Time-lapse fluorescence microscopy is one of the main tools used to image subcellular structures in living cells. Yet for decades it has been applied primarily to in vitro model systems. Thanks to the most recent advancements in intravital microscopy, this approach has finally been extended to live rodents. This represents a major breakthrough that will provide unprecedented new opportunities to study mammalian cell biology in vivo and has already provided new insight in the fields of neurobiology, immunology, and cancer biology.
The discovery of GFP combined with the ability to engineer its expression in living cells has revolutionized mammalian cell biology (Chalfie et al., 1994). Since its introduction, several light microscopy–based techniques have become invaluable tools to investigate intracellular events (Lippincott-Schwartz, 2011). Among them are: time-lapse confocal microscopy, which has been instrumental in studying the dynamics of cellular and subcellular processes (Hirschberg et al., 1998; Jakobs, 2006; Cardarelli and Gratton, 2010); FRAP, which has enabled determining various biophysical properties of proteins in living cells (Berkovich et al., 2011); and fluorescence resonance energy transfer (FRET), which has been used to probe for protein–protein interactions and the local activation of specific signaling pathways (Balla, 2009). The continuous search for improvements in temporal and spatial resolution has led to the development of more sophisticated technologies, such as spinning disk microscopy, which allows the resolution of fast cellular events that occur on the order of milliseconds (Nakano, 2002); total internal reflection microscopy (TIRF), which enables imaging events in close proximity (100 nm) to the plasma membrane (Cocucci et al., 2012); and super-resolution microscopy (SIM, PALM, and STORM), which captures images with resolution higher than the diffraction limit of light (Lippincott-Schwartz, 2011).
Most of these techniques have been primarily applied to in vitro model systems, such as cells grown on solid substrates or in 3D matrices, explanted embryos, and organ cultures. These systems, which are relatively easy to maintain and to manipulate either pharmacologically or genetically, have been instrumental in providing fundamental information about cellular events down to the molecular level. However, they often fail to reconstitute the complex architecture and physiology of multicellular tissues in vivo. Indeed, in a live organism, cells exhibit a 3D organization, interact with different cell types, and are constantly exposed to a multitude of signals originated from the vasculature, the central nervous system, and the extracellular environment. For this reason, scientists have been attracted by the possibility of imaging biological processes in live multicellular organisms (i.e., intravital microscopy [IVM]). The first attempt in this direction was in 1839, when Rudolph Wagner described the interaction of leukocytes with the walls of blood vessels in the webbed feet of a live frog by using bright-field transillumination (Wagner, 1839). Since then, this approach has been used for over a century to study vascular biology in thin areas of surgically exposed organs (Irwin and MacDonald, 1953; Zweifach, 1954) or by implanting optical windows in the skin or the ears (Clark and Clark, 1932). In addition, cell migration has also been investigated using transparent tissues, such as the fin of the teleost (Wood and Thorogood, 1984; Thorogood and Wood, 1987). The introduction of epifluorescence microscopy has enabled following in more detail the dynamics of individual cells in circulation (Nuttall, 1987), in tumors (MacDonald et al., 1992), or in the immune system (von Andrian, 1996), and the spatial resolution has been significantly improved by the use of confocal microscopy, which has made it possible to collect serial optical sections from a given specimen (Villringer et al., 1989; O’Rourke and Fraser, 1990; Jester et al., 1991). However, these techniques can resolve structures only within a few micrometers from the surface of optically opaque tissues (Masedunskas et al., 2012a). It was only in the early nineteen nineties, with the development of multiphoton microscopy, that deep tissue imaging has become possible (Denk et al., 1990; Zipfel et al., 2003b), significantly contributing to several fields, including neurobiology, immunology, and cancer biology (Fig. 1; Svoboda and Yasuda, 2006; Amornphimoltham et al., 2011; Beerling et al., 2011). In the last few years, the development of strategies to minimize the motion artifacts caused by the heartbeat and respiration has made it possible to successfully image subcellular structures with spatial and temporal resolutions comparable to those achieved in in vitro model systems, thus providing the opportunity to study cell biology in live mammalian tissues (Fig. 1; Weigert et al., 2010; Pittet and Weissleder, 2011).
