Instrumentation for Hydrogenative Parahydrogen-Based Hyperpolarization Techniques

Graphical Abstract

INTRODUCTION

The unique properties of the entangled, antisymmetric nuclear spin state of dihydrogen, para-hydrogen (pH₂), has intrigued physicists, chemists, and other scientists for almost a century. pH₂ was used as a model system in the early days of quantum mechanics¹ and is used for fueling rockets as well as combustion-free cars today. In the 1980s, pH₂ was discovered as a convenient and potent source of spin order, allowing the enhancement of the signals of magnetic resonance (MR) by several orders of magnitude.^2–4 In the advent of hyperpolarized (HP) contrast agents (CA) for biomedical MR imaging (MRI) that followed, pH₂-based hyperpolarization methods played an important role in the acquisition of the first HP ¹³C in vivo images (Fig. 1).^5,6

Figure 1.

First hyperpolarized ¹³C in vivo MRI ever published (a). The contrast agent, hydroxyethyl [1-¹³C]propionate-d₃ (HEP), was produced using a prototype commercial millitesla polarizer (Amersham Biosciences, healthcare company), similar to the one shown schematically in (b). Broadly speaking, the polarizers consisted of a unit handling the fluids, the actual hyperpolarization (spin-order transfer), and the coordination of the entire process (numbers 1 – 5 represent valves). Images reproduced from Parahydrogen-Induced Polarization in Imaging: Subsecond 13C Angiography, Golman, K.; Axelsson, O.; Johannesson, H.; Mansson, S.; Olofsson, C.; Petersson, J. S. Magn. Reson. Med., Vol. 46, Issue 1 (ref 5). Copyright 2001 Wiley (a), and reprinted by permission from Springer: Magn Reson Mater Phy, PASADENA Hyperpolarization of 13C Biomolecules: Equipment Design and Installation, Hövener, J.-B.; Chekmenev, E. Y.; Harris, K. C.; Perman, W. H.; Robertson, L. W.; Ross, B. D.; Bhattacharya, P., Vol. 22 (ref 13). Copyright Springer 2009. (b).

Ever since, pH₂ has proven to be highly valuable for analytical investigations and fundamental research, e.g., in analytical and catalytic chemistry or in the physics of singlet spin states.^7–10

pH₂ is produced fast and stored easily. As one of four (nuclear) spin states of dihydrogen, 25 % of H₂ at room temperature are pH₂, while the other 75 % are ortho-Hydrogen (oH₂, following the Boltzmann distribution). At lower temperatures, however, the para-fraction becomes more and more enriched until at approx. 25 K, 100 % pH₂ is obtained. While the para-enrichment is fast using appropriate catalyst, pH₂ can be stored for hours to days at room temperature without significant loss if the catalyst is absent. These unique properties make pH₂ a spin order that can be produced easily, stored conveniently (in a pressurized bottle) and used on demand, e.g., for producing hyperpolarized CAs or observing chemical reactions with enhanced sensitivity.

When it comes to clinical applications, however, dissolution dynamic nuclear polarization (dDNP) evolved faster than pH₂–based HP methods.^11,12 The reasons for this may be that a) at first, only few biologically relevant CAs were available (because of limited chemistry at that time), b) that mastering the complex process of pH₂ induced polarization (PHIP) was not straight forward (involving quantum mechanics, chemical reactions, and magnetic resonance tailored to each individual CA) and that c), no pH₂-polarizers were commercially available, while there were at least three different generations for dDNP.

While pH₂ is easily produced in large quantities, at low cost and with a shelf life of days (depending on the storing conditions), there are some hurdles to overcome before a CA is ready for administration. For one, the pH₂ spin order per se is MR invisible (total spin of 0!) and well hidden inside the dihydrogen molecule. To obtain a hyperpolarized CA from pH₂, typically, the following steps have to be taken: a), bringing pH₂ and the target into contact (by catalytic addition or reversible exchange); b), to transfer or transform the pH₂-derived spin order into a desired form; and c), the purification and quality assurance (QA) prior to an in vivo application. These steps usually involve a chemical reaction at elevated temperatures, pressures, sometimes in aggressive media or extreme pH – synchronized with quantum mechanical spin order transfer (SOT) mediated by evolution at constant or varying magnetic fields and radiofrequency (RF) pulses. To realize this process, various devices have been ingeniously devised; however, a single, unified design has not yet emerged. The lack of such a device may be attributed to the fact that the power and versatility of pH₂ has resulted in quite a few different methods – the magnetic fields alone vary by a factor of 10⁹ (from nanotesla to tesla). Even today, pH₂ hyperpolarization methods keep evolving at a fast pace; among the ground-breaking advances, SABRE,¹⁴ gases,^15,16 continuos HP,^17–19 PHIP-SAH,²⁰ RASER,^21–23 precipitation,²⁴ and relay methods^25,26 are only examples from the last decade. These methods require specific experimental conditions and ultimately dedicated instrumentation, which differs from method to method.

For a standardized, clinical application of a specific contrast agent, however, a consolidated setup is required, that can be certified and approved, provides reliable polarization and quality assurance. In this respect, much can be learned from SEOP²⁷ and DNP^11,12 with respect to polarizers and regulatory approval.

Here, we review the different instrumentations for pH₂ hyperpolarization, with an emphasis on biomedical application. To keep this review concise, we focus on setups for hydrogenative pH₂-based hyperpolarization alone and the most recent literature (~last 5 years); still, we refer to pioneering and game-changing developments whenever appropriate, and when other methods (SABRE, DNP, SEOP, etc.) show similar instrumentational aspects. Dedicated reviews on SABRE-related instrumentation, spin-order transfer, and pH₂ production are expected to be published elsewhere.

ANALYSIS OF PH₂ POLARIZERS

The requirements for a biomedical polarizer may be defined as to i) provide a clean, ii) aqueous solution of iii) highly polarized, iv) appropriately concentrated agents at v) physiological conditions that is preferably produced in a iv) good-manufacturing process (GMP). To make such a contrast agent, the role of hardware may be categorized as follows:

1. making the pH₂ spin order available to the target molecule (hydrogenation);

2. transferring the pH₂-derived spin order into the desired spin hyperpolarization (typically longitudinal X-nuclear magnetization);

3. purification of the solution and assuring quality.

We will elaborate on these steps in the following.

Parahydrogen addition

Before the addition, the lifetime of gaseous^28–30 or dissolved^31–33 pH₂ is typically long (days or many minutes, respectively). However, once bound to the catalyst³⁴ or the precursor,^9,35–39 the lifetime of the spin order is drastically reduced because the added protons, referred to as I₁ and I₂, interact with the environment and the rest of the (now larger) spin system. Thus, the hydrogenation should be conducted as fast as possible to reduce relaxation losses. The hydrogenation kinetics are typically affected by a multitude of coupled reaction parameters, like temperature, pH₂-availability, solvent, catalyst, precursor molecule or pH value. While the catalyst is needed for the addition, it may cause relaxation at the same time and needs to be removed before an injection if it is harmful.

The starting point for the SOT, i.e. the density matrix after the hydrogenation, is strongly dependent on the coupling between I₁ and I₂. It is determined by the molecular structure and the external static magnetic field B₀. If the spins are “strongly coupled” (that is, if their mutual J-coupling J_I1I2 is much larger than the difference of their Larmor precession frequencies δv (δv « J_I1I2)), the eigenstates of the two-spin system are essentially the singlet-triplet (S-T) basis states. In this case, the “singlet spin order” I₁∙I₂ = I_1xI_2x + I_1yI_2y + I_1zI_2z (also known as J-order) is usually the starting point for the SOT. In the opposite case (δv » J_I1I2), the spins are referred to as weakly coupled. Then, I_1ZI_2Z spin order is typically the starting point for SOT after the off-diagonal elements of the density matrix were averaged away during the hydrogenation.⁴⁰ However, I₁∙I₂ order can be preserved in weakly coupled systems by applying sufficiently strong ¹H decoupling,^41,42 or can be encountered in effectively instantaneous reactions as in photo-PHIP experiments,^43,44 where reactions with pH₂ happen within a few microseconds. The exact form of the spin order in the intermediate coupling regime, including the effect from singlet-triplet (S-T) mixing at the hydrogenation catalysts, was described by Bowers, Natterer and others.^39,45–48 Such S-T mixing takes place in high and low fields and is one of the main reasons for reduced hyperpolarization yield.

In so-called PASADENA experiments (pH₂ and synthesis allows dramatically enhanced nuclear alignment), hydrogenation and SOT (or direct proton detection) take place at the same magnetic field.³ Another experimental scheme often applied is referred to as ALTADENA (adiabatic longitudinal transport after dissociation engenders net alignment).⁴⁹ The latter typically features hydrogenation at lower fields (with S-T eigenbasis) followed by an adiabatic (slow) increase of B₀ into the weakly-coupled regime.⁵⁰ This results in population of either the |αβ〉 or the |βα〉 state of the high-field Zeeman eigenbasis, where α is the spin up, and β is the spin down state in the combined spin angular momentum state of both nuclei.

Spin-order transfer

As the literature on SOT may fill an entire review article alone,⁵¹ we are focusing on the parts relevant for the instrumentation only. The available spin order after hydrogenation, the spin system, its eigenstates and its interactions determine the most-effective SOT strategy. While the molecular spin system as such is usually fixed (with workarounds, e.g. PHIP-SAH,²⁰ PHIP-X²⁶), the interactions can be tailored to some extent by varying the (static) external magnetic field (B_SOT) or by applying B₁ fields over a period of time.⁵² Likewise, the spin system can be affected, to some degree, by isotope labeling and reaction parameters such as temperature and pH. Deuteration is a convenient way to reduce relaxation and simplify the spin system, e.g., to an effective 3-spin-½ system in RF-pulsed SOT.^42,53,54

Roughly speaking, a SOT can be achieved by:

• Evolution at one static field (sometimes referred to as “spontaneous” transfer),

• Evolution at different fields – magnetic field cycling (MFC-SOT)^20,24,55,56,

• Evolution plus specific manipulations by RF pulses (RF-SOT)^{9,40,42,51,57–64}.

