Preclinical ¹⁹F MRI cell tracking at 3 Tesla

Abstract

Purpose

To develop methods for fluorine-19 (¹⁹F) MRI cell tracking in mice on a 3 Tesla clinical scanner. Compared to iron-based cell tracking, ¹⁹F MRI has lower sensitivity and, consequently, preclinical ¹⁹F cell tracking has only been performed at relatively high magnetic field strengths (> 3 T). Here, we focus on using ¹⁹F MRI to detect macrophages in tumors; macrophage density is an indication of tumor aggressiveness and, therefore, ¹⁹F MRI could be used as an imaging biomarker.

Methods

Perfluorocarbon (PFC)-labeled macrophages were imaged at 3 T and NMR spectroscopy was performed to validate ¹⁹F spin quantification. In vivo ¹⁹F MRI was performed on tumor-bearing mice, post-PFC at both 9.4 T and 3 T. 3 T MRI utilized varying NEX and ¹⁹F images were analyzed two different ways for ¹⁹F quantification.

Results

As few as 25,000 cells could be detected as cell pellets at 3 T. ¹⁹F quantification in cell pellets by 3 T MRI agreed with NMR spectroscopy. ¹⁹F signal was observed in the liver, spleen and tumor in all mice at 9.4 T and 3 T and there was no significant difference in ¹⁹F spin quantification.

Conclusion

This study demonstrates the ability to detect and quantify ¹⁹F signal in murine tumors using ¹⁹F MRI at 3 T.

Introduction

Fluorine-19 (¹⁹F)-based magnetic resonance imaging (MRI) has recently become an attractive method to image cells in vivo in preclinical studies. This is due to its inherent specificity, quantification ability and chemical inertness of the agent administered. For these reasons, ¹⁹F MRI has been used in many different experimental cell tracking studies including, for example, imaging of immune cells in arthritis [1], bowel disease [2], experimental autoimmune encephalomyelitis [3] and cancer [4,5,6,7]. However, compared to the more widely used iron-based MRI, ¹⁹F is much less sensitive. Limiting factors which contribute to sensitivity have been thoroughly discussed by Srinivas et al. [8] and include the amount of fluorine which can be delivered to a region of interest (i.e., ¹⁹F concentration [9], due to cell loading and/or number of ¹⁹F spins on the label), MR properties of the ¹⁹F label, MRI system and MRI acquisition parameters. One or a combination of these factors may result in low signal to noise (SNR) which could impair ¹⁹F detection. To compensate, it is common to image at a high magnetic field strength to increase ¹⁹F signal. Another simple approach for increasing sensitivity is to use multiple acquisitions, or signal averages, which are summed for improved SNR—this requires longer scan times. All in vivo preclinical ¹⁹F MRI cell tracking studies have been performed at field strengths above 3 Tesla (T) and have used multiple acquisitions [1, 10,11,12,13,14,15].

We have previously used ¹⁹F MRI at 9.4 T to demonstrate the ability to detect and quantify tumor-associated macrophages (TAMs) in a preclinical model of breast cancer [5]. TAM presence and number have been shown to be correlated to breast tumor aggressiveness including growth, progression, metastasis and clinical outcome [16,17,18]. Our study showed that tumors with different levels of aggressiveness could be distinguished using ¹⁹F MRI cell tracking. The ability to detect and monitor the number of TAMs in individual tumors with ¹⁹F MRI could allow for identification of breast tumors with heavy infiltration of TAMs and could be used as a biomarker for decisions about how to best treat these patients as well as for monitoring responses to therapy.

The translation of cellular MRI techniques to the clinic will require the use of human MRI systems, the maximum field strength of which is presently 3 T, in most places. In this paper, we begin to explore the feasibility of performing ¹⁹F MRI cell tracking at 3 T in a murine model. We show that ¹⁹F-labeled cells can be detected in vivo in mice using a clinical 3 T MRI system.

