Reversible Interactions with para-Hydrogen Enhance NMR Sensitivity by Polarization Transfer

See allHide authors and affiliations

Science  27 Mar 2009:
Vol. 323, Issue 5922, pp. 1708-1711
DOI: 10.1126/science.1168877


The sensitivity of both nuclear magnetic resonance spectroscopy and magnetic resonance imaging is very low because the detected signal strength depends on the small population difference between spin states even in high magnetic fields. Hyperpolarization methods can be used to increase this difference and thereby enhance signal strength. This has been achieved previously by incorporating the molecular spin singlet para-hydrogen into hydrogenation reaction products. We show here that a metal complex can facilitate the reversible interaction of para-hydrogen with a suitable organic substrate such that up to an 800-fold increase in proton, carbon, and nitrogen signal strengths are seen for the substrate without its hydrogenation. These polarized signals can be selectively detected when combined with methods that suppress background signals.

The wide variety of applications of nuclear magnetic resonance (NMR) (13) are limited by the technique's extremely low inherent sensitivity. Here we describe an approach that uses hyperpolarized spins derived from parahydrogen (para-H2) (4) to sensitize the NMR experiment without actually incorporating para-H2 into the molecule that is to be probed. Specifically, we show that high-resolution NMR spectra can be collected for a range of molecules and nuclei with detected signal strengths up to 800 times greater than would be normally achievable with an unpolarized sample. This improvement facilitates the collection of diagnostic high-resolution 1H, 13C, 15N, and 19F NMR spectra and magnetic resonance images of selected signals in a fraction of the time that would normally be necessary. When optimized, this route is predicted to increase proton sensitivity by up to four orders of magnitude (5) such that the routine single shot characterization of materials, even at picomole levels, will become possible (6).

The relative weakness of NMR signals exhibited by nuclei with a nonzero magnetic moment results from the way the original energy levels split in a magnetic field (7). The bulk magnetic moment for an ensemble of such nuclei is determined by the Boltzmann population of each energy level. In general, the difference in the energy between these levels is so small that almost-equal spin populations exist across them. For example, in a magnetic field of 9.4 T such as that found in routine high-resolution NMR spectrometers, the difference in spin population will only be around 1 in 32,000 for 1H. Unfortunately, 1H nuclei are the most sensitive, and, for 19F, 31P, 13C, and 15N, the next most common nuclei to be studied, the sensitivity problem is even more acute, with the associated signal decreasing by factors of 1.2, 15, 64, and 104 respectively. The problem is further exacerbated when the natural abundance of 13C (1.108%) and 15N (0.37%) isotopes are taken into account, meaning the effective differences in sensitivity scale from 1 in 32,000 for 1H to 1 in 120 million and 1 in 8.7 billion in these nuclei, respectively. As a result of this, general routine human imaging experiments are restricted to measuring the 1H signals coming from water and lipid in tissues. Furthermore the direct detection of nonproton signals can require many hours of measurement even in NMR spectroscopy.

A number of “hyperpolarization” methods have been developed (818) to enhance signal strength by transferring nonequilibrium nuclear spin polarizations. The method of dynamic nuclear polarization (DNP), as reviewed by Kemsley (10), creates a non-Boltzmann spin population by transfer of polarization from an unpaired electron. Recently this approach has been demonstrated to usefully enhance 13C and 15N signals by factors that exceed 10,000 (1114). Currently, however, this method requires long polarization times (often over 6 hours), normally uses water and methanol as solvents, and is unable to detect enhanced proton signals routinely.

We tackled the sensitivity problem in a different way by using para-H2 as the source of polarization. Para-H2 has the advantage that it can be prepared easily and stored at room temperature for months. Previously, studies with para-H2 were limited to those involving the formation of hydrogenation products containing para-H2–derived protons (15). For example, Pines et al. recently used it to image heterogeneous hydrogenation reactions through the detection of polarized protons of propane (16). More usually, however, the imaging of 13C-based magnetization was targeted because such nuclei can be polarized by para-H2–based hydrogenation reactions in low magnetic field, as demonstrated by Bargon et al. (17) and exploited by Golman et al. (12). In these hydrogenative processes, the newly formed reaction products contain protons originating from a single para-H2 molecule, and they can produce strongly enhanced NMR signals in the reaction product provided the reaction does not change the magnetic arrangement of these coupled atoms (15).