Understanding the SOT requires profound knowledge of the underlying quantum mechanics. While analytical equations can be derived for simple systems, numerical simulations are usually required to determine the optimal parameters of more complex or realistic systems. These simulations can be implemented in any programming environment capable of matrix algebra. Using or building on existing (open source) packages may be convenient.^65–68

Realizing the different variants of SOT usually requires a magnet (superconducting, resistive, permanent; sometimes shims), sometimes a multi-layer mu-metal shield (to reach fields B_SOT in the nanotesla or microtesla range), and an NMR unit to excite and receive MR signal.

Purification and Quality Assurance

While the addition of pH₂ is performed by catalysis at sometimes harsh conditions, a physiological solution devoid of the catalyst is needed for in vivo administration. Various approaches have been described to achieve this goal, including:

• Filtering homogeneous catalysts,⁶⁹

• Using immobilized catalysts that remain in the polarizer,^70–72

• Heterogeneous catalysts that facilitate filtration,^{18,71,73–78}

• Phase separation and precipitation.^24,79–82

Different agents will generally require individual approaches as their chemical properties vary significantly. A QA module similar to that used for DNP will have to encompass (at least) purity, pH, a low bioburden and temperature; no such device has emerged yet. For all approaches, time is of the essence, as precious signal enhancement is rapidly lost once the agent is hyperpolarized.

Components and capabilities of a polarizer

Orchestrating these steps requires a dedicated unit, often referred to as a “polarizer”. Some have been described, with varying goals, properties, and capabilities, e.g., SOT schemes, pressures, temperatures, in situ detection, automation, dosing, etc.^{13,24,36,38,42,55,83–90} For convenience, such a polarizer may be separated in different units, some of which were described in literature:

• A fluid unit, handling the gases and liquids. It is typically composed of electromagnetic, manual, or pneumatic valves and tubes, borrowed, e.g., from high performance liquid chromatography (HPLC), as well as a custom-made reaction chamber or a (high-pressure) NMR tube. Inert materials, pressure and temperature resistance are important factors, as is the option to clean or sterilize.⁹¹

• An NMR unit, taking care of the SOT either by applying a constant magnetic field, a defined field cycle, or RF pulses. Low-field detection facilitates flip angle calibration and quality assurance.^91–93

• A control unit: a software controlling a digital-to-analog converter (DAC) and an analog-to-digital converter (ADC), e.g., for switching valves, B₀ control, temperature readings, NMR. These needs have been addressed by using the hardware of the NMR / MRI,^87,88,94 by PC-controlled digital-acquisition boards (DAQ) with^85,86 and without^{13,42,55,83,89,95} an additional NMR spectrometer. As a minimum, the control software will have to accommodate easy access to the (sometimes overlapping) timings of each step in the polarization procedure, and may extend to acquiring NMR signal in situ to facilitate calibrations and improve polarization (e.g., Paravision, LabVIEW,^{85–87,92,96} MATLAB^38,95).

REVIEW OF PUBLISHED INSTRUMENTS

This section comprises a brief description of instrumentation to produce pH₂, setups for producing HP solutions, HP of gases, purification methods and translation. The section on setups to produce HP liquids is structured by the magnetic field B_SOT, where the SOT takes place.

Parahydrogen generators

As a manifestation of the generalized Pauli exclusion principle, hydrogen molecules exist as two different nuclear spin isomers: the triplet spin states, called orthohydrogen (oH₂), and the singlet spin state, called pH₂. The energy separation between the ground states of pH₂ and oH₂ is 170.6 K⁸³. Hence, in thermal equilibrium at low temperatures (below 25 K), almost all hydrogen molecules are in the singlet state, i.e nearly pure pH₂. Note that the temperature determines the achievable pH₂-fraction f_pH2.

The production of pH₂ is straightforward: letting hydrogen gas flow over an ortho-para conversion catalyst at low temperatures results in enriched pH₂. Common conversion catalysts are granular materials with high surface area, such as Fe(OH)O,^28,29,97 nickel sulfate,⁹⁸ or chromic oxide (CrO₃) supported on silica gel.⁹⁹ Activated charcoal has been used as an ortho-para conversion catalyst in early works, but the efficiency appears to be lower than for the ferromagnetic oxide materials, especially at high flow rates. The catalytic effect originates from (ferro- or para-) magnetic properties, which induce highly inhomogeneous local magnetic fields that accelerate ortho-para conversion.³⁶

The published pH₂ generator designs offer different properties with respect to cost, pH₂-fraction, pressure, flow rate and ease of use. Aside from commercial products (Bruker BPHG 90, XeUS Technologies LTD, HyperSpin Scientific UG, Advanced Research Systems, IDB Budzylek), many designs for home-built generators have arisen over the last decades. In particular, three primary coolant technologies dominate: single-stage or dual-stage closed-cycle Helium cryostats operating at 13.5 to 40 K,^{29,36,100–104} liquid helium dewars at 14 to 30 K^{98,99,105,106} and liquid nitrogen dewars at 77 K.^{18,97,107–111} Consequently, a dual stage or liquid-helium-based pH₂ generator can practically reach enrichments approaching 100% and a liquid-nitrogen-based pH₂ generator enrichments of up to 52 %, respectively. In a liquid N₂ generator pumped below ambient pressure, the enrichment can be increased by lowering the temperature further, e.g., to 63 K at 21 mbar allowing a pH₂ fraction of ≈ 65 %.¹¹²

The quantification of the pH₂ enrichment can be done optically (e.g., Raman), or by NMR spectroscopy. Raman spectroscopy is about 500 times faster than NMR and does not require a reference sample since it detects both oH₂ and pH₂ directly.¹¹⁰ However, NMR spectroscopy is generally the quantification method of choice since it is already available in the MR labs where PHIP experiments are carried out. The details of quantification have been reviewed thoroughly previously.¹¹³

SOT at tesla fields

SOT conditions.

In PASADENA experiments, the hydrogenation reaction, observation, and SOT take place at the high magnetic field of an NMR or MRI system (≈ Tesla). If pH₂ is added at chemically nonequivalent sites, I₁ and I₂ typically become weakly coupled (Figure 2a). The strongly coupled case is typically only present, if δv ≈ 0.^39,114

Figure 2. PHIP setups for SOT at constant magnetic fields > 1 T,

where spin systems are typically in the weak-coupling regime. Schematic view of the hyperpolarization process at high field, where ethyl acetate is polarized by adding pH₂ (a) and spin order transfer with ESOTHERIC (b) or SLIC. The fluidics are usually handled by a unit consisting of switchable magnetic valves and flow regulation for guiding gases or to apply vacuum to a tube or reactor (c). For MRI systems, dedicated reaction chambers were combined with animal beds allowing animal monitoring, anesthesia, life support and fast administration of the HP contrast agent (d). As an example, sub-second in vivo ¹³C angiography of a mouse was demonstrated only seconds after the polarization (e), while compatible temperature, pressure and sterility was assured. Figure 2a reproduced from Pulsed Magnetic Resonance to Signal-Enhance Metabolites within Seconds by Utilizing Para-Hydrogen, Korchak, S.; Yang, S.; Mamone, S.; Glöggler, S., ChemistryOpen, Vol. 7, Issue 5 (ref 62). Copyright 2018 Wiley. Figure 2b reproduced from Over 50 % 1H and 13C Polarization for Generating Hyperpolarized Metabolites—A Para-Hydrogen Approach, Korchak, S.; Mamone, S.; Glöggler, S. ChemistryOpen, Vol. 7, Issue 9 (ref 127). Copyright 2018 Wiley. Figure 2c reprinted from J. Magn. Reson., Vol. 285 (Supplement C), Kiryutin, A. S.; Sauer, G.; Hadjiali, S.; Yurkovskaya, A. V.; Breitzke, H.; Buntkowsky, G. A Highly Versatile Automatized Setup for Quantitative Measurements of PHIP Enhancements, pp. 26 – 36 (ref 94). Copyright 2017, with permission from Elsevier. Figures 2d and 2e reproduced from In Vivo 13C-MRI Using SAMBADENA, Schmidt, A. B.; Berner, S.; Braig, M.; Zimmermann, M.; Hennig, J.; Elverfeldt, D. von; Hövener, J.-B. PLOS ONE, Vol. 13, Issue 7 (ref 90). Copyright 2018 Public Library of Science.

For the former case, PHIP ¹H signal can be observed as anti-phase peaks by a simple 45° excitation. Alternatively, spin order can be converted into in-phase magnetization using RF pulse sequences^115–117 which also enables detection in inhomogeneous B₀ fields.^118,119 However, the first in-phase PHIP ¹H spectra were obtained by Pravica et al. by field cycling (ALTADENA)⁴⁹ – an approach that has also enabled polarization transfer to ¹⁹F.⁵⁰

Several techniques have been published to convert I_1zI_2z into observable magnetization of X-nuclei (often ¹³C). Here, most RF-SOT sequences are adaptations of the insensitive nuclei enhanced by polarization transfer (INEPT) sequence that considers initial I_1zI_2z spin order, Figure 2b.^59,120 After the early work from Haake, Natterer and Bargon, who introduced the pH₂ INEPT+ (phINEPT+) sequence,⁶⁰ l-PHINEPT+,⁵¹ selective excitation of polarization using PASADENA (SEPP) INEPT,¹²¹ selective-90 (s90) phINEPT^54,122, and efficient SOT to heteronuclei via relayed INEPT chains (ESOTHERIC, Fig. 2b)^62,82 sequences were suggested. Note that depending on the spin system, and neglecting relaxation, all RF-SOT schemes can yield ~100 % ¹³C-polarization, in principle, except phINEPT+, which has a theoretical maximum of ~50 % because of the initial 45° ¹H pulse.^38,123 If the initial spin order is I₁∙I₂, different SOT schemes are required, e.g. Goldman’s sequence⁴² for succinate obtained by reacting a fumarate precursor molecule.^39,48 Additional RF-SOT techniques for strongly-coupled protons will be introduced in the next section (SOT at millitesla fields). Adiabatic passages through level anti crossings (also referred to as avoided crossings) were shown to be efficient for the transfer of singlet pH₂ spin order to ¹³C polarization.^124–126

Published setups:

Various setups for performing PHIP in NMR^94,122 and MRI systems have been described.^38,90 Such implementations offer the advantage that parts of the sophisticated MR machines, including NMR excitation and acquisition and highly homogeneous field over a large volume, can be employed in the hyperpolarization process. Typically, the NMR or MRI system offers TTL outputs that allow control of the fluid path directly from within the pulse program.

