Molecular hydrogen and catalytic combustion in the production of hyperpolarized ⁸³Kr and ¹²⁹Xe MRI contrast agents

Significance

The high signal intensity associated with magnetic resonance of hyperpolarized ¹²⁹Xe has enabled countless applications ranging from materials science to biomedical MRI of the lung and brain. New modalities are constantly emerging; for example, hyperpolarized ¹²⁹Xe biosensors are of potential interest for molecular imaging of biomarker distribution in organs. Hyperpolarized ⁸³Kr shows promise as novel surface-sensitive contrast agent for pulmonary MRI but previous work was limited to excised lungs. This work reports a previously unexplored approach in the generation of hyperpolarized contrast agents that makes highly concentrated hyperpolarized ⁸³Kr available for the first time, to our knowledge. The results also provide the basis for on-demand flow production of highly concentrated hyperpolarized ¹²⁹Xe for in situ process monitoring and in vivo molecular imaging.

Abstract

Hyperpolarized (hp) ⁸³Kr is a promising MRI contrast agent for the diagnosis of pulmonary diseases affecting the surface of the respiratory zone. However, the distinct physical properties of ⁸³Kr that enable unique MRI contrast also complicate the production of hp ⁸³Kr. This work presents a previously unexplored approach in the generation of hp ⁸³Kr that can likewise be used for the production of hp ¹²⁹Xe. Molecular nitrogen, typically used as buffer gas in spin-exchange optical pumping (SEOP), was replaced by molecular hydrogen without penalty for the achievable hyperpolarization. In this particular study, the highest obtained nuclear spin polarizations were P = 29% for ⁸³Kr and P = 63% for ¹²⁹Xe. The results were reproduced over many SEOP cycles despite the laser-induced on-resonance formation of rubidium hydride (RbH). Following SEOP, the H₂ was reactively removed via catalytic combustion without measurable losses in hyperpolarized spin state of either ⁸³Kr or ¹²⁹Xe. Highly spin-polarized ⁸³Kr can now be purified for the first time, to our knowledge, to provide high signal intensity for the advancement of in vivo hp ⁸³Kr MRI. More generally, a chemical reaction appears as a viable alternative to the cryogenic separation process, the primary purification method of hp ¹²⁹Xe for the past 2 1/2 decades. The inherent simplicity of the combustion process will facilitate hp ¹²⁹Xe production and should allow for on-demand continuous flow of purified and highly spin-polarized ¹²⁹Xe.

The development of hyperpolarized (hp) noble gas MRI has resulted in a number of excellent protocols to probe different structural and functional aspects of lungs in health and disease (1–6). Technological improvements (6–15) have enabled pulmonary hp ¹²⁹Xe MRI at high spatial resolution, thereby reducing the need for use of the scarcely available ³He isotope. Furthermore, tissue solubility, large chemical shift range, and interaction with specific sensor molecules allow for a variety of hp ¹²⁹Xe applications in biomedical sciences and beyond (1–6, 16, 17). Despite its nuclear electric quadrupole moment that causes rapid relaxation, the isotope ⁸³Kr (nuclear spin I = 9/2) can be hyperpolarized through Rb spin-exchange optical pumping (SEOP) as first demonstrated by Grover (18) and explored in detail by Happer and coworkers (19). Volk et al. first observed ⁸³Kr quadrupolar coupling in the gas phase that originated from the surface of the SEOP cell (20), and Mehring and coworkers described T₂ relaxation as a probe for the cell surfaces (21). The intriguing properties of ⁸³Kr can be more generally used after the removal of the reactive Rb vapor to generate surface-sensitive MRI contrast (22). Most recently, T₁ surface quadrupolar relaxation MRI contrast (23) with hp ⁸³Kr in excised lungs was demonstrated to be indicative of surface-to-volume changes in an animal model of emphysema (24).

