Hyperpolarized ¹²⁹Xe Lung MRI and Spectroscopy in Mechanically Ventilated Mice

Hyperpolarized xenon MRI can quantify regional lung microstructure (air-space dimensions) and physiology (ventilation and gas exchange) in translational research and clinical care. Although challenging, it can provide comparable pulmonary insights in preclinical studies. This protocol describes the infrastructure and procedures needed to perform routine xenon lung MRI in mice.

Hyperpolarized (HP) xenon-129 (¹²⁹Xe) is an inhaled magnetic resonance imaging (MRI) contrast agent with unique spectral and physical properties that can be exploited to quantify pulmonary physiology, including ventilation, restricted diffusion (alveolar-airspace size), and gas exchange. In humans, it has been used to evaluate disease severity and progression in a variety of pulmonary disorders and is approved for clinical use in the United States and United Kingdom. Beyond its clinical applications, the ability of ¹²⁹Xe MRI to noninvasively assess pulmonary pathophysiology and provide spatially resolved information is valuable for preclinical research. Among animal models, mice are the most widely used due to the accessibility of genetically modified disease models. Here, ¹²⁹Xe MRI is promising as a minimally invasive, radiation-free, and sensitive technique to longitudinally monitor lung disease progression and therapy response (e.g., in drug discovery). This technique can extend to preclinical applications by incorporating an MRI-triggered, free-breathing apparatus or mechanical ventilator to deliver gas. Here, we describe the steps and provide checklists to ensure robust data collection and analysis, including creating a thermally polarized xenon gas phantom for quality control, optimizing polarization, animal handling (sedation, intubation, ventilation, and care for mice), and protocols for ventilation, restricted diffusion, and gas exchange data. While preclinical ¹²⁹Xe MRI can be applied in various animal models (e.g., rats, pigs, sheep), this protocol focuses on mice due to the challenges posed by their small anatomy, which are balanced by their affordability and the availability of many disease models.

While pulmonary disorders remain the leading causes of global morbidity and mortality¹, the last decade has seen dramatic improvements in patient outcomes. These improvements are driven in part by two factors. Firstly, Phase III clinical trials now prioritize changes in lung function as endpoints rather than mortality, accelerating drug trials^2,3,4,5. Secondly, advancements in improved animal models have provided insights into disease mechanisms and aided therapy development^6,7. Mouse models are often favored for translational research because they offer physiological parallels to humans, affordability, and rapid disease development. Genetic engineering has expanded the range and quality of available models, with the International Mouse Strain Resource now boasting over 32,000 mouse strains⁸, compared to only 4,218 rat strains (Rat Genome Database⁹). These models have opened new avenues for investigating mechanistic drivers and therapy responses for a range of lung diseases, including chronic obstructive pulmonary disease (COPD)¹⁰, cystic fibrosis (CF)¹¹, pulmonary fibrosis^12,13, pulmonary hypertension^14,15, and asthma¹⁶.

Unfortunately, lung research involving mice is limited by the techniques available to quantify disease burden. Studies often rely on terminal procedures that 1) provide whole-lung information (biochemical assays) or localized information (histology) and 2) demand cross-sectional designs and large sample sizes. Thus, they capture neither spatial nor temporal disease dynamics. In contrast, non-invasive, three-dimensional imaging can assess structure, molecular processes, and function in the lungs over time.

Lung structure (e.g., airway abnormalities and interstitial fibrosis) can be visualized with ultra-short echo-time (UTE) MRI and microcomputed tomography (µCT) at high resolution. Functional and mechanistic information (e.g., ventilation, perfusion, tumor metabolism, and inflammatory processes) can be obtained with exogenous contrast agents (e.g., xenon-enhanced CT and oxygen-enhanced UTE) and ionizing nuclear medicine approaches (i.e., positron emission tomography [PET], and single-photon emission computed tomography [SPECT]). However, functional imaging is challenging due to the modest contrast-to-noise (particularly for oxygen-enhanced UTE at the high magnetic field strengths used for preclinical MRI, where T₁ is lengthened) available without employing ionizing modalities with higher than normal levels of radiation. While imaging with these modalities is well tolerated in animal models using conventional doses, cumulative radiation may confound results in studies on immunology, inflammation, and lung cancer¹⁷. However, hyperpolarized (HP) xenon-129 (¹²⁹Xe) magnetic resonance imaging (MRI) provides minimally invasive, non-irradiating, and highly sensitive structural and functional information. While this technique has been employed in preclinical research to characterize conditions including emphysema^18,19, fibrosis²⁰, lung cancer²¹, COPD²², and radiation-induced lung injury²³ at single or multiple time points, it remains underutilized in the preclinical setting.

To enable routine, preclinical ¹²⁹Xe MRI, several prerequisites are required, including institutional regulatory support, a hyperpolarization device, a ¹²⁹Xe-tuned radiofrequency (RF) coil, and a multi-nuclear-capable scanner. Although advanced applications^{24,25,26,27,28,29,30,31,32,33} require vendor-specific pulse programming that is outside of the scope of this protocol, basic applications can be achieved with modest software modifications. Therefore, we focus on quality control, magnetization handling, data collection, and animal handling procedures — including mechanical ventilation — that are unique to preclinical ¹²⁹Xe MRI (Figure 1).

To date, small animal ¹²⁹Xe imaging has employed three MR-safe gas delivery approaches, each with advantages and disadvantages: free-breathing, piston-driven, and pressure-drop. Free-breathing allows spontaneous inhalation without risk of injury from intubation or tracheostomy but consumes significantly more HP gas and can introduce motion artifacts^34,35. Commercial piston-driven devices are self-calibrating and easy to use out-of-the-box but may be prohibitively expensive³⁶. The pressure-drop-based approach used here is well described in the literature, modular, customizable, and run by open-source code^37,38,39,40. Furthermore, it is cost-effective, typically totaling less than $10k and a few weeks of dedicated build time. The pressure-drop ventilator delivers ¹²⁹Xe from a dose bag within a pressurized cannister while monitoring the airway pressure of an intubated mouse.

Figure 1: Overview of the protocol to collect routine xenon-129 (¹²⁹Xe) magnetic resonance imaging (MRI) in mice. (A) Steps for initial setup. (Note: scanner programming is unique to each vendor and not described in this protocol). (B) Steps to collect daily quality assurance (QA) and animal data. (C) Steps for successful experiment conclusion and data analysis. Please click here to view a larger version of this figure.

Here, we collect and analyze the three common classes of ¹²⁹Xe MRI data: ventilation, diffusion-weighted imaging (alveolar-airspace size), and gas exchange. Ventilation images depict the distribution of inhaled ¹²⁹Xe gas. Regions of the lungs with reduced airflow appear dark in HP gas images, and pathology is quantified by the volume of defective ventilation. In humans, the ventilation defect percentage (VDP) has shown strong repeatability^41,42 and high sensitivity to lung obstruction in diseases like COPD^43,44,45 and asthma^46,47.

The restricted diffusion of the ¹²⁹Xe atoms in the airspace can be measured via the apparent diffusion coefficient (ADC) and serves as a surrogate for air-space size. The ADC is calculated by acquiring a baseline image (b₀) without diffusion weighting and one or more images acquired in the presence of bipolar gradient-induced diffusion weighting (b_N). An elevated ADC reflects an increase in airspace size due to aging or emphysematous remodeling^18,48. Further, using multiple b-value images (≥4) allows more detailed morphometric information (e.g., mean linear intercept) to be calculated^49,50.

