Intravital microscopy (IVM) is a powerful optical imaging technique for collecting dynamic data in vivo. Since its establishment, IVM has been used extensively for the study of cell biology 1, immunology 2, neurobiology 3, and tumor pathophysiology 4. Compared to radiologic imaging modalities, such as positron emission tomography (PET) and magnetic resonance imaging (MRI), the enhanced spatial and temporal resolution of IVM facilitates various mechanistic measurements not possible with the current whole-body imaging techniques. IVM yields structural and functional information with subcellular resolution, sufficient for identifying intracellular organelles 5 and monitoring their trafficking 6,7. With recent advances in non-linear optics and high-speed laser beam scanning mechanisms, it is now possible to quantify in vivo dynamics with high spatial and temporal resolution. Dynamic biological processes for which quantitative IVM procedures have been implemented include the assessment of vascular morphology 8, permeability 9, flow 10, and response to therapy 11. IVM is also well suited for imaging the dynamics of nanoparticles in vivo, and as result, IVM is increasingly leveraged for the design and optimization of nanotherapeutics 12-29.
Direct visualization of nanoparticle transport in live animals can provide invaluable insights into complex events that cannot be obtained by conventional ex vivo analyses. IVM studies of nanotherapeutics delivery in cancer, for example, have revealed spatiotemporal heterogeneities in tissue permeability and convective extravasation, which in turn impact therapeutic efficacy <sup>12</sup>. Real-time monitoring of nanotherapeutics delivery and clearance has been used to investigate tumor-specific targeting <sup>13-15</sup>, quantify mechanisms of particle accumulation <sup>13-21</sup>, optimize particle loading <sup>22</sup>, predict particle delivery <sup>23</sup>, and study cell-particle interactions <sup>13,14,24-26</sup>. When performed longitudinally, IVM studies allow monitoring of neovascularization <sup>27</sup>, disease progression <sup>28</sup>, and treatment outcome <sup>29</sup>. The success of such studies is based on acquisition of dynamic data with subcellular resolution, an approach that can be reliably achieved using optical technologies. Real-time imaging of single nanoparticles provides additional information, notably transport kinetics across multiple cell and tissue compartments <sup>13-16,23-25</sup>.
Here we provide a simple and reproducible approach for imaging the real-time kinetics of individual particles ranging from several tens of nanometers to micrometers in size. Techniques for image acquisition and analysis are described for two tissues: Liver, which is the major organ responsible for uptake and sequestration of circulating nanoparticles, and subcutaneous tumors, which are a common model for the study of nanotherapeutics.
Visualization of nanoparticle transport and uptake in the liver
The mononuclear phagocyte system is a major barrier to the delivery of systemically administered drugs and nanoparticles <sup>30</sup>. The therapeutic efficacy of tumor-specific nanoparticles is largely determined by their ability to delay uptake by the mononuclear phagocyte system. Many nanoparticle-based chemotherapeutics, for example, accumulate in tumors with low efficiency. Instead, a large proportion of circulating nanoparticles are non-specifically captured by macrophages and specialized endothelial cells residing in the liver, spleen, and bone marrow <sup>31</sup>. Early efforts to evade the mononuclear phagocyte system, such as the addition of non-ionic surfactants and polymer macromolecules to the surface of drug carriers, have been shown to extend blood-circulation time while reducing macrophage uptake <sup>31-34</sup>. Such innovations have allowed nanoparticle-based chemotherapeutics to show enhanced efficacy over freely administered drugs. Improving our understanding of how nanoparticles interact with the mononuclear phagocyte system is key to developing more effective delivery strategies, particularly for cancer and other diseases characterized by pathological disruptions of mass transport <sup>35</sup>.
