Detection and Viability of Murine NK Cells In Vivo in a Lymphoma Model Using Fluorine-19 MRI

Abstract

Natural Killer (NK) cell therapies are being increasingly used as an adoptive cell therapy for cancer because they can recognize tumor cells in an antigen-independent manner. While promising, the understanding of NK cell persistence, particularly within a harsh tumor microenvironment, is limited. Fluorine-19 (¹⁹F) MRI is a non-invasive imaging modality that has shown promise in longitudinally tracking cell populations in vivo; however, it has not been studied on murine NK cells. In this study, the impact of ¹⁹F labeling on murine NK cell viability and function was assessed in vitro and then used to quantify NK cell persistence in vivo. While there was no noticeable impact on viability, labeling NK cells with ¹⁹F did attenuate cytotoxicity against lymphoma cells in vitro. Fluorescent microscopy verified ¹⁹F labeling in both the cytoplasm and nucleus of NK cells. Lymphoma bearing mice were given intratumoral injections of ¹⁹F-labeled NK cells in which signal was detectable across the 6-day observation period via ¹⁹F MRI. Quantification from the composite images detected 78–94% of the initially injected NK cells across 6 days, with a significant decrease between days 3 to 6. Postmortem flow cytometry demonstrated retention of ¹⁹F intracellularly within adoptively transferred NK cells with less than 1% of ¹⁹F-containing cells identified as tumor-associated macrophages that presumably ingested nonviable NK cells. This work demonstrates that ¹⁹F MRI offers a specific imaging platform to track and quantify murine NK cells within tumors noninvasively.

Introduction

Over the past two decades, cancer immunotherapies have demonstrated great success against a wide range of cancers.^1,2 However, tumors can evade the immune response by downregulating antigen presentation, reducing immunogenicity, and creating a tumor microenvironment that is exclusive of immune effector cells.³ To overcome these barriers, strategies have focused on expanding and activating immune effector cells in vivo with biologic compounds such as monoclonal antibodies (mAbs), immune checkpoint blockades, and adoptive cell transfers (ACT) of lymphocytes to augment cytotoxic activity and tumor infiltration.⁴

One method of ACT involves using natural killer (NK) cells, members of the innate immune system that preferentially target virally infected and malignant cells without prior sensitization.⁵ NK cells offer distinct advantages over T cell therapies as they do not require human leukocyte antigen matching and do not contribute to graft-versus-host-disease.⁶ NK cells also have the ability to attack malignant cells via antibody-dependent cellular cytotoxicity (ADCC), produce cytokines like interferon-gamma (IFN-g) and tumor necrosis factor-alpha (TNF-a), and have the potential to be highly activated and expanded ex vivo.^7,8 In the last decade, the number of clinical trials using NK cells has increased, but other than a promising recent study showing high remission rates observed with CD19 chimeric antigen receptor (CAR) NK cells⁹, most prior work has shown only modest results.^10–15 This may be partially due to the inability to track and quantify NK cells within tumors after infusion, which may improve the efficacy of treatment.^3,16 In this field of study, animal models have been critical to furthering the understanding of in vivo function and antitumor effects of NK cells.^17,18

Currently, most immunotherapies lack biomarkers to predict the treatment response in patients.¹⁹ A growing body of evidence suggests that the presence of NK cells in solid tumors is associated with more favorable clinical outcomes.^20,21 However, the behavior and persistence of these NK cells after injection are largely unknown although they may hold predictive information²² and inform the timing of cellular infusions.²³ Molecular imaging platforms can provide an avenue to answer these important concerns.²⁴

Magnetic resonance imaging (MRI)-based cell monitoring is free of ionizing radiation and capable of noninvasive, intratumoral imaging throughout the body.²⁵ Fluorine-19 MRI (¹⁹F MRI) is a multinuclear quantitative imaging technique of fluorinated molecules, such as perfluorocarbons (PFC). Among the most commonly imaged PFCs are perfluoropolyether (PFPE) and perfluoro-15-crown-5-ether (PFCE), which can be emulsified to either directly label specific cell populations ex vivo, on indirectly label cells in vivo. When PFCs are introduced in vivo the PFC agent will enter the reticuloendothelial system where phagocytic cells like monocytes and macrophages will take up the agent.²⁶ PFPE, which is of particular interest to our work, is a linear PFC that has a relatively simple NMR spectra and contains >40 magnetically equivalent ¹⁹F nuclei in its main resonance peak.²⁷ Ex vivo ¹⁹F-labeling of cells enables subsequent detection in vivo by ¹⁹F MRI after adoptive transfer.^28,29 Due to the low endogenous concentration of ¹⁹F in vivo, an unambiguous positive signal is detected from the PFPE-labeled cell population without a confounding background signal. Quantification of cell numbers can then be approximated directly from the spin-density weighted ¹⁹F MR images³⁰ assuming PFPE relaxation times are similar for in vivo conditions and regional flip angle variations are small. These assumptions are reasonable for our pre-clinical application in a homogenous tumor environment but may not hold for larger fields of view and heterogeneous local environments expected clinically. Thus, more rigorous development of a rapid workflow for B1-field, and T1, and T2 mapping will be important for clinical translation. Furthermore, conventional proton (¹H) MRI can be acquired in the same imaging session to visualize cell populations in the correct anatomical context. This form of hot-spot ¹⁹F MRI is already in clinical testing in cancer patients with the labeling of dendritic cells³¹; however, ¹⁹F MRI tracking of NK cells is still in preclinical development.^23,32

