In vivo ¹⁹F MRI for Cell Tracking

In vivo cell tracking is essential for the optimization and monitoring of cellular therapeutics¹. Due to its noninvasive nature, imaging offers excellent opportunities to monitor cells in vivo. Magnetic Resonance Imaging (MRI) is independent of ionizing radiation and allows a superior imaging resolution and intrinsic soft tissue contrast. MRI-based cell tracking has already been used clinically to follow dendritic cells in melanoma patients². Conventional clinical MRI is carried out on the ¹H nucleus, present in mobile water in tissues. It is also possible to carry out MRI on other active nuclei, such as ¹³C, ¹⁹F and ²³Na. However, only ¹⁹F MRI has been applied successfully to in vivo cell tracking as it offers the highest sensitivity after ¹H. The absence of MRI-detectable endogenous ¹⁹F in tissues permits high signal selectivity for the detection of exogenous ¹⁹F contrast agents and allows quantification of fluorine concentration directly from the image data. For a detailed discussion on ¹⁹F MRI, see^3-5. A key issue with ¹⁹F MRI is the need to develop and optimize suitable ¹⁹F cell labels, although several labels have been developed, with a trend towards multimodal agents⁶.

The protocol we describe here is based on studies by our groups^7-9, including the first articles that described in vivo quantitative ¹⁹F MRI-based cell tracking^10,11. The general procedure of cell tracking using ¹⁹F MRI is summarized in Figure 1. We describe a general protocol for labeling and imaging of dendritic cells (DCs) using a custom-made perfluorocarbon contrast agent⁸. The imaging protocol is generally applicable to different cell types, labels and animal models. The cell type and animal model described here should only be taken as an example, and thus we do not provide details on the cell isolation and labeling, but rather focus on the imaging protocol. Modifications will be necessary for each label, cell type, animal model, and imaging setup, and these can be found in the literature or may need to be optimized by the researchers. Some common modifications are included in the discussion.

Note: All experiments and procedures involving animals must be carried out in accordance with relevant ethical guidelines and conform with standard animal care and humane requirements.

1. Cell Labeling (Standard Protocol with Coincubation)

1. Add ¹⁹F label⁸ to immature DCs at a concentration of 4 mg/10⁶ cells (in 2 ml medium). Swirl gently to mix.
2. Incubate the cells with the label for 3 days before maturing and harvesting the DCs.
3. Collect the DCs and remove excess label by washing at least 3x in PBS. Count the cells.
4. Resuspend in a small volume of PBS for injection or further testing.

Note: This protocol is relevant for these specific DCs and label. The procedure was optimized in another publication⁸ and included here are only the steps to prepare a sample to image, since the focus of this protocol is on the imaging. Hence, the protocol for isolation, culture and labeling of a specific cell type will either need to be found in the literature or optimized by the researcher. A summary of all other ¹⁹F labels that have been used in the literature is presented elsewhere⁶.

2. Determination of ¹⁹F/cell Using ¹⁹F NMR

1. Place a known number of labeled cells (typically more than 0.5 x 10⁶, or a number that results in sufficient signal) in a NMR tube.
2. Add 10 μl of 5% trifluoroacetic acid (TFA). If the resonance frequency of TFA is unsuitable due to overlap with the label, use an alternative soluble compound, preferably with a single ¹⁹F resonance.
3. Place the sample in an NMR spectrometer and obtain the ¹⁹F spectrum. Ensure that the bandwidth is sufficient for both the TFA (reference) and the agent.

Note: The TR should be long enough for full relaxation of the reference and the sample spins (around 5 times T₁ relaxation time of the longest T₁ component in the tube, typically TR ~5-8 sec). Some spectrometers require D₂O for a frequency lock signal. A standard ¹⁹F spectrum is sufficient. Extensive averaging or higher cell numbers will be required if cell uptake of the label is poor or if cell numbers are low. The spectrum should be centered between the reference and relevant label peaks, unless corrections are made to account for partial excitation.

4. Calculate the relative areas under the peaks of the reference and the sample (or the main peak of the sample), and then calculate the average number of ¹⁹F's/cell in the sample.

Note: The entire process must be repeated and averaged extensively enough to reach statistical significance. This can also be done using spectroscopy directly at the MRI scanner. However, note that this procedure reveals only the average ¹⁹F loading for a large number of cells, not the variability between cells. Checking uniformity of cell uptake requires the presence of a fluorescent component at the ¹⁹F agent, so that cell uptake can be analyzed using microscopy or flow cytometry.

