Development of Dissolution Dynamic Nuclear Polarization of [¹⁵N₃]Metronidazole: a Clinically Approved Antibiotic

Abstract

We report dissolution Dynamic Nuclear Polarization (d-DNP) of [¹⁵N₃]metronidazole ([¹⁵N₃]MNZ) for the first time. Metronidazole is a clinically approved antibiotic, which can be potentially employed as a hypoxia-sensing molecular probe using ¹⁵N hyperpolarized (HP) nucleus. DNP process is very efficient for [¹⁵N₃]MNZ with an exponential build-up constant of 13.8 min using trityl radical. After dissolution and sample transfer to a nearby 4.7 T Magnetic Resonance Imaging scanner, HP [¹⁵N₃]MNZ lasted remarkably long with T₁ values up to 343 s and ¹⁵N polarizations up to 6.4%. A time series of HP [¹⁵N₃]MNZ images was acquired in-vitro using a steady state free precession sequence on the ¹⁵NO₂ peak. The signal lasted over 13 min with notably long T₂ of 20.5 s. HP [¹⁵N₃]MNZ was injected in the tail vein of a healthy rat, and dynamic spectroscopy was performed over the rat brain. The in-vivo HP ¹⁵N signals persisted over 70 s, demonstrating an unprecedented opportunity for in-vivo studies.

Graphical Abstract

[¹⁵N₃]Metronidazole (MNZ) is an emerging hyperpolarized contrast agent. The Dynamic Nuclear Polarization process of this clinically used antibiotic is very fast compared to many other ¹⁵N-labeled molecules and has remarkably long T₁ and T₂. The hyperpolarized signal of MNZ lasted for over a minute in a rat brain. Future work on hyperpolarized MNZ will investigate its potential as an antibiotic or hypoxia-sensing probe in disease models.

Introduction:

Hyperpolarization of nuclear spins increases nuclear spin polarization up to 5 orders of magnitude. Hyperpolarized (HP) molecules employed as exogenous magnetic resonance (MR) contrast agents are non-radioactive and can be used to non-invasively study in-vivo biophysiology, such as metabolism,^[1–15] perfusion,^[6–9] or pH^[10–12] in real time. In recent years, promising in-vivo HP imaging applications have led to increasing interest in the search for new HP molecular probes for magnetic resonance imaging (MRI)^[1,6,13,14] and magnetic resonance spectroscopy (MRS).^[6,15–17] Among the multiple hyperpolarization techniques^[18–22] developed to date, parahydrogen-based hyperpolarization processes, such as Signal Amplification by Reversible Exchange (SABRE)^[20–23], and dissolution dynamic nuclear polarization (d-DNP)^[24–28] have demonstrated excellent signal enhancements of many metabolites and biocompatible molecules supporting a wide range of studies in biological systems, most notably in cancer^[14–29].

The development of HP contrast agents relies mainly on the use of isotopically labeled compounds, where the type of isotope labels (e.g., ¹³C, ¹⁵N) and site of labeling in the molecular structure are specifically chosen for the desired application^[28]. In MR experiments, the use of nuclei with low gyromagnetic ratios, such as ¹³C or ¹⁵N, benefits from wide chemical-shift dispersion of spectral resonances, making it easy to distinguish peaks of different metabolites and quantify results. Moreover, unlike protons, ¹³C and ¹⁵N nuclei have virtually no background signal. In addition, ¹³C and ¹⁵N labeled molecules often possess long longitudinal relaxation times (T₁), which allows them to retain HP state for a long time, therefore, making these nuclei great candidates for utility in HP MRI contrast agents^{[5,14,30–34]}. However, the characteristically low MR ¹⁵N sensitivity, $\sim$1/6 of the ¹³C MR sensitivity, limits ¹⁵N in-vivo applications^[16]. ¹³C labeled compounds have, therefore, been the primary focus on the development of HP contrast agents. To date, many HP ¹³C-labeled molecules have been developed ^[11,35–37] and demonstrated utility in a wide range of in-vivo imaging applications in animal and clinical research.^[3,10,11,36] [1-¹³C]pyruvate in particular has been the key molecule for in-vivo HP ¹³C MR studies because of its central role in cellular energy metabolism^[3,4,38] and also importantly, its long ¹³C carboxyl T₁ ($\sim$68 s in solution and $\sim$ 30 s in-vivo at 3 T).