The need for a suitable hydrogen acceptor, however, reflects an important limitation of the existing approach. Nonetheless, the ability to increase proton signal strengths in such products by 32,000 with para-H2 and hence detect picomoles of material in a single scan has been established (6, 18). In order to generalize the use of para-H2 as a source of polarization, a method for the transfer of polarization without the direct hydrogenation of materials is needed. We show that the temporary association of a substrate and para-H2 via a transition metal center in low magnetic field can achieve just this. Thus, NMR signal amplification by reversible exchange (NMR-SABRE) is achieved without any chemical modification of the hyperpolarized material. As an example, we used the labile complex [Ir(H)2(PCy3)(substrate)3][BF4], which is formed by the reaction of [Ir(COD)(PCy3)(MeCN)][BF4] (where Cy is cyclohexyl and COD is cyclooctadiene) with para-H2 and an excess of the substrate to be polarized (5, 19). Notably, the same observations can be made on a range of materials and metal templates according to the concept illustrated in Fig. 1.

Fig. 1.

Schematic representation of the magnetization transfer process.

We first illustrate this effect by using pyridine as the substrate where the iridium dihydride complex [Ir(H)2(PCy3)(pyridine)3][BF4] is formed. The single scan 1H NMR spectrum shown in Fig. 2A was recorded after the sample was first polarized in a magnetic field of around 2 × 10–2 T and contains signals with enhanced intensity for the three proton sites of the pyridine substrate. All that is necessary to achieve this result is to shake the sample in a low magnetic field (movie S1). This dissolves fresh para-H2 from the head space above the solvent, allowing it to associate with the metal complex and thereby activating the polarization transfer process in the solution. Specifically, the signals for free pyridine at δ 7.84, 8.54, and 7.43 appear in the downward direction that is most simply described as emission. After the sample is then left in the NMR spectrometer for 10 minutes, the resulting NMR spectrum demonstrates that the pyridine's magnetic states have returned to their more usual Boltzmann distribution through relaxation. We note that the resonances illustrated for free pyridine in Fig. 2A can be described as hyperpolarized.

Fig. 2.

Single scan NMR spectra of samples containing a templating medium, the indicated substrate, and para-H2 at 295 K in d4-methanol where the polarization transfer step was achieved in a 2 × 10–2 T field. ppm indicates parts per million. (A) The 1H control trace (top) of 6 nmol of pyridine with 128-fold vertical expansion relative to the bottom 1H trace that was recorded immediately after polarization transfer. (B) Polarized 1H decoupled 13C trace of the same sample after refocusing. (C) Polarized 1H decoupled 15N trace, with refocusing, of a sample containing 25 nmol of 15N labeled pyridine. (D) Polarized 1H decoupled 13C trace after refocusing, in magnitude, of a sample containing 50 μmol of nicotinamide. (E) The 1H NMR spectrum of the same nicotinamide sample showing the control trace (top), with a 32-fold vertical expansion relative to the polarized trace (bottom) with transfer in a 0.5 × 10–4 T field.

When the region of the 1H NMR spectrum containing the pyridine signals is examined in more detail, weaker signals for the bound pyridine ligands within the host-ligand complex [Ir(H)2(PCy3)(pyridine)3][BF4] are also seen to show this emission character. We conclude therefore that, while at low field, spontaneous polarization transfer occurs between para-H2 and the pyridine substrate that is in temporary association with the metal template. Furthermore, substrate exchange with that bound in the host-ligand template during this period leads to the buildup of hyperpolarization in free pyridine. When these enhanced signals are compared with those without enhancement, up to 550-fold increase in signal strength is observed. The enhancement achieved by using this simple process is realized after just a few seconds of contact with the sample in low field (movie S1).

This enhancement effect is not just limited to proton signals because the corresponding 13C (Fig. 2B) and 15N resonances (Fig. 2C) of pyridine are also polarized. Furthermore, these effects can be regenerated by simply removing the sample from the spectrometer and bringing it into contact with fresh para-H2 in low magnetic field. When the hyperpolarized 13C NMR spectrum is compared with that obtained when a standard 13C NMR spectrum is recorded, it would take 670,000 scans to achieve equal signal intensity to that seen with only one scan after the para-H2 enhancement process (5). The time saved through this 823-fold signal enhancement has been estimated to exceed 3 months assuming that the individual measurements are separated by a 20-s recovery delay and use 90° observation pulses. If the NMR spectra of 100% 13C-enriched materials were compared with those obtained from para-H2 enhancement of the unenriched material, an eightfold gain in sensitivity would still be apparent. Of course these enriched materials could themselves be easily polarized and therefore yield even larger signals.