Although hp ¹²⁹Xe can be obtained through dynamic nuclear polarization (25) with high spin-polarization levels of up to P = 30% (26), at present only SEOP can produce hp ¹²⁹Xe with $P\ge 90\%$. Furthermore, SEOP is currently the only method to provide hp ⁸³Kr for biomedical MRI applications (24). During SEOP of ⁸³Kr or ¹²⁹Xe, the noble gas is diluted with a buffer gas, usually either a ⁴He–N₂ mixture or pure N₂ gas. The buffer gas serves a dual purpose as it prevents destructive radiation trapping, originating from radiative electronic relaxation of rubidium (27, 28), but also causes pressure broadening of the Rb D₁ linewidth. Rubidium line broadening maximizes adsorption of laser light emitted by high-power solid devices, even if those are line narrowed. Following SEOP, hp ¹²⁹Xe is cryogenically separated from the gas mixture under carefully chosen conditions to prevent polarization loss (29). Cryogenic separation complicates the operational procedures for clinical and preclinical hp ¹²⁹Xe MRI. For a number of applications a repeat noble gas delivery at high polarization and concentration levels is needed (6, 30–35). Additionally, a very constant level of spin polarization is required for molecular imaging with hp ¹²⁹Xe biosensors using the hyperpolarized xenon chemical exchange saturation transfer (HyperCEST) methodology (17, 36–38). Cryogenic separation enables increased polarization in pure xenon gas but is inherently a “batch mode” production process that disrupts on-demand continuous flow. Whereas cryogenic separation may complicate and limit some hp ¹²⁹Xe applications, it is not an option at all for the hp ⁸³Kr production due to the fast quadrupolar relaxation. To avoid cryogenic separation, SEOP at a high noble gas mole fraction has been explored in the past (6, 10, 14, 39–41). Nevertheless, gas dilution is still required to ensure high spin polarization at the high production volumes required for MRI and the dilution therefore reduces MRI signal intensity per unit volume of inhaled gas. An innovative method developed by Kimura, Imai, Fujiwara, and coworkers (6, 10, 33) produces a constant stream of 70% concentrated hp ¹²⁹Xe but the polarization is typically in the P = 10% regime. In this work, ¹²⁹Xe and ⁸³Kr SEOP is attempted with mixtures containing molecular hydrogen as buffer gas and radiation quenching agent (42–46) that can subsequently be removed through catalytic combustion.

Results and Discussions

Fig. 1A shows the ¹²⁹Xe nuclear spin polarization, P, as a function of gas pressure after 6 min of SEOP of a mixture of 5% Xe and 95% Kr at an external cell temperature of 378 K. The spin polarization is similar to data produced previously with a 5% Xe–95% N₂ mixture under almost identical conditions described and analyzed in detail in ref. 39, also shown in Fig. 1A for comparison. Using H₂ buffer gas, hp ¹²⁹Xe with up to P = 63% was generated within 6 min of SEOP at 70 kPa total gas pressure. Using 5% krypton in 95% buffer gas, the highest ⁸³Kr polarization of P = 29%, obtained with H₂ after 11 min of SEOP at 440 K, is comparable to that obtained previously with N₂ under similar conditions (39).

Fig. 1.

(A) SEOP-generated ¹²⁹Xe and ⁸³Kr spin polarization in mixtures containing 5% noble gas and 95% buffer gas as a function of total SEOP gas pressure. Using H₂ as buffer gas, the ¹²⁹Xe spin polarization (obtained at 378 K) is shown in red filled circles and the ⁸³Kr polarization (obtained at 440 K) is displayed as blue filled squares. For comparison, previous data from ref. 39 with N₂ as buffer gas is shown for ¹²⁹Xe SEOP at 378 K (green open circles) and for ⁸³Kr SEOP at 433 K (green open squares). In contrast, CH₄ is an inefficient buffer gas that led to low spin polarization (orange open triangles, taken from ref. 54). (B) The D₂ fluorescence at 780 nm (in arbitrary units, a.u.) measured at the front of the cell during SEOP indicates that sufficient radiation quenching with H₂ at 378 K (red circles) and at 433 K (blue squares) requires pressures above 50 kPa. Using N₂ (green open circles), the D₂ radiation is efficiently quenched at pressures above ∼5 kPa. Inset (i) shows the high D₂ fluorescence measured for Xe SOEP at 378 K with methane that is a very ineffective quenching gas.

On-Resonance Rubidium Hydride Formation.

The high spin-polarization values for ⁸³Kr with the H₂ in Fig. 1A were achieved repeatedly for more than 30 h of SEOP over a time span of about 1 wk and for ¹²⁹Xe over a duration of at least 6 h of continuous SEOP despite the formation of rubidium hydrides (RbH) during on-resonance D₁ laser irradiation (47). Over time. a white RbH deposition was observed on all SEOP cell surfaces but to a lesser extent at the laser entry window. Furthermore, SEOP cell recycling, causing substantial removal of the RbH, was possible by increasing the temperature to 480 K for about 30 min with ongoing laser irradiation while the cell was connected, via a Rb trap, to a vacuum pump. Although excessive RbH buildup on the cell entry window will reduce laser irradiation, deposition on other cell surfaces may not necessarily adversely affect the achievable noble gas spin polarization. Previous work demonstrated that the ¹³¹Xe T₁ relaxation time increases for this spin I = 3/2 isotope because of RbH buildup (48–50). RbH surface coating may therefore also reduce the ⁸³Kr T₁ relaxation rates and improve SEOP. A detailed study by Cates and coworkers explored the effect of RbH surface deposition on the T₁ relaxation of ¹²⁹Xe as a function of field strength (≥0.08 T) and temperature (≤340 K) in a spherical 25-mm-diameter cell (51). Based on the results one may interpolate ¹²⁹Xe T₁ times in excess of 400 s at the conditions used in the current work, causing little interference during rapid SEOP.