Methods

Cell culture

4T1 murine breast cancer cells (Dr. Fred Miller, Wayne State University, MI, USA) and RAW264.7 murine macrophages (Dr. Greg Dekaban, Western University, ON, CAN) were maintained at 37 °C and 5% CO₂ in Dulbecco’s Modified Eagle’s Medium–high glucose media [4T1 (Thermofisher, ON, CAN)] and RPMI [RAW264.7 (Thermofisher, ON, CAN)] supplemented with 10% fetal bovine serum, penicillin/streptomycin (4T1 and RAW264.7) and l-Glutamine, HEPES, beta-mercaptoethanol [RAW264.7 (Thermofisher, ON, CAN)]. Cells were passaged every 2–3 days (4T1) and 3–4 days (RAW264.7).

In vitro cell labeling

RAW264.7 murine macrophages were labeled with a perfluoropolyether perfluorocarbon (PFC) agent (V-Sense, VS-1000H or a red fluorescent-tagged version, VS-1000H DM Red, CelSense Inc., Pittsburgh, USA). Labeling took place over ~ 24 h at a concentration of 2.5 mg/ml. Cells were washed 3 times with PBS followed by trypsinization. A 10 μl sample was diluted 1:1 with trypan blue and was used for cell counting and viability (Countess Automated Cell Counter; Invitrogen, Carlsbad, CA, USA). PFC-labeled macrophage pellets were then spun down in an Eppendorf tube, supernatant removed, and topped with 1% agarose for MRI or prepared for NMR spectroscopy (see below).

Animal model

Female BALB/c mice (6–7 weeks; Charles River Canada) were obtained and cared for in accordance with the standards of the Canadian Council on Animal Care, under an approved protocol by the Animal Use Subcommittee of Western University’s Council on Animal Care. Mice were anesthetized with isoflurane administered at 2% in oxygen prior to receiving an injection of 300,000 4T1 cells (> 90% viability, measured using the trypan blue exclusion assay) suspended in 50 µl Hanks balanced salt solution (HBSS) into the 4th inguinal mammary fat pad, as previously reported [19, 20]. Animals were observed until alert and active, when they were returned to their cages.

In vitro MRI at 3 T and NMR spectroscopy

Two separate sets of PFC-labeled murine macrophage pellets with high numbers of cells [25, 50, 75, 125, 250, 500, 750 (× 10³), 1, 1.5, 2 and 5 (× 10⁶) cells] and low numbers of cells [5, 10, 25, 125 and 250 (× 10³) cells] were imaged to evaluate the minimum number of cells that could be detected at 3 T. Images were acquired using a clinical GE 3 T MR750 system (General Electric, ON, CAN) and a 4.3 × 4.3 cm dual tuned ¹H/¹⁹F surface coil (Clinical MR Solutions, WI, USA). This coil was originally built for imaging small ROIs on humans. ¹H images were acquired with a 2D fast gradient echo sequence with the following parameters: field of view (FOV) = 50 × 50 mm, matrix = 256 × 256, slice thickness = 5 mm, TR/TE = 100/4.9 ms, flip angle (FA) = 20°, bandwidth (BW) = ± 31.25 kHz, number of signal averages (NEX on a GE scanner) = 4 and scan time = 3.5 min. ¹⁹F images were acquired using a 3D balanced steady-state free precession (bSSFP) sequence with 1 mm³ spatial resolution, FOV = 50 × 50 mm, matrix = 50 × 50, slice thickness = 1 mm, TR/TE = 5.7/2.7 ms, FA = 72°, BW = ± 10 kHz, NEX = 60 and scan time = 30 min.

In a separate experiment, cell pellets containing 450,000 and 900,000 PFC-labeled murine macrophages were used to compare the quantification of ¹⁹F spins using 3 T MRI and NMR spectroscopy. For MRI, there were 3 cell pellets containing 450,000 cells and 3 cell pellets containing 900,000 cells. These were imaged together using the same parameters as above. The samples were imaged 4 times, each time with a different number of NEX: 10, 20, 40 or 80 NEX.

For NMR of PFC-labeled cells, 3 samples containing 450,000 cells and 3 samples containing 900,000 cells were lysed with 10 μl RIPA buffer (VWR, Mississauga, CAN) and sonicated followed by 3 freeze–thaw cycles. This lysate was placed in an NMR tube with 0.1% trifluoroacetic (TFA) acid and D₂0. All ¹⁹F NMR measurements were performed at 376.12 MHz using a Varian Inova 400 spectrometer (Varian Inc, Palo Alto, USA). Spectroscopy parameters were: recycle delay = 6 s, acquisition time = 1 s, spectral width = 18.9 kHz (− 102 ppm to − 62 ppm), tip angle = 60°, number of scans = 100 and scan time = 13 min. The mean intracellular ¹⁹F per cell was calculated from the ratio of the integrated areas of the PFC and TFA spectra.