Despite the low MR sensitivity, some ¹⁵N compounds have the merit of very long T₁ relaxation times in specific ¹⁵N sites, reaching up to 20 min T₁.^[39–41] Despite the substantial potential of ¹⁵N HP molecular probes for in-vivo applications^[16], there have been less than a handful of demonstrated in-vivo applications^[16] including choline,^[42] urea^[34] and carnitine.^[43] Choline and carnitine are tertiary amines with symmetric structures rendering high symmetry of ¹⁵N site and resulting in long T₁. The nitro- moieties present in compounds that are used as, antibiotics,^[44] hypoxia sensors^[45, or cancer radio-sensitizing agents^[46] are also characterized by long T₁ reaching up to 20 min in liquid state.^[47]

There is a large family of antibiotics called the nitroimidazoles known for their actions on several anaerobic bacterial infections.^[44,48] Furthermore, nitroimidazoles can bind to different cellular macromolecules, particularly in hypoxic conditions, allowing them to be used as exogenous hypoxia markers.^[45,46,48] In cancer, most tumors are characterized by a hypoxic micro-environment in which the oxygen concentrations and pH are lower than for normal tissue. The hypoxic status has been demonstrated in various tumors to be an indicator of unfavorable prognosis, which is related to cancer progression with an increased probability of metastasis.^[45,48] Nitroimidazoles’ ability to probe hypoxic conditions has therefore been utilized for cancer screening using positron emission tomography (PET)^[48] and both MRS and MRI.^[45,46] HP nitro- moieties have not been demonstrated in-vivo prior to this report, although several nitroimidazoles have been HP by a method called SABRE in SHield Enables Alignment Transfer to Heteronuclei (SABRE-SHEATH)^[50–52].

Metronidazole (MNZ) is a member of the nitroimidazoles group and is also an antibiotic approved by the United States Food and Drug Administration (FDA) to treat a wide variety of infections. Large dosage is allowed for intravenous injection – a single dose of MNZ injection (known as Flagyl Injection) containing 500 mg metronidazole^[53]. In addition, MNZ’s ¹⁵N sites have shown to have long T₁ relaxation times (several minutes),^[47]. Such a long T₁ greatly complements the biological compatibility of this contrast agent. Furthermore, the hypoxia sensing potential and antibiotic functionality of MNZ makes it a very interesting candidate to be used as a HP contrast agent to probe hypoxia, to test drug delivery or to perform cancer screening. Previously, we have reported HP [¹⁵N₃]metronidazole ([¹⁵N₃]MNZ) using SABRE technique.^[47,52] Although the SABRE technique presents multiple advantages for hyperpolarizing ¹⁵N-labeled compounds, it has not been translated in-vivo yet. However, biocompatible formulations have recently been demonstrated for [1-¹³C]pyruvate^[54] and therefore, could be potentially extended to HP [¹⁵N₃]metronidazole in the near future.

In this work, we employed d-DNP technique to hyperpolarize [¹⁵N₃]MNZ to obtain a biocompatible solution for in-vivo experiments. Using a state-of-the-art d-DNP polarizer (SpinAliner®, Polarize, ApS, Denmark)^[54] that was tuned to ¹⁵N frequency, we were able to monitor the polarization buildup which aided in the optimization of the [¹⁵N₃]MNZ sample formulation (Figure 1). Here we report our methods to hyperpolarize [¹⁵N₃]MNZ in a biocompatible formulation, the liquid state polarization, T₁ and T₂ relaxation times of the injectable solution, and the feasibility for in-vivo [¹⁵N₃]MNZ MRS in a rat.