Previously, we demonstrated the ability to label human NK cells and detect and quantify the cells longitudinally in 3 different xenograft mouse models²³. However, one limitation of xenograft models is the usage of immunodeficient mice that do not inform the impact of the complete endogenous immune system on cell viability and trafficking. The goals of this study were to optimize the labeling for murine NK cells and validate cell-specific label retention in vivo, while longitudinally monitoring and quantifying the persistence of NK cells within a syngeneic, immunocompetent murine lymphoma model. ¹⁹F MRI, while very specific, suffers from relatively low sensitivity compared to nuclear imaging methods²⁴, placing an emphasis on maximizing cellular uptake of the PFPE ¹⁹F label while mitigating overt toxicity to the NK cells. The use of a ¹⁹F MRI-optical dual probe, PFPE-Red, allows for validation of labeling both before infusion and postmortem via flow cytometry and confocal microscopy. A green fluorescent protein (GFP) reporter mouse can be used to specifically co-localize GFP⁺ NK cells in conjunction with the PFPE-red label.

In this study, we performed in vitro experiments to verify PFPE-Red labeling of murine NK cells, monitored NK cells longitudinally with ¹⁹F MRI in vivo in a murine lymphoma model, and verified cell viability 0, 3, and 6 days after injection via flow cytometry, and imaging flow cytometry (ImageStream). Injection of PFPE-labeled GFP+ murine NK cells into a syngeneic lymphoma tumor model indicated that NK cells can be successfully detected and quantified with NK cell-specific ¹⁹F signal persistence up to 6 days post-injection.

Materials & Methods

Study Design

In vitro studies were performed to establish cell viability, labeling efficiency and effects on cytotoxicity after incubation with PFPE. A total of 5 replicates of in vitro cell cultures were incubated with 2,4, and 8 mg/ml of PFPE to find the optimal loading of PFPE per cell while maintaining cell viability. Viability was assessed in duplicate at PFPE-incubation times of 4 and 24 hours at 1, 2, 4, and 8 mg/mL. NK cell cytotoxicity was also evaluated using 6 replicates at 2, 4, and 8 mg/mL to compare to previous work in human NK cells.²³

All protocols were approved by the Institutional Animal Care and Use Committee (IACUC) at the University of Wisconsin-Madison. In vivo studies included a total of 6 male C57BL/6NCr mice studied longitudinally at day 0, 3, and 6 post-injection of PFPE-red labeled murine GFP positive (GFP+) NK cells derived from male C57BL/6-Tg(UBC-GFP)30Scha/J mice with the goal of detecting exogenous NK cells using the PFPE-red label. Cell viability was then verified with the co-incidence of the PFPE-red fluorescence label with the GFP label on flow cytometry after sacrifice. Four animals completed all time points and were sacrificed on day 6 to harvest the tumors. Two additional animals were sacrificed 2 hours after injection and on day 3 to assess cell viability and label specificity at the intermediate imaging time points. Tumors were harvested on day 6 to stay within protocol limits for allowed tumor size. The overall study design has been summarized, graphically, in Figure 1.

Figure 1: Graphical summary of the experimental design.

A graphical overview of the experimental design that starts with the establishment of the syngeneic tumor, purification of NK cells, and subsequent labeling with PFPE-red agent. (A) The in vitro arm of the experiment was used to characterize the cytotoxic effects of the labeling procedure (n=6), characterize the viability (n=2) and label uptake (n=5), and confirm labeling with optical imaging methods. (B) The in vivo arm of the experiment carried out with (n=6) intratumoral injections of either 1.0×10⁶ or 5.3×10⁵ PFPE-red labeled GFP+ NK cells into tumor-bearing mice that were imaged at days 0, 3, and 6 post injection. One mouse was euthanized at the baseline and day 3 timepoint, followed by 4 mice at the conclusion of day 6 of imaging. The tumors (n=6) were excised and analyzed via flow or imaging flow cytometry to assess label retention and cell viability.

Purification, expansion, and labeling of NK cells

Male C57BL/6NCr mice were purchased from the National Cancer Institute (NCI) Animal Production Program and Charles River Laboratories International (Frederick, MD) and used between 24–25 weeks of age. Male C57BL/6-Tg(UBC-GFP)30Scha/J mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and used between the 6–12 weeks of age. Mice were stored in a pathogen-free facility throughout the entirety of the study.

NK cells were isolated by harvesting the spleen and bone marrow and processing them into single-cell suspensions. Erythrocytes were depleted using ACK lysing solution (Lonza, Walkersville, MD). NK cells were isolated by negative selection using magnetic cell separation beads (Miltenyi Biotec, cat#: 130-115-818), sorted using AutoMACS® Separator (Miltenyi Biotec, San Diego, CA), and cultured at 37°C in 5% CO₂ in complete mouse media³³ (CMM) supplemented with 100 IU/mL recombinant human IL-2 (NCI Biological Resources Branch, Frederick, MD).

