Introduction

Non-invasive imaging techniques have been an important tool for diagnosis and stratification of inflammation, both clinically and in animal models. Immune cell probes developed for in vivo imaging modalities such as positron emission tomography (PET) and magnetic resonance imaging (MRI) enable temporal visualizations of immune cell dynamics and inflammatory processes [1]. The evaluation of new in vivo molecular imaging markers requires same-subject signal validation via high-resolution histological methods in tissue slices and assays of necropsied tissues, for example by flow cytometry of single cell suspensions, to confirm target specificity. When comparing whole-body in vivo image data to histological/single cell assays, there remains a ‘resolution gap’ of ~ 10 to 1,000 µm in the ability to visualize and validate labeling of inflammatory immune cells. Additionally, localized populations of immune cells and small dispersed lesion sites may not be adequately sampled in representative histology and cell assay tissue panels. To address these challenges, we explore the use of cryo-fluorescence tomography (CFT), a three-dimensional histology technique [2,3,4] pioneered by Wilson and coworkers [2], in which the entire animal or intact organ/tissue is mounted in optimal cutting temperature (OCT) cryo-embedding compound en bloc and sectioned robotically using a modified cryo-macrotome in 50–100 μm slices. White-light and multi-spectral fluorescence images are acquired after each cut using a camera positioned over the block, and these data are reconstructed into high-resolution three-dimensional (3D) images. CFT has the ability to provide whole animal 3D images at an intermediate-to-high resolution, bridging the resolution gap between in vivo imaging and histology.

Perfluorocarbon (PFC) nanoemulsion tracer probes are a versatile platform for labeling immune cells, including inflammatory macrophages. Nanoemulsions can be used to label cells in culture ex vivo or via ‘in situ’ labeling, where the agent is administered intravenously and nanoemulsion droplets are endocytosed by phagocytic immune cells like macrophages at sites of inflammation [5,6,7]. Nanoemulsions are formulated via emulsification of PFC into an aqueous colloidal suspension, commonly employing phospholipids or synthetic polymers as a surfactant for stabilization [8], to yield small, uniform droplets (typically 100–200 nm in diameter). A variety of moieties may be chemically incorporated into liquid-core PFC probes, enabling sensitive in vivo molecular imaging by 19F MRI [9], PET [10], and fluorescence imaging [11, 12] via chemical attachment to the surfactant, or through conjugation to fluorous substituents that partition into the PFC oil [12, 13]. These 'dual-mode' fluorescent PFC nanoemulsions have proven to be excellent cell labels for optical detection due to their bright fluorescence, low toxicity, and long retention time in viable cells [12]. PFCs are lipophobic, hydrophobic, do not incorporate into cell membranes [14], and are not degraded by enzymes or in low pH environments. They have been rigorously tested for biological safety with no observed adverse effects on cell viability or function [15].

In this preliminary study, we investigate the combined use of PFC nanoemulsions with CFT to evaluate whole-body visualization of probe biodistribution. We perform high-resolution CFT in murine models of acute inflammation, inflammatory bowel disease (IBD), and solid tumor-associated macrophages (TAMs), where prior in vivo 19F MRI and 89Zr PET studies have been performed using functionalized PFC nanoemulsions [10]. We also explore the apparent nanoemulsion detectability across tissues using fluorescent dyes with different emission/absorption wavelengths, and in the context of intrinsic tissue autofluorescence.

Methods

PFC Nanoemulsions

Nanoemulsions were commercially obtained (Celsense, Pittsburgh, PA) and are comprised of emulsified perfluoropolyether (PFPE) oil with ~ 180 nm mean droplet size, suitable for direct intravenous injection. Each manufacturer-supplied nanoemulsion contained fluorophores of different wavelengths: Texas Red (ex/em 589/615 nm; V-sense 1000 DM-Red), green (fluorescein, FITC, ex/em 419/512 nm; V-Sense 1000 DM-Green), or near-infrared (NIR, Cy7, ex/em 756/779 nm; V-sense 1000 DM-NIR). Fluorescent nanoemulsions were prepared for injection as indicated below, using either 200 μl of a single color emulsion or equal proportions by volume of multiple emulsions (100 μl per color) to generate di- or tri-colored cocktails (200 or 300 μl total volume, respectively).