The liver is an important site for studying dynamic changes in nanoparticle transport and cell-particle interactions <sup>13-15,24</sup>. Resident macrophages of the liver, called Kupffer cells, are the predominant cells responsible for phagocytosis of many classes of circulating nanoparticles. These cells can be readily visualized within the liver sinusoids via a small mid-line abdominal incision. In healthy mice, IVM studies of individual nanoparticle dynamics can be used to determine how physical particle properties, such as particle size, influence their capture and uptake by Kupffer cells. As shown in <a href="http://www.nature.com/protocolexchange/system/uploads/2573/original/fig1.png?1367939601" target="_blank">Figure 1</a>, systemically administered nanoparticles can be individually identified and tracked within the liver vasculature. Adhesion of these particles to the vasculature, and subsequent internalization by Kupffer cells can be monitored in real-time between mouse breaths using a sufficiently fast IVM system \(30 fps) with optical sectioning capabilities \(see also <a href=" http://www.nature.com/protocolexchange/system/uploads/2587/original/video1.m4v?1367599329" target="_blank">Videos 1</a>, <a href="http://www.nature.com/protocolexchange/system/uploads/2588/original/video2.m4v?1368118088" target="_blank">2</a>, and <a href="http://www.nature.com/protocolexchange/system/uploads/2589/original/video3.m4v?1368118249" target="_blank">3</a>). Provided that the images are acquired with sufficient lateral and axial resolution, particle internalization can be observed in multi-color overlay images. This is particularly evident for particles that rapidly internalize and traffic to the perinuclear region, such as the silicon particles shown here. Frame-by-frame processing of the IVM movies yields quantitative information that may be used to tailor nanoparticles for specific applications. Coating silicon particles with biomimetic leukocyte-derived membranes, for example, has been shown to transiently delay Kupffer cell-mediated capture and reduce the overall quantity of particles internalized by the liver, resulting in enhanced tumoritropic accumulation <sup>13</sup>. Optimization of silicon particle size, shape, and surface chemistry decreases the rate of particle capture <sup>14</sup>, as well as the quantity of particles captured by Kupffer cells <sup>14,24</sup>. Conversely, addition of hyaluronic acid has been shown to enhance liver-specific targeting for treatment of liver disease <sup>15</sup>.
Figure 1. Imaging single particle dynamics in the liver of live mice. (A) The left lobe of the liver is readily exposed via a small midline incision. A coverslip may be gently placed against the tissue surface to allow high-magnification and high-resolution microscopy without impacting liver movement or blood flow. (B) 600×200 nm plateloid silicon particles (red), injected retro-orbitally, can be individually identified and tracked within the liver microvasculature (green) using a high-speed intravital microscope with equipped with optical sectioning capabilities. Particle accumulation within the vasculature, as well as particle uptake by Kupffer cells (blue) are visualized from the time of injection. When the color images are overlaid, particles appear to shift in color from red to violet upon internalization by Kupffer cells (inset). Images shown here were averaged across 4 frames for clarity. The associated movies may be found in the supplementary information. Scale bar, 50 µm. (C) Particle accumulation is quantified at regular time-intervals (shown here as every 10 seconds).
Endogenous and exogenous sources of contrast may be used to provide structural and functional references for studying cell-particle interactions \(<a href=" http://www.nature.com/protocolexchange/system/uploads/2574/original/fig2.png?1367939624" target="_blank">Fig 2</a>). Transgenic mouse strains that express fluorescent reporters under the control of endothelial promoters, such as TIE2-GFP mice <sup>36</sup> and their athymic equivalent <sup>37</sup>, are useful for delineating the liver vasculature. A variety of transgenic mice with fluorescent Kupffer cells have also been developed <sup>38</sup>. Auto-fluorescence of the liver <sup>39</sup> and blood <sup>40</sup> can be used to provide endogenous contrast, and are particularly useful for identifying fields-of-view \(FOVs) to be compared across animal models of health and disease <sup>41</sup>. A variety of exogenous contrast sources, including fluorescent albumin and dextran tracers of different sizes, molecular-specific nanoparticles, nucleic acid dyes, and antibodies may be systemically administered to delineate structures of interest. Autologous or syngenic red blood cells \(RBCs), labeled with lipophilic dyes in vitro, facilitate the imaging of blood flow dynamics and indirect labeling of the Kupffer cells \(described in further detail below).