Figure 1.

Schematic of a d-DNP experiment. The [¹⁵N₃]metronidazole ([¹⁵N₃]MNZ) samples were HP (1) at cryogenic temperature and under microwave irradiation in a d-DNP polarizer, where the build-up (BU) of the polarization was monitored. After reaching maximum polarization, the sample was dissolved (2) with a warm solvent and collected into a syringe. The syringe was rapidly transferred to an MRI scanner where the MRS or MRI experiment was performed (3). (*) [¹⁵N₂]imidazole resonance of a phantom positioned on the top of the surface coil.

Results and Discussion

The DNP sample of [¹⁵N₃]MNZ was optimized in several steps. A maximum MNZ concentration of 1.5 M was achieved by dissolving MNZ in >99.7% dimethyl sulfoxide (DMSO), which also avoided otherwise undesirable crystallization of the sample at cryogenic temperatures. Moreover, trityl (AH111501) is very soluble in DMSO, which resulted in a homogeneous radical distribution (with 30 mM AH111501). Although other sample mixtures containing water, glycerol, and DMSO were tested, no other solution could achieve a MNZ concentration above 1 M. For example, in a mixture of 1:1 glycerol:water, the maximum achievable MNZ concentration was only $\sim$ 0.25 M, which would have been diluted to $\sim$ 12 mM after dissolution and exceedingly challenging to detect in animal studies. The concentration of 1.5 M achievable in DMSO yielded $\sim$ 75 mM concentration after dissolution, which is comparable to the concentration of hyperpolarized ¹³C-labeled substrates used in animal research today^[26,55]. Regarding the biocompatibility of DMSO, the 21-fold dilution (see Supporting Information or SI) occurred during the dissolution process reduced the concentration of DMSO to a level that is comparable to the concentration used in several therapeutical treatments for interstitial cystitis^[56] and amyloidosis^[57,58], among other diseases^[59]. Therefore, the DMSO toxicity is not a concern.

The ¹⁵N polarization build-up performed at an optimal microwave frequency (see SI) revealed the unusually fast mono-exponential build-up time constant (T_b) of 13.8 ± 0.2 min (see SI). Previously reported build-up rates for long-T₁ ¹⁵N sites have been shown substantially greater: e.g., ¹⁵N choline T_b=45±3 mins.^[42] This result is important because short T_b values allow for full polarization build-up substantially faster—for example, near steady-state ¹⁵N polarization was achieved in less than one hour for [¹⁵N₃]MNZ. In comparison, typical ¹⁵N DNP experiments are performed for several hours because of the high T_b values.^[43]

An example of the first ¹⁵N spectrum of a dynamic spectroscopy experiment is displayed in Figure 2a. The signal of the three ¹⁵N sites can be observed in the first scan with enhancements of $\sim$ 20560, $\sim$30647, and $\sim$ 7144 for ¹⁵N₁, ¹⁵NO_2, and ¹⁵N₃, respectively. The thermal equilibrium spectrum is also displayed in Figure 2c. The ¹⁵N T₁ relaxation times for ¹⁵N₁ and ¹⁵NO₂ at 4.7 T were long, 262±2 s and 343±3 s, respectively. The T₁ of ¹⁵N₃ was 41.2±0.2 s, which is substantially shorter than that of the other two ¹⁵N sites, yet it is still comparable to that of many ¹³C-labeled HP contrast agents. The vast difference in the ¹⁵N T₁ values may be due to different electronic properties of the sites (and thus, by the Chemical Shift Anisotropy relaxation contribution), and the fact that this ¹⁵N site participates in the proton exchange, and thus, maybe more susceptible to other solute-induced sources of T₁ relaxation. Based on the measured enhancement levels (Figure 2b), the polarization at the time of dissolution was determined to be 5.2%, 6.4%, and 4.6 % for ¹⁵N₁, ¹⁵NO_2, and ¹⁵N₃, respectively.