In vivo MRI

Mice bearing 4T1 tumors at 3 weeks post-implantation were administered 200 µl of the PFC agent intravenously (IV) via the tail vein 24 h prior to imaging. While imaging, mice were anesthetized with 2% isoflurane in oxygen. Two cell pellets containing 1 and 2 (× 10⁶) PFC-labeled macrophages and two reference tubes of known ¹⁹F concentration (3.33 × 10^{16 19}F/μl) were placed alongside the mice for quantification purposes (see below). Mice (n = 2) were imaged with ¹H and ¹⁹F first on a small animal 9.4 T MRI (Varian Inc., Palo Alto, USA) using a custom built dual ¹H/¹⁹F birdcage coil with a 3 cm diameter and then immediately after at 3 T. In vivo ¹H and ¹⁹F images were both acquired with 3D bSSFP pulse sequences.

At 9.4 T, ¹H imaging parameters were: 500 μm³ spatial resolution, FOV = 80 × 40 mm, matrix = 160 × 80, slice thickness = 0.5 mm, FA = 20°, BW = ± 62.5 kHz, TR/TE = 2.9/1.5 ms, 4 signal averages and 8 phase cycles (PC) with a scan time of 10 min. ¹⁹F imaging parameters were: 1 mm³ spatial resolution, FOV = 80 × 40 mm, matrix = 80 × 40, slice thickness = 1 mm, FA = 63°, BW = ± 25 kHz, TR/TE = 3.1/1.6 ms, 4 signal averages and 30 PC resulting in a scan time of 30 min.

At 3 T, ¹H imaging parameters were: 0.5 × 0.5 × 0.6 mm spatial resolution, FOV = 80 × 40 mm, matrix = 160 × 80, slice thickness = 0.6 mm, FA = 35°, BW = ± 31.25 kHz, TR/TE = 6.3/3.2 ms, 6 NEX and 8 PC with a scan time of 35 min. ¹⁹F imaging parameters were: 1 mm³ spatial resolution, FOV = 80 × 40 mm, matrix = 80 × 40, slice thickness = 1 mm, FA = 72°, BW = ± 10 kHz, TR/TE = 5.6/2.8 ms and 100 NEX resulting in a scan time of 30 min.

Using a separate group of tumor-bearing mice (n = 4), consecutive ¹⁹F images were acquired at 3 T using 2, 5, 10, 19, 38, 75 and 150 NEX (resulting in 0.5, 1, 2, 4, 8, 16, 30 min scans). ¹H parameters were: 200 μm³ spatial resolution, FOV = 70 × 35 mm, matrix = 350 × 175, slice thickness = 0.2 mm, FA = 20°, BW = ± 31.25 kHz, TR/TE = 11.9/5.9 ms, 1 NEX and 4 phase cycles with a scan time of 25 min and ¹⁹F images were acquired with the above 3 T parameters with an FOV = 70 × 35 mm and matrix = 70 × 35.

Histological analysis

Mice were euthanized via overdose of isoflurane following the last MRI exam. Mice were transcardially perfused with saline followed by 4% paraformaldehyde. All primary tumors were excised and cryoprotected by passaging through a sucrose gradient of 10, 20, and 30% for 24 h each. Samples were then frozen in optimal cutting temperature compound and cryostat sections were collected (10 μm). Tumor tissue was examined using fluorescence microscopy to detect the Texas Red fluorescence of the PFC and subsequently stained using rat anti-mouse F4/80 primary antibody (Bio-Rad AbD Serotec Inc., Raleigh, NC, CI:A3-1, MCA497) to identify macrophages.

A sample of the PFC-labeled macrophages containing ~ 300,000 cells in ~ 300 μl PBS was centrifuged to make cells adhere to a slide. After fixation, slides were stained with nuclear fast red to enhance visualization of the PFC droplets with bright field microscopy.