The aim of this review is to highlight the power of IVM in addressing cell biological questions that cannot be otherwise answered in vitro, due to the intrinsic limitations of reductionist models, or by other more classical approaches. Furthermore, we discuss limitations and areas for improvement of this imaging technique, hoping to provide cell biologists with the basis to assess whether IVM is the appropriate choice to address their scientific questions.
Imaging techniques currently used to perform intravital microscopy
Confocal and two-photon microscopy are the most widely used techniques to perform IVM. Confocal microscopy, which is based on single photon excitation, is a well-established technique (Fig. 2 A) that has been extensively discussed elsewhere (Wilson, 2002); hence we will only briefly describe some of the main features of two-photon microscopy and other nonlinear optical techniques.
The first two-photon microscope (Denk et al., 1990) was based on the principle of two-photon excitation postulated by Maria Göppert-Mayer in her PhD thesis (Göppert-Mayer, 1931). In this process a fluorophore is excited by the simultaneous absorption of two photons with wavelengths in the near-infrared (IR) or IR spectrum (from 690 to 1,600 nm; Fig. 2 B). Two-photon excitation requires high-intensity light that is provided by lasers generating very short pulses (in the femtosecond range) and is focused on the excitation spot by high numerical aperture lenses (Zipfel et al., 2003b). There are three main advantages in using two-photon excitation for IVM. First, IR light has a deeper tissue penetration than UV or visible light (Theer and Denk, 2006). Indeed, two-photon microscopy can resolve structures up to a depth of 300–500 µm in most of the tissues (Fig. 2 B), and up to 1.5 mm in the brain (Theer et al., 2003; Masedunskas et al., 2012a), whereas confocal microscopy is limited to 80–100 µm (Fig. 2 A). Second, the excitation is restricted to a very small volume (1.5 fl; Fig. 2 B). This implies that in two-photon microscopy there is no need to eliminate off-focus signals, and that under the appropriate conditions photobleaching and phototoxicity are negligible (Zipfel et al., 2003b). However, confocal microscopy induces out-of-focus photodamage, and thus is less suited for long-term imaging. Third, selected endogenous molecules can be excited, thus providing the contrast to visualize specific biological structures without the need for exogenous labeling (Zipfel et al., 2003a). Some of these molecules can also be excited by confocal microscopy using UV light, although with the risk of inducing photodamage.
More recently, other nonlinear optical techniques have been used for IVM, and among them are three-photon excitation, and second and third harmonic generation (SHG and THG; Campagnola and Loew, 2003; Zipfel et al., 2003b; Oheim et al., 2006). Three-photon excitation follows the same principle as two-photon (Fig. 2 B), and can reveal endogenous molecules such as serotonin and melatonin (Zipfel et al., 2003a; Ritsma et al., 2013). In SHG and THG, photons interact with the specimen and combine to form new photons that are emitted with two or three times their initial energy (Fig. 2 C). SHG reveals collagen (Fig. 2 C) and myosin fibers (Campagnola and Loew, 2003), whereas THG reveals lipid droplets and myelin fibers (Débarre et al., 2006; Weigelin et al., 2012). Recently, two other techniques have been used for IVM: coherent anti-Stokes Raman scattering (CARS) and fluorescence lifetime imaging (FLIM). CARS that is based on two laser beams combined to match the energy gap between two vibrational levels of the molecule of interest, has been used to image lipids and myelin fibers (Müller and Zumbusch, 2007; Fu et al., 2008; Le et al., 2010). FLIM, which measures the lifetime that a molecule spends in the excited state, provides quantitative information on cellular parameters such as pH, oxygen levels, ion concentration, and the metabolic state of various biomolecules (Levitt et al., 2009; Provenzano et al., 2009; Bakker et al., 2012).
We want to emphasize that two-photon microscopy and the other nonlinear techniques are the obligatory choice when the imaging area is located deep inside the tissue, endogenous molecules have to be imaged, or long-term imaging with frequent sampling is required. However, confocal microscopy is more suited to resolve structures in the micrometer range, because of the possibility of modulating the optical slice (Masedunskas et al., 2012a).