Figure 2. Endogenous and exogenous sources of contrast. The vasculature (top) and Kupffer cells (bottom) of the liver are readily delineated using fluorescent reporter genes (Tie2-GFP, MHCII-GFP) or systemically injected contrast agents including fluorescent antibodies (CD 204 mAb), macromolecule tracers (FITC-dextran), and labeled RBCs. The RBC track projection shows a summation of 900 sequential frames (30s). Scale bar, 50 µm.
Visualization of nanotherapeutics delivery to tumors
Complementary approaches may be utilized to study nanoparticle delivery to tumors. The challenge in imaging tumors lies largely in sample preparation and FOV selection. Subcutaneous tumors exposed by skin-flap, for example, are generally nodular and have high interstitial pressure. This makes tumor positioning and coverslip placement more technically difficult. RBC pre-labeling is highly useful, since the motion of these cells may be used to confirm that the vessels are not compressed and to select appropriate FOVs for automated acquisition prior to nanoparticle injection. The heterogeneous structure of tumors, combined with the relative inefficiency nanoparticle tumor targeting, makes it important to sample a large area of tissue. Ideally, multiple high-magnification FOVs are sampled in parallel over time, requiring a careful balance between image magnification, nanoparticle density, and FOV number. Properly mounted skin flap tumors generally show less motion than the liver, and thus high-speed microscopes can be leveraged to rapidly cycle between multiple positions in an automated manner. This allows users, for example, to quantify particle accumulation in tumors <sup>13,14,23</sup>, test the influence of different particle surface coatings <sup>13,14</sup>, and compare targeting efficiency across different tumor models <sup>23</sup>. When performed using IVM, such studies allow spatial and temporal differences in particle accumulation to be related to specific local structural and functional features such as inflammation <sup>13</sup>, surface biomarker expression <sup>14</sup>, and vessel perfusion <sup>23</sup>. This quantitative information can then be integrated with multi-scale tumor growth models to predict drug delivery and treatment response <sup>22,23</sup>.
Overview. Here we present a technique to orthotopically image and quantify the dynamics of systemically administered nanoparticles in the liver and subcutaneous tumors of live mice following acute surgery. The techniques described here are also compatible with surgically placed window chambers 8,42,43 for longitudinal studies. The Procedure section consists of a step-by-step protocol for the study of nanoparticle transport and cell-particle interactions in the liver, followed by an Application Notes section that describes how this protocol may be adapted for subcutaneous tumor imaging. The methods presented can serve as guideline for the analysis of a variety of nanoparticles and tissues.
Experiments are performed as follows: The tissue of interest is surgically exposed via a small mid-line incision, gently stabilized with a coverslip, and then scanned at video-rate \(30 fps) using a laser scanning confocal microscope. Step-by-step visual guides for <a href="http://www.nature.com/protocolexchange/system/uploads/2582/original/suppl_fig1.png?1367939988" target="_blank">animal preparation</a>, <a href="http://www.nature.com/protocolexchange/system/uploads/2585/original/suppl_fig3.png?1367940071" target="_blank">coverslip assembly</a>, and <a href="http://www.nature.com/protocolexchange/system/uploads/2586/original/suppl_fig4.png?1367940104" target="_blank">nanoparticle preparation</a> may be found in Supplementary Figures 1-4. To locally monitor the blood flow, mice are injected with fluorescently labeled autologous RBCs as shown in <a href="http://www.nature.com/protocolexchange/system/uploads/2575/original/fig3.png?1367939650" target="_blank">Figure 3</a> and <a href="http://www.nature.com/protocolexchange/system/uploads/2590/original/video4.m4v?1368118176" target="_blank">Video 4</a>. This approach has the added benefit of staining Kupffer cells when labeled RBCs are administered at least 24 hours prior to nanoparticle injection. Multi-color confocal images are acquired at 7 fps \(tumor) to 30 fps \(liver) from the time of injection in order to record nanoparticle accumulation and uptake in real-time \(<a href="http://www.nature.com/protocolexchange/system/uploads/2573/original/fig1.png?1367939601" target="_blank">Figure 1c</a>). Schematics of the surgeries, as well as workflow diagrams outlining recommended image acquisition parameters are provided in the Procedures section. Step-by-step diagrams for quantifying dynamic particle behaviors are also provided.
