Hyperpolarized Xenon for NMR and MRI Applications

1. Preparation of the SEOP Setup

Rubidium must be brought into the optical pumping cell, to facilitate the transfer of polarization from the laser light to xenon. Due to its high reactivity this process must occur without the Rb coming into contact with oxygen or water, otherwise it will become oxidized and won’t polarize Xe. Extra caution should be taken as Rb reacts violently with water.

1. If the optical cell has previously been used it will be coated with a layer of Rb and Rb oxide, as can be seen in Figure 4b. The cell must first be clean before use. Close the inlet and outlet tubes of the optical cell. While keeping it pressurized, transport the cell to a chemical hood. Under the hood using appropriate personal protective equipment, open the cell to atmosphere and wait approximately one hr to allow the Rb surface to oxidize.
2. Gently pipette pure isopropanol into the cell. This will dissolve the Rb oxide layer, and shiny Rb droplets will move over the surface of the isopropanol like water droplets on a hot plate. Pour the isopropanol (and any Rb that comes with it) into a beaker. Repeat until all Rb is removed.
3. If this still doesn’t remove all Rb, make a solution of 10 % water and 90 % isopropanol and repeat step 1.2) increasing the percentage of water (in steps of 10 %) until all Rb is removed.
4. Once all Rb is removed, rinse the optical cell with acetone.
5. Bring a previously evacuated and then argon-filled optical pumping cell into a glove box with argon atmosphere. Also include an ampoule of rubidium, a tool to break the ampoule, Pasteur pipettes, Kimwipes, and a heat gun. In order to maintain a dry atmosphere in the glove box, place a Petri dish with phosphorous pentoxide as a desiccant. The presence of unwanted traces of oxygen can be monitored with a light bulb where the glass bulb is opened to expose the filament to the glove box atmosphere. Conditions are fine as long as no smoke arises with the light turned on.
6. Open the filling port of the pumping cell, break the Rb ampoule and melt the alkali metal with the heat gun. Soak up some liquid Rb with a pipette and inject it into the pumping cell.
7. Close the filling port after increasing the argon pressure in the glove box to maintain a slight overpressure in the pumping cell for the transport to the polarizer setup. Take the cell out of the glove box.
8. Connect the cell to the polarizer manifold, ensuring that it is aligned with the laser beam line illuminating the cell during the pumping process (this can be done with the visible light aiming beam, Figure 7) and check that the heating device with its thermocouple has proper thermal contact with the cell (as in Figure 4a). Attach a thermocouple to the top of the cell.
9. Evacuate the tube connections up to the inlet and outlet valve of the pumping cell. After reaching a pressure of < 30×10^-3 mbar, purge the lines with high purity Ar (or nitrogen). Repeat this three times.
10. With the Ar tank open to the pumping cell inlet, slowly open the inlet and outlet valve of the cell. Carefully open the polarizer outlet valve a tiny bit to establish an Ar flow of ca. 1 SLM through the manifold. Maintain this flow for 2 min. By now, oxygen impurities should be substantially eliminated to avoid Rb oxidization. Close the polarizer outlet valve and the inlet connection to the Ar tank.
11. Turn on the heater of the pumping cell (set temperature ca. 180-190 °C for a heating strip mounted underneath the cell). This will vaporize part of the Rb droplet.
12. Open the Xe gas mixture connection into the polarizer setup. The tank regulator should be set to ca. 3.5 bar overpressure.
13. Turn the laser on and tune its emitting wavelength to ca. 794.8 nm by adjusting the set temperature of the diode coolant. Monitor the laser profile through an optical spectrometer.
14. Continuous vaporization of Rb causes increasing laser absorption. Make sure the laser emission profile is absorbed symmetrically (readjust coolant temperature if necessary). Once the temperature sensor on top of the cell reads ca. 100 °C, you should observe a significantly reduced laser transmission (see Figure 8).
15. Laser absorption also causes additional heating, therefore increasing the pressure in the cell. Monitor the cell conditions and carefully vent gas from the polarizer outlet (as in normal operation) to release some pressure whenever the value approaches the limit the pumping cell is rated for (5 bar abs. in our setup).
16. Turn on the magnetic field (ca. 2-3 mT) around the pumping cell while monitoring the laser profile. Transmission should go up immediately as the field causes selective optical pumping (see Figure 8).
17. Wait for all temperatures to stabilize. The polarizer is now ready to use.

2. Preparation of the NMR Setup

1. Insert a test tube with water into the NMR probe head and perform tuning and matching of the radio frequency (RF) circuit for the proton and the xenon channel.
2. Shim on the water signal with the automated shim routine of the MRI user interface.