Gas exchange can be characterized due to 1) the solubility of ¹²⁹Xe in the capillary membrane tissue, plasma, and RBCs (red blood cells) and 2) the >200 ppm downfield chemical shift of ¹²⁹Xe when dissolved in these compartments. Both spectroscopic and imaging data provide insight into cardiopulmonary diseases (e.g., pulmonary hypertension and left heart failure^51,52,53). While many species (humans, canines, and rats) display unique spectral peaks originating from each compartment, mice lack a unique RBC signal due to differences in hemoglobin-xenon binding site interactions. Instead, all dissolved components are combined into a single signal in mice⁵⁴. However, it is possible to observe a distinct RBC resonance in transgenic mice expressing human hemoglobin, such as those used in models of sickle cell disease⁵⁴. Overall, dissolved ¹²⁹Xe spectroscopy and imaging provide unique insights into cardiopulmonary pathophysiology in mice^55,56.

Before attempting this protocol, it is necessary to understand background information about the MRI scanner, mechanical ventilation, and mouse handling techniques required for mouse studies. Prior to initiating animal studies, all procedures must be approved by the local Institutional Animal Care and Use Committee (IACUC)⁵⁷. Because the total magnetic moment available in the mouse lung is intrinsically low (i.e., tidal volume ~250 µL), voxel size must be 1000-fold smaller than in humans to achieve anatomically equivalent resolution. The murine breathing rate is also exceedingly rapid (>100 breaths/minute). As such, the single-breath-hold procedures typically used for human imaging are not feasible. Instead, only a few RF excitations can be applied within each breath, so ¹²⁹Xe images must be encoded over tens to hundreds of breaths. Pulse programming may be required to permit external triggering of acquisitions and to properly loop slices, phase encodes, and/or diffusion-weighted images while balancing signal-to-noise ratio (SNR), resolution, and scan duration. Here, the ventilator outputs a transistor-transistor logic (TTL) pulse once per breath to trigger data acquisition (Figure 2).

Figure 2: Representative mechanical ventilation and data acquisition timing. (A) User-controlled ventilation can trigger data acquisition at end-inspiration, during the breath hold, or at end-expiration. (B) For this 3D radial ventilation sequence, the user defines the total number of projections acquired and the number of projections per breath. (C) For a slice-selective, 2D diffusion-weighted image, the user defines the order of slices, b-value images, and phase encodes. Please click here to view a larger version of this figure.

To enable reliable ventilation and ¹²⁹Xe delivery, robust sedation and intubation procedures are required. For each study, the downstream effects of each anesthetic must be considered - including changes to minute ventilation, heart rate (HR), and blood pressure^{58,59,60,61,62,63,64,65,66}. While a variety of sedatives have been used for preclinical HP gas MRI, we employ a mixture of ketamine, xylazine, and acepromazine, due to its availability, cost-effectiveness, reliability, and duration^67,68. Once sedated, animals must be intubated for effective mechanical ventilation. Intubation of mice is difficult due to the small size of their anatomy, and thus, it is important to train thoroughly in this technique. We encourage investigators to review published video protocols^69,70. Because most commercial intubation cannulas contain stainless steel, we introduce a technique to craft metal-free (i.e., MRI and HP-gas compatible), wedge-shaped cannulas that can be customized to match airway diameter to create an airtight seal with the mouse tracheal wall.

Because ¹²⁹Xe images are collected over many breaths, ventilator settings are critical. Protective ventilation strategies must be carefully considered to prevent lung injury^71,72,73,74. In particular, the use of low tidal volume (TV), moderate positive end-expiratory pressure (PEEP), and alveolar recruitment maneuvers (RMs) reduce the risk of ventilator-induced lung injury in human patients and animal models^{75,76,77,78,79,80,81}. Here, we recommend a simple technique that is compatible with pressure-drop ¹²⁹Xe mechanical ventilation that is protective and provides sufficient ¹²⁹Xe image SNR. Specifically, we apply PEEP by adding a commercial PEEP valve to the exhale line of the ventilator. To perform RMs, the exhalation line must be closed so that the animal receives multiple inhalations without exhalation until a target pressure and duration have been reached.

Throughout, we provide general ventilation settings, but it is advised to review the literature to address specific study goals^82,83. In addition to monitoring the peak inspiratory pressure during mechanical ventilation, it is important to monitor the animal's temperature, which can be done using standard mouse temperature monitoring methods. While not required for imaging, monitoring heart rate via electrocardiogram (ECG) can be advantageous; ECG can indicate if an animal is waking from sedation, overdosed, or distressed, allowing the researcher to intervene.

The protocol we describe is designed to collect ¹²⁹Xe 3D radial ventilation data⁶¹, 2D GRE diffusion-weighted data⁷⁶, and dynamic pulse-acquire spectroscopy gas exchange data. This protocol aims to bridge the gap between preclinical research in small animal models and the potential for ¹²⁹Xe MRI to advance our understanding of pulmonary disorders.

All methods described here were approved by the Institutional Animal Care and Use Committee (IACUC) of Cincinnati Children's Hospital Medical Center.

1. Initial site preparation

1. Create and test a thermally polarized ¹²⁹Xe gas phantom (Figure 3).
   1. Obtain a borosilicate glass vessel (~60 mL), a plunger valve with a front-sealing O-ring, and a ground borosilicate glass stem, all rated up to 150 psig. Ensure there are no magnetic parts. Attach a compression fitting to the glass stem. Tighten to produce a gas-tight seal.
   2. Connect the vessel to a vacuum pump and oxygen reservoir according to Figure 3A. Vacuum vessel to less than 100 mTorr absolute pressure.
   3. Fill the vessel with oxygen to the pressure of 1.5 atm to reduce the ¹²⁹Xe T₁ from > 30 min to ~2 s (at 7 T magnetic field strength; For a 9.4 T scanner, use 1.6 atm oxygen). Seal vessel.
      NOTE: For higher field strengths, somewhat higher oxygen partial pressure will be required to achieve T₁≤ 2 s⁸⁴.
   4. Fill a gas-impermeable reservoir with 400 mL of isotopically enriched xenon (85% ¹²⁹Xe).
      NOTE: Natural abundance xenon (26% ¹²⁹Xe) can also be used, but signal taveraging will need to be increased to mitigate the ~3-fold reduction in SNR.
   5. Connect the vessel to the reservoir of ¹²⁹Xe. Vacuum tubing to less than 100 mTorr absolute pressure.
   6. Fill an open-mouth, benchtop liquid N₂ Dewar to ~90%. Submerge the bottom of the vessel (~5 cm) in liquid nitrogen to condense O₂ and create a vacuum (Figure 3B). While submerged, open the valve to allow ¹²⁹Xe from the reservoir to flow into the vessel (Figure 3C).
   7. Seal the plunger valve by slowly pulling the stem until the inflow hole in the plunger is pulled past the O-ring. Immediately after the hole has passed the O-ring, hand tighten to seal the vessel. Remove the vessel from liquid nitrogen and allow it to thaw.
      NOTE: Once thawed, the vessel will pressurize to ~4.5 atm (2 atm O₂ + 2.5 atm ¹²⁹Xe).
   8. Protect glassware (e.g., insert vessel in padded polyvinyl chloride (PVC) pipe container, Figure 3D).
      NOTE: If properly maintained, the phantom can maintain pressure for a decade or longer.
   9. Measure the T₁ of the phantom (e.g., using a spectroscopic inversion recovery sequence). Confirm T₁ < 2 s for 7 T scanners. Track signal and T₁ over time for quality assurance (QA).

