisions with O₂. Second, each dFT on the nanoparticle is separated from its next nearest neighbor, and encapsulated nanoparticles prevent interaction between dFT molecules on different nanoparticles. Both of these eliminated the Heisenberg exchange interaction which decrease the usable relaxation time of the dFT. The Heisenberg exchange reaction (the effect of trityl concentration) is also a confounding variable to the interpretation of 3D pO₂ imaging maps made with free paramagnetic probes. Thus, in addition to providing increased safety and biocompatibility, the RADI-Sense agent also creates an EPR imaging probe which should have a higher sensitivity to oxygen concentration compared to free trityl.