3. Hyperpolarization Quantification

1. Connect the polarizer outlet tube to the inlet of the test phantom with its ca. 5 capillaries to inject the Xe and the gas vent tube to the connection with the mass flow controller.
2. Make sure the gas flow controllers are set to ‘closed’ and slowly open the polarizer outlet valve to pressurize the phantom. Set the flow rate to ca. 0.5 SLM to start a continuous flow through the phantom. Estimate from the phantom volume and the gas flow rate how long it takes to entirely replace the gas volume. In our setup, this is ca. 2 sec.
3. For determining the required excitation pulse length, acquire a series of free induction decay (FID) signals with the NMR spectrometer using a hard pulse of 10dB attenuation and various excitation lengths (e.g. 5-100 μsec). Further parameters are: spectral width of sw = 10 kHz, acquisition time ta = 1 sec and a repetition time TR that is longer than the replacement time calculated in step 3.2. The frequency of the excitation pulse for Xe gas at 9.4 T is ca. 110.683 MHz. The FID with the strongest signal will give you the correct combination of pulse power and length for maximum signal.
4. After decreasing the flow to 0.1 SLM, increasing TR to 15 sec (to be comparable with step 3.7), and leaving the other parameters unchanged, acquire a dataset with 16 FID scans while the hyperpolarized Xe is flowing through the sample. Perform the Fourier transformation and measure the peak amplitude in the spectrum. This is the signal intensity for the hyperpolarized xenon gas mixture. Also, note the resonance frequency of the gas peak in Hz.
5. Evacuate a heavy wall NMR tube equipped with a valve for low pressure sealing and fill it with ca. 2 bar overpressure of pure xenon.
6. Evacuate the gas manifold holding the NMR tube and fill ca. 0.2 bar of pure oxygen on top of the Xe into the NMR tube (i.e., adjusting the O₂ pressure to 2.2 bar overpressure). The oxygen will enhance the relaxation of the Xe magnetization after RF excitation (it allows us to work with TR = 15 sec; the process otherwise requires very long TR if the gas is not replaced for the next excitation like in step 3.4).
7. Replace the previously used gas flow phantom in the NMR magnet with this low pressure NMR tube and perform the NMR pulse sequence as in 3.4. This will give you the NMR signal intensity for thermally polarized high-concentration Xe.
8. Compare the signal intensities of thermally and hyperpolarized Xe and calculate the signal enhancement taking the different concentrations and pressures into account. Calculate the spin polarization as follows:

The thermal spin polarization P_th needs to be determined first as a reference. It is defined as the population difference of the two spin states over the sum of the populations, i.e.,

At room temperature, this is given by the high temperature approximation and the population ratio R as

(k is the Boltzmann constant, T the absolute temperature, and γ the magnetogyric ratio). Since the thermal energy kT is by far the dominant factor, R is close to 1, i.e. 0.999982232 for Xe at B₀ = 9.4 T. This yields P_th (9.4 T) = 8.9 10^-7.

Next, the normalized signal enhancement factor ε has to be calculated from the ratio of the hyperpolarized signal S_hp and the signal from thermal polarization S_th (assuming that all NMR pulse sequence parameters were identical for both applications):

Where c and p represent the concentration of Xe in the gas mixture (in %) and the pressure of the gas mixture for both the experiments with thermally and hyperpolarized Xe, respectively. The achieved hyperpolarization is then given by the product εP_th.

4. Functionalized Xenon Solution State Spectroscopy

1. Prepare a 50-200 μM solution of a (functionalized) xenon host (e.g., cryptophane-A with a targeting unit). Depending on the hydrophobicity of the cage construct, add more or less DMSO to water as a solvent. In our demonstration with a cryptophane-A monoacid cage, it is easiest to use pure DMSO. Take ca. 1.5 ml of this solution and fill it into the gas flow phantom, ensuring that the 5 fused silica capillaries allow sufficient bubbling of the solution with the Xe gas mixture. Perform a bubbling test outside the NMR magnet with 0.1 SLM and check for unwanted excessive foaming.
2. Insert the phantom into the NMR probe and tune and match on both the proton and X-channels and perform an automated shim like in step 2.2.
3. Use an FID acquisition with appropriate delays and trigger pulses from the spectrometer to open and close the mass flow controllers. Allow for ca. 15-20 sec bubbling with 0.1 SLM and subsequent 5-8 sec waiting delay for the bubbles to disappear, followed by RF excitation and FID readout.
4. Perform 16 or 32 repetitions (depending on your cage concentration) with sw = 40 kHz, centered at ca. 11 kHz downfield from the gas resonance frequency determined in step 3.4. FID readout should be 500-1,000 m. Fourier transform the FID to get the spectrum.
5. Set the chemical shift value for most right signal (gas phase) to 0 ppm. Write down the frequency of the intense solution signal (most left signal) in Hz and ppm. Also note the offset between this signal at δ_solution and the signal of encapsulated Xe at δ_cage ~ 60 – 80 ppm in ppm. This offset is called Δω (see also representative results).