3. Considerations for Experimental Design - Cell Injection

Due to the relatively poor sensitivity of ¹⁹F MRI for cell tracking, an animal model where the cells will localize in large numbers is necessary for successful in vivo imaging. Typical sensitivity values range from 1,000-100,000 cells/voxel/1 hr scan time⁶, with a cell loading in the range of 10¹¹-10¹³ fluorine atoms.

The number and frequency of imaging sessions must be carefully planned, as this influences the choice of anesthetic. Note that isoflurane may not be suitable for ¹⁹F MRI if the ¹⁹F resonance frequency of isoflurane is close to that of the label compound used. isoflurane may still be suitable if the imaging sessions are short enough to prevent significant build-up. Several alternative anesthetics have been used in the literature^10,11; an example is included in the imaging protocol. Anesthetics may need to be optimized for different mouse strains and disease models.

4. Design of an External ¹⁹F Reference

1. Use an external reference containing a known amount of ¹⁹F signal for quantification¹². Choose the signal intensity of the reference to be roughly comparable to that expected from the subject.

Note: Generally, it is simplest if the reference compound is the same as the cell label, to match the relaxation parameters. If spectroscopic sequences or sequences without spatial localization are used, then a compound with a different chemical shift may be necessary.

2. Modify the placement, size and shape of the reference as necessary, dependent on the region of interest (ROI). Note: It is generally easiest to use tubes with a uniform cross-section for the reference.

5. Imaging and Mouse Preparation

1. Dilute commercially available ketamine and xylazine 1:10 v/v with physiological saline to yield stock solutions of 10 mg/ml ketamine and 2 mg/ml xylazine.
2. Turn on the heating system in the scanner (typically warm air) to ensure the bore will be warm before the mouse is imaged.
3. Inject 100 mg/kg ketamine and 10 mg/kg xylazine intraperitoneal to initiate anesthesia. Note: These doses have been tested for CD1(ICR) and NU-Foxn1^nu 8-week old male mice. Other strains may need slightly different doses, which should be optimized in a separate experiment. This dose will keep the mouse anesthetized for about 45 min. Note: The depth of anesthesia is normally sufficient if the animal shows no reaction to a foot pinch.
4. Once anesthetized, place the animal in an animal cradle and immobilize to prevent motion during MRI. If required, position the external reference next to the animal. Both the reference and the ROI (expected location of the injected cells) should be in the center of the coil. The animal should be connected to temperature and breathing control for physiological monitoring during scanning.
5. Cover the eyes of the mouse with sterile eye ointment to prevent drying out during long imaging sessions.
6. If the imaging time exceeds 40 min, prepare additional catheters for injection of additional ketamine/xylazine before the animal may wake up. Position two separate subcutaneous catheters with the working solutions of ketamine and xylazine. Use this catheter to give the mouse an additional 2.5 mg of ketamine every 20 min after the first 40 min. If necessary, further prolong anesthesia by injection of additional 0.5 mg xylazine through the catheter, 100 min after the initial injection. In all cases, experimental time should not exceed 3 hr, except possibly under terminal anesthesia.
7. Ensure that the mouse is warmed directly after the first injection and maintained at 37 °C body temperature. With this protocol, <80% blood oxygenation was detected under atmosphere air breathing. To prevent side effects from the lowered blood oxygenation, use 100% oxygen (0.3 L/min). Body temperature and breathing should be controlled throughout the whole experiment and until full recovery of the animal. Note that the spontaneous breathing rate under ketamine/xylazine is very high (200-250/min). Irregularities in the breathing pattern, rather than the breathing rate, can indicate that a higher dose is necessary to maintain a suitable state of anesthesia.  The mouse will awaken spontaneously within 20-30 min after the last dose of anesthetic.

Note: For cell tracking, it is necessary to image the same animal more than once. In that case, the imaging sessions should be scheduled with consideration to the time necessary for clearance of anesthesia from the system and animal use guidelines of the institute.

6. Imaging

¹H adjustments and imaging

1. Position the animal inside the scanner so that the ROI is in the isocenter using a low-resolution, anatomical scan with different slice orientations (a localizer).
2. Use the regular shimming protocol on ¹H, e.g. a global shim on the whole volume inside the coil.
3. Adjust the reference pulse gain (RG), i.e. the transmitter power measured in dB for a 1 msec Hermite 90° RF pulse, using the adjustment scan provided by the vendor.