These methods can be extended beyond pyridine, and in Fig. 2D we illustrate a 13C spectrum of nicotinamide that is collected in the same way; 345-fold 1H signal enhancements have been quantified for this system on the basis of spectra such as that shown in Fig. 2E. All of the 13C resonances for nicotinamide are enhanced, although not to the same degree (5). We note that 3-fluoropyridine, nicotine, pyridazine, quinoline, quinazoline, quinoxaline, and dibenzothiophene also show enhancement and that 19F and 31P signals can also be detected (5). These substrates associate weakly with the metal complex through their basic donor sites, thereby facilitating the polarization transfer step. NMR spectra representative of these materials can be found in figs. S2 and S3.

We recently described a gradient-based NMR method called only parahydrogen spectroscopy (OPSY) that suppresses signals derived from nuclei with thermally equilibrated spin state populations while allowing the observation of signals from para-H2–derived protons found in hydrogenation products (20). This method works by selectively probing the longitudinal two-spin order term that is generated for coupled spins derived from para-H2 while dephasing terms from the usual thermal magnetization through the use of pulsed field gradients. When it is applied, all thermal signals are successfully suppressed and those for the hyperpolarized molecules still remain. As a result, NMR spectra of polarized substrates can be recorded in the presence of a protio solvent. The corresponding 1H-selected OPSY spectral trace showing signals for polarized pyridine, collected in protio methylenechloride, is shown in Fig. 3A.

Fig. 3.

(A) Unlocked 1H NMR spectrum recorded in protio-CH2Cl2 by using the OPSY filtration sequence to suppress background signals, which illustrates the detection of 60 μmol of polarized pyridine in a single scan. (B) The 1H NMR spectrum recorded on the same sample by using the same pulse sequence but recorded 150 s after the low field polarization step.

This polarization transfer process is predicted to generate spin states that are unaffected by dipole-dipole relaxation mechanisms and as such to have long lifetimes in low field (5). The potential impact of long-lived states in NMR has been highlighted by Levitt in a different context (21). The OPSY-based 1H NMR spectrum in Fig. 3B was recorded 150 s after polarization. This illustrates the presence of the longitudinal two-spin order magnetic state associated with a pair of coupled hydrogen atoms within polarized pyridine.

These results suggest a route for the use of naturally occurring molecules as contrast agents in magnetic resonance imaging (MRI). In this regard, Fig. 4 shows two separate single-average true fast imaging with steady state precession (True-FISP)–based MRI images of pyridine collected over 0.7 s on an 8-mm sample tube containing cylinders of 1-mm internal diameter (22). We collected the first trace on a 0.5-mm slice by using polarized pyridine in d4-methanol, whereas the second was collected on the same sample after the pyridine polarization had decayed, albeit over a 20-mm slice thickness with half the in-plane resolution in both directions; this introduces an inherent 160-fold increase in signal strength between the two measurements before any additional signal enhancement is considered. Even under the latter conditions, no image is visible, illustrating both the viability of this approach in MRI and that the sensitivity gain is over 160-fold.

Fig. 4.

Single average 1H True-FISP MRI images of an 8-mm sample tube containing glass cylinders with 1-mm internal diameter showing (A) signals from polarized pyridine in d4-methanol within a 0.5-mm slice and (B) signals from the same sample after the decay of polarization for a 20-mm slice thickness.

The procedures described in this paper are relatively cheap to implement and can produce large amounts of hyperpolarized materials in a short time. Furthermore, the video sequence (movie S1) illustrates the simplicity of the measurement process that we have described. The method can be used on routine proton-based MRI instruments without the need to exploit other magnetically active nuclei that provide much weaker signals. Given the opportunities that hyperpolarized methods have been shown to offer for the development of real-time metabolic imaging applications, the results presented here take us a step further toward the goal of responsive high-sensitivity MR-based methods for the diagnosis of disease (10).

Supporting Online Material

Materials and Methods

SOM Text

Figs. S1 to S7

Movie S1


References and Notes

Stay Connected to Science

Navigate This Article