To perform the hyperpolarization in an NMR system, Kiryutin et al. introduced a setup consisting of several valves for supplying gases or liquids, e.g., a cleaving agent to convert a precursor into the desired metabolite.⁹⁴ Using standard 5 mm or 10 mm NMR tubes for hydrogenation, SOT, detection, and purification, the system can be implemented in all standard and benchtop NMR spectrometers and has been done so by many labs (Figure 2c).^{62,119,122,127,128} For in vivo imaging, the HP contrast agent has to be transferred to an imaging system.

Performing hyperpolarization within the bore of an MRI system allows in vivo imaging only seconds after the hydrogenation (SAMBADENA). Here, the hydrogenation takes place in a reactor enclosed in a dual tune, transmit-receive ¹H/¹³C volume coil, equipped with a local receive coil, for RF-SOT and in vivo MRI.³⁸ The reactor was mounted on the animal bed (with animal warming, vital sign monitoring, anesthesia) such that the contrast agent was delivered directly into a syringe for injection at high field, Figure 2e.⁹⁰ In vivo imaging of a sterile solution of hydroxyethyl [1-¹³C]propionate-d₃ within 15 s after hyperpolarization was demonstrated (Figure 2f), although no purification (i.e. catalyst filtering) was performed.⁹⁰

Published agents:

High ¹³C polarizations for highly concentrated molecules have been demonstrated with phINEPT+ and ESOTHERIC sequences and SAMBADENA.^38,119,127 Promising PHIP agents that were produced at high field with polarization above 10% are tetrafluoropropyl [1-¹³C]propionate-d₃ (TFPP),¹²⁹ [1-¹³C]succinate-d₂,^39,48 and [2-¹³C]pyruvate.¹³⁰ The latter was achieved using PHIP-SAH, ESOTHERIC and subsequent cleavage of the side arm of cinnamyl [2-¹³C]pyruvate. Remarkable polarizations were also achieved for other promising PHIP-SAH molecules, namely ethyl [1-¹³C]acetate^119,127 and cinnamyl [1-¹³C]pyruvate.^62,82

Challenges:

The main challenges of in situ polarizer setups such as SAMBADENA are due to the limited space within the bore of the MRI or NMR magnet: accommodating production, purification, QA, and the administration within a small volume is not an easy task. RF-SOT sequences - theoretically - achieve ≈ 100 % ¹³C-polarization, but only in fully ²H-labeled molecules as additional J-couplings will interfere (neglecting relaxation).^{38,54,122,127,128} In NMR spectrometers, radiation damping can disturb the SOT, while the limited RF power is a serious concern in MRI setups^{119,131–133} because of the large excitation bandwidth needed.¹³⁴ Moreover, translational motion of the molecules in the inhomogeneous field of the relatively large reaction chamber in an MRI during SOT can reduce the polarization.¹³⁵

SOT at millitesla fields

SOT conditions:

The characteristic feature of SOT in the millitesla range is the strongly-coupled regime (δv « J_I1I2). Typically, the pairwise pH₂ addition takes place under proton decoupling at B_SOT, in part to reduce relaxation, but also to preserve the I₁∙I₂ spin order (Figure 3f and Figure 3h). After the decoupling, RF-SOT sequences are used to transfer the spin order to a ¹³C nucleus. A wide range of efficient pulse sequences have been reported^51,57,58 to accomplish polarization transfer in millitesla field range including the pioneering sequence developed by Goldman and co-workers (Figure 3e).^42,83

Figure 3. Setups and examples for PHIP at millitesla fields:

Overall schematic (a), graphical user interface (b) and rendering (c) of a 5.75 mT automated, pre-clinical, open-source PHIP polarizer. A dual-channel, ¹H/¹³C RF coil was placed inside a B₀ solenoid electromagnet (c, left), and a spray-injection reactor was nested inside the coil. In situ NMR detection at ~62 kHz enabled RF calibration and quantification of ¹³C polarization with respect to thermally polarized water (d). During the hydrogenation reaction of a few seconds, ¹H-decoupling was applied to preserve I·S spin order, and before the polarization was transferred to ¹³C using the sequence proposed by Goldman et al. (e) – note, that there are only three ‘effective’ pulses (white) used while the others (gray) are for refocusing of the Zeeman evolution only. Reaction scheme (f) and in vivo ¹³C-NMR (g) of contrast agent SUX detected at low magnetic fields (~0.048 T). Reaction scheme (h) and in vivo ¹³C MRI imaging (i) of contrast agent PLAC at 4.7 T: ¹³C-MRI (color) was acquired approximately 5 – 10 s after the injection of ~0.2 mL, ~30 mM hyperpolarized PLAC into the tail vein of a healthy nude mouse (prior tumor implantation) and overlaid on representative ¹H images (gray scale, ¹³C: GRE, 3 × 3 mm² in-plane resolution, 6 mm slice thickness, FOV = 96 × 96 mm²). Figures 3a–d, 3h and 3i reproduced from Coffey, A. M.; Shchepin, R. V.; Truong, M. L.; Wilkens, K.; Pham, W.; Chekmenev, E. Y. Open-Source Automated Parahydrogen Hyperpolarizer for Molecular Imaging Using 13C Metabolic Contrast Agents. Anal. Chem. 2016, 88, 8279–8288 (ref. 86). Copyright 2016 American Chemical Society. Copyright 2006, with permission from Elsevier. Figures 3f and 3g reprinted from J. Magn. Reson., Vol. 281 (Supplement C), Coffey, A. M.; Feldman, M. A.; Shchepin, R. V.; Barskiy, D. A.; Truong, M. L.; Pham, W.; Chekmenev, E. Y. High-Resolution Hyperpolarized in Vivo Metabolic 13C Spectroscopy at Low Magnetic Field (48.7mT) Following Murine Tail-Vein Injection, pp. 246 – 252 (ref 147). Copyright 2017, with permission from Elsevier.

Published setups:

As the (¹³C) hyperpolarization can persist for several minutes, the contrast agent can be transferred from the polarizer to the detector without overwhelming loss, e.g., for polarimetry (i.e., measuring the degree of induced polarization) or ultimate application (i.e., in vivo imaging). As a result, several millitesla setups with the main static magnetic field ranging from 1.7 mT^13,89 to 50 mT were set up and reported.⁸⁵

The use of the low magnetic field offers certain advantages: (1) the field can be generated by inexpensive electromagnets, resulting in an overall cost-efficient, portable setup;⁹² (2) susceptibility effects are low, and B₀ inhomogeneities can be compensated for by magnet design,⁹² shims, and sufficiently strong RF pulses. Utilizing an electromagnet has allowed SOT over large reaction volumes – up to 100 mL were reported¹³ – allowing the production of clinical-scale doses of ¹³C-hyperpolarized contrast agents.⁸³

These translational advantages were likely decisive for Amersham Biosciences, a healthcare company, to choose this approach for the first commercial prototype – in vivo feasibility studies using the Amersham polarizer have been reported extensively.^42,69,83,136 The overall design of the Amersham polarizer was never reported in detail,^42,83,84,137 but several closely related PHIP millitesla polarizer designs were subsequently reported.^{13,36,85–87,89,91–93,96}

As described above, the lifetime of the pH₂-derived ¹H spin order is of the order of seconds and hence, for effective ¹³C HP, the hydrogenation needs to be completed quickly.^36,138 In practice, fast hydrogenation is achieved through the use of a reactor pressurized to ca. 10 bar with pH₂ followed by a spray injection of a hot stream of precursor solution into the chamber.^13,89 For biomedical applications, the reaction is performed in aqueous media employing water-soluble PHIP catalyst at elevated temperature (70–95 °C).^{13,36,86,89,137} As a result, the entire bolus of the precursor molecule can be reacted quickly (in a few seconds), i.e., on a time scale that is faster or similar to the decay of the I₁∙I₂ spin order.

The experimental hardware for the hyperpolarization process is relatively similar among reported millitesla PHIP polarizers (Fig. 1). By definition, such system employs a B₀ magnet operating in the millitesla range with typical field strength of 2 – 9 mT^13,36,86,89 although the use of higher field (48 mT) permanent magnets have been reported.⁸⁵ A large-volume (ca. 300 mL) dual-channel RF probe is placed inside the magnet to deliver ¹H and ¹³C RF pulses for the SOT sequence. The high-pressure reactor is nested inside the dual-channel RF coil – Figure 3c shows an example of such an electromagnet-based polarizer. A millitesla-PHIP polarizer is typically connected to (or contains) cylinders of compressed ultrahigh-purity (>99.999% or 5.0) pH₂ and inert propellant gas (N₂ or Ar). A series of high-pressure valves and tubes form a manifold (Fig. 3a) to fill the reaction chamber with pH₂.¹³⁹ This step is followed by the injection of hot precursor solution into the chamber through a nozzle using an inert propellant gas. Various setups were designed to inject a defined amount of the precursor solution into the reaction chamber, and to eject the HP contrast agent into a receiver ready for transfer and in vivo MRI. Although specialized heaters were employed to control the reaction temperature, the design shown in Figure 3c employs heating generated by a >100 W electromagnet.

Several approaches were developed to orchestrate the interplay of actuating multiple valves and playing out the RF-SOT, usually including software and a hardware interface (LabVIEW-,^13,36,95,96 Arduino-,⁸⁶ Matlab,⁹⁵ and NMR spectrometer based⁸⁷). A typical graphical user interface (GUI) allows setting various parameters (reaction time, precursor dose), choosing the SOT sequence, actuating individual valves, and executing various polarization or maintenance routines and sometimes performing NMR (Figure 3b, open-source Arduino-based controller software).⁸⁶ All published controller designs have their own merits.