Microscopy was performed using a Zeiss Axio Imager A1 microscope (Zeiss Canada, Toronto, ON, Canada) equipped with a Retiga EXi (QImaging Scientific Research Cameras, Surrey, BC, Canada) digital camera.

Analysis of MRI data

¹⁹F images were overlaid onto the ¹H images (Osirix, Pixmeo SARL, Bernex, Switzerland) for anatomical reference of the ¹⁹F signal within either the macrophage pellets or mice. Manual ROI delineation for ¹⁹F quantification was performed in two different ways: (1) By choosing only voxels which are judged to contain ¹⁹F signal in each pellet or tumor, after a window/level algorithm is applied (“¹⁹F only” ROI) and (2) By outlining a tumor slice by slice in the ¹H images, creating a whole tumor 3D ROI and then copying the ROI onto the ¹⁹F images (“tumor” ROI).

Average magnitude ¹⁹F signal from the tumors (using “tumor” and “¹⁹F only” ROIs) or pellets and reference tubes (S) and standard deviation of the noise (σ) were obtained from each data set to calculate SNR $\left( {{\text{SNR}} = 0.655 \times \frac{S}{\sigma }} \right)$ of tumors or pellets.

In brief, the total number of ¹⁹F spins was determined by comparing the total ¹⁹F signal within a chosen ROI to the signal generated by a reference tube containing a known amount of ¹⁹F spins (3.33 × 10^16 19F/μl). It is known that in low SNR situations there is a rician distribution of signal in the noise [21]. Many of these data sets will have low SNR due to low NEX, so therefore when tumor or pellet SNR < 5, a correction of 0.655 was made to the magnitude signal before calculating ¹⁹F spins, as described previously [14].

Statistical analysis

Statistical analyses were performed using PRISM software (Graphpad, Version 7.0a). To compare the number of ¹⁹F spins quantified from MR images and the known cell number in the PFC-labeled macrophage pellets, the Pearson correlation coefficient was used. A one-way ANOVA was performed to determine differences in ¹⁹F quantification by NMR and all MRI exams (80, 40, 20, 10 NEX). A paired, two-tailed t test was used to compare ¹⁹F quantification between 9.4 T and 3 T MRI exams.

Results

In vitro ¹⁹F quantification at 3 T MRI and NMR spectroscopy

PFC-labeled murine macrophages imaged at 3 T showed that the ¹⁹F signal from pellets containing as few as 25,000 cells could be detected. No ¹⁹F signal was detected from the pellets containing 5000 or 10,000 cells (not shown). PFC-labeled macrophage pellets were visible as a homogeneous region of ¹⁹F signal. A ¹H/¹⁹F overlay from the “high number” PFC-labeled cell pellets is shown in Fig. 1a. Low signal is visualized in 75, 50 and 25 (× 10³) cell pellets in the single image slice shown in Fig. 1a; however, there was additional signal in adjacent slices and a sagittal view of the 25,000 pellet (inset) demonstrates more obvious ¹⁹F signal. A linear correlation was found between the known cell number in each pellet and the number of ¹⁹F spins (R = 0.983, p < 0.0001) (Fig. 1b). A nuclear fast red stained cytospin slide containing a sample of PFC-labeled cells is shown in Fig. 1c and the PFC droplets can be clearly seen within cells. The average number of ¹⁹F spins/cell as estimated from this experiment is 7.93 × 10¹¹.

Fig. 1

¹⁹F MRI of in vitro PFC-labeled murine macrophages at 3 T. a¹H/¹⁹F MRI overlay of 25,000–5,000,000 PFC-labeled macrophage pellets. This ¹⁹F image acquisition used 60 NEX resulting in a 30 min scan. ¹⁹F signal detection limit was 25,000 cells (inset: sagittal image of 25,000 cell pellet). b¹⁹F spin quantification is linearly correlated with number of cells (R = 0.983, p < 0.0001) and average ¹⁹F/cell calculated is 7.93 × 10¹¹. c PFC droplets can be seen in cells stained with nuclear fast red, without a marker under bright field microscopy (inset: zoomed). k thousand, M million