Note: This adjustment will vary with the type of RF coil and is essential for ¹⁹F MRI since the signal on the ¹⁹F frequency is usually too low for direct adjustments. As a consequence, an estimate of the RG must be made from measurements on the ¹H signal. For RF transmission with a surface coil, the ¹H RG should be adjusted to a plane parallel to the coil through the part of interest, for a volume coil with homogenous B₁ profile, the exact settings of the adjustment are less important.

4. The following protocol of ¹⁹F reference pulse gain adjustment works only if both ¹H and ¹⁹F RF transmission is carried out with the same (single or double-tuned) RF coil. For such an imaging setup, the coil profile (B₁ transmit field) on ¹H should be proportional to the ¹⁹F coil profile, when a fixed RG is applied, i.e. B_1,1H=C*B_1,19F with a proportionality constant C.
   1. To confirm that coil profiles are proportional for a specific setup, acquire a ¹H and a ¹⁹F flip angle map (see following section on B₁ correction) on a tube with highly concentrated ¹⁹F, e.g. TFA in water (1:1 v/v) using identical fields-of-view (FOVs), beforehand (Figure 3).
   2. Using the same RG, check that both maps are identical except for the global factor C.
   3. Once the ¹H RG is determined, note this number down.
5. Carry out a conventional ¹H MRI scan as an anatomical reference for the ¹⁹F imaging. Tune the system to the ¹⁹F frequency of the cell marker and carry out a ¹⁹F MR scan. Ideally the ¹⁹F MRI scan will have the same field-of-view and slice selection as the ¹H MRI, although usually with lower resolution. The animal can be revived as soon as the imaging is complete. Image processing can then be carried out offline. 
6. Determine the ¹⁹F RG via the relation dB_19F=dB_1H-20*log₁₀(C). Estimate the ¹⁹F agent frequency in a separate in vitro experiment.

Note: In vivo, the exact frequency can vary slightly from experiment to experiment due to different shimming conditions. To prevent chemical shift artifacts the exact frequency should therefore be determined in each experiment by ¹⁹F MRS, if possible. A typical scan for very low ¹⁹F signal would be global (pulse-acquire) spectroscopy (TR=200 msec, NEX=3,000, TA=10 min, BW=50 kHz). NEX, thus TA, can be dramatically reduced for high ¹⁹F signal. The ¹⁹F agent frequency is determined from the spectrum after Fourier transformation and phase correction. NEX refers to the number of excitations, TA is the acquisition time and BW is the bandwidth.

Note: Typical imaging parameters⁹ on a 11.7T/16 cm dedicated animal scanner are: An anatomic ¹H imaging scan using a turbo spin echo (TSE) sequence with TR/effective echo time (TE_eff) = 2,200 msec/42.8 msec, 8 echoes per excitation, NEX = 2, 10 consecutive, 1 mm thick slices, FOV = 1.92 x 1.92 cm², 128 x 128 matrix, i.e. a resolution of 150 x 150 x 1,000 μm³, TA = 1 min, BW = 50 kHz and a linear phase encoding scheme. ¹⁹F images were acquired with the same sequence and matching geometry, but at lower in-plane resolution and lower BW (NEX = 256, 48 x 48 matrix, i.e. a resolution of 400 x 400 x 1,000 μm³, TA = 57 min, BW = 10 kHz).

7. B₁ correction
   Using an RF coil with inhomogeneous B₁ profile, e.g. a surface coil, can hamper quantification of ¹⁹F data since the signal depends not only on ¹⁹F concentration but also on the distance from the coil. Acquire a B₁ map on the ¹H channel in order to retrospectively correct ¹⁹F data. This requires that ¹H and ¹⁹F coil profiles are identical, except for the proportionality factor C, and that sufficient ¹H background signal is provided in regions of ¹⁹F. An example protocol for correction of a ¹⁹F spin echo sequence is presented, using a single-tuned Tx/Rx surface coil tunable to both ¹H and ¹⁹F and with proportional B₁ profiles¹³.
   1. Acquire a ¹H flip angle map with the two flip angle method ¹⁴. Acquire a gradient echo scan with very long TR (>3 times ¹H T₁) with an estimated flip angle of just below 90°. Close to the coil. Acquire a second gradient echo scan with a RG=dB_1H-6, i.e. twice the flip angle of the first scan.
   2. Calculate the ¹H flip angle, α, at voxel (x,y,z) via the signal intensities (SI) of the first and second gradient echo scans: α(x,y,z)=arcos(SI_GE,1(x,y,z)/2SI_GE,2(x,y,z))¹⁴. The ¹⁹F flip angle is identical (except for the factor 1/C) due to proportional B₁ profiles. According to Faraday's principle of reciprocity¹⁵, the attenuation of ¹⁹F signal intensity from a spin echo sequence (excitation flip angle α, refocusing flip angle 2α) during signal reception is att_Rx(x,y,z)~α(x,y,z). The attenuation due to imperfect excitation is att_Tx(x,y,z)~sin³(α(x,y,z))¹⁶. The overall attenuation is written as att=attRx*attTx= α*sin³(α)/90°. Note that it is normalized to 1 in case of perfect 90°/180° excitation/refocusing flip angle. The B₁ corrected ¹⁹F image signal intensities SI_19F,corr are calculated via SI_19F,corr(x,y,z)= SI_19F(x,y,z)/att(x,y,z).
   3. In order to compare signal intensities from different experiments, place a reference tube with defined ¹⁹F concentration in the field of view, and normalize the SI_19F,corr to the mean signal in the reference.