To achieve high and reliable polarization through robust device operation, the accurate and precise application of the RF-SOT is critical.⁵¹ Optimal RF-SOT performance requires calibrated B₁ power and transmission resonance frequencies for both RF channels in particular. Performing these calibrations in situ by detecting NMR signal in the reaction chamber using a transmit-receive polarizer design drastically facilitates these procedures. In situ calibrations, however, are not necessarily needed to produce the HP agents, because an external (i.e., ex situ) NMR spectrometer, MRI scanner or a low-field polarimetry station may be employed for the calibrations,⁸⁹ and for probing the achieved ¹³C polarization level.⁶⁹ This so-called transmit-only design is less complex and typically result in a lower device cost as no extra NMR receive hardware is needed. The transmit RF pulses can be generated using simple waveform generators (e.g., NI, Austin, TX, USA), consumer-grade RF amplifiers (WRAT, Onkyo, Osaka, Japan), and untuned RF coils with low Q factor, mitigating radiation damping issues noted in the above section.

The hyperpolarizers employing transmit-receive design were realized by using a commercial, low-frequency, dual-channel NMR spectrometer and dual tune transmit-receive coil (e.g., Kea2, Magritek).^86,87 In another approach, a geometrically decoupled, single or dual tuned receive coil was added to the untuned transmit coil, using the same low-cost waveform generators, amplifiers and ADC/DAC hardware as for the transmit-only design described above.^92,95

These transmit-receive designs allow using the signal of thermally polarized water to calibrate B₀ and B₁.^{86,87,92,95,140} As the polarization of HP samples is independent of the detection field, a high signal-to-noise ratio (SNR) spectrum was readily obtained in in-situ low-field polarimetry as shown for the HP contrast agent [1-¹³C]phospholactate-d₂ (PLAC) at ~62 kHz (Figure 3d).

While transmit-receive designs are more complex and can be more expensive (by ca. $30,000 for the Kea2), calibration is relatively straightforward, and good reliability and reproducibility of the hyperpolarization yield is achieved.

The efficiency of SOT can be improved further by reducing the complexity of the spin system. For example, deuteration was used to simplify HEP, SUC and PLAC to an effective 3-spin-1/2 system (two nascent protons and one ¹³C nucleus), which increased the final ¹³C polarization for PLAC to 15% versus 1% for the fully protonated variant, Figure 3h.^53,141 While deuterium (and phosphorus-31) nuclei possess a spin, they are not excited by the RF and hence not effectively involved in the SOT.⁴² In favorable cases, P_13C of more than 20 % was achieved with millitesla polarizers,^36,42,137 e.g. P_13C = 28% for [1-¹³C]succinate-d₂ (SUX, Figure 3f,). Despite the enormous signal enhancement provided by hyperpolarization, metabolic imaging with HP contrast agents usually results in low SNR images. Thus, it is not surprising that only deuterated precursors with high P_13C were translated to in vivo studies using millitesla PHIP polarizers so far.^{13,69,83,86,89,142–148}

To date, millitesla-polarizers were employed for in vivo ¹³C MRI or MRS with SUX^147,149, SUX esters¹⁴⁵, HEP,^5,69,150 and TFPP,¹⁴³ and PLAC⁸⁶, Figures 3g,i. Thorough reviews covering the in vivo applications can be found elsewhere.^113,151

Two interesting developments, which broaden the scope of polarizable molecules drastically, are PHIP-SAH²⁰ and PHIP-X.²⁶ The experimental realizations of these techniques will be described in more detail below, but the millitesla polarizers described here appear to be well positioned for these emerging protocols.

Challenges:

The millitesla PHIP polarizers have been shown to be successful devices for the efficient production of HP ¹³C contrast agents in aqueous media with high P_13C exceeding 25% for in vivo applications. Using PHIP-SAH molecules may dramatically enlarge the pool of agents, e.g., to ¹³C-labeled pyruvate. Despite this success, however, all in vivo translated precursors so far require deuterated substrates in addition to ¹³C labeling. While deuteration offers the benefit of a prolonged polarization lifetime, it also increases the cost and the complexity of synthesis of the precursors, which is a clear drawback compared to MFC-SOT. No millitesla polarizer incorporating a purification unit was presented so far.

SOT at ultralow fields (ULF)

SOT conditions:

At ultra-low fields (micro- or nanotesla), the frequency differences between heteronuclei and protons (γ_1H-γ_X) are reduced to be in the order of the J-couplings or below. At such low fields, a spin bath between all spin-spin coupled nuclei is effectively established through which polarization can easily propagate spontaneously. In other words, if a part of a nuclear spin network is initialized in the singlet state via pH₂, and the Zeeman interactions are negligible compared to the J-couplings, then the polarization spreads through the network.^152,153 The beauty of this “spontaneous” polarization transfer approach is that it is general and does not require specialized pulse sequences that are highly dependent on the specific spin system under study. Note that this concept was realized early on, when the so-called ALTADENA approach was established,⁴⁹ where pH₂ was introduced into the spin system through hydrogenation and polarization “flows” to other ¹H spins at millitesla fields or below,¹⁵⁴ as mentioned in the section above and is theoretically analysed in the following section. However, to polarize a spin-1/2 heteronucleus, such as ¹³C, ¹⁵N, ³¹P, etc., ultra-low microtesla fields are required to strongly couple the protons of the spin system with the target X-nucleus to allow spontaneous polarization transfer. Over the last decade, a wide variety of experiments have been demonstrated taking advantage of this heteronuclear spin bath at microtesla fields, including hydrogenative^{20,124,155–160} and non-hydrogenative PHIP.^161–164 Newest approaches have used pulse sequences at microtesla fields,^{125,165–167} which can focus polarization transfer in a more targeted way to specific nuclei, but become highly dependent on the specific spin system under study.

Published setups:

While dedicated setups to produce diagnostically relevant contrast agent have not yet emerged, some interesting setups were described that allow exploiting the unique properties at these fields. Among these are (a) the distribution of the polarization across an entire molecule and different coherences,¹⁰⁶ (b), the unique sensors to detect signals in the Hz – kHz range,^168–170 and (c) the identification of molecules by their J-couplings rather than their chemical shift.^171–173

To establish magnetic fields below the Earth’s field, entering nanotesla to microtesla field regimes, typically mu-metal shields are used.^168,174 Inside such shielding, well-defined magnetic fields are commonly generated with conventional resistive coils in Helmholtz, solenoid or other configurations driven by low noise current sources.

To acquire HP spectra (typically from small molecules in solution), a variety of ULF MR setups have been designed, Figure 4. The detection sensitivity for conventional RF coils, as used for high field MR experiments, decreases with frequency.¹⁷⁵ Therefore, at ultralow fields different magnetic field detectors such as atomic magnetometers (Figure 4a),^{152,153,169,174,176,177} or superconducting quantum interference devices (SQUIDs) (Figure 4b)^168,178 may be more sensitive than inductive detection. Such sensors are even sensitive enough to perform magnetoencephalography,¹⁷⁹ but can also be used to detect nuclear spins. They can operate as broad band detectors (SQUIDs: DC – ≈1 GHz, atomic magnetometers: DC – ≈1 kHz), where no matching and tuning is required and the MR signal of different nuclei such as ¹³C, ¹⁵N, ¹⁹F and ¹H can be detected simultaneously.¹⁰⁶ More recently, fluorescent nitrogen vacancy (NV) centers in diamonds (Figure 4c) have also been used to detect PHIP signal. While the NV centers typically have less absolute magnetic field sensitivity, the advantage is that they can be brought in direct contact with the solution (unlike SQUIDS or atomic magnetometers).^170,180 These can also be used for magnetometry at zero field.¹⁸¹ The linewidth of the detected MR signals for ULF setups can be less than 1 Hz down to tens of mHz,¹⁷¹ so that J-couplings can be easily resolved. Narrow linewidths are obtained since field inhomogeneities decrease with the field strength, which enables chemically specific, high resolution J-coupling spectroscopy.

Figure 4. Setups and examples suitable for PHIP at ULF.

(a) atomic magnetometer system based on a Rb vapor cell. (b) PHIP enhanced zero field NMR spectrum of dimethyl maleate acquired with a Rb vapor magnetometer. (c) SQUID based system. (d) optically probed nitrogen-vacancy (NV)-NMR spectrometer. Figure 4a reproduced from Towards Large-Scale Steady-State Enhanced Nuclear Magnetization with in Situ Detection, Blanchard, J. W.; Ripka, B.; Suslick, B. A.; Gelevski, D.; Wu, T.; Münnemann, K.; Barskiy, D. A.; Budker, D.Magn. Reson. Chem., Vol. -, Issue - (ref 169). Copyright 2021 Wiley. Figure 4b reprinted from Parahydrogen-Induced Polarization at Zero Magnetic Field, Butler, M. C.; Kervern, G.; Theis, T.; Ledbetter, M. P.; Ganssle, P. J.; Blanchard, J. W.; Budker, D.; Pines, A., J. Chem. Phys., 2013, Vol. 138, Issue 23 (ref 152), with permission of AIP Publishing. Figure 4c reprinted from Mutual Benefit Achieved by Combining Ultralow-Field Magnetic Resonance and Hyperpolarizing Techniques,Buckenmaier, K.; Rudolph, M.; Fehling, P.; Steffen, T.; Back, C.; Bernard, R.; Pohmann, R.; Bernarding, J.; Kleiner, R.; Koelle, D.; Plaumann, M.; Scheffler, K., J. Rev. Sci. Instrum., 2018, Vol. 89, Issue 12 (ref 168), with permission of AIP Publishing. Figure 4d reproduced from Micron-Scale NV-NMR Spectroscopy with Signal Amplification by Reversible Exchange, Arunkumar, N.; Bucher, D. B.; Turner, M. J.; TomHon, P.; Glenn, D.; Lehmkuhl, S.; Lukin, M. D.; Park, H.; Rosen, M. S.; Theis, T.; Walsworth, R. L., PRX Quantum, Vol. 2, Issue 1 (ref 170). Copyright 2021 American Physical Society.