Figure 2 shows the average number of ¹⁹F spins determined by NMR spectroscopy and by 3 T MRI for PFC-labeled cell pellets containing 450,000 and 900,000 macrophages. The mean number of ¹⁹F spins measured by NMR was 9.76 × 10¹⁶ for the six pellets containing 450,000 cells and 1.87 × 10¹⁷ for the six pellets containing 900,000 cells. From these data, the mean number of ¹⁹F spins/cell was calculated as 2.12 × 10¹¹. The average number of ¹⁹F spins determined by MRI was not significantly different (p = 0.9991) from that measured by NMR, regardless of the NEX. Average pellet SNR was 5.89 for the six pellets containing 900,000 cells and 4.57 for the six pellets containing 450,000 cells. Mean ¹⁹F spins/cell was calculated as 1.97 × 10¹¹ by MRI.

Fig. 2

In vitro PFC-labeled macrophages quantified by ¹⁹F MRI (using varying NEX) and NMR. PFC-labeled macrophages (450,000 or 900,000 cells) were quantified by either MRI (n = 6) using 80, 40, 20 and 10 NEX or NMR (n = 6). There was no statistically significant difference in the mean number of ¹⁹F spins measured by NMR versus MRI using the different MRI acquisitions (p = 0.9991). The average ¹⁹F/cell was 1.97 × 10¹¹ by MRI and 2.12 × 10¹¹ by NMR. MRI scan times were: 80 NEX = 18:05 min, 40 NEX = 9:03 min, 20 NEX = 4:32 min and 10 NEX = 2:08 min

Comparing in vivo mouse body ¹⁹F MRI at 9.4 T and 3 T

Figure 3 shows 9.4 T (Fig. 3a) and 3 T (Fig. 3b) images of the same 4T1 tumor-bearing mouse, acquired ~ 24 h post-IV PFC. ¹⁹F signal was observed in the liver, spleen, and tumor in both mice. In the representative image slice shown in Fig. 3, ¹⁹F signal is visible in the liver and in the periphery of the tumor (as well as in the cell pellets included in the FOV). The ¹⁹F signal appeared similar in both the 9.4 T and 3 T images and the average tumor SNR at 9.4 T (10.6) and 3 T (7.9) was not significantly different between the MRI exams at different field strengths; (p = 0.7126). The table presented in Fig. 3c describes the number of ¹⁹F spins quantified in the tumor and liver for each mouse at 3 T and 9.4 T, using “tumor” ROI for ¹⁹F quantification. The tumor and liver ¹⁹F spin values are not significantly different when imaged at 3 T or 9.4 T (p = 0.5828 and p = 0.2911, respectively). If we use the value for ¹⁹F/cell obtained from NMR, we can estimate that there is an average of 4.05 × 10⁷ PFC-labeled cells in a 4T1 tumor and 8.28 × 10⁸ PFC-labeled cells in the liver of these mice.

Fig. 3

¹⁹F MRI at 9.4 T and 3 T of a 4T1 tumor-bearing mouse. ¹H/¹⁹F MRI overlays of the same 4T1 tumor-bearing mouse (Mouse A) at a 9.4 T and b 3 T, imaged consecutively. c There was no statistically significant difference in the number of ¹⁹F spins measured for the tumor or liver, at 3 T versus 9.4 T, for each mouse. Using the value of ¹⁹F/cell obtained from NMR of cell pellets (2.12 × 10¹¹¹⁹F/cell), an estimate of the number of PFC-labeled cells in this tumor is 64 × 10⁶ from 3 T or 87 × 10⁶ from 9.4 T. This analysis used the “tumor” ROI method. Reference tubes are not visible in these slices. L liver

In vivo ¹⁹F quantification at 3 T MRI

A window/level algorithm was applied to all ¹⁹F images to enhance visualization of the tumor signal. In Fig. 4, representative images of the full dynamic range of signal in the whole mouse body are compared to the adjusted signal for visualization of tumor with high and low SNR acquisitions (150 NEX and 10 NEX, respectively). The relative minimum/maximum values of signal intensity for each image are included in figure legends.