Figure 3: Creation of a thermally polarized ¹²⁹Xe gas phantom guided by the protocol detailed in Step 1.1. The O₂ and ¹²⁹Xe partial pressures can be altered to customize the T₁ to yield appropriate ¹²⁹Xe T₁ times and signal strength at a given field strength⁸⁴. Please click here to view a larger version of this figure.

2. Perform quality assurance with the thermally polarized gas phantom (Table 1 and Figure 4).
   1. Place the ¹²⁹Xe coil at the isocenter of the magnet and center the ¹²⁹Xe gas phantom within the coil. In a single pulse-and-acquire sequence ("single pulse"), set the working frequency to match the approximate ¹²⁹Xe gas frequency in the phantom (~83.07 MHz at 7T).
   2. Center the acquisition and excitation frequencies to the ¹²⁹Xe Larmor frequency, and use this frequency for all phantom calibration and quality control ¹²⁹Xe scans. See Table 1 for experimental parameters for all QA scans. Confirm that the phantom is centered with a phantom localizer (Figure 4A).
   3. Perform nutation experiment to calibrate flip angle: Assuming SNR is sufficient, use single RF pulses with repetition time (TR) spacing > 5 x T₁. For each acquisition, incrementally increase RF power until the signal nulls and begins to invert. The standard used here is: number of pulses = 65; TR = 10 s; pulse duration = 125 µs; RF power = 1-13.8 W, incremented by 0.2 W
   4. Fourier transform and phase the first spectrum (i.e., the spectrum acquired with the lowest RF power). Apply the same phasing for all spectra. Plot the real spectra as a function of the RF pulse power (Figure 4B).
   5. The power that produces a null peak (i.e., minimum peak height) corresponds to the 180° flip angle. Achieve a 90˚ flip angle by using the same power at half the pulse length needed to produce the 180˚ flip angle. Assuming the scanner software allows, set this 90° reference power and pulse length for subsequent flip angle scaling.
   6. Use a single pulse to shim by minimizing the full width at half maximum of the ¹²⁹Xe spectrum (TR ~ 1 s). If necessary, recenter the frequency after shimming. Record the full-width half max.
   7. Run the ¹²⁹Xe QA scan (Table 1 and Figure 4C). Record QA data: SNR, mean phantom signal, and standard deviation of the noise.

Figure 4: Pre-scan quality assurance. (A) A low-resolution 2D GRE coronal phantom localizer ensures the phantom is centered in the magnet. (B) A nutation experiment to set a 90˚ pulse shows a null peak at the 180° pulse. (C) After localizing and calibrating the flip angle, acquire a higher resolution 2D GRE QA image. Please click here to view a larger version of this figure.

Table 1: Phantom calibration quality assurance sequence parameters. TR = repetition time, TE = echo time, N_pts = number of points, FOV = field of view, BW = bandwidth. Please click here to download this Table.

3. Plan polarization (Figure 5A, B).
   1. Select polarized ¹²⁹Xe volume and accumulation time: 400 mL in 15 min is optimal for this protocol (Figure 5) but can easily be adjusted for other applications and equipment.
   2. Assuming a known dead volume within the hyperpolarizer (e.g., V_{dead vol} = 80 mL), calculate the flow rate (F_Xe in SLM) for a dispensed volume V_desired, ¹²⁹Xe gas fraction f, and accumulation time t_acc:
           (1)
      NOTE: While longer production times will typically yield higher polarization, they may not be practical for in vivo imaging. Use an appropriate model for the hyperpolarizer^85,86,87,88 to determine a flow rate that will balance production time and polarization. Here, the model of J. W. Plummer et al.⁸⁹ (Figure 5A) was used. This applies to continuous-flow polarizers and is not applicable to stopped-flow hyperpolarizers⁹⁰.
   3. Polarize gas according to these parameters, measure the polarization with a commercial or homebuilt device and compare it to the predicted polarization for QA.
   4. Measure the polarization loss during transportation. If polarization decreases by a sufficiently large amount (e.g., >10%), build a magnetic carrying case to protect polarization during transportation. See Supplementary File 1: Managing polarization during transportation and Supplementary Figure 1.

Figure 5: Polarization management. (A) Polarization and produced volume are a function of accumulation time and flow rate. A 400 mL bag of gas provides high initial polarization (~35%) over 20 min. While using 1 L of gas might seem attractive, it will have a lower initial polarization (~20%). (B) After ~15 minutes of ventilation, a 1 liter batch of HP ¹²⁹Xe would deplete to <10% polarization while 600 mL of gas remained¹¹⁶. Thus, using multiple 400 mL bags of ¹²⁹Xe maintains higher average delivered polarization. C) Locations where the primary field and active shielding field intersect (red box at position (N,N,N)) can cause rapid relaxation of HP ¹²⁹Xe. Characterizing the magnet's fringe field helps identify safe zones where reservoirs of HP ¹²⁹Xe can be placed without rapid relaxation (green box at position (0,0,n)). Please click here to view a larger version of this figure.

Supplementary File 1: Managing polarization during transportation. Please click here to download this File.

4. Measure the position-dependent T₁ of HP ¹²⁹Xe within the fringe field (Figure 5C).
   1. Create reference points with known distances and positions relative to the magnet isocenter along the X, Y, and Z dimensions. Label isocenter and label the other positions n through N. The number of positions to investigate will depend on the space available.
   2. Hyperpolarize a small volume of ¹²⁹Xe (~250 mL) and transport it to the MRI control room. Fill a large syringe (50-100 mL) with ¹²⁹Xe and place it at the isocenter within the magnet (Position 1 in Figure 5C). Play out a ~1˚ flip angle pulse to measure the signal.
   3. Leave the syringe in position for ~10 min, then acquire another spectrum. Remeasure the signal every 10 min until the signal has decayed at least 1 T₁ (i.e., the signal has decayed to ~1/3rd its initial value).
   4. Start a new T₁ experiment with a new syringe of ¹²⁹Xe by repeating Step 1.4.2.
   5. Move the syringe to a new position (e.g., Position n in Figure 5C) and leave it there for 10 min. Return the syringe to the isocenter to acquire an additional ~1˚ flip angle spectrum.
   6. Repeat this process: move the syringe to Position n, wait for 10 min, put it back to the isocenter, and remeasure the signal until it has decayed at least 1 T₁.
   7. Repeat Steps 1.4.4 - 1.4.6 for the remaining reference positions covering the X, Y, and Z directions.
   8. The initial signal (S₀) will decay monoexponentially over n RF pulses with a θ flip angle. Fit the signal (S) as a function of time (t) in the fringe field to calculate the T₁ in each position:
           (2)
   9. Identify a position with sufficient T₁ (>20 min) for ¹²⁹Xe reservoir placement.
5. Create metal-free intubation cannulas (Figure 6).
   1. Obtain two veinous indwelling polytetrafluoroethylene (PTFE) catheters with Luer connectors. Discard needles into a sharps disposal container.
      ​NOTE: For mice >25 g, use 18 G and 20 G catheters. For smaller animals, use 20 G and 22 G catheters.
   2. Cut the Luer connector off the catheters. Feed the smaller catheter into the upper end of the larger catheter to create a more sharply tapered and longer cannula. Cut the composite cannula to ~4.6 cm with a beveled end, not including the Luer base (Figure 6A, B).
   3. Cut the wider end off of a 200 µL pipette tip to a length of ~2.6 cm (Figure 6C).
   4. Coat the inside of the pipette tip with general mold release lubricant. Use another pipette tip inserted inside to thinly distribute the lubricant. Fill the pipette tip with acetoxy silicone vulcanizing paste epoxy (Figure 6D, E).
   5. Plug the cannula with 22 G wire extending out of both ends. Feed the cannula tube through silicone in the pipette tip. Extend the tube ~7 mm past the end of the pipette tip (Figure 6F, G).
   6. Slide the cannula tube on the wider side of the pipette tip through a plastic male slip Luer connector, gluing the pieces together with the epoxy. Trim tubing that extends past the Luer connector (Figure 6H).
   7. Wait for the epoxy to dry (>24 h), then carefully remove the silicone intubation cannula from the pipette tip mold. Remove the wire from the cannula, ensuring the tube has not been occluded (Figure 6I).
   8. To make a handle for easy intubation, connect tubing (1/16" or 1/8") to a female Luer connector. When ready for intubation, connect this female Luer connector to the male Luer intubation cannula. This piece can be easily detached post-intubation (Figure 6J).
   9. Sanitize before each use on animals: wipe the outside of the cannula with 70% alcohol. Wipe 20 G wire with sanitizer then feed the wire through the cannula to sanitize the inside and ensure no blockages.