Control Mice

All animal experiments were performed in accordance with the guidelines provided by the UC San Diego Institutional Animal Care and Use Committee (IACUC) and the National Institute of Health Guide for the Care and Use of Laboratory Animals. Wild-type C57BL/6 mice (N = 3), 8–10 weeks old, were obtained from Jackson Laboratories (Bar Harbor, ME). All mice were maintained on low-fluorescence chow (5LJ5, LabDiet, St. Louis, MO) for a minimum of one week prior to initiation of imaging studies. A 300 μl injection of tri-color nanoemulsion (red/green/NIR) was administered via tail vein and mice were euthanized for CFT at 24 h post-injection. As an additional control cohort, uninjected control mice were prepared and embedded in OCT alongside the mice injected with nanoemulsion, using the methods described in Sect. 2.3. Quantification of intestinal fluorescence/auto-fluorescence in the injected and uninjected control mice, respectively, was performed using FIJI software.

Acute Inflammation Model

We employed a model for acute inflammation using carageenan injected into the hind paw, which has previously been used for in vivo 19F MRI, 89Zr PET, and fluorescence imaging using nanoemulsion [10, 16]. Female CD-1 mice (N = 3), 8–10 weeks old, were obtained from Envigo (Indianapolis, IN). A 1% λ-carrageenan (Sigma-Aldrich, St. Louis, MO) solution was prepared in saline for subplantar injection (50 µl) into the right hind paw [17]. Paw width and thickness were measured using calipers to confirm swelling due to inflammation. Intravenous di-color nanoemulsion was injected at 24 h post-induction, followed by euthanasia for CFT at 48 h.

Inflammatory Bowel Disease Model

C57BL/6 mice (N = 3), 8–10 weeks old, were obtained from Jackson Laboratories. Drinking water was supplemented with 3% dextran sulfate sodium salt (DSS, Spectrum Chemical, New Brunswick, NJ) on Day 0 and offered ad libitum for 10 days to induce acute IBD [18]. Animals were weighed daily and observed for stool consistency and hemoccult. Mice in this model were injected with a tri-color nanoemulsion cocktail and euthanized 24 h thereafter for CFT.

Solid Tumor Model

Two different solid tumor models were studied. C57BL/6 mice bearing colon adenocarcinoma MC38 tumors were administered single color nanoemulsion (NIR, 200 μl) when tumors reached ~ 1 cm diameter and sacrificed 24 h thereafter. In a second model, murine breast cancer 4T1 cells were implanted into the fourth mammary fat pad of 6–8 week old, female Balb/c mice (N = 3, Jackson Laboratory). Tumor volumes were monitored using calipers. A 200 μl di-color nanoemulsion cocktail was injected at 35 days post implantation, once tumors reached 1,500–2,000 mm3. Mice were sacrificed for CFT 24 h thereafter.

Sample Preparation

Animals were euthanized by CO2 gas and frozen by submersion in a bath of hexanes in dry ice. A block mold was assembled with squid-ink spaghetti fiducials placed perpendicular to cutting plane for image alignments. The mold was chilled on dry ice, and mice were then embedded for coronal sectioning in OCT compound poured into the chilled mold, followed by storage overnight at -20 °C. Some mice were frozen in an orientation suitable for axial sectioning (Controls and IBD model) in a similar manner as above. The total sample prep time from euthanasia to frozen block is ~ 12 h, due in large part to the overnight freezing process. Up to ~ 6 whole mice can be embedded in a single block.