Challenges:

Generating and detecting HP signals at ULF offers unique insights into the investigated spin systems and provides NMR and MRI without requiring superconducting magnets. However, a disadvantage of all ULF detection methods remains, which is the frequency-dependent noise of the sensors (SQUIDs ≈ 1 fT/Hz^−1/2, atomic magnetometers ≈ 10 fT/Hz^−1/2). Accordingly, there is an ongoing quest to decrease the noise level below the sample noise. Moreover, medically viable polarizers or sensors that take advantage of PHIP at ULF have not emerged yet. The aspiration remains that the described discoveries and techniques will soon be translated into broad application.

Magnetic field cycling (MFC)

SOT conditions:

MFC is another method that is widely used for the polarization transfer from I₁∙I₂ spin order to heteronuclei. In contrast to the previous sections, which dealt with SOT at a constant magnetic field B₀, MFC exploits varying the magnetic field from values close to geomagnetic field (30–50 μT) to nearly zero field, and a few T. MFC-SOT was the first method to produce net ¹H magnetization⁴⁹ and used to produce the first ¹³C in vivo MRI in 2001.⁵ The method is not to be confused with the fast field cycling relaxometry where the magnetic field is varied to investigate relaxation or spin-state evolution at a low magnetic field in order to gain information about physical or chemical properties of the system under study.¹⁸² For PHIP, a MFC process does not explicitly require in situ signal detection, although some setups do this.⁸⁸

To give an idea of the effect of MFC on the spin state populations in PHIP-polarized molecules, we consider a three-spin system formed by two pH₂ protons (H and H’) and a heteronucleus (X). The spin states can be conveniently described using the singlet-triplet-Zeeman basis. The eight basis states are:

$$|T_{+}\alpha \rangle =|\alpha \alpha \alpha \rangle ,|T_{+}\beta \rangle =|\alpha \alpha \beta \rangle ,$$

$$|T_0\alpha \rangle =\frac{1}{\sqrt{2}}|\left(\alpha \beta +\beta \alpha \right)\alpha \rangle ,|T_0\beta \rangle =\frac{1}{\sqrt{2}}|\left(\alpha \beta +\beta \alpha \right)\beta \rangle ,$$

$$|S_0\alpha \rangle =\frac{1}{\sqrt{2}}|\left(\alpha \beta -\beta \alpha \right)\alpha \rangle ,|S_0\beta \rangle =\frac{1}{\sqrt{2}}|\left(\alpha \beta -\beta \alpha \right)\beta \rangle ,$$

$$|T_{-}\alpha \rangle =|\beta \beta \alpha \rangle ,|T_{-}\beta \rangle =|\beta \beta \beta \rangle ,$$

where the first letter refers to the proton spin state, and the second refers to the heteronuclear spin state.

Upon pH₂ addition to a precursor at geomagnetic fields the states |S₀α〉 and |S₀β〉 are populated almost equally. At these “higher” fields, the ¹H and heteronuclei are weakly coupled and no heteronuclear magnetization is obtained, because to a good approximation only the |T₀α〉-|S₀α〉 states and the |T₀β〉-|S₀β〉 states are mixed. However, when the magnetic field is low enough such that the difference in proton and heteronuclear Larmor frequencies approximately matches the J-coupling frequencies, the states with equal angular momentum projection along the field axis (z): |T₊β〉, |T₀α〉, |S₀α〉, and |T₋α〉, |T₀β〉, |S₀β〉 are mixed. This leads to level anticrossings (LACs) between the relevant states, and by varying the magnetic field either rapidly (diabatic passage) or slowly (adiabatic passage), it is possible to transfer spin state populations between the eigenstates.¹⁸³ The spin state energies and populations are shown in Figure 5a, and the relevant Hamiltonians are reported in ref.¹⁵⁶.

Figure 5. Diagrams and Setups for PHIP by MFC.

(a) Energies of the eigenstates of [1-¹³C]fumarate, plotted as a function of magnetic field strength, with states relevant to MFC-SOT highlighted in black (top). In an ideal MFC, the pH₂-derived singlet order I₁·I₂ is converted into heteronuclear magnetization by first diabatic passage to near-zero-field, followed by adiabatic passage back to microtesla field (bottom). The diabatic passage preserves the state of the system with the protons in a singlet state (i.e., |S₀α〉 and |S₀β〉), but during the subsequent adiabatic return to high field the state |S₀α〉 evolves into |T₊β〉, meaning some degree of proton singlet order is lost, but heteronuclear polarization in the β state is gained. (b) Illustration of a polarizer where a sample is polarized by MFC or field sweeping and transferred to an NMR system for detection. Figure reprinted from Polarization Transfer via Field Sweeping in Parahydrogen-Enhanced Nuclear Magnetic Resonance, Eills, J.; Blanchard, J. W.; Wu, T.; Bengs, C.; Hollenbach, J.; Budker, D.; Levitt, M. H., J. Chem. Phys., 2019, Vol. 150, Issue 17 (ref 156), with permission of AIP Publishing.

In the most common example of MFC, the hydrogenation reaction using pH₂ is carried out in the laboratory field (usually tens of microtesla), which is high enough to prevent leakage of proton spin order onto heteronuclear spins, but low enough that proton chemical shift differences, which would lead to loss of spin order, are suppressed, i.e., the protons remain strongly coupled. The field is then diabatically (rapidly) reduced to near-zero field, and then adiabatically (slowly) increased to the tesla range, such that the spin state populations follow the eigenstates as the field passes through the LACs. Overall, this process leads to I₁∙I₂ order being transformed into both proton and heteronuclear magnetization. For biomedical applications, it is the heteronuclear Zeeman order that is of interest, although we note that the protons are also polarized by this process, Figure 5a.

Published setups:

The most basic MFC experiment is a hydrogenation in a few millitesla followed by transfer to a few tesla for detection.⁴⁹ To transfer polarization to X-nuclei, a mu-metal chamber can be used to reach near-zero field conditions (nanotesla). The two passages can be achieved manually, by dropping the sample after hydrogenation into the shield (diabatic passage) and lifting it out slowly (adiabatic passage).⁵ In more sophisticated setups, the speed of passages is controlled by means of an electromagnet inside the mu-metal shield.^157,184 The use of coils allows more-precise control over the magnetic fields and has become a routine approach for MFC. A schematic of an experimental apparatus used for MFC is shown in Figure 5b.

The approach was first introduced experimentally by K. Golman et al. in 2001.^5,55 After pairwise addition of pH₂, the sample was (rapidly) dropped into a three-layer mu-metal shield at < 100 nT and then slowly lifted at approximately 10 cm/s rate. Using this method, the authors generated 4 % ¹³C polarization on [1-¹³C]maleic acid dimethyl ester. A more thorough theoretical explanation of the technique and description of the setup followed,^55,137 now including coils inside the magnetic shield to provide time-dependent variable fields. The magnetic field was initially held at 100 μT as the hydrogenated sample was inserted into the shield, then diabatically reduced to 30 nT in about 1 ms, and then exponentially ramped back to 100 μT on the order of seconds. This approach, in combination with a spray-injection chamber for the hydrogenation with pH₂, led to ¹³C polarization on hydroxyethyl-propionate of ≈21 %. In later work, this experimental apparatus was used to produce HP ¹³C contrast agents for coronary angiography imaging in pigs.¹⁵⁰

There is merit to the approach of shuttling samples in and out of a magnetic shield by hand in the simplicity and low experimental requirements, and this approach was employed to hyperpolarize different heteronuclei such as ¹³C,^{185 15}N,¹⁸⁶ and ¹²⁹Si.¹⁸⁷ A few years later, this approach was employed to hyperpolarize the ¹³C spins in a number of molecules, including pyruvate and acetate, by means of PHIP-SAH.^20,56 A detailed description of a coil-based magnetic field cycling setup was provided by Shchepin et al. where they studied the dependence of ¹³C polarization in [1-¹³C]ethyl acetate on the minimum field used during the MFC.¹⁸⁴ They demonstrated that the MFC should reach fields below 1 μT for efficient ¹³C polarization. Using this optimized coil-based approach, the ¹³C polarization of [1-¹³C]pyruvate generated via side-arm hydrogenation was improved from 2.3 %^20,56 to 8.3 %,¹⁵⁷ and the method was also applied for hyperpolarizing [1-¹³C]fumarate (FUM).^24,188

In FUM, derived from hydrogenation of ADC (acetylene dicarboxylic acid), an AA’X spin system is formed.¹⁸⁹ Because of its simplicity, the system is useful for studying and optimizing MFC methods. It was on this molecular system that a MFC variant known as magnetic field sweeping was tested.¹⁵⁶ In this method, rather than a diabatic field reduction followed by an adiabatic increase, the magnetic field is inverted adiabatically, i.e., passing through zero field. The field sweeping method was then applied to hyperpolarize vinyl-acetate, and the effect of the magnetic field sweep step size and rate was investigated.⁸⁸ The authors found that for MFCs using a field range of a few microtesla, a sweep rate of 50 steps/μT/s was sufficient, with higher values leading to no improvement. Magnetic field sweeping does not need rapid field changes, and hence seems more amenable for application to liquids under continuous flow. Its lower efficiency with respect to field cycling,¹⁵⁶ is likely due to the deleterious effects of transverse magnetic fields when passing through the zero-field point, and hence in future work an MFC approach was used for generating HP FUM.^24,188

This molecular system was also used to study the benefits of constant-adiabaticity methods for MFC and field sweeping.¹⁹⁰ Under the constant-adiabaticity constraint, the magnetic field variations are slow at the LAC field, but the field can be varied more rapidly away from this key feature. This is particularly useful for spin systems that relax rapidly during the MFC process.

Challenges:

More complicated molecular systems can present some pitfall to the application of MFC, in particular when quadrupolar nuclei such as ²H¹⁹¹ or ¹⁴N¹⁸⁶ are J-coupled to the pH₂–derived protons. The strong-coupling condition that is reached at nearly zero field brings these heteronuclei in contact with the spin order of pH₂ and quadrupolar relaxation can work as a hyperpolarization sink.