Figure 6: Creating MRI and HP ¹²⁹Xe compatible mouse intubation cannulas. These cannulas are constructed of venous catheters, pipette tips, and silicon sealant, as described in Step 1.5. Please click here to view a larger version of this figure.

2. Daily data collection

NOTE: See Supplementary File 2: Preclinical scan QA checklist.

Supplementary File 2: Preclinical scan QA checklist. Please click here to download this File.

1. Complete daily scanner quality control and setup ventilator (Figure 4).
   1. With phantoms, run QA protocols on the scanner (see Table 1 for QA scan parameters and Step 1.2 for daily quality control steps).
   2. Calibrate Ventilator according to the method of J. Nouls et al.³⁸. See Supplementary File 3: Ventilator calibration, Supplementary Figure 2, and Supplementary Figure 3.
   3. Set the ventilator settings for imaging at end-inspiration (Table 2). Place the animal bed on the scanner rack and the life support module (i.e., mechanical ventilator parts) on the table next to the scanner.
   4. Activate the site-specific animal heating system. Set a heater to 35.5 - 40 °C, turn on circulating air, and place the air hose within ~5" of where the animal's head will rest to pre-warm the bore of the scanner.

Table 2: Recommended ventilator settings for ¹²⁹Xe imaging. Parameters can be fine-tuned for specific study objectives and experimental conditions^{117,118,119,120,121,122,123,124}. Please click here to download this Table.

Supplementary File 3: Ventilator calibration. Please click here to download this File.

2. Sedate and intubate animal.
   1. Turn on the incubator to 27.7 °C and/or the electric heating pad to 37.7 °C. Measure and record the body mass of the animal. Calculate sedative dosing based on mass. See Table 3 for a typical dosing regimen.
   2. Inject the sedative intraperitoneally. Note the time of injection and set the timer for the next dose of sedative.
      NOTE: Complete the remaining steps in Section 2 (Daily data collection) as quickly as possible to minimize the time under sedation and the risk of overdose.
   3. Apply eye lubricant to the animal's eyes and place the animal in a cage on the heating pad or within an incubator to prevent hypothermia.
   4. Confirm the animal is fully sedated by performing a toe pinch test 10-15 min after sedative injection⁶⁸. Intubate following procedures outlined in Das et al.⁶⁹.
      NOTE: The article by Das et al.⁶⁹ is accompanied by a comprehensive video demonstration of the technique. The steps are as follows:
   5. Hang the animal supine by the teeth on a slanted board. Use a rodent tongue depressor to pull out the tongue.
   6. To ensure vocal cords are visible, supply white light via a fiber optic cable within the intubation cannula or a bright light placed on the outside of the throat. Insert the cannula less than 5 mm past the vocal cords.
   7. Ensure the cannula is in the trachea, not the esophagus, by connecting it to a piece of tubing with a small droplet of water inside. If the water droplet moves in time with the animal's breaths, the positioning is likely correct.

Table 3: Common anesthetic formulary for mice. Please click here to download this Table.

3. Ventilate animal (Table 2).
   1. Attach the animal to ventilator via Luer connector on the intubation cannula. Monitor diaphragmatic motion and peak inspiratory pressure (~10-12 cm H₂O for a tidal volume of 10 mL/kg of body weight). If the pressure or breathing motion is abnormal, carefully adjust the neck angle and depth of the cannula as needed.
   2. Ensure the intubation cannula is airtight by performing a recruitment maneuver: prevent exhalation (e.g., with a finger, block the exhale port) such that the animal inhales multiple times without exhaling.
      NOTE: If the airway pressure reaches 35 cmH₂O peak inspiratory pressure over ~6 s, the airway seal is sufficiently tight. If not, see the discussion for troubleshooting.
   3. Allow normal exhalation to resume. Perform recruitment maneuvers between scans and every ~5 min when not scanning to maintain lung compliance and prevent atelectasis. Connect the PEEP valve to the exhale line. Set PEEP to 4 cmH₂O. Watch the peak inspiratory pressure increase by that amount.
   4. After successful intubation, plan and initialize the production of HP ¹²⁹Xe around the sedative redose schedule to prevent the animal from waking during a scan. Monitor body temperature throughout the experiment.
4. Data acquisition: Collect ventilation images.
   NOTE: Data Acquisition Steps 2.4, 2.5, and 2.6 can be done in any order
   1. Set the ventilator according to Table 2 for imaging at end-inspiration.
   2. Load the following setup protocols: proton animal localizer, single pulse to center on gas frequency in the mouse lungs, and ¹²⁹Xe animal localizer. See Table 4 for scan parameters.
   3. Position the animal at the isocenter and confirm that the thoracic cavity is in the center of the field of view with proton localization. If using single-frequency coils, replace the proton coil with a ¹²⁹Xe-tuned RF coil.
   4. Record ¹²⁹Xe polarization and transport it to the MRI scanner. See Supplementary File 4: Xenon polarization QA checklist.
   5. Place a bag of ¹²⁹Xe within the ventilator cannister and seal. Connect the canister in line with the ventilator and allow the cannister to pressurize (3 - 6 psig).
   6. Initiate ¹²⁹Xe mechanical ventilation. Each time ¹²⁹Xe ventilation is activated, allow the animal to complete ~5 breaths prior to starting a scan to turn over the functional residual capacity of the lungs.
      NOTE: Switch to the N₂/O₂ mixture between ¹²⁹Xe scans to conserve hyperpolarized gas.
   7. Using a single pulse, adjust the working frequency to match the in vivo resonance frequency of gaseous ¹²⁹Xe (~83.07 MHz at 7 T). Copy the frequency to all subsequent gas phase ¹²⁹Xe scans. Perform ¹²⁹Xe localization to confirm lungs are at isocenter.
   8. Load and run the ¹²⁹Xe radial ventilation sequence. Monitor peak inspiratory pressure.
      NOTE: If the ¹²⁹Xe gas runs out before the protocol is finished, the peak inspiratory pressure will decline rapidly. A 400 mL bag of ¹²⁹Xe can ventilate a 30 g mouse for ~24 min when ventilated with 70% ¹²⁹Xe at 80 breaths per minute with a tidal volume of 10 mL/kg of ideal body weight.
   9. When the scan is finished, switch to ventilating with the N₂/O₂ mixture and remove the empty bag of ¹²⁹Xe.
   10. For imaging at end-expiration, change the breath hold duration and trigger delay according to Table 2 and repeat Steps 2.4.2 - 2.4.9.
   11. Export the raw data from the scanner.