CFT Data Acquisition

For CFT, the frozen block was transferred to the cutting stage inside the cryo-chamber (-15 °C) of the CFT cryostat (Xerra™, Emit Imaging, Baltimore, MD) and the temperature was equilibrated for 2 h. Initial pre-cuts were used to surface the block, and a camera autofocus feature was applied to laboratory pen markings placed on block face. Gain and exposure times were determined automatically and recorded by the instrument on a per-slice basis. The CFT employed a 12 mega-pixel camera with maximum of 4,096 × 3,008 pixel images, capable of resolving isotropic voxels as small as 20 μm3 for small fields of view. Whole mice were robotically sectioned at 50 μm increments, imaging the block face between cuts. Brightfield images were acquired sequentially, followed by laser illumination at 470 nm (green), 555 nm (red), or 730 nm (NIR), with discrete emission filters inserted robotically into light path (511/20 nm for green, 620/14 nm for red, and 794/32 nm for NIR) prior to camera capture. The process of sectioning and imaging is entirely automated. For data acquired using a single fluorescence channel, with 44 μm2 x–y resolution and a 50 μm slice thickness, a total of ~ 8 h are required to section and digitize the block.

Image Data Processing

CFT data were initially processed using vendor supplied software (Emit Imaging, Recon Version 1) running on a Dell Precision 7820 with dual 2.3 GHz Intel Xeon Gold 5218 processors and 128 GB random access memory. Within this software, image slices are automatically aligned via reference to the spaghetti fiducials. Regions of interest (ROIs) were selected (e.g., whole-mouse) for image reconstruction using the anatomical information provided by brightfield images as a guide. Subsurface correction and intensity normalization were performed relative to a reference ROI located in a region of OCT devoid of tissue using software-supplied algorithms. Three-dimensional images of each subject were reconstructed and exported as MHD/JPEG image stacks (brightfield) or MHD/RAW and TIFF image stacks (fluorescent images). Images were rendered using VivoQuant (Invicro, Needham Heights, MA.). For maximum image projects (MIPs), a 3 × 3 pixel median filter was applied. Where required, further alignment was performed using FIJI [19].

Results

CFT paired with PFC nanoemulsions are highly effective for visualization of RES compartments and lesion inflammatory infiltrates. Wavelength-dependent autofluorescence and light absorption/scattering may confound probe localization but can also enhance anatomical interpretations.

Wild-Type Control Mice

In wild-type subjects (Fig. 1), PFC nanoemulsions incorporating NIR, red, and green dyes accumulate in RES tissues including the liver, spleen, bone marrow and lymph nodes, as previously observed in 19F MRI and PET studies [6, 10, 11, 20]. Visible differences are observed in the fluorescence signals for each channel, attributed to a combination of structural differences between tissues and attenuation of particular wavelengths of incident and emitted light due to tissue-specific chromophores [21, 22]. Although fluorescence signal from the emulsion is clearly visible with NIR (Fig. 1A) and weakly visible in the red channel (Fig. 1B), signal in the green channel (Fig. 1C) is not observed in blood-rich tissues, specifically the liver and bone marrow, presumably due to interaction of these wavelengths with hemoglobin [23]. Overall, the NIR channel is most representative of the true nanoemulsion biodistribution, resembling signals acquired by 19F MRI [11]. Accumulated populations of labeled phagocytes are visible within the periphery of the lymph notes for all wavelengths studied, potentially attributable to subcapsular macrophages or follicular dendritic cells.

Fig. 1
figure 1

CFT of wild-type mice showing fluorescent PFC nanoemulsion biodistribution and intrinsic tissue autofluorescence. Shown are maximum intensity projections (MIPS) of NIR (A,D; 730 nm laser, 794/32 filter), red (B,E; 555 nm laser, 620/14 filter), and green (C,F; 470 nm laser, 511/20 filter) channels. In A-C, mouse (C57BL/6) was injected with a triple cocktail of NIR, red, and green PFC nanoemulsion and imaged 24 h thereafter. Panels D-F display MIPS in un-injected mouse exhibiting intrinsic autofluorescence in NIR (D), red (E), and green (F) channels. Reference anatomical structures are labeled. Supplemental data provides 3D MIP animations of Fig. 1 datasets. Scale bars represent 1 cm

Full size image

In control mice without nanoemulsion injection (Figs. 1D-F), autofluorescence is clearly observed in the gall bladder, stomach, and intestines, and can provide vivid anatomical context, particularly in the red and green channels (Fig. 1E,F). Prominent light scattering from fur is also observed in the green channel (Fig. 1F). 3D maximum intensity projection (MIP) animations of Fig. 1 datasets are provided in Supplemental Data.