5. Hyper-CEST Imaging

1. In order to test the contrast agent capability of a xenon host molecule, an experiment with a two-compartment phantom can be performed. To do so, take ca. 50% of the test solution from section 4 and fill it into a 5mm NMR tube. Insert this tube into the 10 mm bubbling setup from section 4. Fill the outer compartment with only the solvent and no cage up to the same level as the inner compartment. Insert 3 of the bubbling capillaries into the outer and 2 capillaries into the inner compartment.
2. Reconnect the tubing to the bubbling setup and repeat step 4.2.
3. Select a single-shot EPI sequence for fast imaging. This sequence possibly needs to be modified to include delays and trigger pulses from the spectrometer to open and close the mass flow controllers. Allow for ca. 15 – 20 sec bubbling with 0.1 SLM and subsequent 5 – 8 sec waiting delay for the bubbles to disappear, followed by MRI encoding.
4. Set the detection nucleus to ¹²⁹Xe on the X-channel and the transmitter/observer frequency to the value determined for the solution signal in step 4.5. Using the RF pulse calculator tool, convert the pulse parameters (amplitude and duration) from step 3.3 into the excitation used in your imaging sequence.
5. The imaging geometry in our example is as follows: 10 – 20 mm slice thickness, transverse orientation; 20 x 20 mm field of view; matrix size 32×32; double sampling (to avoid artifacts) and partial Fourier encoding factor set to 1.68 for accelerated acquisitions (i.e., only 19 of the 32 phase encoding steps are actually acquired).
6. Open the CEST module (a modified magnetization transfer module) for signal preparation and enable a cw presaturation pulse (parameters, e.g., 2 sec duration, 5 μT amplitude). Perform 2 scans in transverse orientation with the carrier frequency of this saturation pulse once being set to δ_cage = δ_solution – Δω and once to δ_control = δ_solution + Δω.
7. Using an image post-processing tool, generate the Hyper-CEST difference image by subtracting the image with saturation at δ_cage from the one with saturation at δ_control. The result should only highlight areas where the Xe host was present (see also representative results).

6. Representative Results

The laser absorption can be monitored by switching the magnetic field around the cell on and off. Depending on the laser power and cell temperature, almost complete absorption is observed with the magnetic field switched off and ca. 30% transmission occurs with the field on (a comparison is shown in Figure 8).

For an NMR system operating at 9.4 T (400 MHz for ¹H, 110 MHz for ¹²⁹Xe), the signal enhancement should be ca. 16000-fold when comparing thermally polarized xenon with hyperpolarized xenon. According to step 3.8, this corresponds to a spin polarization of ca. 15%. Values > 10% should be achievable when using a line narrowed diode laser with cw output of > 100 W.

The ¹²⁹Xe NMR spectrum of a DMSO solution containing 213 μM of a molecular host should exhibit a signal of caged xenon with a signal-to-noise ratio of ca. 10 for 16 acquisitions (Figure 9; at room temperature, line broadening of 10 Hz used).

The Hyper-CEST MRI data set shows full signal intensity for the off-resonant control image and signal depletion in areas containing the Xe host molecule in the on-resonant saturation image. The difference image exclusively displays the areas that responded to the saturation pulse (Figure 10).

Figure 1. Side view of optical components for achieving circularly polarized light. The laser light is coupled into the system through the optical fiber on the left. Both the polarizing beam splitter cube (PBC) and the λ/4 wave plate are installed on rotating mounts to adjust the fast axes for producing circularly polarized light (see movie 1). The ordinary beam reflected by the PBC can be diverted by a mirror to end up in a beam dump (not shown).

Figure 2. Top view of optical components for achieving circularly polarized light. This view includes the beam dump for the ordinary beam. As a safety measure, thermocouples are monitoring the temperature of the primary beam expander, the beam dump, and the polarizing beam splitter cube.

Figure 3. Side view of pumping cell with the side wall of the oven box opened. The laser light enters the box from the left through a parallel glass window. The pinhole on the right end attenuates the transmitted laser power to protect the optical spectrometer which receives the light through a collimator and optical fiber. The Xe gas mixture travels opposite to the laser light direction: it enters the cell through the right leg and exits on the left side.

Figure 4. a) Close-up view of Rb droplet inside the pumping cell. The orange silicon heater (controlled by a PID regulator) is attached to the bottom of the glass cell. A thermocouple on top monitors the cell temperature. b) Close-up view of the gas inlet area of a medium aged pumping cell with increasing condensate build-up on the glass wall. c) Remaining Rb droplet in same pumping cell as in b), as seen by illuminating the cell from the back and with short exposure time to suppress visibility of the glass wall coating.