Figure 2.

a) ¹⁵N spectrum obtained from the first scan of the HP dynamic MRS experiment. b) Signal enhancement as a function of time for the three [¹⁵N₃]MNZ ¹⁵N resonances, with color coding indicated in a). c) 1D spectrum at thermal equilibrium, acquired from the same sample after the HP experiment, with 20,000 averages for a total scan time of $\sim$6 hours. d) Stack plot showing the signal evolution in time obtained in the HP dynamic MRS experiment.

A modified balanced steady state free precession (bSSFP) sequence was used to acquire a series of 1000 images of a syringe containing HP [¹⁵N₃]MNZ (Figure 3). The sequence was modified so that an $\text{α}/2$ pulse was applied prior to the first repetition only, and so that no additional $\text{α}/2$ pulses were applied between subsequent averages or repetitions. Moreover, frequency-selective RF excitation pulses were employed to selectively image HP ¹⁵NO₂ resonance. ¹⁵NO₂ HP images were obtained with a max signal-to-noise ratio (SNR) of 67 in the first image of the bSSFP experiment (Figure 3d). The SNR obtained by averaging the first 200 images is 305.

Figure 3.

¹⁵N imaging of a 5-mL syringe filled with HP MNZ solution. a) Time series of ¹⁵NO₂ bSFFP images after 3x zero-fill, the repetition number is displayed in green. b) Proton image of the syringe. c) Overlay of the proton image and the ¹⁵N image in the first repetition of the acquisition. d) Long-lasting ¹⁵N HP signal measured in bSSFP acquisition. The fitted curve used to determine the T₂ is displayed in red. The blue dotted line represents the repetition acquired 10 min after the dissolution experiment (image shown in a, repetition n^o 300).

The bSSFP signal was used to model the T₂ of ¹⁵NO₂ (Figure 3d) using the equation developed by Scheffler^[60,61], adapted for HP imaging.^[62,63] The ¹⁵NO₂ T₂ relaxation time at 4.7 T was long, 20.5±0.4 s. To confirm the T₂ measurement, a bSSFP experiment was done with a HP [1-¹³C]acetate phantom (see SI). The determined ¹³C T₂ of HP [1-¹³C]acetate was 4.6 s, which is similar to the previously reported value.^[1–15][17] The long-lasting HP [¹⁵N₃]MNZ imaging signal in the bSSFP acquisition was remarkable, which is rationalized by the correspondingly long T₁ and T₂ values.^[17,63] The long-lasting HP signal indicates a good potential for in-vivo ¹⁵N imaging applications.

Results of the pilot in-vivo study using tail-vein bolus injection of HP [¹⁵N₃]MNZ are displayed in Figure 4. Note that the spatial localization in this experiment is achieved using a surface coil placed over the rat head. Since the blood flow to the brain is vastly greater than that to the rest of the head tissues, we attribute the detected ¹⁵N signals to those in cerebral vasculature and brain tissues. The in-vivo dynamic MRS acquisition started approximately 5 seconds after the start of injection. The signal time curves (Figure 4d) showed that [¹⁵N₃]MNZ reached the brain within 5 seconds after the start of injection and the signal reached maximum in 15–20 seconds. The HP signal lasted for over a minute after injection providing an excellent time window for neuroimaging applications. The highest SNR values are obtained by adding scans 3–9, $\sim$227, $\sim$270, and $\sim$60 for ¹⁵N₁, ¹⁵NO₂, and ¹⁵N₃, respectively (spectrum displayed in Figure 4a). SNR values are $\sim$ 57, $\sim$66, and $\sim$ 13 for ¹⁵N₁, ¹⁵NO₂, and ¹⁵N₃, respectively, for scan 7. The SNR of the pilot experiment can be substantially improved through the use of singular value decomposition to suppress the noise^[64].