IVM to investigate biological processes at the tissue and the single cell level
The main strength of IVM is to provide information on the dynamics of biological processes that otherwise cannot be reconstituted in vitro or ex vivo. Indeed, IVM has been instrumental in studying several aspects of tissue physiopathology (Table 1). A clear example is the response of the vasculature to physiological and pathological stimuli (Fig. 3, A and B). Although other approaches such as classical immunohistochemistry, electron microscopy, and indirect immunofluorescence may provide detailed structural and quantitative information on blood vessels, IVM enables measuring events such as variations of blood flow at the level of the capillaries or local changes in blood vessel permeability. These data have been instrumental in understanding the mechanisms of ischemic diseases and tumor progression, and in designing effective anticancer treatments.
IVM has also been used successfully to study the dynamics and the morphological changes of individual cells within a tissue (Table 2). These aspects are strongly influenced by a multitude of factors that include interactions with: (1) other cell types within the tissue, (2) components of the extracellular matrix (e.g., collagen fibers), and (3) molecules coming from the vasculature (e.g., glucose, oxygen, hormones), the central and the peripheral nervous system (e.g., neurotransmitters), or the immune system (e.g., chemokines).
In neurobiology, for example, the development of approaches to perform long-term in vivo imaging has permitted the correlation of changes in neuronal morphology and neuronal circuitry to pathological conditions such as stroke (Zhang and Murphy, 2007), tumors (Barretto et al., 2011), neurodegenerative diseases (Merlini et al., 2012), and infections (McGavern and Kang, 2011). This has been accomplished by the establishment of surgical procedures to expose the brain cortex, and the implantation of chronic ports of observations such as cranial windows and imaging guide tubes for micro-optical probes (Svoboda and Yasuda, 2006; Xu et al., 2007; Barretto et al., 2011). In addition, this field has thrived thanks to the development of several transgenic mouse models harboring specific neuronal populations expressing either one or multiple fluorescent molecules (Svoboda and Yasuda, 2006; Livet et al., 2007).
In tumor biology, the ability to visualize the motility of cancer cells within a tumor in vivo has provided tremendous information on the mechanisms regulating invasion and metastasis (Fig. 3 C; Beerling et al., 2011). Tumor cells metastasize to distal sites by using a combination of processes, which include tumor outgrowth, vascular intravasation, lymphatic invasion, or migration along components of the extracellular matrix and nerve fibers. Although classical histological analysis and indirect immunofluorescence have been routinely used to study these processes, the ability to perform long-term IVM through the optimization of optical windows (Alexander et al., 2008; Kedrin et al., 2008; Gligorijevic et al., 2009; Ritsma et al., 2012b) has provided unique insights. For example, a longitudinal study performed by using a combination of two-photon microscopy, SHG, and THG has highlighted the fact that various tissue components associated with melanomas may play either a migration-enhancing or migration-impeding role during collective cell invasion (Weigelin et al., 2012). In mammary tumors, the intravasation of metastatic cells has been shown to require macrophages (Wang et al., 2007; Wyckoff et al., 2007). In head and neck cancer, cells have been shown to migrate from specific sites at the edge of the tumor, and to colonize the cervical lymph nodes by migrating though the lymphatic vessels (Fig. 3 C; Amornphimoltham et al., 2013). In highly invasive melanomas, the migratory ability of cells has been correlated with their differentiation state, as determined by the expression of a reporter for melanin expression (Pinner et al., 2009).
Imaging the cells of the immune system in a live animal has revealed novel qualitative and quantitative aspects of the dynamics of cellular immunity (Fig. 2 C and Video 1; Germain et al., 2005; Cahalan and Parker, 2008; Nitschke et al., 2008). Indeed, the very complex nature of the immune response, the involvement of a multitude of tissue components, and its tight spatial and temporal coordination clearly indicate that IVM is the most suited approach to study cellular immunity. This is highlighted in studies either in lymphoid tissues, where the exquisite coordination between cell–cell interactions and cell signaling has been studied during the interactions of B lymphocytes and T cell lymphoid tissues (Qi et al., 2006), T cell activation (Hickman et al., 2008; Friedman et al., 2010), and migration of dendritic cells (Nitschké et al., 2012), or outside lymphoid tissues, such as, for example, brain during pathogen infections (Nayak et al., 2012), heart during inflammation (Li et al., 2012), and solid tumors (Deguine et al., 2010).