Axial sections from both un-injected control mice (Figs. 2A-C) and mice injected with PFC nanoemulsion (Figs. 2D-F) imaged in the same block appear similar. Thus, autofluorescence originates primarily from the intestinal contents at all three wavelengths, especially in the lumen of the intestine. In this region green autofluorescence is particularly prominent (Fig. 2C,F), presumably due to collagen and elastin.

Fig. 2
figure 2

Axial CFT sections through the intestine of C57BL/6 mouse with no PFC nanoemulsion injected (A-C) and with nanoemulsions injected (D-F). Panels show autofluorescence in lumen (A; 730 nm laser, 794/32 filter, B; 555 nm laser, 620/14 filter, and C; 470 nm laser, 511/20 filter), and injected and uninjected animals appear similar. Scale bars represent 1 cm

Full size image

Acute Inflammation Model

In the carrageenan model, inflammation is readily visible in the red channel in the injected right paw (inset, Fig. 3D) and in the upper hindlimbs, as well as within lymph nodes throughout the body (Fig. 3A,C). Autofluorescence is visible in the intestinal tract, stomach, and gall bladder. In this model, green signal is only observed throughout the coat, intestines, stomach, and gall bladder, and is not detectable in inflammatory lesions (Fig. 3B,C), possibly due to fluorescence quenching of the green fluorophore.

Fig. 3
figure 3

CFT in acute inflammation model. Carageenan was delivered to right hind paw in CD1 mouse. In (A), inflammation is apparent in right footpad extending into the upper hindlimb in the red channel. Panel (B) displays significant autofluorescence and scattering from the coat in the green channel, (C) displays an overlay of the red/green channels, and (D) displays a zoomed-in view of inflammation in the right hind paw, boxed in panel (C). Scale bars represent 1 cm

Full size image

Inflammatory Bowel Disease Model

The IBD model displays clear NIR signal (Fig. 4A) in the descending colon, due to bowel inflammation, and liver, both of which are consistent with prior studies employing in vivo 19F MRI with immunohistochemistry validation [24]. Due to autofluorescence in the intestinal contents, the signal from the red/green nanoemulsion cannot be easily discerned from the background (Fig. 4B,C). Notably, the signal is more clearly localized in the lumen (Fig. 4D-I) in axial and coronal slices. MIPs of autofluorescence are informative, rendering the anatomy clearly visible throughout the gastro-intestinal tract and providing a semi-transparent reference on which to overlay the 3D rendering of the inflammatory lesion. Interestingly, red/green nanoemulsion fluorescence quenching in the liver is readily observed compared to the NIR images (Fig. 4A-C, G-I).

Fig. 4
figure 4

CFT in IBD mouse model. Shown are MIPS of a C57BL/6 mouse with DSS-induced intestinal inflammation and injected with tricolor PFC nanoemulsion cocktail (A; 730 nm laser, 794/32 filter, B; 555 nm laser, 620/14 filter, and C; 470 nm laser, 511/20 filter). Panels (D,E,F) and (G,H,I) display axial and coronal sections, respectively, for the same mouse with white light underlay and displays the wall thickening (arrows) in the descending colon, and increased signal compared to the wild-type mouse (Fig. 1). Scale bars represent 1 cm

Full size image

Solid Tumor Models

Mice injected with NIR nanoemulsion show strong signal in the MC38 tumor (Fig. 5A), with minimal signal from intestinal contents. Strong signal is also observed in the liver, spleen, lymph nodes and bone marrow. These observations are consistent with non-invasive 19F MRI and PET TAM imaging studies in different solid tumor models [5, 10, 11, 25].

Fig. 5
figure 5

CFT in solid tumor models. Panels (A) display MIPs of MC38 tumor in flank (asterisk) in a mouse injected with NIR PFC nanoemulsion, where the bottom panel shows segmented tumor. In (B)-(D), MIPs from a mouse with 4T1 flank tumor (asterisk) injected with di-color (red/green) are shown (B; 555 nm laser, 620/14 filter, C; 470 nm laser, 511/20 filter, D; overlay), where the lower panels are the segmented tumor. Scale bars represent 1 cm for whole mouse images in the top panels and 0.1 cm for segmented tumor images in the bottom panels

Full size image

The red/green nanoemulsion proved less effective at visualizing tumor phagocytes in the 4T1 example (Fig. 5B-D), with autofluorescence from the intestinal contents dominating the red and green channels. When the tumor itself is segmented (bottom panels, Fig. 5B-D), however, high resolution subtumoral signals localized in the tumor margin, invasive front, and tumor center are observed in the red/green channels.