Fig. 4

¹H/¹⁹F MRI overlays of a 4T1 tumor-bearing mouse, showing the full dynamic range and adjusted ¹⁹F signal at 3 T. Images were acquired with 150 and 10 NEX. Full dynamic range presents all signal intensities in each slice as lowest = yellow and highest = pink. There is a smaller window width in the adjusted images used to visualize the ¹⁹F signal in the tumor (outlined in white) which does not represent all signal intensities. Minimum/maximum signal intensities: a 0/765, b 103/353, c 0/725, and d 109/375. L liver, R reference tubes

Figure 5 shows a ¹H image of a tumor-bearing mouse along with ¹⁹F images acquired with decreasing NEX/scan time, using “tumor” (Fig. 5a) and “¹⁹F only” (Fig. 5b) ROIs. Examples of ROIs used for each are drawn over the tumor in white, in the ¹⁹F images. The ROI volumes measured for this representative mouse are shown above each tumor; this is the ROI volume used for ¹⁹F spin quantification. ¹⁹F signal was detectable in the tumor, liver, spleen and reference tubes in all exams.

Fig. 5

MRI of a representative 4T1 tumor-bearing mouse at 3 T. ¹H images are shown in the far left column. ¹⁹F images are shown, with decreasing number of averages (NEX); the values for NEX and scan time are listed along the bottom. a Using the “Tumor” ROI method, the tumor is outlined in the proton image and this ROI is used for all ¹⁹F images; the volume of the ROI is the same (590 mm³). Window/level values are the same for all ¹⁹F images (min/max: 70/362). b Using the “¹⁹F only” ROI method, the drawing of the ROI is subjective and depends on ability to see ¹⁹F signal. This increases with SNR as does the volume of the ROI selected; volumes are shown above the ROIs. Minimum/maximum signal intensities: 2 NEX: 233/421, 5 NEX: 134/310, 10 NEX: 96/364 and 19 & 150 NEX: 70/362. L liver, R reference tubes

When using the “tumor” ROI (where the entire tumor was outlined), the same ROI was applied to every ¹⁹F image (Fig. 5a). In the ¹⁹F images, there is an obvious increase in noise as NEX/SNR decreases, when window/level values are the same in all images.

When using the “¹⁹F only” ROI (where only obvious ¹⁹F signal within the tumor was outlined), the size of the ROI chosen was different for each image acquired with different NEX. Window/level values were changed for each scan to optimize the visible ¹⁹F signal in the tumor. ROI size increased as NEX/SNR increased because the ¹⁹F signal was more obvious with less background noise. The quantification of ¹⁹F spins from these two analysis methods is shown in Fig. 6.

Fig. 6

Scatter plots of ¹⁹F spin number for each tumor, quantified using each of the NEX acquisitions by either a “tumor” ROI or b “¹⁹F only” ROI analysis method. The range of ¹⁹F spin number is smaller when using a “tumor” ROI versus “¹⁹F only” ROI. There is a trend of higher numbers of ¹⁹F spins with higher NEX using a “¹⁹F only” ROI. Bar is the mean value

In Fig. 6, scatter plots show the number of ¹⁹F spins measured for each tumor, from each of the different acquisitions, using either the “tumor” ROI (Fig. 6a) or “¹⁹F only” ROI (Fig. 6b) method. The average number of ¹⁹F spins measured for each tumor, regardless of NEX, was similar for each analysis method. There was a smaller range of values for the number of ¹⁹F spins measured for the different NEX when using a “tumor” ROI compared to the “¹⁹F only” ROI. When using the “¹⁹F only” ROI, there was a trend of higher numbers of ¹⁹F spins measured with higher NEX.

Microscopy was used to confirm the presence of F4/80+ macrophages [brown staining, revealed by diaminobenzidine (DAB)] and the red fluorescent PFC agent. Figure 7 shows an example of a 4T1 tumor, 3 weeks post-cancer cell implantation, taken at the tumor periphery. The brown F4/80+ cells (Fig. 6a, d) correspond with the location of the red fluorescence from the PFC agent (Fig. 6b, e). Overlays (Fig. 6c, f) demonstrate that the location of the F4/80+/PFC+ cells correlates well with the ¹⁹F signal, which is in the outer periphery of the tumor in the MR images in Fig. 3.