Figure 4.

Dynamic ¹⁵N MRS of HP [¹⁵N₃]MNZ in rat brain with a phantom of 3M [¹⁵N₂]imidazole on the top of the coil. a) Averaged spectrum of the scans 3 to 9 in the dynamic MRS series resulting in maximum SNR. b) Surface plot of the in-vivo dynamic spectroscopy experiment showing [¹⁵N₃]MNZ signal evolution in time. c) Anatomical image of raťs head with the green line representing the position of the homemade ¹⁵N transmit/receive coil. d) Signal time curves of the three ¹⁵N sites. Colors follow the same color scheme as in a). The injection period is outlined by the green box. The Tm was extracted on the region after the black dotted line.

The dynamics of in-vivo ¹⁵N MRS with respect to the injection duration and the arrival of the MNZ bolus are depicted in Figure 4d. The reader should be reminded that the signal rise is attributed to the arrival of the HP [¹⁵N₃]MNZ bolus to the brain, and the signal decay is associated with both the bolus washout from the brain and the ¹⁵N T₁ relaxation. The apparent ¹⁵N signal decay time constant (Tm) during the HP [¹⁵N₃]MNZ bolus washout was estimated by fitting to a mono-exponential decay of the signal time curves (region to the right of the black dotted line in Figure 4d): 24.3 ± 0.4 s and 24.7 ± 0.7 for ¹⁵N₁ and ¹⁵NO₂ sites, respectively (the respective time constant for ¹⁵N₃ site was challenging to determine due to low SNR).

To understand the fast signal decays observed from the rat head, we measured the T₁ decay time constants of the three ¹⁵N labels of [¹⁵N₃]MNZ in blood serum and in D₂O (as a reference solution) at 1.4 T (SpinSolve 60 MHz benchtop spectrometer by Magritek, Wellington, New Zealand) under the same experimental conditions (see SI). The resulting ¹⁵N T₁ for ¹⁵N₁, ¹⁵NO₂, and ¹⁵N₃ sites in serum were 47 ± 2 s, 81 ± 3 s, and 36 ± 2 s, respectively, compared to 119 ± 4 s, 212 ± 6 s, and 20 ± 5 s in D₂O. The T₁ values in serum were smaller than those in D₂O, as expected. The T₁ values in serum were much larger than the apparent ¹⁵N signal decay time constants (Tm) measured in the rat head for both the ¹⁵N₁ and ¹⁵NO₂ sites (Figure 4b). Furthermore, the T₁ value of the ¹⁵NO₂ site in serum (81 ± 3 s) was significantly larger than the T₁ of the ¹⁵N₁ site (47 ± 2 s) and yet the Tm values in the rat head were both reduced to very comparable values (24.3 ± 0.4 s and 24.7 ± 0.7 s for ¹⁵N₁ and ¹⁵NO₂, respectively). These observations suggested that the signals of ¹⁵N₁ and ¹⁵NO₂ in vivo were both affected by a common and strong mechanism responsible for the fast signal decay in vivo. This mechanism in vivo is likely to be washout from the bloodstream in the brain.