Another caveat to be considered is that exposure to magnetic fields magnetizes the mu-metal, and so the magnetic shielding should be periodically degaussed to ensure the near-zero field condition is met inside it. In cases where the magnetic shield is near a high-field magnet, it is common to use active shimming with coils inside the shield to achieve the near-zero field condition.

PHIP OF GASES

While low sensitivity of NMR often is an issue, it is particularly severe for gases as their densities at ambient pressure are ca. 1000-fold lower compared to liquids. PHIP has been applied to produce HP molecules in the gas phase^15,16,192 - mostly gases, but also vapors of volatile liquids and even solids. One approach is to use a solution of a metal complex in a homogeneous hydrogenation with subsequent transfer of the target species to gas phase. For instance, aqueous stoichiometric hydrogenation of norbornadiene resulted in the release of water-insoluble hyperpolarized norbornene to the gas phase.¹⁹³ Bubbling of a mixture of pH₂ with propylene (propyne, etc.) through a solution of a rhodium or iridium complex was used to produce polarized gases.¹⁹⁴ Another approach relies on the use of solid catalysts to produce PHIP in heterogeneous hydrogenations (HET-PHIP). By bubbling pH₂ through a suitable volatile liquid, pH₂ is saturated with its vapor, and this gaseous mixture is then supplied to a cell with a solid catalyst for hydrogenation. Often, unsaturated gases premixed with pH₂ are used in HET-PHIP experiments.

Published setups:

Published examples include hydrogenation of vinyl acetate vapors over a Rh/TiO₂ catalyst with subsequent dissolution and hydrolysis of ethyl acetate to hyperpolarized ethanol and acetate,¹⁹⁵ and hydrogenation of vinylethyl ether to hyperpolarized diethyl ether, a known inhalable anesthetic.¹⁹⁶ Unsaturated gases are simply premixed with pH₂ and supplied to a catalytic reactor to yield a continuous stream of hyperpolarized gas.^{15,16,192,197}

The experiments with gases can be performed under PASADENA or ALTADENA condition. For propane (H₃C-CH₂-CH₃) produced upon hydrogenation of propylene (H₂C=CH-CH₃), in PASADENA experiments, a pair of enhanced antiphase multiplets is observed with an admixture of in-phase contributions of opposite sign for the two signals. In ALTADENA experiments with hydrogenation at the Earth’s field, where all ¹H spins in a product molecule are strongly coupled, transfer of the gas to the NMR probe results in the observation of polarization for all coupled ¹H nuclei. For propyne hydrogenation to propylene, it is possible to estimate the stereoselectivity of the hydrogenation by fitting the experimental spectra to ALTADENA numerical simulations, which would be otherwise impossible for this reaction.¹⁹⁸ The adiabatic condition of low-to-high field transfer is seldom met fully with gases as T₁ times are short, and slow transfer leads to major polarization losses. Short relaxation times in the gas phase are due to spin-rotation interaction, which also makes the T₁ of heteronuclei shorter than for protons. As a result, polarization transfer from ¹H to heteronuclei is impractical even though its feasibility has been demonstrated.¹⁹⁹

PASADENA experiments with gases are particularly easy to implement – the substrate gas and pH₂ (and, if required, a diluent gas, e.g., N₂) can be premixed in a gas cylinder or supplied by combining the outputs of mass flow controllers, and the hydrogenation reaction performed in an NMR tube containing a solid catalyst, its suspension in a liquid, or a solution of a suitable transition metal complex. ALTADENA experiments provide a much broader flexibility in experimental conditions, Figure 6a. A reactor can be as simple as a temperature-controlled section of stainless steel or copper tubing or a quartz U-tube containing solid catalyst powder which is held in place by glass wool plugs. It is important to monitor the reactor bed temperature as hydrogenation is exothermic and run-away heating effects are possible. The gas is supplied via gas lines from a gas cylinder to the reactor and then from the reactor to the NMR tube and eventually towards the exhaust; all connections should be gas-tight to avoid gas leaks, and all components should withstand the required gas pressures. Commercially available flow NMR probes, popularized with the advent of liquid chromatography (LC) NMR, are well suited for acquiring the spectra on aliquots of the flowing hyperpolarized gases. However, gas flow rates through the smaller coil volumes of these probes need to be reduced accordingly to avoid residence-time line-broadening. Such effects can also be averted with an interrupted-flow system where the gas flow is allowed to bypass the probe during NMR acquisition while not perturbing the steady-state of the reactor bed. This is also a useful way to isolate a gas sample in the probe to acquire thermally equilibrated spectra. A provision for heating the reactor and catalyst pretreatment in a stream of H₂ or gas mixtures is useful. A somewhat more sophisticated setup was designed²⁰⁰ for a controlled clinical-scale (>300 mL in 2 s) batch production of HP propane (Figure 6c), which makes provisions for efficient preheating of reactants and subsequent dissipation of heat produced in the highly exothermic hydrogenation reaction, as well as for operation at elevated pressures (~ 8 bar).

Figure 6. Setups for PHIP of gases.

a) Experimental setup employed for producing hyperpolarized gases or vapors via heterogeneous hydrogenation of unsaturated precursors premixed with pH₂. b) Schematics of an experimental setup (left) for biphasic hydrogenation of gases upon bubbling their mixture with pH₂ through a homogeneous solution of metal complex catalyst, with the hyperpolarized product continuously escaping into the gas phase. The spectra shown as an example (right) are for biphasic hydrogenation of propyne to propylene. c) Schematics of a more advanced experimental setup used for a rapid batch-mode production and MRI detection of hyperpolarized propane gas. The insets show the reaction scheme (top-right) and a gradient echo 2D MR image of ~200 mL of hyperpolarized propane gas in an ~56 mL container acquired at 4.7 T (bottom-left). d) The diagram of the experimental setup used to produce hyperpolarized vapor of norbornene by bubbling pH₂ through an aqueous solution of a Rh(I) complex possessing a norbornadiene ligand (bottom), the reaction scheme (top), and the resulting gas-phase ¹H NMR spectra (middle). Open circles label the signals of norbornane. Plastic spheres in the reactor were used to reduce its volume. e) The diagram of the experimental setup used for injecting Rh(I) complex solution into a 56-mL volume containing pH₂ at ≈7 bar pressure for subsequent in situ ¹H NMR spectroscopy of hyperpolarized norbornene at 47.5 mT. Figure 6a reproduced from Heterogeneous Parahydrogen-Induced Polarization of Diethyl Ether for Magnetic Resonance Imaging Applications, Salnikov, O. G.; Svyatova, A.; Kovtunova, L. M.; Chukanov, N. V.; Bukhtiyarov, V. I.; Kovtunov, K. V.; Chekmenev, E. Y.; Koptyug, I. V. Angew. Chem. Int. Ed. Engl., Vol. 27, Issue 4 (ref 196). Copyright 2021 Wiley. Figure 6b reproduced from Kovtunov, K. V.; Zhivonitko, V. V.; Skovpin, I. V.; Barskiy, D. A.; Salnikov, O. G.; Koptyug, I. V. Toward Continuous Production of Catalyst-Free Hyperpolarized Fluids Based on Biphasic and Heterogeneous Hydrogenations with Parahydrogen. J. Phys. Chem. C 2013, 117 (44), 22887–22893 (ref. 194). Copyright 2013 American Chemical Society. Figure 6c reproduced from Salnikov, O. G.; Nikolaou, P.; Ariyasingha, N. M.; Kovtunov, K. V.; Koptyug, I. V.; Chekmenev, E. Y. Clinical-Scale Batch-Mode Production of Hyperpolarized Propane Gas for MRI. Anal. Chem. 2019, 91 (7), 4741–4746 (ref. 200). Copyright 2019 American Chemical Society. Figures 6d and 6e reproduced from Kovtunov, K. V.; Barskiy, D. A.; Shchepin, R. V.; Coffey, A. M.; Waddell, K. W.; Koptyug, I. V.; Chekmenev, E. Y. Demonstration of Heterogeneous Parahydrogen Induced Polarization Using Hyperpolarized Agent Migration from Dissolved Rh(I) Complex to Gas Phase. Anal. Chem. 2014, 86 (13), 6192–6196 (ref. 193). Copyright 2014 American Chemical Society.

HP norbornene vapor was produced from an aqueous solution of Rh(I) complex incorporating a norbornadiene ligand by either bubbling pH₂ through it (for high-field NMR; Figure 6d) or by spraying it into a chamber pressurized with pH₂ (for NMR at 47.5 mT; Figure 6e).¹⁹³ Gas-liquid biphasic hydrogenations employed a dissolved catalyst, and gaseous reactants were bubbled through the solution, Figure 6b. The reaction product returned to the gas phase and retained a significant level of hyperpolarization, providing a complete separation of the hyperpolarized substance from the catalyst.¹⁹⁴

Prepolarized propane was used in many studies to image voids in various objects^201–203 including microfluidic devices.^204,205 Imaging of HP reaction products formed in an operating model catalytic reactor facilitates mapping of their spatial distribution within the catalyst bed.^206–209

Challenges:

The main challenge with gases is the short T₁ time; for propane it is below 1 s at 9.4 T and 1 bar, but depends strongly on pressure and gas mixture composition.²¹⁰ T₁ times tend to be longer for shorter molecular rotational correlation times, i.e., for larger and/or heavier molecules, higher gas pressures, or gases in small pores. The T₁ values of small gas molecules generally increase upon dissolution in liquids. For example, T₁ ≈ 1 ms for H₂ gas at 1 bar but it increases by ca. 1000-fold upon dissolution in methanol-d₄. Condensation of hyperpolarized diethyl ether vapor was also shown to prolong hyperpolarization lifetime.¹⁹⁶ Furthermore, because of the strong coupling for protons at low fields, they can exhibit properties of long-lived spin states (LLSS).²¹¹ For instance, measurements at 0.05 T for propane gas gave T_LLSS/T₁ ≈ 3.1 at 3 – 7.6 bar, with T₁ ≈ 4 s and T_LLSS ≈13 s at 7.6 bar.²¹² Similar trends were reported for diethyl ether vapor.²¹³ The spin-lock induced crossing (SLIC) pulse sequence can be conveniently used to convert LLSS to an observable signal.^212,213 Another challenge with heterogeneous hydrogenation is that it is outperformed by its homogeneous counterpart in the polarization levels achieved due to low selectivity to pairwise H₂ addition, often P_1H = 1 – 3 % or even lower. Values of P_1H = 7 – 10 %^{203,214–216} or even 60 %²¹⁷ have been reported using heterogeneous catalysis, but in most cases accompanied by low levels of catalytic conversion.