Quantification of PFPE-label uptake via ¹⁹F-NMR

To determine ¹⁹F uptake, NK cells were labeled (n=5 replicates) with PFPE (Celsense, Pittsburgh, PA) added directly into the cell culture medium (1 ×10⁶ cells/mL) at 2, 4, or 8 mg/mL. After a 24-hour incubation at 37°C with 5% CO₂, cells were harvested and washed 3 times with 1x PBS (Fisher Scientific, Pittsburgh, PA). Cells were counted using a Coulter Counter (Beckman Coulter, Indianapolis, IN) and aliquoted into microcentrifuge tubes for lysis. Cells were centrifuged at 1200 RPM for 6 minutes. Dry cell pellets were lysed using a lysis buffer containing deuterium oxide (D₂O), 1% Triton X-100 (Sigma Aldrich, St. Louis, MO), and 0.1% trifluoroacetic acid (TFA, Sigma Aldrich). TFA served as a ¹⁹F NMR chemical shift reference and fluorine quantification reference to determine the uptake of the PFPE cellular label. Samples were transferred into NMR tubes (Wilmad Lab Glass, Vineland, NJ) and further diluted with D₂O such that the cell lysate solution fit the entire detection region of the NMR coil.

NMR spectroscopy was performed on a Varian Unity-Inova 500 MHz NMR spectrometer (Agilent Technologies, Santa Clara, CA). A 1D FID was acquired with a $TR$ = 9 s, $\theta$ = 90°, 14.075 kHz spectral range, and $NEX$ = 64. TFA and PFPE resonant peaks were integrated at −76 ppm and −91 ppm, respectively. The spectral areas from the cells ($I_C$) and reference ($I_R$) peaks, moles of reference ($M_R$), and the number of NK cells ($N_C$) along with Avogadro’s number ($N_A$), were used to determine the cellular loading density, or the number of ¹⁹F nuclei per cell ($F_C$), according to Eq. 1 below.

$$F_C=\frac{3\cdot I_C\cdot M_R\cdot N_A}{I_R\cdot N_C}$$ (1)

The mean $F_C$ was taken as the average of each sample (N = 5) and the uncertainty was estimated as the standard deviation of $F_C$. NMR peak integration was performed in VNMR 6.1C (Agilent, CA, USA) with plotting performed in Matlab (MathWorks, Natick, MA).

To confirm intracellular PFPE uptake, NK cells isolated from GFP+ mice were incubated with 4 mg/mL of a dual MRI-optical PFPE-red cellular label (Celsense, Pittsburgh, PA) for 24 hours. PFPE-Red contains the linear PFPE polymer covalently conjugated to the red fluorescent dye BODIPy-TR with 588/616 nm excitation/emission. PFPE-red labeled NK cells were stained for 5 minutes with 2mL of a 300 μM solution of DAPI (ThermoFisher, cat#: D1306) in 1x Dulbecco’s PBS (DPBS), washed with 1x PBS, and plated on a collagen-coated 35 mm glass-bottom dish (MatTex Corp., Ashland, MA). Live-cell fluorescence microscopy was performed on a Nikon A1RS 4-channel inverted point scanning confocal microscope system (Nikon Inc., Japan) with a 60x objective lens. This system can utilize 4 lasers to simultaneously excite blue (408 nm), green (488 nm), red (561 nm), and far-red fluorophores (638 nm). Image processing was performed in ImageJ (v1.6.0, NIH, USA).

NK cell viability and cytotoxicity post-labeling

NK cell viability was tested by trypan blue (Thermo Scientific, cat#: SV300084.01) exclusion assay using a hemocytometer. Samples of NK cells were incubated at concentrations of 1, 2, 4, and 8 mg/mL PFPE-red, in duplicate, for a total of 4 and 24 hours with control groups that had no PFPE label added.

NK cell cytotoxicity was determined by a 4 hour [⁵¹Cr]-release assay against the NK cell–sensitive lymphoma cell line Yac-1 (Sigma-Aldrich, cat#86022801) using either ¹⁹F-labeled or unlabeled NK cells (effector). Yac-1 (target) cells were labeled with 50 μCi of ⁵¹Cr (NEZ030S; Perkin Elmer, Hopkinton, MA) per 10⁶ cells and incubated for 90 minutes at 37°C. Cells were then washed to remove extra ⁵¹Cr and target cells were resuspended in media. Yac-1 targets (5 ×10³) were added at different effector to target ratios in a 96-well plate and incubated at 37°C for 4 hours. Each group was measured using 6 replicates (n=6). The γ-scintillation of supernatant was measured by a γ-counter (PerkinElmer). Maximum release was determined by adding 100μL of 1X-Triton X-100 detergent (Sigma-Aldrich, cat#: 9002-93-1) to target cells. Spontaneous release was determined by adding 100μL of media to 100μL of target cells. The percent lysis was determined by ⁵¹Cr release calculated according to equation 2.

$$\text{\% Lysis}=100\%\times \frac{(\text{Experimental}-\text{Spontaneous})}{(\text{Maximum}-\text{Spontaneous})}$$ (2)

Murine syngeneic tumor model

EL4 lymphoma cells (ATCC® TIB-39, Manassas, VA) were grown in CMM at 37°C in 5% CO₂ and used after 3–5 passages in culture after thawing. Cell authentication was performed using short tandem repeat analysis (Idexx BioAnalytics, Westbrook, ME) and per ATCC® guidelines using morphology, growth curves, and Mycoplasma testing within 6 months of use using the e-Myco mycoplasma PCR detection kit (iNtRON Biotechnology Inc, Boca Raton, FL). To establish a syngeneic lymphoma model, EL4 cells were injected subcutaneously at 10 ×10⁶ cells in 0.1mL into the flank of (n=6) C57BL/6 male mice. Once tumors reached 5mm in diameter, GFP+ NK cells were injected intratumorally as described below.