H₂ as Radiation Quenching Agent.

The high polarization values obtained with H₂ in Fig. 1A, i.e., in the absence of any N₂, are remarkable. Depending on SEOP temperature, the gas mixtures usually contain at least a partial pressure of 5 kPa of molecular nitrogen for radiation quenching to prevent radiation trapping of arbitrarily polarized fluorescence photons that severely reduce Rb polarization (27, 28). Previous work investigating the efficacy of H₂ to quench Rb radiation found that the effective cross-sections for quenching of the D₁ and D₂ transitions are about 1 order of magnitude smaller for H₂ than for N₂ (42). An empirical model of radiation quenching with H₂ is reported in ref. 52 but is limited to buffer gas pressures below 7 kPa. Generally, H₂ has found little consideration as quenching agent at high gas density and temperature conditions typically used for SEOP production of hp ¹²⁹Xe or hp ⁸³Kr for MRI applications.

The measurement of D₁ fluorescence at 795 nm, obscured by high-power laser radiation, is experimentally very demanding. Instead, the Rb D₂ fluorescence during SEOP was straightforwardly monitored in this work. The mixing cross-sections between the electronic states Rb*5²P_1/2 and Rb*5²P_3/2 are comparable for H₂ and N₂ (42) and the D₂ fluorescence should therefore provide at least qualitative insights into D₁ radiation quenching efficiency of the molecular hydrogen. Fig. 1B demonstrates that H₂ serves as a remarkably efficient radiation quenching agent at partial pressures above ∼40–60 kPa, where little fluorescence was detected during SEOP even at the very high temperature of 433 K. Therefore, H₂ can be used as radiation quenching agent during SEOP but the required partial pressure is about 10 times higher than that for N₂, in agreement with previous measurements of quenching cross-section in ref. 42. Below this critical partial pressure, radiation quenching is largely incomplete and a substantial difference in the D₂ fluorescence is observed between SEOP of ¹²⁹Xe and ⁸³Kr. Fig. 1B depicts a higher fluorescence count at the elevated temperature used for ⁸³Kr SEOP (in the 433–440-K range) where the Rb density is almost 20-fold increased compared with that for ¹²⁹Xe SEOP at 378 K. At low buffer gas pressure, where radiation quenching is incomplete, the high Rb density for ⁸³Kr SEOP produces significantly more radiation trapping. The consequences of the increased radiation trapping can be observed in Fig. 1A where, at buffer gas pressures below ∼50 KPa, the noble gas polarization obtained through SEOP with H₂ falls short compared with that with N₂. The effect of Rb density on radiation trapping was previously studied at extremely low H₂ partial pressure (<0.1Pa) for the purpose of SEOP of dissociated atomic hydrogen (43–46). Under these conditions, radiation quenching is effectively nonexistent and radiation trapping can be limited only by keeping the Rb densities very low through SEOP temperatures around or below 318 K. In the current work, no adverse effect of H₂ as buffer gas at SEOP pressures above 40–50 kPa (with the exception of reversible RbH formation, discussed above) was found compared with SEOP with N₂ under otherwise identical conditions, regardless of the temperature. Note that the high IR quenching cross-section of H₂ is a rare exception among small molecules without chemical double or triple bonds and Fig. 1 includes data obtained with CH₄ as buffer gas for comparison. The inefficient small quenching cross-section of CH₄ (53) causes a high IR count (Fig. 1B) and a correspondingly low ¹²⁹Xe spin polarization below that of hp ⁸³Kr (Fig. 1A), a problem that was previously solved using a ternary mixture of 5% Xe, 85% CH₄, and 10% N₂ (54).

Buffer Gas Removal Through Catalytic Combustion.