Note: Other ¹H B₁/flip angle mapping methods can be used and different ¹⁹F pulse sequences require modifications of att_Tx,19F. Any kind of B1 correction scheme will introduce additional error in the cell quantification procedure. If accurate cell quantification is the main premise, an RF coil with homogenous B1 profile can be used to circumvent this part of the protocol.

7. Image Processing

1. Making ¹H and ¹⁹F overlay images
   1. Export the image data for the final ¹H and ¹⁹F magnitude images from the scanner.
   2. Open the files in an image processing program, such as ImageJ (freeware, NIH).
   3. Carry out image adjustments as necessary (cropping, brightness, contrast) keeping the adjustments the same for all images. The ¹⁹F images will typically need to be upscaled to a higher resolution to match the corresponding ¹H images.
   4. Render the ¹⁹F images in false color (select a different look-up table, under the "image" menu in Image J) and then overlay on the corresponding ¹H images in order to localize the signal.
2. ¹⁹F Quantification
   1. Select a ¹⁹F image (magnitude image) with the relevant cells and the reference visible.
   2. In a program such as Image J, draw ROIs over the relevant cells, the reference and a region outside the subject (noise).
   3. Use the command Analyze and then Measure (or simply Ctrl+M) to calculate the mean pixel intensity within these regions, and their area. Let S(cells) refer to the mean pixel intensity in the ROI over the cells. Multiply that by the volume (= area x slice thickness) to calculate the total signal.
   4. Calculate the SNR, for example using the formula SNR = 0.65 x S(cells) / σ, where σ is the standard deviation of pixel intensity in an ROI outside the sample which is free of any chemical shift artifact (typically in the background air)¹⁷. If the SNR is under 5, a correction for the signal due to the noise may be necessary and can be carried out as described elsewhere^{11 10}.
   5. Otherwise, calculate the total amount of ¹⁹F responsible for the signal in the ROI over the reference by multiplying the area of the reference in the slice by slice thickness (to calculate the volume) by the concentration of ¹⁹F in the reference, which is known and assumed to be homogeneous.
   6. Multiply that value by the total signal in the ROI over the cells, divided by the total signal over the reference to calculate the amount of ¹⁹F in the cells.
   7. Calculate the number of cells by dividing that number by the value for ¹⁹F/cell, which was calculated in section 2.

Here we show the typical results for a protocol involving the transfer of ¹⁹F-labeled cells homing to a draining lymph node. Figure 2 shows a ¹⁹F NMR spectrum of 10⁶ labeled cells using a TFA reference. The imaging setup was carried out as described in the protocol (Figure 3). For the in vivo imaging, we used a reference consisting of a sealed cylinder of the same label used for the cells placed between the feet of the mouse. A representative processed image is shown in Figure 4. By applying the protocol described in section 7.2, we calculated a cell number of approx. 1.2 x 10⁶ based on the raw ¹⁹F magnitude image.

Figure 1. Key steps for ¹⁹F MRI-based cell tracking. These include the appropriate selection of label and labeling protocol is crucial for success of the experiment (Step 1). Cell labeling (Step 2), may require enhancement through the use of coatings or transfection agents. After suitable preparation, including removal of any excess label, (Step 3) the cells can be imaged. Post-processing can be used to generate quantitative data from the images, if acquired suitably (Step 4). Finally, various ex vivo analysis can be carried out (Step 5) to corroborate the in vivo data. Figure reproduced with permission⁶. Click here to view larger figure.

Figure 2. Representative ¹⁹F NMR spectrum. The spectrum shows the ¹⁹F signal obtained due to the known amount of reference, in this case TFA, and the known number of cells, labeled "sample". This can be used to calculate the amount of fluorine per cell, which is necessary for in vivo quantification from MR image data.