In vivo ¹H/¹⁹F MRI

GFP+ NK cells were labeled with PFPE-Texas Red (Celsense) at 4 mg/mL for approximately 24 hours. A total of n=6 EL4 T cell lymphoma-bearing mice were injected with 5.3×10⁵ (n=3) and 1.0×10⁶ (n=3) viable labeled NK cells in 50 μL. Injected mice were imaged directly after injection (day 0), day 3, and day 6. Mice were induced with 3% isoflurane (Abbott Laboratories, Abbott, IL), monitored with a respiration pad, and maintained at 37°C using a warm-air blower. Mice were placed prone into a home-built ¹H/¹⁹F quadrature volumetric RF coil. An external PFPE phantom with known ¹⁹F spin density (ρ_R) was placed within the imaging volume and used as a ¹⁹F signal reference during image acquisition. MRI was performed using a horizontal bore 4.7T small animal MRI (Agilent Technologies, Santa Clara, CA). Anatomic T₁-weighted images were acquired with a spoiled gradient-echo with a flip angle (FA) of 20°, 132.6/4.2 ms TR/TE, a field of view (FOV) of 72×36 mm² with 2 mm slice thickness, acquisition matrix of 256×128, 50 kHz receiver bandwidth (RBW), NEX = 3, and a scan time of 50 seconds. T₂-weighted images were acquired with a fast spin-echo (FSE), 3000/20 ms TR/TE, 16 echo train length (ETL), matching FOV and resolution, and a scan time of 1 minute 12 seconds. The coil was then tuned to match the resonant frequency of ¹⁹F (188 MHz). ¹⁹F images were acquired using an FSE with a 1250/20 ms TR/TE, ETL of 8, matching FOV and slice thickness, acquisition matrix of 64×32, in-plane resolution of 1.1×1.1 mm², 16 kHz RBW, 400 NEX, and an added chemical saturation pulse at +2150 Hz to mitigate any ¹⁹F signal contamination from isoflurane, for a total scan time of 25 minutes.

¹⁹F MRI Postprocessing and Cell Quantification

Image postprocessing and cell quantification were performed using FIJI ImageJ (National Institutes of Health, USA, http://imagej.nih.gov/ij) and MATLAB 2018B. The first postprocessing step was to scale the images against their noise, on a per slice basis. Given that the magnitude operation is nonlinear, the noise will follow a Rician distribution instead of a Gaussian distribution; therefore, magnitude images were first corrected to their expected Gaussian value. Using MATLAB, regions of interest (ROIs) were placed in areas devoid of signal in each slice of the magnitude images, where the mean value was corrected to give the standard deviation of the expected Gaussian distribution (s_N). Voxel values were then corrected using the scheme outlined in Gudbjartsson and Patz 1995:³⁴

$$\tilde{A}=\sqrt{{|I}^2-2{\sigma }_N^2|}$$ (3)

where I is the measured signal intensity and $\tilde{A}$ is the expected value. To limit noise propagation, images were thresholded at 4s_N to ensure an adequate signal-to-noise ratio (SNR). After correcting and scaling the images against their own noise, the quantification process can now take place using ImageJ. For each slice, pixels within the manually segmented tumor were summed to represent the total ¹⁹F signal within the volume (S_V). An ROI was then placed within the boundaries of the ¹⁹F reference vial where the mean value (${\stackrel{-}{S}}_R$) was recorded along with its standard deviation. The known spin density of reference vials (¹⁹F atoms/mL) was normalized against the voxel volume to give ¹⁹F atoms/voxel. Using the mean loading density, F_C, (¹⁹F atoms/cell) the apparent number of cells in the MR images was calculated as seen in Equation 4.

$$N_C=S_V\cdot \frac{{\rho }_R}{\begin{array}{c}{F}_C\cdot {\stackrel{-}{S}}_R\\ \end{array}}$$ (4)

The associated uncertainty in the apparent number of cells (s_Nc) was calculated according to Equation 5:

$${\sigma }_{N_C}=N_C\cdot \sqrt{{\left(\frac{{\sigma }_{S_V}}{S_V}\right)}^2+{\left(\frac{{\sigma }_{F_C}}{F_C}\right)}^2+{\left(\frac{{\sigma }_{{\stackrel{-}{S}}_R}}{{\stackrel{-}{S}}_R}\right)}^2}$$ (5)

Where ${\sigma }_{S_V}={\sigma }_N\cdot \sqrt{2\cdot N}$ and N is the number of nonzero pixels in the tumor region. For display purposes, ¹⁹F images were resampled (bicubic interpolation) via MATLAB to match the image size of the ¹H images and composite images were created for anatomic reference.