PURIFICATION

As soon as the development of PHIP as a hyperpolarization method for metabolic MRI began,^5,137 it became apparent that purification of the reaction solution would be necessary for preclinical and clinical in vivo applications. In this regard, PHIP presents some challenges. As outlined above, the hyperpolarization often requires rather non-physiological conditions (organic solvents, pH value, temperature) and high concentrations of organometallic catalysts.^{5,13,14,20,24,89,127,128,137,149} Moreover, the hydrogenation reaction may be incomplete or lead to side products.²⁴ Hence, purifying the polarized solution is important and - keeping in mind the short lifetime of liquid-state hyperpolarization - needs to be performed as quickly as possible, ideally within a few tens of seconds. Many purification strategies exist in chemistry already and some were adapted to PHIP as described below.

Catalyst scavenging

One way to remove the catalyst is metal scavenging. In this approach, the scavenger is either used in a filtration column or mixed with the reaction solution and subsequently filtered out.^{81,114,137,218} A commercially available metal scavenger (QuadraPure TU) was employed to clean a solution of HP fumarate prior to in vivo MRI.¹¹⁴ Near complete removal of ruthenium was achieved by slowly passing the solution through the scavenger in about a minute, but a reduction of concentration from ~7 g/L to 100 mg/L was reached within seconds.¹¹⁴ Similar scavenging approaches have been introduced by Kidd et al.²¹⁸ and Barskiy et al.⁸¹ to remove an iridium-based SABRE catalyst.

A shortcoming of metal scavenging is that although the catalyst is removed, unreacted starting materials and side products may remain. Nevertheless, it was found to be a useful step for reducing the remaining metal contamination further in already purified solutions.⁸¹

Liquid-liquid phase separation:

Liquid-liquid (LL) phase extraction was found to provide near catalyst-free aqueous solutions of PHIP contrast agents. This method relies on the following steps: a) hydrogenation of a labile, lipophilic precursor of the target substrate in a hydrophobic organic solvent (e.g. chloroform); b) SOT; c) hydrolysis of the hyperpolarized product by means of fast reaction with an aqueous base - for instance, carboxylate sodium salts (hydrophilic) can be obtained by hydrolysis of the corresponding esters (lipophilic); d) extraction of the aqueous phase that contains the hydrophilic product, Figure 7.⁷⁹ To obtain a biocompatible pH, an acidic buffer was added.^56,80,82,219 The technique was shown to work well with PHIP-SAH²⁰ and has provided aqueous solution of hyperpolarized sodium [1-¹³C]pyruvate and acetate, e.g., by hydrogenating their propargylic or vinyl esters.

Figure 7.

Upper panel: Continuous-flow liquid-state ALTADENA HET-PHIP polarizer. a) The system incorporated a packed-bed heterogeneous reactor containing a slid Rh/TiO₂ nano-rod catalyst and a tube-in-tube membrane system for bubble-free pH₂ dissolution. The liquid is drawn into the syringe from the left liquid reservoir and the 3-way valve is then changed to allow the liquid to flow through the tube-in-tube and then on to the heated catalyst cartridge. b) Rendering of the tube-in-tube device. C) Close up of the liquid inlet port showing the PFA (clear) tubing, 316 stainless steel needle (grey), and AF2400 membrane ‘inner tube’ (black). Figure reproduced from Toward Continuous-Flow Hyperpolarisation of Metabolites via Heterogenous Catalysis, Side-Arm-Hydrogenation, and Membrane Dissolution of Parahydrogen, Hale, W. G.; Zhao, T. Y.; Choi, D.; Ferrer, M.-J.; Song, B.; Zhao, H.; Hagelin-Weaver, H. E.; Bowers, C. R. ChemPhysChem, Vol. 22, Issue 9 (ref 18). Copyright 2021 Wiley. Lower panel: PHIP-SAH LL-separation (left) and HP in vivo ¹³C-MRI (right). The scheme on the left shows the LL separation of a carboxylate sodium salt in the water phase from its ester in the organic phase. HP [1-¹³C]pyruvate obtained using this method has been applied in vivo for metabolic studies (right): ¹³C-CSI image of [1-¹³C]pyruvate overlaid with an anatomical image (¹H-RARE) of a healthy mouse (image acquired at 3T).

The LL phase extraction can be performed in any solvent-resistant container, e.g., in NMR tubes or a dedicated reactor. Its efficiency depends on the level of dispersion of small droplets of the organic solution into the aqueous phase, as hydrolysis likely occurs at the LL interface. As the dispersion should happen as fast as possible, injecting heated and pressurized base solution into the organic phase was suggested (a video showing the procedure can be found in ref.²²⁰). Instead of using pure hydrophobic chloroform,²²¹ mixtures containing some (few %) hydrophilic solvents (i.e., ethanol, methanol) or toluene have been suggested^82,220 to further improve the LL mixing and hydrolysis, as well as bubbling N₂ gas through the solution.⁸² Ultimately, thanks to the instability of the LL mixture, the two phases separate within a few seconds.

Notably, this approach was used for the first in cellulo and in vivo metabolic studies with PHIP-polarized pyruvate,^80,219,221 after a first demonstration of hyperpolarized [1-¹³C]succinate (from maleic anhydride).⁷⁹ Losses of polarization during hydrolysis and phase transfer of the ester derivatives have been observed and could not be explained by T₁ relaxation alone.¹⁵⁷ Instead, the effect has been attributed to the presence of paramagnetic impurities derived from catalyst degradation in the organic phase and a beneficial effect on hyperpolarization has been obtained through the addition of a radical scavenger (sodium ascorbate) to the aqueous base. The concentration of the metal (rhodium) in the aqueous phase has been determined to be 30 μM.²²²

Solvents such as methanol, ethanol, and acetone mix well with water and are transferred to the aqueous phase during phase extraction. Therefore, their application in the hyperpolarization of substrates for biological use (in cells and in vivo) must be considered carefully, due to their toxicity, especially for methanol. Nevertheless, it must be mentioned that, even when pure hydrophobic solvents (chloroform and toluene) were used, their dissolution and ultimately the concentration in the aqueous phase is non-negligible and a cytotoxicity effect has been observed.²¹⁹ In order to solve this issue, filtration of the aqueous solution through a lipophilic resin (Tenax TA, Porous Polymer Adsorbent, 60–80 mesh, Supelco) was recently shown to lead to a reduction of the concentration of these solvents well below the concentration recommended by Environmental Protection Agencies (EPA).^69,220

Heterogeneous Catalysts

In contrast to homogeneous catalysis, heterogeneous catalysis affords straightforward separation of the solution-state hyperpolarized hydrogenation adducts from the solid catalyst due to the insolubility of the latter. Moreover, heterogeneous catalysis is inherently compatible with continuous-flow production of HP gases and liquids. These advantages were already recognized in the initial demonstration of HET-PHIP⁷³ and continue to drive the development of HET-PHIP catalysts and reactor systems.

In ref.⁷³ silica and a polymer were functionalized with phosphine groups to chelate the Rh complex (e.g. Wilkinson’s catalyst). The linkage proved to be resilient, but the catalysts still suffered from low stability, possible leaching into solution upon oxidation of the phosphine moieties, reduction during the reaction, and metal complex dimerization.¹⁹² A recent article described an alternative linkage scheme, where a silica supported polymer incorporating pyridyl groups was used to tether Wilkinson’s catalyst.²²³ This catalyst showed better stability and resistance to leaching. Modest signal enhancements of up to 200 were reported for the hydrogenation of styrene in acetone-d₆.

In the quest to realize efficient, stable and robust HET-PHIP catalysts, metal-oxide-supported nanoparticle catalysts consisting of Pt, Pd, Rh, Ru, Ir as well as bimetallic compositions (e.g. Pd-In)²²⁴ and intermetallic nanoparticles (e.g. PtSn, Pt₃Sn)²¹⁴ with varying shapes, sizes, and support materials have been explored.¹⁹² Most of the published solution-state HET-PHIP studies were performed using a batch reactor configuration,^{214,224–227} where hydrogen is bubbled through a heated NMR tube containing the insoluble solid catalyst and the unsaturated substrate in solution. For larger catalyst particles (millimeter size), settling out of the catalyst after cessation of bubbling can occur within seconds. Under the relatively mild conditions of solution-state hydrogenation, supported metal nanoparticles were found to resist leaching. For example, after hydrogenation of 2-hydroxyethyl acrylate over 25 mg of Pt₃Sn intermetallic nanoparticles in 2 mL D₂O at 120 °C and 5.7 bar, Pt and Sn levels in the decanted solution were found to be well below 100 ppb (by mass).²¹⁴ Transfer of hyperpolarization to heteronuclei in aqueous media for biologically relevant compounds has been addressed as well.²²⁸

As noted above, MRI of HP ¹³C pyruvate provides a means for detection of abnormal metabolism in malignant tumors and other pathologies. There is evidence, however, that administration of a continuous stream of the HP pyruvate over longer periods is preferable to a single large bolus for some applications.²²⁹ Production of purified continuous-flow streams of HP pyruvate by either dissolution DNP or PHIP is challenging and has not yet been demonstrated.