Table 4: In vivo sequence parameters. The 3D multi-echo radial ventilation sequence described previously³⁹ acquires images at 6 echo times. Results are shown for the first echo image (TE = 1.12 ms, Figure 7). Please click here to download this Table.

Supplementary File 4: Xenon polarization QA checklist. Please click here to download this File.

5. Data acquisition: Run dynamic dissolved phase spectroscopy.
   1. Set the ventilator according to Table 2 for imaging during breath hold. Set the BR to 100 breaths/min. Prepare for a new bag of ¹²⁹Xe as in Steps 2.4.2 - 2.4.5.
   2. Initiate ¹²⁹Xe mechanical ventilation. Load and run a single pulse to adjust the working frequency to match the dissolved frequency (~83.084 MHz at 7 T). Copy the working frequency to the dissolved phase dynamic spectroscopy.
      NOTE: This will be a single peak in mice with wild-type hemoglobin⁵⁴.
   3. Load and run the dynamic spectroscopy sequence centered on the dissolved frequency in animal lungs. When the scan is finished, switch to ventilating with the N₂/O₂ mixture and remove the empty bag of ¹²⁹Xe. Export the raw data from the scanner.
6. Data acquisition: Collect diffusion-weighted images.
   1. Set the ventilator according to Table 2 for imaging during breath hold. Prepare for a new bag of ¹²⁹Xe as in Steps 2.4.2 - 2.4.7.
   2. Load and run the diffusion-weighted sequence. When the scan is finished, switch to ventilating with the N₂/O₂ mixture and remove the empty bag of ¹²⁹Xe. Export the raw data from the scanner.

3. Concluding the experiment

1. Recover animal.
   1. Pull the intubation cannula straight out of the mouth of the animal. If the animal does not immediately begin breathing spontaneously, administer light chest compressions. If available, administer a light flow of medical-grade oxygen near the animal's face to support sedation recovery.
   2. Once the animal is steadily breathing on its own and if the animal was sedated for >2 h, administer 0.5 - 1 mL of normal saline subcutaneously to prevent dehydration.
   3. Return the animal to a cage by itself. Place the cage in an incubator or on a heating pad.
      NOTE: Sedated animals are vulnerable to cannibalism and cannot be placed in a cage with cage mates until fully recovered (i.e., ambulating independently). Sedated animals cannot regulate their body temperature. Use the back of the hand to feel the temperature of the animal every few minutes.
   4. Monitor the animal closely until their righting reflex has returned (i.e., it can independently flip from a supine to a prone position).
   5. Remove the animal from heat support once it can ambulate independently. Return the animal to a cage with its cage mates.
      NOTE: if the animal has no cage mates, it is more susceptible to sedation-induced hypothermia. Provide the animal with extra bedding, and if available, leave the animal in an incubator overnight.
   6. Record the animal's weight once per week for 2 weeks to monitor its health.
      ​NOTE: If the animal sustained an injury to the mouth or esophagus from intubation, it may stop eating. If the animal loses >20% of its initial body weight, consult a veterinarian about euthanasia.
2. Analyze ventilation images (Figure 7).
   1. Load raw data into a programming platform. Download the open-source reconstruction framework for non-Cartesian Images⁹¹.
   2. Reconstruct images according to open-source framework instructions. Normalize the first point on each radial projection³⁹.
   3. Segment the lung parenchyma in the images, including voxels with low or no ¹²⁹Xe signal. Do not include large airways. Segment the background noise in the image, excluding lungs, airways, and imaging artifacts.
      NOTE: The proton localizer image can aid in determining parenchymal boundaries.
   4. Calculate the SNR using the formula:
           (3)
   5. Quantify defective ventilation.
      NOTE: Various methods have been proposed to quantify impaired ventilation in small animal models. Analysis methods remain an open area of inquiry, but easily implemented approaches include: (i) Semi-quantitative manual segmentation⁹², (ii) Histogram approach using the signal from the trachea to normalize parenchymal signal⁴⁷, and (iii) Segmenting the total lung volume (TLV) and defining a signal threshold (e.g., <60% of the whole lung mean) to divide the lungs into defective volume and ventilated lung volume (VV). Quantify VDP^93,94 according to:
           (4)
3. Analyze dissolved phase dynamic spectroscopy (Figure 8).
   1. Load raw data into a programming platform. Perform a fast Fourier transform and phase the spectra (simultaneous manual, zero-order phasing of spectra is sufficient for this application.)
   2. Calculate standard nuclear magnetic resonance (NMR) data^95,96: SNR, full width half max, integrated area, chemical shift, and phase of both peaks.
   3. From the magnitude data, divide the signal amplitude of the dissolved spectrum by that of the gas spectrum for each repetition to find the dissolved-to-gas ratio over time.
4. Analyze diffusion-weighted images (Figure 9).
   1. Load raw data into a programming platform. Segment the lung parenchyma of the b₀ image as in Step 3.2.3. Calculate the SNR for each b-value image.
   2. Calculate the signal-value-to-noise ratio, SVNR_0, by dividing the signal in each voxel of the b₀ image by the standard deviation of the image noise. Exclude voxels with SVNR₀ < 2.5 times the image noise⁹⁷.
      NOTE: The SVNR₀ is an individual voxel metric, while the parenchymal SNR is a whole-lung metric.
   3. Calculate ADC by fitting the signal decay over the b-values (b_i) according to Equation 5^98,99:
           (5)

Ventilation Images

If animal preparation and ventilation procedures are implemented properly, 3D radial imaging can successfully capture ventilation patterns when data acquisition is performed at either inspiration or expiration (Figure 7). While these images are collected over many breaths, the method described here is similar to the single-breath imaging method used in humans. This is because all lines of k-space are collected at a specific time during the breath cycle (e.g., all at end-inspiration), cumulatively creating an image representative of a single inhalation. In these images, the major airways (trachea and bronchi) are clearly visible, no obvious image artifacts are present, and the lung parenchyma appears fully inflated with high SNR (~15 or higher, depending on initial polarization). While counterintuitive, the images acquired at inspiration (Figure 7A, SNR = 15) display lower SNR than those acquired at expiration (Figure 7B, SNR = 27). This is related to the modest driving pressure and slow flow used when ventilating this mouse. For inspiration images, the HP gas had not yet fully washed into the parenchyma of the lungs, causing the ¹²⁹Xe (and thus signal) to be concentrated in larger airways. This effect is visually exacerbated by k-space scaling during image reconstruction because the high airway signal dominates the dynamic range of the data. Nonetheless, the analysis should be straightforward if the parenchyma is segmented such that the large airways are omitted.

Figure 7: Representative 3D radial ventilation images in an adult C57BL/6J mouse acquired using previously published methods³⁹. While both images have relatively high SNR, images at (A) inspiration have lower SNR than images at (B) expiration largely because the ¹²⁹Xe gas had not yet had the time to wash into the lungs. Please click here to view a larger version of this figure.