Postmortem Flow Cytometry

Immediately after imaging on day 6, mice were euthanized, and their tumors were excised for either postmortem flow cytometry (n=3) or image flow cytometry (n=1) analysis. Tumors were cut into 2–3mm pieces and digested with 2 ml of 2 mg/ml collagenase D (Worthington Biochemical, Cat#: LS004188) for 10 minutes at 37° C. 10 ml of 1x PBS was added to quench the collagenase D reaction, at which point the tumors were mechanically dissociated and passed through a 70μm cell strainer (Corning®, New York, New York). Tumor cell suspensions were then spun down in a centrifuge at 1200 RPM for 10 minutes. The cell suspensions were then digested with ACK lysing buffer. Cell suspensions were then washed and spun down again at 1200 RPM, and then resuspended in 10ml 1x PBS, 10μL of which was added to 10mL of Isoton® diluent and counted on a Coulter Counter® Z1 Series Particle Counter (Beckman Coulter, Brea, CA). From this count, 10⁶ cells were added to flow cytometry tubes in staining buffer (PBS with 2% FBS, 0.05% sodium azide, and 0.05% EDTA) and stained with 5μL of Alexa Fluor 647 anti-mouse F4/80 antibody (Biolegend® catalog #) for 30 minutes at 4° C. The cells were then washed with 1x PBS, spun down, and 0.5μL of Ghost Dye™ Red 780 viability dye (Tonbo Biosciences) was added for 20 minutes at room temperature. Cells were then washed with staining buffer, spun down, and resuspended in 400μL of staining buffer. The cells were then run on either the Attune™ NxT Flow Cytometer (Thermo Fisher Scientific) or the Amnis™ ImageStream MKII (Luminex Corp, Madison, WI), with subsequent analysis being performed with FlowJo™ software (BD, Ashland, OR). The addition of the viability and antibody dyes to the 3 resected tumors, allow us to first isolate the viable from non-viable cells, then further classify the cells as either adoptively transferred (GFP+) or endogenous (GFP−), labeled (PFPE-red+) or unlabeled (PFPE-red-), and endogenous macrophage (GFP−, F4–80+) or not.

Statistics

Statistics were performed using GraphPad Prism version 9.0 (GraphPad Software, San Diego, CA). MRI-based cellular quantification data were expressed as mean ± SD, whereas all other data were expressed as mean ± SEM. For analysis of the three cell quantification groups, a mixed-effects model was implemented. For significant results, a post hoc Tukey’s multiple comparison test was performed. A P-value (p) less than 0.05 was considered statistically significant.

Results

PFPE labeling optimization of NK cells

PFPE labeled NK cells showed similar viability to that of the unlabeled controls at both 4- and 24-hours using concentrations of 1, 2, 4, and 8 mg/mL PFPE (Figure 2A). A reduced NK-mediated cytotoxicity was detectable at high effector: target ratios in the presence of 2, 4, and 8 mg/mL of PFPE compared to unlabeled control at 24-hours post-labeling (Figure 2B). NMR results (Figure 2C) indicated a mean PFPE labeling on the order of 10^{12 19}F atoms/cell for all three labeling concentrations. Thus, while the ¹⁹F-PFPE label did not impact the viability of murine NK cells and led to a high number of ¹⁹F atoms on a per cell basis, there was a measurable reduction of cytotoxic potential across the labeled groups compared to control. For this reason, 4 mg/mL PFPE-Red was chosen as our labeling concentration for in vivo experiments, which achieved an average PFPE-red loading of 2.0±0.5 ×10^{12 19}F atoms/cell.

Figure 2: Viability, cytotoxic potential, and PFPE loading of murine NK cells after labeling.

(A) Trypan blue exclusion assay results of mouse-derived NK cellular viability post-labeling with 4 different concentrations of PFPE compared to control, in duplicate (n=2). All bars were normalized to a viability measurement at 0 hours for all groups. (B) A [⁵¹Cr] release assay for NK (effector) cell cytotoxic function against a NK cell sensitive tumor cell line, Yac-1, is shown after incubation with different concentrations of PFPE label. Each group was performed with six replicates (n=6). (C) NK cell labeling density in terms of ¹⁹F atoms per cell ($F_c$) as measured by ¹⁹F NMR from five replicates (n=5). Statistical difference indicators: * = p<0.05, ** = p<0.01, *** = p<0.001, **** = p<0.0001.

Live-cell confocal microscopy validation of cellular labeling

Live-cell confocal microscope images of DAPI-stained, PFPE-red labeled GFP⁺ NK cells demonstrated intracellular labeling could be detected in the presence of an endogenous fluorescent transgene (Figure 3). This data indicated that ¹⁹F is taken up by murine NK cells, can be detected intracellularly, and can be detected simultaneously in the presence of another fluorescent marker.

Figure 3: Live cell fluorescence microscopy images shows NK cell labeling.