Imaging subcellular structures in vivo and its application to cell biology
The examples described so far convey that IVM has contributed to unraveling how the unique properties of the tissue environment in vivo significantly regulate the dynamics of individual cells and ultimately tissue physiology. Is IVM suitable to determine (1) how subcellular events occur in vivo, (2) whether they differ in in vitro settings, and (3), finally, the nature of their contribution to tissue physiology?
IVM has been extensively used to image subcellular structures in smaller organisms (i.e., zebrafish, Caenorhabditis elegans) that are transparent and can be easily immobilized (Rohde and Yanik, 2011; Tserevelakis et al., 2011; Hove and Craig, 2012). In addition, the ability to easily perform genetic manipulations has made these systems extremely attractive to study several aspects of developmental and cell biology. However, their differences in term of organ physiology with respect to rodents do not make them suitable models for human diseases. For a long time, subcellular imaging in live rodents has been hampered by the motion artifacts derived from the heartbeat and respiration. Indeed, small shifts along the three axes make it practically impossible to visualize structures whose sizes are in the micrometer or submicrometer range, whereas it marginally affects larger structures. This issue has been only recently addressed by using a combination of strategies, which include: (1) the development of specific surgical procedures that allow the exposure and proper positioning of the organ of interest (Masedunskas et al., 2013), (2) the improvement of specific organ holders (Cao et al., 2012; Masedunskas et al., 2012a), and (3) the synchronization of the imaging acquisition with the heartbeat and respiration (Presson et al., 2011; Li et al., 2012). Very importantly, these approaches have been successfully implemented without compromising the integrity and the physiology of the tissue, thus opening the door to study cell biology in a live animal.
For example, large subcellular structures such as the nuclei have been easily imaged, making it possible to study processes such as cell division and apoptosis (Fig. 4 A; Goetz et al., 2011; Orth et al., 2011; Rompolas et al., 2012). Interestingly, these studies have highlighted the fact that the in vivo microenvironment substantially affects nuclear dynamics. Indeed, mitosis and the structure of the mitotic spindle were followed over time in a xenograft model of human cancer expressing the histone marker mCherry-H2B and GFP-tubulin (Orth et al., 2011). Specifically, the effects of the anticancer drug Paclitaxel were studied, revealing that the tumor cells in vivo have a higher mitotic index and lower pro-apoptotic propensity than in vitro (Orth et al., 2011). FRET has been used in subcutaneous tumors to image cytotoxic T lymphocyte–induced apoptosis and highlighted that the kinetics of this process are much slower than those reported for nontumor cells in vivo that are exposed to a different microenvironment (Breart et al., 2008). Cell division has also been followed in the hair-follicle stem cells of transgenic mice expressing GFP-H2B. This study determined that epithelial–mesenchymal interactions are essential for stem cell activation and regeneration, and that nuclear divisions occur in a specific area of the hair follicles and are oriented toward the axis of growth (Rompolas et al., 2012). These processes show an extremely high level of temporal and spatial organization that can only be appreciated in vivo and by using time-lapse imaging.