Efficient SEOP of noble gases in H₂ mixtures opens a path for oxidative buffer gas elimination that produces H₂O as the only reaction product. Most of the generated water vapor can straightforwardly be removed through condensation at ambient temperature. The resulting gas phase is composed of the purified hp noble gas in addition to a small remaining quantity of the water vapor. Previous work demonstrated qualitatively the feasibility of hp ¹²⁹Xe MRI (54) of a methane combustion zone. To quantify the effect of catalytic hydrogen combustion on the noble gas spin polarization, an experimental setup was devised as sketched in Fig. 2A. Shuttling of hp noble gas–H₂ mixture into the reactor chamber that contained a small quantity of Pt and monitoring the signal decay over time through NMR spectroscopy at 9.4 T provided the baseline data, shown as open circles in Fig. 3B, for hp ¹²⁹Xe and in Fig. 4B for hp ⁸³Kr. For oxidative H₂ removal, O₂ was added, which led to hydrogen combustion within <100 ms due to the presence of the Pt catalyst, as depicted in Fig. 2B. The reactor pressure during this process was monitored (Figs. 3A and 4A) but potential short pressure increases during the reaction were not detected at the time resolution of the pressure gauge. Upon adding O₂, slightly above the stoichiometric ratio, the pressure decreased within 15 s as the sole reaction product, H₂O, had condensed rapidly upon cooling. The temperature increase measured outside the reactor was limited to 5 K and the final reactor pressure of 4.7 ± 0.5 kPa was close to that expected from the combined water vapor pressure at ambient temperature (3.2 kPa at 298 K) and the noble gas partial pressure (1.2–1.3 kPa).

Fig. 2.

(A) Experimental setup used for combustion experiments. Using pressure equalization, the hp gas/H₂ mixture was shuttled to the reactor/detection cell containing the Pt catalyst. Oxygen was then injected from the O₂ reservoir. (B) Exported pictures from 100 frame per second video of the reactor upon O₂ delivery. Approximately 100 ms after opening of valve C, minor combustion zones became visible (see white arrows), indicating that the catalytic oxidation of the H₂ gas had commenced. The major combustion event took place 20–30 ms thereafter and the combustion appeared to be largely completed within 80 ms. The reactor outside temperature increase was limited to approximately 5 K.

Fig. 3.

(A) In situ reactor pressure measurements (blue diamonds) during the combustion reaction of a 5% Xe/95% H₂ gas mixture, with oxygen reservoir tap (tap C, Fig. 2A) opened at t = 40 s adding 11.2 ± 0.4 kPa partial pressure of O₂. Average initial partial pressure of H₂ = 22.0 ± 1 kPa and the oxygen addition caused the pressure drop due to the combustion reaction (a short term pressure increase, if present, was not observed at the time resolution of the gauge). (B) NMR measurements performed in separate experiments without pressure gauge (i.e., tap B, Fig. 2A was kept shut to reduce the dead volume of the experiment). Normalized integrated hp ¹²⁹Xe NMR signals following a 9° pulse during the catalytic oxidation with 13.4 ± 0.4 kPa oxygen (red circles) or 20.5 ± 0.5 kPa (green triangles) added at t = 40 s. Average initial partial pressure of H₂ = 25.5 ± 1 kPa. Control signals with no oxygen/combustion (tap C, Fig. 2A, remained closed) shown as black open circles. Signal intensity data were corrected for flip-angle attenuation and are the average from three repeat measurements.

Fig. 4.

(A) Pressure curve of combustion reaction of a 5% Kr/95% H₂ gas mixture with 11.2 ± 4 kPa O₂ added after 12 s causing the reaction and observed pressure drop. Average initial partial pressure of H₂ = 23.5 ± 1 kPa. (B) Normalized integrated hp ⁸³Kr NMR signals following a 12° pulse during the combustion reaction adding 13.4 ± 0.4 kPa (red circles) and 20.5 ± 0.5 kPa (green triangles) partial pressure of oxygen. Average initial partial pressure of H₂ = 27.5 ± 2.5 kPa. Black open circles show control signals for a 5% Kr/95% H₂ gas mixture with no combustion. Signal intensity data were corrected for flip-angle attenuation and are the average from three repeat measurements.

Monitoring the hp ¹²⁹Xe signal intensity (Fig. 3B), an initial signal increase was observed upon O₂ delivery, caused by hp ¹²⁹Xe influx from the connecting tubing [i.e., perfluoroalkoxy polymer (PFA) tubing with 1.6-mm inner diameter] that was pushed into the reaction chamber by the oxygen gas. Within 20 s, gas convection and diffusion returned the signal intensity approximately to the baseline value, indicating that the nuclear spin state had experienced no significant depolarization. However, after the reaction, the ¹²⁹Xe relaxation was accelerated due to a small excess of paramagnetic O₂ (approximately 0.7 kPa partial pressure). Increasing the O₂ excess to 7.5 ± 0.5 kPa led to further accelerated ¹²⁹Xe signal decay.