Live cell fluorescence microscopy images of PFPE-Red labeled GFP+ NK cells, showing the fluorescent signal from (A) DAPI staining, (B) PFPE-Red, (C) GFP, and (D) white light individually, where (E) the merged image shows colocalization of the various labels. A 100 μm scale bar is included.

In vivo ¹H/¹⁹F MRI and cellular quantification

To assess the detectability of PFPE-labeled GFP⁺ murine NK cells in a syngeneic tumor model, EL4 lymphoma cells were injected into the flank and allowed to establish solid tumors with a minimum size of 5×5 mm². Then intratumoral injections of ¹⁹F labeled GFP⁺ NK cells were performed and ¹H/¹⁹F MRI was acquired at 0, 3, and 6 days post NK cell transfer. As seen in a representative mouse (Figure 4), labeled NK cells were successfully detectable in all mice at each imaging time point. To assess the remaining NK cells in the tumor at each time point, quantification was performed from the composite images, indicating an average of 91% of the initially injected cells were detected on day 0 (1–2 hours post-intratumoral injection), 94% on day 3, and 78% were detected at day 6 (Figure 5), where a statistically significant decrease was detected in NK cell number from days 1 to 6 and days 3 to 6 (p < 0.05).

Figure 4: ¹H/¹⁹F composite images showing signal across the 3 imaging timepoints.

Representative magnitude images across the 3 imaging time points. ¹⁹F Images were scaled on a per slice basis against their own corrected noise. The ¹⁹F signal (“hot metal” color and color bar) from the labeled cells is visible within the tumor in all three images with an average SNR within the tumor of 11±2, 14±3, and 13±4. White arrow indicates the reference vial of known ¹⁹F spin density.

Figure 5: ¹⁹F MRI murine NK cell quantification verifies NK cell persistence over 6 days.

In vivo MRI NK cell quantification results indicate the percentage of initially injected GFP+ NK cells within the boundaries of the mice tumors. The percentage of initially injected cells from days 0 to day 6 is shown where a significant decrease is detected from days 3 to 6 (p<0.05). Error bars are represented as the standard deviation of the NK cell quantification calculation.

Postmortem flow cytometry

At days 0, 3, and 6 imaging time points, 1 mouse was euthanized, and its tumor resected for analysis by image flow cytometry. As depicted in Figure 6A, cross-sectional images for brightfield, green, red, and green-red composite channels of a GFP+ PFPE-red+ cell show internalization of the imaging agent at each timepoint. ImageStream also demonstrated (Figure 6B–C) that greater than 60% of the GFP+ NK cells and less than 7% of the endogenous cells (GFP−) contained PFPE-red label at all imaging time points. However, between days 3 and 6, we see a rapid decline in the percentage GFP−, PFPE-red+ and an increase in the percentage of labeled GFP+, PFPE-red+ present within the tumor, suggesting clearance of non-viable cells and free imaging agent between these time points.

Figure 6: Postmortem ImageStream flow cytometry validation of cell label retention across 3 time points.

(A) ImageStream flow cytometry showing the brightfield, green, red, and composite channels for a representative GFP+ PFPE-red labeled murine NK cell that was detected in the resected tumors at days 0, 3, and 6 post injection, where each time point represents a tumor from a single mouse. (B) The resultant percentage of GFP+ that were ¹⁹F+ (labeled adoptively transferred population) and (C) GFP− that were ¹⁹F+ (endogenous labeled population) have been plotted in blue and red, respectively. A peak in non-specific labeling is seen at day 3, followed by both a sharp increase in percentage of labeled GFP NK cells and a decrease in percentage of endogenously labeled cells.

In addition to the imaging flow cytometry, on Day 6, the remaining 3 tumors were resected and analyzed by conventional flow cytometry, which demonstrated a mean of 52±8% of the remaining adoptively transferred GFP⁺ NK cells contained ¹⁹F PFPE-red (Figure 7A). The mean GFP+ viability for these tumors was approximately 39.7±17%. In contrast, less than 0.3±0.1% of the endogenous tumor GFP⁻ cells contained ¹⁹F PFPE-red. Within the 3 tumors analyzed a mean of 29±3% of the endogenous ¹⁹F PFPE-red⁺ cells were F4/80⁺ tumor-associated macrophages (Figure 7B). The median fluorescent intensity (MFI) of the ¹⁹F PFPE-Red was decreased (p < 0.01) in endogenous cells within the tumor by a factor of 2.3 compared to the adoptively transferred NK cells (Figure 7C) indicating the primary cell population that is detected and quantified by ¹⁹F-MRI are viable, adoptively transferred NK cells.

Figure 7: Postmortem flow cytometry indicates viable, PFPE-labeled NK cells at day 6 post-injection.

Flow cytometric analysis at day 6 post-intratumoral injection of NK cells from 3 excised murine syngeneic EL4 lymphoma tumors showing (A) the percentage of cells from GFP+ and GFP− that contain the PFPE-red label for each mouse, (B) the quantified MFI for GFP+ and GFP− PFPE-red containing cells, where the error bars represent the standard deviation in the measure, (C) the percentage of GFP− PFPE-red containing cells that are macrophages, and (D) a summary of the mean values found by flow cytometry across the tumors. The post-processing GFP+ viability was 64.7%, 29.0%, and 25.5% for mice 1, 2, and 3, respectively. Mean GFP+ viability was 39.7%. Note: ** = p-value < 0.01.