Fig. 7

Microscopy of a tumor 3 weeks post 4T1 cancer cell implantation. a, d F4/80 DAB stains for macrophages (brown), here evident in the periphery of the tumor. b, e Red fluorescence from the PFC agent is observed in the same location as the F4/80+ macrophages. The overlay is shown in c, f. Yellow box in a represents the area shown in d–f. Two magnifications of ×5 (a–c) and ×10 (d–f), scale bars represent 200 micron

Discussion

This study demonstrates the use of a clinical 3 T MRI to image and quantify PFC-labeled cells in a preclinical murine model of breast cancer. We first used in vitro PFC-labeled macrophages to validate ¹⁹F quantification by MRI and NMR. For in vivo studies, we analyzed the ¹⁹F images two different ways, as we believe that there may be some user bias when manually delineating ¹⁹F voxels with obvious signal (“¹⁹F only” ROI), especially at a lower SNR with increased noise. When using a “tumor” ROI, where ROIs from anatomical images are placed onto ¹⁹F images, this bias may be eliminated. Here, we used a range of number of NEX (resulting in a range of SNR), with ¹⁹F spin number remaining more consistent when using “tumor” ROI versus “¹⁹F only” ROI.

Cellular MRI is a powerful tool to track cells in vivo. Groups have used the IV administration of a cell label and subsequent MRI to track immune cells in vivo, overtime at multiple time points to evaluate the dynamics of cell infiltration [22] and response to treatments or interventions [1]. This technique relies on the labeling of macrophages in situ, including macrophages which are part of the reticuloendothelial system (i.e., liver, spleen, bone marrow, lymph nodes) [23,24,25,26]. Additionally, macrophages associated with cancer have been shown to take up both ¹⁹F [4, 5, 27,28,29] and iron [30,31,32] agents. The advantage of cell tracking with ¹⁹F MRI is the ability to quantify the signal; ¹⁹F signal intensity is directly proportional to the amount of ¹⁹F spins present in an ROI. When cells are labeled in vitro, cell loading can be determined by NMR and ¹⁹F signal can be related to cell number. When cells are labeled in vivo, ¹⁹F signal can be linearly related to a degree of inflammatory cell presence [33].

We first used in vitro PFC-labeled macrophage cell pellets to validate ¹⁹F quantification and to get an estimate of ¹⁹F spins/cell. ¹⁹F MRI of cell pellets at 3 T showed a strong correlation between the number of cells in a pellet and the number of ¹⁹F spins. A relatively low labeling concentration of 2.5 mg/ml (similar to other published methods [15, 34]) resulted in an estimate of 7.93 × 10¹¹¹⁹F spins/cell. From another sample of PFC-labeled cells, an estimate of 2.04 × 10¹¹¹⁹F spins/cell was obtained from MRI (using 80, 40, 20 and 10 NEX) and an average 2.12 × 10^{11 19}F spins/cell was calculated by NMR. The lowest number of PFC-labeled cells detected by MRI at 3 T was 25,000, or 8.2 × 10¹⁵¹⁹F spins.

We then examined the use of ¹⁹F MRI at 3 T to track immune cells in vivo. In 3 T images of 4T1 tumor-bearing mice, 24 h post-PFC injection, we observed ¹⁹F signal in the liver, spleen, tumor and reference tubes. We were able to quantify the number of ¹⁹F spins in each 4T1 tumor. The average value for ¹⁹F spins measured in 4T1 tumors at 3 T was 9.5 × 10¹⁸, with the spatial distribution of ¹⁹F signal around the tumor periphery, which agrees with our previous findings; that study was performed at 9.4 T [5]. The accumulation of TAMs and subsequent ¹⁹F signal is sometimes referred to as “hot spots” [28]. Macrophages have been identified by multiphoton microscopy to localize as clusters in the periphery, or the “invasive edge”, and their role here is to allow for tumor expansion and to aid in tumor cell intravasation [35, 36].