Discussion

In this paper we explore the use of fluorescent PFC nanoemulsions for producing high-resolution three dimensional images of probe biodistribution by CFT. Studies were conducted in murine models of acute inflammation, IBD, and solid tumor, all of which demonstrate significant macrophage involvement. Nanometer-sized probes such as PFC nanoemulsions exploit highly evolved cell phagocytic functions for efficient intracellular probe labeling in situ. Upon PFC cell labeling, the fluorous droplets coalesce into lysosomal vesicles and macropinosomes [9], yielding durable labeling without exocytosis for the life of the cell. The biocompatibility and pharmacokinetics of PFC nanoemulsions are well characterized over many decades of research, and the biological inertness and high gas solubility of liquid-core PFC emulsions have made them candidates for oxygen-carrying blood substitutes [15, 26]. Perfluorocarbons are simultaneously lipophobic and hydrophobic, do not incorporate into cell membranes [14], do not degrade in vivo due to enzymes or low pH [14], and clear the body without chemical modification via lung exhalation [27] with a half-life of > 250 days for the nanoemulsion PFC molecule (perfluoropolyether) used in this study.

In situ macrophage labeling with PFC nanoemulsions has been used in a multitude of in vivo preclinical imaging studies. For example, 19F MRI studies have been conducted in myocarditis [28], solid organ transplant rejection [29], inflammatory bowel disease (IBD) [24, 30], ischemic stroke [31], acute infection [32], peripheral nerve inflammation [33], multiple sclerosis [6, 20], and TAMs [5, 25, 34] using the same nanoemulsion formulation employed in this study. Importantly, these and other studies provide a histological characterization of the cell phenotypes that take up nanoemulsions, with Ly6C + monocytes/macrophages comprising the dominant phenotype in lesions yielding detectable signal. In situ uptake is ‘pan-macrophage’, analogous to CD68 + immuno-staining, with no known uptake specificity among the spectrum of “reparative” and “pro-inflammatory” macrophage subtypes [35]. Phagocytic neutrophils are also known to take up nanoemulsion, but are physically smaller than macrophages with a ~ fourfold lower volume, and thus have less intrinsic ability to take up tracer mass compared to macrophages/monocytes. As a result, their contribution to the observed signal on a per cell basis is significantly lower compared to macrophages. Similarly, lymphocytes will take up trace amounts of agent in situ, but provide no significant signal contributions.

Our study suggests that CFT can provide a powerful histological tool for validating putative in vivo ‘hot-spot’ signals from PFC nanoemulsions using 19F MRI and PET [10] detection. The CFT method is highly complementary to these in vivo imaging techniques, providing near-cellular, sub-tissue localization. For example, in previous studies employing PFC nanoemulsions for visualization of inflammation in a dextran sulfate sodium-induced IBD model using PET and MRI [10], it was difficult to discern whether the tracer was localized in the intestinal lumen or the intestinal contents due to the lack of 3D resolution. The former can be easily confirmed using CFT (Fig. 4).

For applications such as CFT, nanoemulsions are complexed with fluorophores to create ‘dual-mode’ PFC-fluorescent agents for 19F MRI and optical detection. Direct conjugation of dye to fluorous molecules prior to nanoemulsion formulation ensures that the fluorescent signal’s stability remains associated with the nanoemulsion inside PFC labeled cells [12, 36]. Alternatively, a lipophilic fluorescence dye (e.g., C59H97ClN2O4 or ‘DiI’) can be incorporated into nanoemulsions formulated with lipid surfactant [6, 9]. Moreover, convenient ‘click’ chemistry attachment of azide-functionalized fluorophores to the nanoemulsion surfactant has recently been demonstrated [16]. The flexibility of click tracers enables rapid adaptation to suit a wide variety of models and experimental use cases.