Discussion

NK cells continue to be an emerging cell therapy for the treatment of relapsed and refractory cancers, including lymphoma.⁹ Developing cellular imaging methods to monitor and assess NK cells in vivo will address one of the challenges associated with the translation of immunotherapies from preclinical animal models to humans. Without quantitative image guidance, the clinician is blinded as to where the effector cells migrate to, how long they persist, and if they reach the target tissue²², which can make treatment timing and assessment challenging.^8,35 The field also does not know what minimum threshold of NK cells is needed to reside in a tumor to elicit a clinical response. Thus, pre-clinical development of a platform incorporating ¹⁹F-MRI with murine NK cells can leverage immunocompetent tumor models to test the feasibility for in vivo cell detection, the persistence of ¹⁹F-label, and quantification using MRI to further develop this technique for future clinical trials.

Cell tracking with ¹⁹F MRI, in general, will benefit from data acquisition strategies that improve the detection sensitivity to monitor cell behavior in vivo. Here, we focused on improving the quantification and labeling strategy. The signal strength is determined in part by the number of ¹⁹F nuclei infused into the cells during incubation with ¹⁹F-PFPE. We found a balance between ¹⁹F-PFPE concentration and labeling efficiency in terms of cell viability and killing effectiveness, which allowed for a greater average signal per cell and thus, a more MRI-sensitive acquisition while maintaining the cell-killing capacity. Previous work by us²³ and others³² have shown PFPE labeling of human NK cells is feasible, non-toxic, and does not appear to affect viability, cytotoxicity, or cytokine secretion by NK cells. Here we show that the cellular uptake or labeling efficiency can then be quantified by ¹⁹F NMR, where the mean PFPE uptake showed a linear increase in loading (¹⁹F nuclei/cell) with increasing PFPE labeling concentration. Moreover, the observed cellular labeling efficiency (¹⁹F nuclei/NK cell) for murine NK cells in this study by NMR is nearly 10-fold greater than those reported previously by our group using human NK cells²³ which is advantageous for detection by MRI. However, ¹⁹F-PFPE labeling of murine NK cells showed a modest but statistically significant impact on the cytotoxic function of the labeled NK cells, unlike our previous analysis of human NK cells²³, suggesting a species-specific, cell subset sensitivity to ¹⁹F as no impact on cytotoxicity has been observed with murine³⁶ or human³⁷ T cells.

Our study demonstrates viable NK cell detection delivered via intratumoral injection in vivo over 6 days, where a significant decrease in labeled NK cells was detected after the day 3 imaging timepoint. The use of the ¹⁹F-PFPE-red and the secondary GFP label confirmed cellular uptake before infusion via confocal microscopy and assessment of label retention and cellular viability via postmortem flow cytometry. The postmortem analysis indicated that an overwhelming majority of the ¹⁹F MRI signal was specific to viable adoptively transferred ¹⁹F-labeled NK cells retained in the tumor after 6 days. The results from the imaging flow cytometry indicate that at days 0 and 3, the majority of GFP+ cells within the tumor still contained their label and provides visual confirmation that the PFPE-red agent is contained within the cell membrane. Furthermore, the reduction in the percentage of endogenously labeled cells and increase in the percentage of labeled adoptively transferred cells from days 3 to 6 suggest clearance of endogenously labeled cells from the tumor. Less than 0.5% of the endogenous cells within the tumor presented with the ¹⁹F label, supporting the conclusion that the signal detected by ¹⁹F-MRI remains specific to the adoptively transferred NK cells. Overall, GFP+ NK cell viability at Day 6 was widely variable, with an average of 39.7±17%. While the source of this large range is unknown, we suspect it could be related to inherent biological variability and/or differences in tumor processing time after collagenase digestion for the respective tumors. This result may warrant further research. The presence of ¹⁹F⁺ tumor-associated macrophages is consistent with the expected makeup of the tumor microenvironment in that any nonviable ¹⁹F⁺ NK cells would be likely phagocytosed by tumor-associated macrophages but given the relatively small amount detected and their lower mean fluorescent intensity, the presence of phagocytic cells within the tumor had minimal effect on the appearance of ¹⁹F images and their quantification.

Several imaging modalities with higher cell detection sensitivity than ¹⁹F MRI have previously utilized in pre-clinical settings to monitor NK cell infusions, including positron emission tomography (PET), single-photon emission computed tomography (SPECT), optical imaging (OI) (bioluminescence and fluorescence imaging) and MRI.^24,38 While superparamagnetic iron oxide nanoparticle (SPIO) based MRI methods provide greater sensitivity, they rely on indirect detection of the cells by measuring negative contrast.^39,40 These methods are therefore less specific and due to the non-linear relationship between iron concentration and signal loss, quantification is much more complicated.⁴¹ PET/SPECT isotopes such as Fluorine-18 (¹⁸F), Carbon-11 (¹¹C), and Indium-111 (¹¹¹In) offer high detection sensitivity to allow detection of smaller cell populations labeled with receptor-targeted radioactive isotopes; however, labeling NK cells with long half-life radioactive isotopes exposes patients to ionizing radiation, and could have toxic effects on NK cell viability. Additionally, PET and SPECT lack the high resolution of anatomical structures and only offer a limited imaging window due to radioactive decay of the isotope, preventing detection of long-term NK cell persistence.^16,42 Furthermore, detachment of the cell-bound radioisotope from the cell may give false positives for the true location of NK cells.⁴³ Optical imaging using fluorescent and bioluminescent techniques has previously provided insight into NK cell trafficking patterns in preclinical models^44,45, but optical methods have limited light penetration, severely impeding the clinical applicability of these techniques.^24,46