Figure 5. Energy transitions in alkali metal vapor. a) Without external B-field, the magnetic sub-levels are not defined (illustrated in grey only); hence any atom in the ground state absorbs light. b) Turning on an external field defines the Zeeman levels and causes pumping of only one transition according to the dipole selection rules. This causes accumulation of atoms in one of the sub-levels while a reduced number of atoms in the other ground state sub-level absorbs laser light.

Figure 6. Functionalized cryptophane cage for detecting a specific target of biochemical interest. The Xe NMR signal will change upon the binding event of the specific targeting unit.

Figure 7. Visible aiming beam (red light) for alignment of the pumping cell to ensure complete illumination of the pumping volume.

Figure 8. Laser profiles for different pumping cell conditions. No absorption is observed for the cold cell (room temperature) when no Rb vapor is present. We observe two emission lines from our diode laser (together with a FWHM of 0.5 nm that is within the manufacturer’s specification). When the cell reaches its set temperature (180 °C) and the magnetic field is turned off, general D₁ excitation causes almost complete absorption of the laser light. Switching the magnetic field on induces selective pumping of only one transition and increases the transmission intensity.

Figure 9. ¹²⁹Xe NMR spectrum of a DMSO solution containing cryptophane-A monoacid (structure also shown) as a Xe cage. The gas peak is referenced to 0 ppm. Free Xe in solution appears at δ_solution = 245.7 ppm and the caged Xe at δ_cage = 79.2 ppm. For the Hyper-CEST experiment, the saturation pulse is once set to δ_cage for enabling saturation transfer to decrease the solution peak and once set to δ_control = 412.2 ppm to collect the reference signal for subtraction. Experimental parameters: 213 μM cage in DMSO at 295 K, 16 acquisitions with 32.3 kHz band width, 772 ms FID read out, Xe bubbled into solution at 0.1 SLM for 20 sec.

Figure 10. ¹²⁹Xe MR images of xenon dissolved in DMSO. The phantom consists of two separate compartments with only the inner compartment containing cryptophane-A monoacid (at a concentration of 50 μM). Before each EPI image is taken, a 5 μT continuous-wave saturation pulse is applied for 2 sec. a) The saturation pulse is at δ_control, i.e., off resonant with the Xe@cage peak and we observe strong signal from both compartments. b) The saturation is on resonant with the Xe@cage peak at δ_cage, almost completely destroying the signal from the inner compartment. The subtraction image a) – b) reveals the location of the Xe host molecule. The images were acquired with a FOV of 20 x 20 mm, a slice thickness of 10 mm and 32 x 32 pixels. They were then thresholded and interpolated to 256 x 256 pixels.

Movie 1. Animation of assembling a setup for SEOP. The laser beam is first increased in diameter by a primary beam expander and passes through a polarizing beam splitter cube (PBC). Rotation of this cube changes the relative intensities of the ordinary and extra-ordinary beam. For the position with maximum transmission, the fast axis of the PBC is aligned with the dominant polarization axis of the incoming light. The linear polarization of the transmitted light – which is influenced by the quality/extinction ratio of the PBC – can be tested using a second PBC as an analyzer. Aligning its fast axis with the fast axis of the first cube should give maximum transmission whereas further rotation by 90 ° should give zero transmission and full reflection. Insertion of a λ/4 wave plate converts the linear into circular polarization if its fast axis is rotated 45 ° against the fast axis of the first PBC. The intensity of the transmitted light should now be independent of the rotation of the second cube. Removing the analyzing components and replacing them with a secondary beam expander yields the right beam diameter to illuminate the pumping cell. A rubidium droplet sitting in this cell is partially vaporized once a heater outside the cell is turned on. The xenon gas mixture flowing through the setup in opposite direction to the laser beam distributes this vapor all over the cell. Without a magnetic field, this causes general D₁ excitation of the Rb atoms and strong absorption of the laser light. Turning the magnetic on allows for selective pumping of only one transition between the now defined magnetic sub-levels. As a consequence, only a reduced number of atoms absorb the laser light and transmission is increased again. Click here to view movie.

Movie 2. Animation explaining the CEST effect. Cryptophane cages serve as molecular hosts to trap Xe atoms which change their resonance frequency upon this binding event (transition blue -> green). A first NMR acquisition determines the amount of unbound Xe as a reference signal. Next, a selective saturation pulse affecting only the caged atoms destroys their magnetization. Since the Xe binding is a reversible process, a long pulse cancels the magnetization of many atoms and a second NMR acquisition reveals a significant signal decrease from free Xe compared to the reference signal. Click here to view movie.