Specifically, we envision that this FDA-approved antibiotic could be employed as hypoxia sensing molecular probe. As the first step towards this goal, we performed computational studies to determine the ¹⁵N chemical shifts of putative metabolites of downstream [¹⁵N₃]MNZ metabolism in a hypoxic environment. The calculated ¹⁵N chemical shifts (157 ppm, 249 ppm, and 343 ppm for ¹⁵N₁, ¹⁵N₃ and ¹⁵NO₂, respectively) match remarkably well to the measured chemical shifts (166 ppm, 249 ppm, and 357 ppm, respectively) of the three ¹⁵N in [¹⁵N₃]metronidazole (Figure 5). These simulation results gave us confidence for the predicted large ¹⁵N chemical shifts of the reduced [¹⁵N₃]metronidazole (Figure 5), which should be well differentiated from the injected [¹⁵N₃]metronidazole in hypoxic conditions. The stepwise reduction of the ¹⁵NO₂ moiety results in substantial changes (of up to 390 ppm) of all ¹⁵N sites, i.e., all three ¹⁵N sites can sense the reduction of nitro group through the changes in their chemical shifts. This represents a clear translational advantage of employing a multi-chromatic HP probe. However, the reader should be reminded that the directly attached protons in the downstream metabolites would likely render short ¹⁵NO₂ T₁ in hydroxylamino- and amino-derivatives making them likely undetectable – this is not a fundamental limitation of this potential metabolic sensor as ¹⁵N₁ and ¹⁵N₂ sites will likely not experience any T₁ changes through these chemical transformations, and they have sufficient chemical shift dispersion for spectroscopic sensing of putative metabolites. No metabolites were detected via ¹⁵N MRS in this pilot study because we employed healthy rat without any brain hypoxia or infection. All in all, the presented pilot results here represent a substantial advance over the previous report to isotopically enrich [¹⁵N₃]metronidazole and perform mechanistic SABRE studies. While ¹⁵N polarization in the SABRE studies was higher (P_15N $\sim$ 16%) than the one reported here ($\sim$6%), it should be noted that the product of polarization and the concentration (i.e., polarization payload) is nearly the same in this study employing d-DNP versus previous SABRE studies. However, most importantly the d-DNP prepared ¹⁵N polarization payload is already biologically compatible, which we demonstrated here, whereas SABRE technology has not been translated in vivo with any contrast agent (ca. 12/2022), to the best of our knowledge. We have also reported the computational studies of ¹⁵N chemical shifts that can act as the potential reporters to sense hypoxia in vivo. The pilot results reported here indeed bode well for future systematic in-vivo studies. Specifically, we envision the applications related to hypoxia sensing in stroke, brain trauma, neurodegenerative diseases, resistance to radiotherapy in solid cancer tumors, and others.

Figure 5.

Putative metabolism of [¹⁵N₃]metronidazole in hypoxic biological environment^[48]: the color-coded numbers denote ¹⁵N chemical shifts of the corresponding ¹⁵N sites.

Conclusion:

In this study, we utilized d-DNP for efficient ¹⁵N hyperpolarization of uniformly ¹⁵N-labeled metronidazole. The DNP sample formulation that we developed resulted in an exceptionally fast ¹⁵N polarization build-up constant of 13.8 ± 0.2 mins. ¹⁵N polarization of over 6% was obtained after approximately a 30-second-long process to transfer the sample from the polarizer to the MRI scanner. Dissolution yielded $\sim$ 75 mM HP [¹⁵N₃]MNZ in aqueous biocompatible media suitable for in-vivo applications. Very long ¹⁵N T₁, exceeding 5 minutes in solution after dissolution, was observed for the ¹⁵NO₂ site at 4.7 T. We have successfully demonstrated the feasibility of in-vitro ¹⁵N MRI using bSSFP imaging sequence and obtained remarkably long-lasting bSSFP signal on ¹⁵NO₂ due to its very long T₂. Finally, we have also performed a pilot in-vivo demonstration using tail-vein intravenous injection of HP [¹⁵N₃]MNZ in a healthy rat. HP ¹⁵N signals were successfully detected over 1 minute after the bolus of HP [¹⁵N₃]MNZ was delivered to the brain. Fast DNP ¹⁵N polarization buildup, good levels of ¹⁵N polarization, and exceedingly long ¹⁵N T₁ and T₂ of HP [¹⁵N₃]MNZ bode well for HP MRI experiments in general, which will be developed in future studies to explore potential biomedical applications of this FDA-approved antibiotic as an antibiotic or hypoxia probe.