However, there are three important limitations in the work presented here. The first limitation is the small sample size, although each animal serves as its own control increasing the statistical power and the in-depth study to evaluate NK cell viability using flow cytometry is unprecedented. Second, we have not optimized the pulse sequence used suggesting further improvement in sensitivity is possible. Specifically, balanced steady-state free precession (bSSFP) and ultra-short echo time (UTE) have been explored for PFC-based ¹⁹F MRI to increase sensitivity.^47,48 With PFPE relaxation times at 4.7 T, Faber and Schmid deduced that a bSSFP approach would produce the optimal sensitivity or SNR-efficiency.⁴⁹ However, in this study we chose to use a 2D FSE to reduce potential artifact presence. Other methods using T₁-shortening agents like gadolinium-based contrast agents or superparamagnetic iron oxide (SPIO) nanoparticles to boost SNR-efficiency have also been studied.^39,50,51 While these improvements to the ¹⁹F contrast agents are promising, cellular toxicity and corresponding safety profiles have yet to be fully explored. The work presented here, and ¹⁹F MRI in general, could also benefit from acceleration techniques like parallel imaging⁵² or compressed sensing (CS) approaches.^53,54 CS techniques, which undersample k-space to reduce the amount of data required, have achieved acceleration factors of up to 8 without loss in SNR in ¹⁹F MRI applications.^55,56 These acceleration techniques can be used to decrease scan times for translational applications or allow for high signal averaging to increase maximum achievable SNR. The third limitation is that most clinical trials using adoptively transferred cells are administered intravenously. However, clinical trials testing intratumoral injection of immune cells like dendritic cells or CAR T cells have emerged in the last decade.^57–63 In the majority of cases, lymphomas occur in deep lymph nodes within the patient, where intratumoral injection is not feasible. For clinical translation, intratumoral injection of NK cells needs to be applied on large, superficial lymph nodes (e.g. cervical or supraclavicular) or advance to intravenous injection of NK cells with homing of ¹⁹F-labeled NK cells to deep regions of lymphomatous involvement. Nonetheless, the presented work establishes a timeline for adoptively transferred NK cell viability within the tumor microenvironment of an immunocompetent mouse, and thus is a first step to quantifying longitudinal cell-specific labeling in vivo for future IV injection experiments.

In this study, we optimized cellular labeling and quantification of murine NK cells with ¹⁹F in vitro and unambiguously detected and quantified ¹⁹F-labeled NK cells within an established syngeneic lymphoma tumor in vivo longitudinally across 6 days. Furthermore, we validated ¹⁹F label persistence and viability within adoptively transferred murine NK cells and found minimal potential false-positive readings from endogenous phagocytic cells within the tumor over the same time period. This work demonstrates the feasibility of tracking murine NK cells as a therapeutic adoptive cell therapy for cancer and will inform preclinical models of cancer immunotherapy.

List of Abbreviations

¹⁹F

Fluorine-19

⁵¹C

Chromium-51

ACK

Ammonium-chloride-potassium

ACT

Adoptive cell transfer

ADCC

Antibody-dependent cellular cytotoxicity

CAR

Chimeric antigen receptor

CMM

Complete mouse media

CS

Compressed sensing

CSE

Chemical shift encoding

D₂O

Deuterium Oxide

DPBS

Dulbecco’s phosphate-buffered saline

ETL

Echo train length

FA

Flip angle

FBS

Fetal bovine serum

FC

Cellular loading density

FOV

Field of view

FSE

Fast spin-echo

GFP

Green fluorescent protein

GVHD

Graft versus host disease

IACUC

Institutional Animal Care and Use Committee

IFN-γ

Interferon-gamma

IL-2

Interleukin 2

mAbs

Monoclonal antibodies

MFI

Median fluorescent intensity

NEX

Number of excitations

NK

Natural killer

PBS

Phosphate-buffered saline

PCR

Polymerase chain reaction

PET

Positron emission tomography

PFC

Perfluorocarbon

PFCE

Perfluoro-15-crown-5-ether

PFPE

Perfluoropolyether

ROI

Region of interest

SD

Standard deviation

SEM

Standard error of the mean

SENSE

SENSitivity encoding

SNR

Signal-to-noise

SPECT

Single photon emission computed tomography

SPIO

Superparamagnetic iron oxide

SSFP

Steady-state free precession

TFA

Trifluoroacetic acid

TNF-α

Tumor necrosis factor alpha

UTE

Ultrashort echo time

Data Availability

The data that support the findings of this study are available from the corresponding author, LML, upon reasonable request.