Figure 3. Fluorescent labeling of RBCs and Kupffer cells. (A) Step-by-step procedure for ex vivo labeling of autologous RBCs using lipophilic DiD. (B) A small proportion of RBCs (blue) are taken up by Kupffer cells immediately after re-injection, providing an indirect way to identify Kupffer cells in vivo. RBC internalization is time-dependent, with the Kupffer cell morphology being fully delineated ~ 24 hours after RBC injection.¬
For the examples presented here, we utilized plateloid porous silicon particles of 600×200 and 1000×400 nm dimensions <sup>14</sup> \(shown in <a href=" http://www.nature.com/protocolexchange/system/uploads/2587/original/video1.m4v?1367599329" target="_blank">Video 1</a> and <a href="http://www.nature.com/protocolexchange/system/uploads/2591/original/video5.m4v?1368118223" target="_blank">Video 5</a> respectively). Larger than traditional nanoparticles, they nevertheless possess unique physical features that can be modulated on the nanoscale <sup>44</sup> to overcome sequential biophysical transport barriers in order to reach and locally release therapeutics in close proximity to tumor cells. This multi-stage delivery concept was first proposed by our laboratory <sup>30</sup> and demonstrated in vivo by Tanaka et al. <sup>45</sup>. The geometry of these particles is readily modulated to control their behavior within the bloodstream <sup>46,47</sup> and ability to accumulate in tumors <sup>47,48</sup>. Using the IVM techniques detailed here, we have studied the real-time transport and cellular association of particles of different size, shape, and surface properties across a variety of tissues and tumor models <sup>13,14,22,23</sup>. Importantly, observed particle dynamics were found to be highly reproducible from animal to animal, and were independently validated using a combination of ICP-AES, SEM, TEM, flow chamber experiments, transwell migration assays, in vitro time-lapse microscopy, immunocytochemistry, and immunohistochemistry.
Controls and calibration standards. Different nanoparticles and conjugated dyes are detected with varying efficiency. Use of a new nanoparticle generally requires at least one practice mouse to determine the optimum laser power, gain, and amplifier offset. While it is possible to calibrate such settings prior to imaging, in our experience, the best image quality is achieved by performing the optimization directly within the organ-of-interest in order to correct for tissue absorbance, tissue autofluorescence, and exogenous fluorophore cross-talk. Following this optimization step, the same imaging parameters may be reused from mouse to mouse, provided nanoparticle fluorescence is stable with storage and consistent between lots.
Controls for nanoparticle experiments generally consist of different nanoparticle formulations such as targeted versus untargeted particles. A major challenge in nanoparticle studies is that small changes in nanoparticle structure and chemistry can substantially alter nanoparticle biodistribution and transport kinetics, thus it is important to identify controls that are physiologically relevant. Differentially labeled experimental and control nanoparticles may be injected into the same mouse to reduce animals numbers, provided that there is no competition between different particles. It is also good experimental practice to swap dyes between the control and experimental particles to ensure that different dye labeling protocols do not influence particle transport and uptake. Ideally experimental and control studies are done on mouse littermates, with mouse age dictated by experimental needs. We have successfully imaged mice up to 14 months of age, fed on a high fat diet to promote atherosclerosis.
As a negative control, it is important to image at least one or two mice that have not been injected with nanoparticles to compare the nanoparticle signal against tissue autofluorescence and other exogenous contrast agent emissions. In instances where fluorophore cross-talk is observed, it is important for nanoparticles to be sufficiently bright in order to resolve them from the background. Provided that imaging settings are properly optimized prior to nanoparticle injection, mice imaged from time zero can act as their own internal controls by comparing before and after images.