Imaging membrane trafficking has been more challenging because of its dynamic nature and the size of the structures to image. The first successful attempt to visualize membrane traffic events was achieved in the kidney of live rats by using two-photon microscopy where the endocytosis of fluid-phase markers, such as dextrans, or the receptor-mediated uptake of folate, albumin, and the aminoglycoside gentamicin were followed in the proximal tubuli (Fig. 4 D and Video 2; Dunn et al., 2002; Sandoval et al., 2004; Russo et al., 2007). These pioneering studies showed for the first time that apical uptake is involved in the filtration of large molecules in the kidney, whereas previously it was believed to be exclusively due to a barrier in the glomerular capillary wall. However, in the kidney the residual motion artifacts limited the imaging to short periods of time. Recently, the salivary glands have proven to be a suitable organ to study the dynamics of membrane trafficking by using either two-photon or confocal microscopy. Systemically injected dextrans, BSA, and transferrin were observed to rapidly internalize in the stromal cells surrounding the salivary gland epithelium in a process dependent on the actin cytoskeleton (Masedunskas and Weigert, 2008; Masedunskas et al., 2012b). Moreover, the trafficking of these molecules through the endo-lysosomal system was documented, providing interesting insights on early endosomal fusion (Fig. 4 E; Masedunskas and Weigert, 2008; Masedunskas et al., 2012b). Notably, significant differences were observed in the kinetics of internalization of transferrin and dextran. In vivo, dextran was rapidly internalized by stromal cells, whereas transferrin appeared in endosomal structures after 10–15 min. However, in freshly explanted stromal cells adherent on glass, transferrin was internalized within 1 min, whereas dextran appeared in endosomal structures after 10–15 min. Although the reasons for this difference were not addressed, it is clear that the environment in vivo has profound effects on the regulation of intracellular processes (Masedunskas et al., 2012b). Similar differences have been reported for the caveolae that in vivo are more dynamic than in cell cultures (Thomsen et al., 2002; Oh et al., 2007). Endocytosis has also been investigated in the epithelium of the salivary glands (Sramkova et al., 2009). Specifically, plasmid DNA was shown to be internalized by a clathrin-independent pathway from the apical plasma membrane of acinar and ductal cells, and to subsequently escape from the endo-lysosomal system, thus providing useful information on the mechanisms of nonviral gene delivery in vivo (Sramkova et al., 2012). Receptor-mediated endocytosis has also been studied in cancer models. Indeed, the uptake of a fluorescent EGF conjugated to carbon nanotubes has been followed in xenografts of head and neck cancer cells revealing that the internalization occurs primarily in cells that express high levels of EGFR (Bhirde et al., 2009). The role of endosomal recycling has also been investigated during tumor progression. Indeed, the small GTPase Rab25 was found to regulate the ability of head neck cancer cells to migrate to lymph nodes by controlling the dynamic assembly of plasma membrane actin reach protrusion in vivo (Amornphimoltham et al., 2013). Interestingly, this activity of Rab25 was reconstituted in cells migrating through a 3D collagen matrix but not in cells grown adherent to a solid substrate.
IVM has been a powerful tool in investigating the molecular machinery controlling regulated exocytosis in various organs. In salivary glands, the use of selected transgenic mice expressing either soluble GFP or a membrane-targeted peptide has permitted the characterization of the dynamics of exocytosis of the secretory granules after fusion with the plasma membrane (Fig. 4 E; Masedunskas et al., 2011a, 2012d). These studies revealed that the regulation and the modality of exocytosis differ between in vivo and in vitro systems. Indeed, in vivo, regulated exocytosis is controlled by stimulation of the β-adrenergic receptor, and secretory granules undergo a gradual collapse after fusion with the apical plasma membrane, whereas, in vitro, regulated exocytosis is also controlled by the muscarinic receptor and the secretory granules fuse to each other, forming strings of interconnected vesicles at the plasma membrane (compound exocytosis; Masedunskas et al., 2011a, 2012d). Moreover, the transient expression of reporter molecules for F-actin has revealed the requirement for the assembly of an actomyosin complex to facilitate the completion of the exocytic process (Masedunskas et al., 2011a, 2012d). This result underscores the fact that the dynamics of the assembly of the actin cytoskeleton can be studied both qualitatively and quantitatively in live animals at the level of individual secretory granules. In addition, this approach has highlighted some of the mechanisms that contribute to regulate the apical plasma membrane homeostasis in vivo that cannot be recapitulated in an in vitro model systems (Masedunskas et al., 2011b, 2012c; Porat-Shliom et al., 2013). Indeed, the hydrostatic pressure that is built inside the ductal system by the secretion of fluids that accompanies exocytosis plays a significant role in controlling the dynamics of secretory granules at the apical plasma membrane. This aspect has never been appreciated in organ explants where the integrity of the ductal system is compromised. Finally, a very promising model has been developed in the skeletal muscle, where the transient transfection of a GFP-tagged version of the glucose transporter type 4 (GLUT4) has made possible to characterize the kinetics of the GLUT4-containing vesicles in resting conditions and their insulin-dependent translocation to the plasma membrane (Fig. 4 B; Lauritzen et al., 2008, 2010). This represents a very powerful experimental model that bridges together physiology and cell biology and has the potential to provide fundamental information on metabolic diseases.