The hp ⁸³Kr data in Fig. 4B demonstrate that combustion caused no signal loss and, in contrast to ¹²⁹Xe, the ⁸³Kr gas-phase relaxation slowed down after the reaction. The decrease of the ⁸³Kr T₁ relaxation was likely a consequence of competitive coadsorption of water molecules on the reactor surface that have been found previously to reduce krypton surface interactions (55). The very low gyromagnetic ratio makes ⁸³Kr insensitive to paramagnetic relaxation even at higher (7.5 ± 0.5 kPa) oxygen partial pressure (Fig. 4B) (40). An initial short-term signal increase upon O₂ gas delivery, present in the ¹²⁹Xe data, was not observed because hp ⁸³Kr located in the connecting PFA tubing depolarized due to fast quadrupolar T₁ relaxation in the presence of the fluorocarbon surface (56).

Conclusion

The capability to almost instantaneously remove a buffer gas that serves as an efficient radiation quenching agent during SEOP allows, for the first time to our knowledge, for the generation of purified, highly spin-polarized ⁸³Kr. Furthermore, the method may be of interest for a fully automated hp ¹²⁹Xe production at reduced costs without the need for cryogenic operations. Because of the low total gas pressure after catalytic buffer gas removal, the hp gases will require recompression to ambient pressure for biomedical application. Recompression has previously been demonstrated with little polarization loss for hp ¹²⁹Xe (10, 40) and acceptable 1/4 polarization loss for hp ⁸³Kr (40). Based on the result in this work, this would suggest that purified hp ⁸³Kr with 21% spin polarization is now feasible, a sevenfold improvement over recent hp ⁸³Kr work with 3% apparent polarization that already allowed for MRI with 0.795 × 0.635-mm² resolution in ex vivo lungs (24).

Note that the experimental setting was for the sole purpose of proof of concept demonstration and can be substituted by combustion under continuous flow within a catalytic converter. Furthermore, direct contact between H₂ and O₂ may be avoidable through the use of recyclable metal oxides similar to those used in chemical looping combustion (57). Reactive H₂ removal may take place in a small flow-through catalytic converter placed downstream from the SEOP cell that would facilitate gas transport through the generated pressure gradient. Kimura, Imai, Fujiwara, and coworkers used a peristaltic pump for continuous hp ¹²⁹Xe recompression after continuous flow SEOP at low pressures (6, 10, 33), a concept that could be further extended by flow-through catalytic buffer gas removal before recompression. Enabling on-demand production of an uninterrupted, continuous flow of hp ¹²⁹Xe at high spin polarization and purity, this concept should be of high value for a variety of applications ranging from material science to biomedical MRI. If required, hp ¹²⁹Xe could also be stored before recompression (and even before H₂ removal), using the extraordinary slow T₁ relaxation of ¹²⁹Xe described in the past for gas phase xenon under certain conditions (58–60). As a final note, the presented concept does not involve any toxic reagents or reaction products and the complete absence of nitrogen and carbon prevents any accidental generation of nitrogen oxides (NO_x) and carbon monoxide (CO) during combustion. Any H₂ passing through the “buffer gas removal step,” however unlikely, can straightforwardly be detected through reliable hydrogen microsensors.

Materials and Methods

SEOP was conducted at 0.04-T field strength in a 75-cm³ cylindrical borosilicate cell (∼120 mm long, 28-mm inner diameter) using a Comet module (Spectra Physics) laser with a 0.25-nm linewidth producing 23-W incident power at the SEOP cell entrance window under very similar conditions as described in detail in ref. 39. A mixture of 5% Xe (or 5% Kr) in H₂ was used for all experiment in this work with natural abundance noble gas isotope distribution (i.e., 26.4% ¹²⁹Xe and 11.5% ⁸³Kr). Spin polarization was determined as described in ref. 39 (see also ref. 6) and a general expression for polarization of I ≥ 1/2 spin systems is discussed in ref. 61. To monitor the Rb D₂ fluorescence during SEOP at 780 nm, an HR2000+ Ocean Optics spectrometer was used.

All catalytic combustion experiments took place in a 1.5-mm glass wall vessel that contained 25 mg Pt/Al₂O₃ catalyst powder (i.e., 5 wt % dry loading Pt on alumina). Standard safety precautions were followed and the H₂–hp noble gas mixture was delivered to the reactor at pressures below 30 kPa to limit temperature and pressure bursts that could compromise reactor integrity. The reactor pressure was metered using a gas composition independent diaphragm gauge (OmegaDYNE Inc.).