Hale et al. recently presented a novel apparatus that allowed continuous production of hyperpolarized allyl acetate by hydrogenation of propargyl acetate with pH₂, Figure 7. The apparatus incorporated a packed-bed catalytic reactor, side-arm hydrogenation, and pH₂ membrane dissolution.¹⁸ The polarizer continuously achieved a conversion of 30 % and ¹H signal enhancements up to 300 (relative to thermal equilibrium at 9.4 Tesla) were shown to be feasible. However, the polarization transfer to ¹³C, side-arm cleavage, and transfer to the aqueous phase have yet to be addressed.

Challenges for heterogeneous catalysis include lower product concentrations and lower polarizations compared to homogeneous approaches. To solve these shortcomings, emphasis is made on improved flow reactor design as well as the rational design of catalysts to obtain higher pairwise selectivity without sacrificing yield.

Precipitation

Very recently, a scheme to purify HP molecules based on precipitation was introduced.^24,230 HP fumarate was generated in an aqueous solution using a trans-selective hydrogenation catalyst. After hydrogenation and polarization transfer, the reaction solution contained the ruthenium-based catalyst, unreacted starting material, and side products in addition to the desired HP fumarate salt. To purify the solution, acid and non-polarized fumarate were added to the solution so that fumaric acid precipitated out of the solution almost immediately.

Because fumarate precipitated very efficiently and the catalyst, starting material, and side products remained mostly dissolved, separation was easily achieved by decanting the solution. Subsequently, the solid, HP fumaric acid was re-dissolved in aqueous solution. It was important to keep the solid fumarate (from precipitation to re-dissolution) at a sufficiently high magnetic field to avoid fast relaxation in the solid state. The remaining metal concentration was found to be 16 μM after a washing procedure.²⁴ While this method worked well for fumarate, the generality of this approach remains an open question, and further research towards this promising approach is certainly warranted.

OTHER INTERESTING DEVELOPMENTS

PHIP-on-a-chip:

An emerging research area that is mainly pursued at high magnetic fields is the integration of PHIP into lab-on-a-chip devices.^18,231–234 The advantage offered by working at a microfluidic scale (i.e., with sample volumes on the order of microliters) is in the much higher degree of experimental control that can be leveraged, in terms of sample contact time with pH₂, molecular diffusion, temperature and pressure. The other important benefit of microfluidic implementation of PHIP experiments is in the short transport paths (and hence time) between the point of the PHIP process occurring and signal detection. This is particularly important when nuclear spin relaxation times are short.

RASER:

Radio amplification by stimulated emission (RASER) using PHIP was recently discovered^21,235 in PASADENA and ALTADENA conditions.^22,236 RASER emission was detected at low (millitesla)²³⁵ and high fields (tesla) and reproduced by simulations.²² Although the RASER effect was first observed using other HP techniques,^237–239 PHIP has the advantage that the polarization can be continuously refreshed by a constant supply of pH₂ as long as substrate is not depleted. While a biomedical application of RASER was not described, newly emerged applications included observation of NMR spectra with very narrow signals,^22,235 background-free proton NMR spectroscopy,²³⁶ and polarization transfer to other molecules via intermolecular dipole-dipole interaction.²³

PHIP-X:

pH₂ hyperpolarization relayed via chemical exchange (PHIP-X)²⁶ facilitates polarization of molecules that undergo suitable proton exchange like glucose.^25,240–242 First, pH₂ is added to an intermediary molecule by homogeneous catalysis, where the polarization is then transferred to an exchanging proton. Secondly, the polarization is incorporated by the target molecule by exchange of a polarized proton from the reacted intermediary molecule. Experimentally, this process was implemented by placing a high-pressure reaction chamber in millitesla fields. Upon hydrogenation, the sample was transferred to high field NMR for detection. If necessary, the target molecule can be added after the hydrogenation to avoid interference with the hydrogenation step. Both millitesla and MFC setups appear to be well suited to host this process, which involves SOT in the intermediary and target molecule, for which the spin physics has yet to be fully elucidated.

TOWARDS CLINICAL APPLICATION

The results described above impressively demonstrate the power, versatility, and maturity of pH₂-based hyperpolarization approaches and instrumentation. Indeed, HP [1-¹³C]pyruvate produced by hydrogenative PHIP has been recently employed to detect the response of the heart to altered metabolism in real time.⁸⁰

While preclinical imaging with PHIP agents was demonstrated in more than 15 papers since 2001, this method has not been translated to human imaging yet.^{5,13,36,69,80,86,87,90,114,142–146,150}

Still, we may have reached a tipping point, as all ingredients for producing clean, aqueous solutions of interesting (not only biocompatible) agents were described in the literature. As we speak, work continues at multiple locations to make studies with PHIP-polarized agents a reality.

Let us assume that all relevant technical hurdles were to be addressed, and that there was a device that produces a clean and pure solution or solid of highly polarized contrast agent with a fast and relevant function in vivo. Still, there will be regulatory aspects to meet before human studies may commence. Here, much can be learned from the path taken so impressively by the DNP community.

Contrast agents are considered drugs by most regulatory bodies. As no injectable HP contrast agents are approved as drugs (yet), the guidelines for the application and evaluation of non-approved drugs apply. On the other hand, propane (also known as E944) is approved for unlimited use in foods and is already regulated by FDA. The above regulatory requirements may vary between countries, but are likely to include

1. GMP manufacturing of ingredients by vendor or in-house – pH₂, catalyst, solvents, precursors

2. ISO 5 clean bench (or higher depending on regulations) preparation of ingredients on site – mixing precursor solution, pH₂

3. Polarization and rapid QC – measure sample concentration, temperature, pH and polarization; residual catalyst concentration, perform sterile filter integrity test.

These steps would be followed by transfer, administration by an MD or qualified person, and imaging. However, prior to injection the hyperpolarized CA should be passed through a 0.2 μm pore size filter (the integrity of the filter would need to be tested in accordance with manufacturer regulations to ensure sterility). Given the limited lifetime of hyperpolarization, the QC needs to be rapid (< 30 s) and is recommended to be performed in parallel with the sample transfer and filter integrity test to minimize the time to injection. The transfer distance will be determined by the lifetime, and long-lived samples may be transported between sites. Attempts for individual treatment of patients on a small scale may require less stringent regulations. Safety and dose escalation trials would be followed by efficacy tests.

CONCLUSIONS

pH₂-based hyperpolarization has come a long way since its inception in the 1980s, and the first demonstration of hyperpolarized in vivo ¹³C-MRI in 2001. For hyperpolarizing small molecules in solution for in vivo MRI application, dDNP is currently the state-of-the-art methodology, but PHIP is beginning to emerge as a viable alternative. Arguably, there is no hyperpolarization technique that is more flexible and that offers more variants than PHIP. In addition, PHIP is inexpensive, (often) easy to implement, fast and scalable. Consequently, many applications have emerged, ranging from basic physics to chemistry and biomedicine. The latter, however, has not advanced as fast as, e.g., with dDNP.

The main advantages of PHIP lie in the inherent low cost, ease of implementation, rate of hyperpolarized sample turnover, and scalability. These advantages stem mainly from the chemical nature of the technique; hyperpolarization is delivered via chemical reactions rather than requiring cryogenic cooling of the to-be-polarized substance and/or irradiation with high power microwaves, as is the case for dDNP. The chemical basis of the process unfortunately carries some drawbacks: (1) the method is limited to polarizing molecules that can be generated through hydrogenation reactions; (2) the HP solutions are contaminated with other chemicals from the reaction, and; (3) polarization levels are often only modest since the nuclear spins relax during the hyperpolarization process.

All three drawbacks are being overcome through state-of-the-art research, much of which is covered in this review: Many biologically and diagnostically promising agents are available by direct hydrogenation (SUX, FUM, TFPP, PLAC) and variants such as PHIP-SAH and PHIP-X (pyruvate, acetate, glucose, lactate). Physicochemical methods for purifying the HP solutions through phase separation have been shown for a few PHIP-polarized metabolites, and scavenging can be used to remove the catalyst from solutions. Advances in instrumentation for hydrogenation, polarization transfer, and sample transport, have helped to improve the resulting polarization levels in PHIP-polarized molecules. PHIP seems to be a promising hyperpolarization method for preclinical MRI and in vitro studies where experiment turnover is high but requirements for high polarization and solution sterility/nontoxicity are lower.

Nonetheless, it remains to be seen whether PHIP has a future role to play in clinical imaging. To date no device is commercially available – not even for in vitro and preclinical in vivo studies. Over the years, some initiatives were and still are active (Amersham, Stelar, Spindynamics, Bruker, XEUS Technologies LTD, NVision), and numerous patents were filed. The technical challenges towards clinical applications are accompanied by regulatory aspects, which must be met for a use in human. Here, the contrast agents do not only need to be pure and highly polarized, but also produced effectively and compliant to GMP. The reactive nature of PHIP may pose a challenge, too. Ideally, agents must have a long enough lifetime for testing and transferring after polarization.

Although the focus of this review has largely been on producing PHIP-polarized biomolecules for in vivo applications, there are many more uses of PHIP. Examples include studying chemical reaction mechanisms and as a source of signal enhancement for high-resolution NMR. Advances in instrumentation have underpinned much of the method development made in recent years, and will continue to pave the way into the future. Even if PHIP (or hyperpolarization in general) would eventually turn out to be ineligible for clinics, many more applications are lurking behind the corner – the marriage of pH₂ and MR are far from over yet, and the best has yet to come.

Figure 8.

¹³C-chemical shift imaging reporting of pyruvate and lactate distribution acquired on living mice, obtained upon the injection of a dose of [1-¹³C] pyruvate hyperpolarized using pH₂. The spatial localization of each metabolite is shown upon overlapping the ¹³C-CSI results to the anatomical proton image (T₂ weighted fast spin echo image). Each metabolite map is scaled individually and is displayed on a fire color scale so that the region of highest metabolite signal appears white and the lowest appears black. Reprinted by permission from Nature: Sci. Rep., The 13C hyperpolarized pyruvate generated by ParaHydrogen detects the response of the heart to altered metabolism in real time, Cavallari, E.; Carrera, C.; Sorge, M.; Bonne, G.; Muchir, A.; Aime, S.; Reineri, F., Sci. Rep., Vol. 8, Issue 1 (ref 80). Copyright Nature 2018.