Figure 3. Flip angle maps of tubes with TFA in water were acquired at the ¹H and ¹⁹F channel. The proportionality factor was calculated to be 1.03±0.08 for this setup. FA: flip angle.

Figure 4. In vivo ¹H/¹⁹F image of dendritic cells homing to a draining lymph node. The figure shows a representative image of a mouse, with ¹⁹F signal visible, in false color, in the reference (R) and the labeled cells. Only the legs of the mouse were imaged here. In this example, the injected labeled cells migrated to the draining lymph node. The imaging parameters are summarized in the main text. The figure is for illustration purposes only.

The protocol outlined here describes the general procedure for in vivo ¹⁹F MRI cell tracking. Despite the described co-incubation method, there are several different protocols for labeling cells with a ¹⁹F agent. However, the co-incubation is most often used and can be further optimized e.g. by addition of transfection agents⁶. The actual cell labeling protocol will depend on the cell type. Only key labeling steps are described in the protocol here. Generally, the label is prepared and added to the cells for a selected time ranging from a few hours to a few days. Finally, the cells must be washed carefully to remove any excess of label, especially if quantification from the image data will be carried out³. If the ¹⁹F agent includes a fluorescence tag, the presence of label outside (or not bound to) the cells can be detected using microscopy. After labeling, it is necessary to study the cell viability and functionality, in relation to nonlabeled cells (e.g. by trypan blue exclusion assay). Cell-specific functionality tests, such as the ability to activate T cells in the case of DCs, must also be carried out. Finally, as some non-¹⁹F MRI labels have an effect on cell migration¹⁸, such tests should also be included for ¹⁹F labels, especially with high cell loading.

The protocol for the animal model is also dependent on the user´s needs. However, it is important to remember the strict sensitivity limits of ¹⁹F MRI when planning experiments. Published sensitivity values range from 1,000-00,000 cells/voxel/1 hr time⁶, with a cell loading in the range of 10¹¹-10^{13 19}F. Sensitivity and expected label concentration in the region of interest also impact the imaging parameters. For example, the slice thickness used for the ¹⁹F image can be adjusted to match that of the ¹H images or to cover a much thicker area (such as the entire lymph node or region of interest in a single slice). Typically, the ¹⁹F imaging is carried out at lower resolution to maximize signal detection, especially since high resolution is not usually necessary to distinguish structures such as lymph nodes. If the signal is sufficient, a larger number of thinner slices can be acquired and the total signal over the volume of interest calculated from these. However, the optimal parameters will need to be derived for each application individually.

We describe an injection anesthesia protocol with ketamine/xylazine avoiding fluorinated inhalation gases like isoflurane. Other protocols have been used in the literature^10,11. However, in all cases, these protocols will need to be adjusted for the mouse strain, age, gender, health and the length and frequency of the imaging sessions.

Several ¹⁹F MRI imaging sequences have been used for cell tracking, for a review see⁶. Our imaging protocol can be readily modified to include other imaging sequences. However, the quantification method described here does not take into account any partial volume effects, discrepancies in selection of ROIs, changes in relaxation parameters of the label in vivo, or changes in cellular ¹⁹F loading. These issues are described elsewhere³. In brief, the error has been found to be within tolerable limits for longitudinal cell tracking¹⁰. Compared to these earlier works we additionally propose an image correction scheme for the use of surface RF coils by mapping the B1 field. Depending on the specific RF coil setup it is important to be aware of limitations of B1 mapping techniques, which have been well-studied in the context of ¹H MRI¹⁶. For example, for the double angle method presented here the B1 map is most accurate for flip angle pairs just below 90°-180° ¹⁴ and significant error can be introduced in other regions. Still, after proper validation in vitro, such retrospective correction can further improve cell quantification. In general, we recommend corroboration of the data with more established techniques, at least initially. This is typically done in combination with other in vivo imaging techniques, e.g. optical imaging, to help exclude large errors. In most cases, histological evaluation is necessary to assess ¹⁹F label location precisely and to allow correlation with the imaging data. This may require the addition of a fluorescent dye to the ¹⁹F agent.

Finally, although ¹⁹F MRI has not yet been applied to clinical cell tracking, it would be possible to modify this protocol for human use. The main modifications necessary would be to carry out as much of the optimization and setup beforehand as possible (to minimize the length of time the subject is in the scanner), and to adjust imaging parameters to keep within clinical specific absorption rate (SAR) restrictions.