For mice imaged at both 3 T and 9.4 T, the number of ¹⁹F spins measured in liver and tumor was not significantly different between the two magnetic field strengths. Despite the higher magnetic field strength, ¹⁹F SNR was only slightly higher at 9.4 T compared to 3 T, due to the use of a surface coil at 3 T. While cell detection limit will be affected by many factors, including, for example, ¹⁹F cell loading, MRI sequence acquisition parameters, type of radiofrequency (RF) coil used and MRI field strength, in this study the different RF coils used at the two field strengths likely played the major role in our ability to detect cells at 3 T. We utilized an in-house built volume birdcage RF coil at 9.4 T, while at 3 T we utilized a commercially built surface coil. Sensitivity and SNR from each coil are dependent on both design and use. For example, we can optimize birdcage coil diameter and length and “filling factor” [37] to improve sensitivity. The birdcage coil we used was built for imaging mice and, therefore, a mouse fits well into the coil (i.e., mouse size dimension equals coil dimension). Surface coils provide superior sensitivity versus volume coils and this is because of a smaller coil diameter and proximity to the region of interest [38]. These coils were both built to maximize sensitivity, with respect to their configurations and use, but it is impossible to compare the two considering their inherent differences.

Our study and others have demonstrated that ¹⁹F signal is related to PFC-positive TAMs identified by microscopy [4, 5, 7, 27]. If we use the number of ¹⁹F spins/cell measured from the NMR of macrophages labeled in vitro, we can estimate that we are detecting roughly 30–55 × 10⁶ PFC-labeled cells in the 4T1 tumors only imaged at 3 T. Since cells labeled in vitro will have a higher ¹⁹F loading per cell, compared with cells labeled by an IV injection of PFC, this is likely an underestimation.

¹⁹F quantification is typically performed using manual delineation of only the perceived ¹⁹F signal, with ¹H images used for anatomical reference (“¹⁹F only” ROI) [2, 3, 27]. However, the distribution of TAMs within a tumor may be sparse and heterogeneous; this has been observed by both clinical histological examination [18, 39] and with preclinical ¹⁹F MRI [5, 7]. This, along with low SNR data, will cause regions of ¹⁹F signal to be less obvious and may lead to user bias with respect to the ROI chosen for ¹⁹F quantification. Here, we used two different methods for ¹⁹F quantification. When “¹⁹F only” ROI was utilized, a wider range of values for the number of ¹⁹F spins was found between the low and high NEX acquisitions and there was a trend of higher numbers of ¹⁹F spins measured as NEX/SNR increased. We observed that in low SNR ¹⁹F images, when choosing the “¹⁹F only” ROI, the size of the ROI was smaller. When using the “tumor” ROI, there was more consistent ¹⁹F spin quantification between low versus high SNR ¹⁹F images. This method may produce more consistent results in vivo, as the ROI is taken from an anatomical image and not from perceived ¹⁹F signal. However, in ¹⁹F images where the signal is obvious and localized to a homogeneous area, ROI can be taken from the ¹⁹F data, such as what we have done for the images of PFC-labeled cell pellets.

The ¹⁹F MRI methods described could also be adopted for use in other disease models. Not only could an IV injection of a PFC agent label monocyte/macrophage populations in situ to quantify inflammatory responses, but also, for example, ¹⁹F MRI could track and quantify a specific, in vitro PFC-labeled cell population, over time. This would be of use in cellular therapy, where ¹⁹F MRI could assess the location and success of migration. This has been implemented by Ahrens et al. to study PFC-labeled dendritic cells (DCs), which were introduced into colorectal adenocarcinoma patients as an immunotherapy in a clinical trial [34]. In this instance, they were able to image and quantify the PFC-labeled DCs at the injection site and noted a large decrease in ¹⁹F signal at 24 h, which they attribute to DC migration to lymph nodes, or cell death resulting in clearance of the PFC. For future clinical translation of our work, the surface coil utilized in our study was built to image small regions on a human patient and is available commercially. A surface coil like this would be useful for imaging a superficial tumor to study TAMs or a lymph node for immunotherapy studies to confirm successful migration.

Despite strong evidence for a link between TAM content in breast cancer and patient outcome, there are few strategies for measuring TAMs in breast tumors and currently no in vivo approach. With continued developments of this imaging technology on clinical MRI systems, ¹⁹F MRI cell tracking has the potential to be used routinely to provide a picture of TAM distribution and a measure of TAM density in breast tumors; this information can be used in a meaningful way to predict response to therapy and monitor treatment.