It is important to distinguish between pathologically obstructed ventilation and other causes of low SNR. In Figure 7A, no ventilation obstruction is expected (i.e., low VDP) because the animal was a healthy, wild-type mouse. If ventilation were obstructed, defects would typically appear as regions with low to no ¹²⁹Xe signal. Nonpathological causes of low signal commonly include:

(i)  Low ¹²⁹Xe polarization: This causes low SNR in lung parenchyma and airways without accompanying artifacts. Low polarization can typically be attributed to incorrectly following polarization procedures, depolarization during transport, or excessive O₂ gas content in the HP ¹²⁹Xe delivery tubing (e.g., from an incorrectly calibrated ventilator).

(ii)  Intubation cannula inserted too deeply: If inserted too deeply into the trachea, the cannula may deflect down the left or right bronchus. If lodged tightly enough, the whole-lung tidal volume may be confined to only a few lobes and airways while the rest of the lung receives none. Should this occur, the hyperinflated lobes could be injured, and it may be necessary to euthanize the animal. Alternatively, the tip of the cannula may come in direct contact with tissue. In this case, ¹²⁹Xe will be primarily confined to the interior of the cannula, resulting in a bright signal from a visibly straight tube and minimal signal within the parenchyma. In this case, the cannula position should be corrected promptly, and the animal should be monitored for signs of distress (e.g., abnormal, absent, or asynchronous chest motion compared to the ventilator).

(iii)  Intubation cannula placed too shallowly: When a bolus of gas is delivered too shallowly in the oropharynx, the resistance to that bolus flowing into the lungs can be stronger than the resistance to that bolus flowing around the exterior of the cannula and out of the mouth. This can result in poor lung inflation. Tracheal SNR may be high, while parenchymal SNR will be low. Motion artifacts may be present due to gas flowing up the trachea during data acquisition. The character of the artifact will vary depending on sequence type (GRE, radial, etc.).

(iv)  Incorrect acquisition timing: If RF pulsing and data acquisition occurred during periods of inspiratory or expiratory gas flow, severe motion artifacts will usually be present — particularly when using Cartesian sequences — and parenchymal SNR will be low.

(v)  Spontaneous respiration: If the animal is insufficiently sedated during data acquisition, the spontaneous respiratory motion will be superimposed asynchronously on top of the mechanical ventilation pattern. This will result in variable breath-to-breath ¹²⁹Xe delivery, moderate to severe motion artifacts, and poor parenchymal SNR. Careful tracking of sedative dose and timing is critical.

(vi)  Animal death: In rare instances, mice may die (e.g., due to incorrect cannula position, insufficient O₂ delivery, sedative overdose, or severe experimental pathology). If this occurs, lung tissue stiffens rapidly, and the lung parenchyma will inflate poorly. The upper trachea may remain bright, but lung parenchyma will be very low in signal.

Gas Exchange Spectroscopy

In human subjects, spectroscopy-derived parameters (e.g., signal intensity, spectral width, chemical shift, and phase) offer insights into cardiopulmonary dynamics and disease state^52,53. Unlike in humans⁸⁶, dogs¹⁰⁰, and rats^20,101, whole-lung spectra from mice obtained when ¹²⁹Xe magnetization is confined to the lung parenchyma display a single, broad, dissolved phase resonance corresponding to ¹²⁹Xe dissolved in the capillary membrane tissue, blood plasma, and RBCs⁵⁴ (Figure 8A, 197 ppm). When transported by blood flow to more distal tissues, ¹²⁹Xe produces distinct resonance frequencies from various tissues, including adipose tissue⁵⁴, muscle, gray matter, and white matter^102,103. However, due to relatively long transit times (several seconds) and modest solubility¹⁰⁴ (Ostwald solubility = 0.1 - 0.3), these signals are typically weak and only observed during dedicated, careful experiments. The exception is ¹²⁹Xe dissolved in fatty tissues (Figure 8A, 189 ppm). From these tissues, a ¹²⁹Xe signal can appear within a few seconds (e.g., over the initial few breaths during animal setup). Moreover, signal intensity can be high, because the Ostwald solubility of ¹²⁹Xe is nearly 20-fold greater in adipose tissues than in blood plasma (i.e., 1.7 vs 0.094, respectively, at 37 ˚C¹⁰⁴). The longitudinal relaxation times are also expected to be relatively long (in vivo T₁ is ~20 s in lipids¹⁰⁵ vs. 3-8 s in blood¹⁰⁶). As a result, the ¹²⁹Xe signal from adipose tissue can be relatively high compared to other tissues.

Due to differing signal dynamics, ¹²⁹Xe spectroscopy experiments can be tailored to interrogate adipose tissue (e.g., to target brown adipose tissue thermometry¹⁰⁷) or gas uptake in the lungs by selecting the correct acquisition parameters. When ¹²⁹Xe spectra from the lungs are acquired with the settings described in Table 4 (i.e., TR = 50 ms and a 90˚ flip angle centered on the dissolved phase peak), both gas phase and dissolved phase peaks are observed (Figure 8A,B). Notably, the gas volume in the mouse lung is 2-3x higher than the tissue volume, and the solubility of ¹²⁹Xe in this tissue is only ~10%^{104,108,109,110,111}. Despite this, the amplitude of the dissolved signal exceeds that of the gas phase by approximately 2-fold because the gas phase excitation is far off-resonance (~15 kHz at 7 T; Figure 8B, SNR = 102 vs. 43, respectively). Therefore, the 20-to-30-fold greater magnetization pool in the gas phase can be approximated as a stable (i.e., non-depleted) reservoir for the dissolve-phase signal during static portions of the breath cycle.

With a new breath of magnetization every ~600 ms and a TR of 50 ms, the signal intensity from both gas phase and dissolved phase ¹²⁹Xe oscillates. For the gas signal, this is attributable to the change in total gas volume (i.e., high at inspiration and low at expiration). The situation is more complex for the dissolved signal because the source gas magnetization, T₂*, and capillary blood volume may fluctuate with the inflation state. As a result, the dissolved-to-gas signal amplitude ratio also oscillates with respect to the breath cycle (Figure 8C, mean dissolved:gas ratio = 1.9 ± 0.39).

Figure 8: Representative dynamic spectroscopy in an adult C57BL/6J mouse. (A) A single pulse spectrum shows signal originating from ¹²⁹Xe in the gas phase, dissolved in adipose tissue (189 ppm), and dissolved in both capillary membrane and blood tissues (197 ppm). (B) Dynamic spectroscopy showing the time-varying signal intensity of both dissolved and gaseous ¹²⁹Xe (repetition time = 50 ms). To aid visualization, only five spectra are displayed. Diagonal lines on the time axis represent scale breaks. (C) The ratio of dissolved-to-gas signal amplitude over a 2 s interval (corresponding to the red scale break in panel B shows clear variation with the breathing cycle (0.6 s per breath). Please click here to view a larger version of this figure.

The following is a list of factors that could compromise ¹²⁹Xe gas exchange spectroscopy results. In each case, the likely result is low SNR, which may make the spectral fitting and resulting metrics (e.g., dissolved-to-gas signal amplitude ratio, chemical shift, spectral width, and phase) inaccurate.

(i)  Wrong working frequency: Because of the significantly higher gas volume and low tissue solubility, RF pulses centered on the gas phase frequency will produce no dissolved phase signal. RF pulses should be centered at or near the dissolved phase frequency.