For systems prone to pinhole or other hardware drift, it is recommended to calibrate the hardware before each set experiments by imaging a small drop of dilute fluorescent flow cytometry beads. It is also helpful to image mice from experimental and control groups on the same day, rather than splitting these groups across different days. This allows for the correction of day-to-day fluctuations in hardware and contrast agent preparation, if needed.
Compatible nanoparticles and other sources of contrast. A variety of nanoparticle formulations may be studied, provided they are sufficiently bright for single nanoparticle detection. We find it is not necessary to resolve individual nanoparticles at actual size; instead, sufficiently diluted nanoparticles can be distinguished based on specific optical properties (color, dot size and shape, intensity, etc.). We have successfully imaged individual liposomes (50 – 100 nm), silica beads (40 – 100 nm), polystyrene beads (100 – 1700 nm), and silicon particles of various shapes (200 – 3200 nm) in a variety of tissues. Highlights are shown in Figure 4. With the availability of a large number of fluorescent reporters that may be used concomitantly to study tissue structure and function, it is now possible to study a wide range of kinetic parameters that drive nanotherapeutics delivery, uptake, and treatment response.
Figure 4. Examples of single nanoparticles resolved by IVM. Nanoparticles successfully imaged in vivo range in size and shape. Large particles containing many fluorophores, such as 600×200 nm silicon particles are readily observed at low magnification (e.g. 12×) as single particles flowing within tumors. This approach works equally well for particles beyond the resolution of the IVM system, such as 100 nm AlexaFluor 555-conjugated silica beads and 50 nm rhodamine-liposomes. Provided the nanoparticles are sufficiently bright, they appear as one or more pixels within the FOV. Scale bar = 100, 50, 30 μm respectively.
Microscopy sessions. IVM experiments usually require some optimization of both mouse positioning and hardware setup; therefore, beginners are recommended to initially image no more than 1-2 mice in each experiment. As users gain sufficient expertise and speed, experiments can be scaled up. We have observed highly reproducible results when performing experimental replicates on separate days, thus we recommend obtaining statistically significant data through repeated smaller experiments on sequential days rather than larger, less consistent ones.
Users are encouraged to develop an efficient, reproducible imaging routine. Reproducibility in terms of tissue excitation, emission collection, image magnification, acquisition start time, scan speed, imaging duration, and FOV number is key to comparing quantitative data across animals. The workflow diagrams provided here \(Fig 5, 6) can be used as a starting point for designing reproducible, experiment-specific microscopy sessions. The individual parameters listed may be adapted for specific preparations.
Minimum equipment requirements. IVM data acquisition and analysis techniques are readily applied using a high-speed microscope with optical sectioning capabilities. Experiments equivalent to those described here have been successfully performed using a custom-built 30 fps confocal laser scanning microscope equipped with an x-axis polygon scanner and y-axis galvanometer 14,15,49,50 as an alternative to the resonance scanner. These studies are best suited for systems capable of multicolor image acquisition at video-rate (30 fps) or faster. The high frame rate is essential to measure the flow speeds of particles and their distribution within the blood vessels. For liver imaging, the acquisition frame rate is dictated by the mouse respiratory rate, which is generally 60 – 180 breaths per min (1 – 3 bps) under anesthesia. To reproducibly acquire frames in the same plane of focus during the constant motion of the liver, 30 fps is ideal. We have successfully acquired images as slow as 4 fps by combining trigger-acquisition hardware with a deeper plane of anesthesia, suggesting the possibility of using standard confocal galvanometers for liver image acquisition. For the tumor imaging workflow presented here, speed is less critical and generally does not require corrections for breathing. Spinning disk systems may be substituted for the resonance-based system, with a compromise in axial resolution and a temporal lag between sequential color channels. For such an approach to be effective, single color imaging at 30 fps is recommended, since multi-color imaging (achieved through the use of automated excitation/emission wheels at frame-rates of 60-90 fps) is likely to be insufficiently sensitive for detecting dim fluorescence. Conventional epi-fluorescence microscopes are not recommended, since their limited axial-resolution make it difficult to localize nanoparticles within specific structures or cells of interest.