Limitations to the interpretation of data generated using fluorescent probes include the presence of tissue autofluorescence, fluorescent light absorbance, and light scattering, each which can give rise to false-positive and false-negative signals [37]. Prominent autofluorescence from endogenous fluorophores is observed in gut and skin [38]. This effect is wavelength dependent, with endogenous fluorophores like chlorophyll, breakdown products, flavins, lipofuscin, collagen and elastin in the intestinal lumen fluorescing in the green channel [39], as clearly observed in Fig. 2C. In the case of gut, chlorophyll autofluorescence can be ameliorated to some extent via specialized diets to reduce fluorescence over the spectral ranges of interest. Gallbladder autofluorescence is also commonly observed, as in Fig. 1, due to presence of bilirubin and biliverdin [40]. Endogenous tissue fluorescence alone is not necessarily detrimental, however, and can have diagnostic value for the identification of disease pathologies, particularly in the liver [38].

Strategies to discern probe fluorescence signals include use of wavelengths that do not interact significantly with biological tissues and comparsion of datasets collected with fluorophres spanning multiple wavelengths [41]. Importantly, a strength of CFT lies in its capacity for multiplex detection of commercially available fluorescent probes thereby permitting flexibility in probe wavelength. To that end, our data confirm that intrinsic autofluorescence may be reduced by utilization of NIR fluorophores within the nanoemulsion, which avoid the spectral range of naturally abundant fluorophores in biological tissues.

Moreover, attenuation of fluorescence signal through scatter of incident or emitted light is a major problem for fluorescence imaging techniques. Absorptive and refractive properties of endogenous biomolecules such as hemoglobin, collagens, and flavins may result in attenuation of incident radiation and emitted fluorescence signals. This is observed in liver and bone marrow, as evident from comparisons of the NIR, red, and green images (Fig. 1). Light scattering off skin surface and fur may also contribute to background fluorescence signals, as observed in the green channel (Fig. 5B).

In current practice, CFT does not have the ability to support whole mouse immunostaining for tracer co-localization studies, nor does the technique have sufficient resolution to resolve individual cells, preventing visual co-localization. Although we have experimented with stopping the cryotome mid-sample and transferring a single slice onto histology tape for subsequent histology processing, recovering intact slices in this manner is challenging with current instrumentation.

Currently, this CFT approach is intended for qualitative fluorescence, analogous to widely used in vivo fluorescence imaging methods [42]. Quantitative fluorescence imaging of tissues in situ is challenged by non-linear signal effects due to dye aggregation and self-quenching inside intracellular vesicles which may reduce bulk tissue fluorescence, as well as the wavelength-dependent impact of the particular tissue’s absorbance/scattering properties. Quantitative fluorescence measurements (e.g., dye or nanoemulsion mass) in a tissue region of interest based on CFT data would require a future validation study, perhaps aided by external fluorescence quantification standards. Moreover, algorithmic methods for quantitative comparisons among individual mice, or among cohorts with varying treatments, do not yet exist and would be computationally demanding due to the size of CFT data sets (~ 100 GB/mouse per fluorescence channel). Moreover, by combing CFT data with more quantative whole-body imaging methods in the same subjects, such as PET [10] or 19F MRI methods that use PFC nanoemulsion tracers (abiet at lower resolution), one could calibrate and benchmark absolute dye mass in fluorescent CFT imagings. A recent study displays the quantitative biodistribution of PFC nanoemulsion in C57BL/6 J mouse using quantative 19F magnetic resonance spectroscopy [43]; this study reports absolute F-levels and should represent ground-truth values.

Conclusions

Fluorescent PFC nanoemulsions represent a powerful multimodal probe platform for preclinical imaging of inflammation which, in conjunction with CFT, presents a valuable, high-resolution complement to in vivo imaging methodologies. In the course of developing new generations of in vivo imaging probes, it is essential to validate image specificity, interpretation, and whole-body biodistribution. Current methods, such as conventional histology, are laborious, prone to sampling error, and render it difficult to observe localized concentrations of cells in the context of whole organs and systems. CFT therefore provides a valuable intermediate-resolution technique to bridge the gap between existing in vivo imaging techniques and histological micrographs.