(ii)  Too broad of a bandwidth: The excitation bandwidth should be broad enough that some gas phase ¹²⁹Xe is excited in order to measure the ratio of the dissolved-to-gas signal amplitude. But if the bandwidth is too broad, the excitation pulse will destroy magnetization in the gas phase before it dissolves into the tissue. This can be seen as a very high gas signal and low to no dissolved phase signal despite the pulse being centered on the dissolved phase frequency.

(iii)  Unoptimized TR: To interrogate signal dynamics over a breath, Nyquist-Shannon sampling theorem requires the signal to be sampled at a minimum of twice its frequency (here, 100 breaths per minute were sampled every 50 ms, or ~12x per breath). Depending on experimental conditions, further interrogation of cardiopulmonary signal dynamics at a faster sampling rate is possible. The results of these experiments would depend on many factors, including excitation bandwidth, breathing rate, and heart rate. The xenon exchange time (i.e., the time for xenon to dissolve into tissue; ~56 ms⁵⁴) and pulmonary transit time (i.e., the time for blood to flow from right to left ventricle; ~830 ms for a heart rate of ~436 beats/min¹¹²) must also be considered. Sampling too often will result in the early destruction of magnetization and low signal, but it will also reduce the influence of blood flow. Sampling less frequently will allow the magnetization to dissolve further and build up more signal, but it will become more dependent on blood flow and less localized to the gas exchange regions in the alveoli.

Diffusion-Weighted Images

The diffusion-weighted images, used to quantify alveolar-airspace size, have been rigorously validated in small animal models through traditional histomorphometry methods^98,113,114. Here, we restricted discussion to 2D, slice-selective imaging to minimize scan time and gas usage while matching previously validated protocols⁹⁸. A minimum of 2 b-value images is sufficient for calculating the ADC, and 4 or more b-value images are necessary for advanced morphometry calculations (such as mean linear intercept). The acquisition of multiple b-value images offers a significant advantage by enhancing the accuracy of the calculations. Unlike imaging humans during a single breath hold, preclinical imaging over many breaths provides continuously replenished signal. Because of this, it is possible to acquire many high SNR b-value images, limited only by the volume of HP gas available and animal sedation duration. Here, we acquired 7 b-value images, each with SNR >15, using 400 mL of ¹²⁹Xe gas over 18 min of scan time (Figure 9). The signal in each image decays with increased diffusion gradient strength (b-value). This signal decay is proportional to the ADC of the ¹²⁹Xe within the lung. Large airways can be distinguished from acinar tissue by comparing b₀ and b₆ images. In regions that are anatomically likely to contain airways, pixels will appear very bright in the b₀ image but appear dark in the b₆ image (red arrows, Figure 9). These pixels should be excluded from the analysis of lung parenchyma.

Figure 9: Representative diffusion-weighted images in an adult C57BL/6J mouse. Large airways (red arrows), through which ¹²⁹Xe freely diffuses while diluted with oxygen, display increased signal attenuation relative to the lung parenchyma. This results in bright airways in the b₀ image and low signal airways in the b₆ image. Mice and rats have similarly small ADC values relative to humans, but the values will depend on biological and experimental factors like age, tidal volume, pressure, and diffusion parameters. Please click here to view a larger version of this figure.

The ADC of this healthy, adult, wild-type mouse follows the expected result based on the literature, its age, SNR, acquisition parameters, and more. Factors that may deleteriously impact the precision and accuracy of the ADC results include:

(i)  Low SNR: Importantly, low SNR in the images bias the ADC calculation towards lower values^99,115. For this reason, we exclude voxels with SVNR₀ < 2.5 times the image noise⁹⁷.

(ii)  Improper segmentation: Poor segmentation (e.g., using signal thresholding without visual inspection for quality control) can skew results. If the images contain artifacts (like those caused by motion), they may exceed the threshold and be incorrectly categorized as parenchyma. Further, it is particularly important to exclude pixels near large airways, as partial volume effects may bias their ADC toward higher values.

(iii)  Miscalibrated ventilator: In general, dilution of ¹²⁹Xe with lighter gases (e.g., O₂ or N₂) can shift the ADC toward higher values, so ventilation with known gas mixtures is necessary for quantitative comparisons. If ventilating with a ¹²⁹Xe/O₂ mixture, an increase in oxygen concentration prior to delivery will increase T₁ relaxation and reduce ¹²⁹Xe magnetization in the gas delivery tubing, thus decreasing SNR.

(iv)  Lung derecruitment: In anesthetized and mechanically ventilated animals, alveolar collapse (atelectasis) can occur over time. When this happens during constant volume ventilation, the same tidal volume is necessarily redistributed to the alveoli that remain recruited. As moderate to severe atelectasis develops, the rest of the lung can become hyperinflated, causing injury and increased ADC in those regions. Regular recruitment maneuvers and PEEP prevent alveolar collapse, avoid decreased lung compliance, and minimize the risk of lung injury that can skew ADC results.

(v)  Poor data fitting: The method of ADC estimation must be taken into consideration. Log-linear fitting is the most computationally efficient but introduces bias in low SNR images. Therefore, it is best applied to high SNR images (>20). The weighted linear method reduces this bias and is appropriate in images with SNR > 15. Bayesian estimation can yield estimates of ADC with low uncertainty even with SNR <10 but is computationally expensive⁹⁹.

(vi)  Suboptimal diffusion time: The optimal diffusion time is influenced by the maximum gradient strength of the MRI scanner and the acinar airway radius in mice (radius ~100 µm⁴⁹), which is approximately three times smaller than that in humans. To ensure optimal conditions, the diffusion time is set to achieve a diffusion length larger than the average alveolar radius but smaller than the mean length of the alveolar ducts. The optimal diffusion time for the site-specific maximum gradient strength can be determined using the methods outlined by Sukstanskii and Yablonskiy⁴⁹. Using a suboptimal diffusion time may lead to nonlinear decreases in ¹²⁹Xe ADC, complicating the interpretation of results.

(vii)  Suboptimal b-values: There is a tradeoff between maximizing the SNR and maximizing the contrast-to-noise ratio (CNR) generated across b-value images. If CNR is sacrificed by the use of only low b-values (i.e., diffusive signal decay as if acquiring only the first 3 b-value images in Figure 9), then the ADC estimation will be inaccurate due to poor fitting. If the diffusion weighting is too strong, the resulting SNR will be insufficient, and the ADC will be biased toward lower values.

Hyperpolarized ¹²⁹Xe MRI is emerging as a sophisticated and powerful technique to study lung microstructure and function in small animal models. This protocol is intended to guide initial site preparation and describe experimental procedures needed to quantify ventilation, diffusion, and gas exchange in mouse lungs with HP ¹²⁹Xe. Key prerequisites for experiments include establishing a ¹²⁹Xe gas phantom and ensuring high gas polarization is delivered to the animal. The latter requires careful planning of polarization parameters, mitigating polarization losses during gas transport (e.g., building a magnetic carrying case), and characterizing (and ideally avoiding) T₁ relaxation within the fringe field of the preclinical MRI magnet^116,125. In parallel to developing the necessary HP ¹²⁹Xe infrastructure, it is imperative to establish robust animal handling protocols and procedures. These include effective sedation to enable robust mechanical ventilation, non-injurious intubation to enable longitudinal studies, and ventilation strategies that balance the need to deliver high ¹²⁹Xe magnetization with safe and physiologically relevant ventilation conditions.