Multi-photon systems, which offer enhanced sensitivity and depth penetration, may be utilized but are not required. A major advantage of the confocal approach is that multiple fluorophores are readily imaged simultaneously, making it possible to study cell, molecule, and particle interactions in real-time. High-speed multicolor two-photon microscopy may be achieved by leveraging tissue harmonics and selecting the right combination of contrast sources while mode-locking the two-photon laser at a single wavelength, as highlighted in <a href="http://www.nature.com/protocolexchange/system/uploads/2580/original/fig8.png?1367939844" target="_blank">Figure 5</a>. The availability of infrared-compatible cell lines, transgenic animals, and contrast agents makes this approach increasing feasible. To label RBCs, for example, lipophilic dye DilC<sub>18</sub>\(7) \(‘DiR’) may be directly substituted for DiD.
Figure 5. Multicolor two-photon microscopy of fluorescent beads in the mouse ear. (A) Individual FITC-loaded 200nm polystyrene beads (50 µl) are readily identified within the vasculature using a 20×, 1.0 NA water immersion objective lens. Scale bar = 100 μm. Three-color microscopy was achieved by counterstaining the vasculature with 70 kDa rhodamine-dextran (3% w/v) and collecting the 2nd harmonic signal from collagen following excitation with a Mai-Tai DeepSee laser mode-locked at 810 nm ( ~ 25mW). Fluorescence emissions were collected at 405 ± 10 nm (collagen), 525 ± 25 nm (FITC beads), and 600 ± 50 (rhodamine-dextran) using photomultiplier tubes. (B) Analysis of consecutive frames (recorded at 60 fps) allow the trajectories of individual beads be resolved and measured. Here 1000 nm beads are shown in a few select frames for clarity. When the particle displacement between frames greatly exceeds particle size (see 76 μm/s trace), it is recommended to inject of a small number of particles (less than or equal to 5×108) to avoid confusion between particles. Scale bar = 50 μm.
Other recommended hardware includes a motorized stage and multiple plan-apochromat objectives. While it is still relatively uncommon to find motorized stages on IVM systems, we find that this hardware is critical for sampling a sufficiently large area at high-magnification to allow statistical analysis of nanoparticle dynamics. In a single liver FOV captured with a 40× magnification objective, for example, we expect to see ==== 100 nanoparticles when a mouse is treated with 5×108 particles. Sampling over 10 FOVs increases this to ==== 1000 particles, allowing for more statistically stringent calculations. Increasing image magnification reduces the number of particles sampled per FOV, and thus image acquisition becomes a balance between image magnification, particle density, and FOV number. Depending on the given experiment, we generally use one of three plan-apochromat air objectives (4×/NA 0.2; 20×/NA 0.75; 40×/NA 0.95) with up to 3× digital zoom for sampling at Nyquist resolution. Both air and water immersion objectives may be used, with the immersion objectives providing enhanced resolution, depending on availability and personal preference. While higher magnification objectives may be used, we find it is not necessary to resolve sufficiently diluted individual nanoparticles.
Limitations. This protocol is suitable for visualizing and quantifying the kinetics of particles several tens of nanometers in size or larger. The ability to resolve individual particles is dictated by the lateral and axial resolution of the optical system used, thus smaller particles are difficult to track. Particles sized below the diffraction limit of light will appear larger than their actual size, thus it is important that nanoparticles be monodisperse and sufficiently diluted at the time of injection. Invasive surgery is generally required to expose the organ of interest for both confocal and multi-photon laser-scanning microscopy. The imaging penetration depth is limited to the top several layers of cells ( ==== 50-100 μm) for confocal systems and deeper ( ==== 200-500 μm) for multi-photon systems, thus it is good general practice to independently validate IVM findings using whole-organ and/or whole-animal quantification techniques.