To some degree, the methods, procedures, and results described here depend on the specific hardware used (e.g., ventilator, polarizer, and MRI scanner type). If alternative equipment is available, experimental techniques can be modified relatively easily. Further, the daily quality assurance procedures may be optimized (Step 1.2, Figure 4, and Table 1). Meticulous record keeping during the ~30 minutes of QA is critical during the establishment of a new site. It allows early detection of hardware issues and the ability to rule them out during troubleshooting. Once the day-to-day variations in QA metrics are well characterized, the flip angle calibration and ~21-min QA scan may be scaled back for efficiency. Additionally, the general approach can be extended to other common small animal species (e.g., rats and rabbits) by adjusting the anesthetic dosages, ventilator settings, and acquisition parameters. The basic protocols can be used to explore ventilator-dependent physiology like the impact of respiratory rate, volume, and PEEP levels on gas exchange or alveolar recruitment. If longitudinal data are not required or a longitudinal study is completed, the intubation cannula allows for bronchoalveolar lavage or rapid lung fixation and harvest for conventional analyses (e.g., histology or morphometric measurements). Moreover, the potential for modification of this protocol extends to MRI sequences and parameters, representing a dynamic aspect of ¹²⁹Xe MRI research.

Cannula placement is critical to animal safety and image quality. Because the intubation cannula is not rigidly affixed — for example, by suturing around the trachea — relatively small displacements (e.g., movements when connecting physiological monitoring equipment) can break the airtight seal with the trachea. This airtight seal is necessary for recruitment maneuvers (~35 cm H₂O peak inspiratory pressure), which are crucial for maintaining lung compliance and preventing atelectasis. Should the recruitment maneuver not reach the target pressure, the investigator must weigh the corrective strategies in the context of the study goals (e.g., diffusion-weighted imaging vs. ventilation imaging; single time-point vs. longitudinal). Mitigation strategies can include 1) removing the cannula and reintubating, 2) adjusting the cannula position to increase contact with the tracheal walls while still inserted, and 3) proceeding with the experiment without obtaining a gas-tight seal. From the standpoint of data fidelity, reintubation is the most robust option, but it can be time-consuming and lead to unwanted ¹²⁹Xe magnetization decay. Adjusting the cannula position is more rapid, but it should be undertaken cautiously, as it carries a greater risk of injury. Continuing the experiment without RMs is best used when the mouse will be ventilated for <20 min to avoid atelectasis-induced changes in lung dynamics¹¹⁷.

Beyond the obvious need for safe animal handling, the key to developing a robust and reliable HP ¹²⁹Xe mouse MRI program — and to troubleshooting its challenges — is meticulous record keeping. For HP ¹²⁹Xe production, it is advisable to track laser power, accumulation time, flow rate, and polarization. During phantom imaging and spectroscopy, it is best to record image SNR, peak position, spectral linewidth, and the RF power needed to obtain a 90˚ pulse. For in vivo imaging, similar QA data should be recorded along with physiological monitoring parameters (peak inspiratory pressure, body temperature, etc.) and a detailed timeline of animal handling events. A few key pieces of equipment can make ventilator troubleshooting easier. These include an O₂ sensor, a small animal spirometer, and a manometer to calibrate ventilation and PEEP pressure. An oscilloscope and voltmeter can assist in confirming that the ventilator timing is precise and matched to MRI scanner gating. Finally, a hand-held gaussmeter is indispensable for confirming the integrity of a magnetic carrying case if it is required for transport. Regarding MRI sequence troubleshooting, it is recommended to use both the gas phantom described in this protocol and a syringe filled with HP gas. The use of a syringe containing HP ¹²⁹Xe gas is particularly beneficial for troubleshooting diffusion-weighted sequences, as the accuracy of the data can be assessed based on whether it reflects a known ground truth: the free diffusion coefficient of pure ¹²⁹Xe within the syringe.

Several limitations must be acknowledged when applying this protocol. Firstly, it depends on specialized equipment and expertise, which reduces accessibility for new sites. However, with recent regulatory approval for clinical use in the U.S., the number of sites equipped with ¹²⁹Xe MRI capabilities is expected to increase. The technique is also constrained by inherent limitations of hyperpolarized gas, including a relatively short polarization lifetime and the need for rapid data acquisition. The use of a mechanical ventilator typically improves image SNR and intrinsically avoids motion artifacts, but it masks natural breathing dynamics that could be captured by employing a free-breathing gas delivery device. Furthermore, findings from small animal models using this protocol should be interpreted cautiously when extrapolating to human physiology, given significant species variations. Another limitation is the challenging nature of oral intubation and mechanical ventilation, which are difficult in mice weighing less than 25 g. This may exclude males younger than ~8 weeks and females younger than ~20 weeks unless dedicated equipment and procedures are used. The most glaring limitation of this protocol is the absence of pulse programming and advanced image reconstruction techniques. The inclusion of such methods in a generalized protocol is impractical due to their intricate technical details and specialized knowledge. Finally, this protocol describes basic quantitative metrics like SNR, VDP, and ADC but not more advanced measures like regional fractional ventilation, diffusion morphometry, and gas exchange imaging. These advanced metrics will be more easily implemented once the basic infrastructure described here is set up.

Although HP ¹²⁹Xe gas MRI of small animals is resource-intensive and experimentally challenging, it has the potential to significantly advance preclinical lung research. It offers a minimally invasive and sensitive means to quantify regional ventilation, alveolar-airspace size, and gas exchange dynamics. As such, it can provide a unique means to assess lung physiology, disease progression, and therapeutic interventions over time. It, therefore, represents an improvement over traditional preclinical techniques, including histology, biochemical analyses, ionizing imaging modalities, and invasive lung mechanics measurements. Moreover, small animal ¹²⁹Xe MRI remains an actively developing subfield of research, and balancing the relevant imaging physics with preclinical requirements represents substantial research opportunities in its own right. Finally, as clinical ¹²⁹Xe MRI becomes more widely accepted and used clinically, the opportunities for mouse-to-human and back translational studies will continue to grow with time.

Supplementary Figure 1: A circuit diagram for the magnetic carrying case. A solenoid made of enameled copper wire, L1, is connected in parallel with V1, a 12 Volt rechargeable sealed lead acid battery. The resistance of resistor R1 can be selected to control the current through V1 and L1. A panel indicator LED lamp, R2 and D1, is also wired in parallel with V1. A toggle on-off switch, S1, is connected in series with V1. Please click here to download this File.

Supplementary Figure 2: Ventilator calibration results for nitrogen. The volume output from 30 breaths was measured in triplicate. A linear regression was performed to determine the driving pressure necessary to produce relevant gas volumes. Please click here to download this File.

Supplementary Figure 3: Ventilator pressure settings cheat sheet. The slope and intercept results from the linear regression of each gas are used in Equation S1. This equation was used to determine the driving pressure necessary to produce a tidal volume of 10 ml/kg of ideal body weight for gas mixtures of 70% N₂ + 30% O₂ and 70% ¹²⁹Xe + 30% O₂ for pertinent animal weights. Please click here to download this File.

Peter Niedbalski is a consultant for Polarean Imaging, Plc.

The authors extend their heartfelt gratitude to Jerry Dalke for being a guiding light in ventilator construction. We'd like to thank Carter McMaster for brewing HP ¹²⁹Xe gas. We'd also like to thank Dr. Matthew Willmering and Dr. Juan Parra-Robles for their thought-provoking scientific discussions. Figures created with BioRender.com. This work was funded by the National Institutes of Health (Grant Nos: NHLBI R01HL143011, R01HL151588)