These examples underscore the merits of subcellular IVM to investigate specific areas of cell biology such as membrane trafficking, the cell cycle, apoptosis, and cytoskeletal organization. However, IVM is rapidly extending to other areas, such as cell signaling (Stockholm et al., 2005; Rudolf et al., 2006; Ritsma et al., 2012a), metabolism (Fig. 4 C; Débarre et al., 2006; Cao et al., 2012), mitochondrial dynamics (Fig. 4 F; Sun et al., 2005; Hall et al., 2013), or gene and protein expression (Pinner et al., 2009) that have just begun to be explored.
IVM has become a powerful tool to study biological processes in live animals that is destined to have an enormous impact on cell biology. The examples described here give a clear picture of the broad applicability of this approach. In essence, we foresee that IVM is going to be the obligatory choice to study highly dynamic subcellular processes that cannot be reconstituted in vitro or ex vivo, or when a link between cellular events and tissue physiopathology is being pursued. In addition, IVM will provide the opportunity to complement and confirm data generated from in vitro studies. Importantly, the fact that in several instances confocal microscopy can be effectively used for subcellular IVM makes this approach immediately accessible to several investigators.
In terms of future directions, we envision that other light microscopy techniques will soon become standard tools for in vivo studies, as shown by the recent application of FRET to study signaling (Stockholm et al., 2005; Rudolf et al., 2006; Breart et al., 2008; Ritsma et al., 2012a), and FRAP, which has been used in the live brain to measure the diffusion of α synuclein, thus opening the door to studying the biophysical properties of proteins in vivo (Unni et al., 2010). Moreover, super-resolution microscopy may be applied for imaging live animals, although this task may pose some challenges. Indeed, these techniques require: (1) the complete stability of the specimen, (2) extended periods of time for light collection, (3) substantial modifications to the existing microscopes, and (4) the generation of transgenic mice expressing photoactivatable probes.
To reach its full potential, IVM has to further develop two main aspects: animal models and instrumentations. Indeed, a significant effort has to be invested in developing novel transgenic mouse models, which express fluorescently labeled reporter molecules. One example is the recently developed mouse that expresses fluorescently tagged lifeact. This model will provide the unique opportunity to study F-actin dynamics in vivo in the context of processes such as cell migration and membrane trafficking (Riedl et al., 2010). Moreover, the possibility of crossing these reporter mice with knockout animals will provide the means to further study cellular processes at a molecular level. Alternatively, reporter molecules or other transgenes that may perturb a specific cellular pathway can be transiently transfected into live animals in several ways. Indeed, the remarkable advancements in gene therapy have contributed to the development of several nonviral- and viral-mediated strategies for gene delivery to selected target organs. In this respect, the salivary glands and the skeletal muscle are two formidable model systems because either transgenes or siRNAs can be successfully delivered without any adverse reaction and expressed in a few hours. In terms of the current technical limitations of IVM, the main areas of improvement are the temporal resolution, the ability to access the organ of interest with minimal invasion, and the ability to perform long-term imaging. As for the temporal resolution, the issue has begun to be addressed by using two different approaches: (1) the use of spinning disk microscopy, as shown by its recent application to image platelet dynamics in live mice (Jenne et al., 2011); and (2) the development of confocal and two-photon microscopes equipped with resonant scanners that permit increasing the scanning speed to 30 frames per second (Kirkpatrick et al., 2012). As for accessing the organs, recently several microlenses (350 µm in diameter) have been inserted or permanently implanted into live animals, minimizing the exposure of the organs and the risk of affecting their physiology (Llewellyn et al., 2008). Finally, although some approaches for the long-term imaging of the brain, the mammary glands, and the liver have been developed, additional effort has to be devoted to establish chronic ports of observations in other organs.
In conclusion, these are truly exciting times, and a new era full of novel discoveries is just around the corner. The ability to see processes inside the cells of a live animal is no longer a dream.
Online supplemental material
Video 1 shows time-lapse confocal microscopy of a granulocyte moving inside a blood vessel in the mammary gland of a mouse expressing GFP-tagged myosin IIb (green) and labeled with MitoTracker (red). Video 2 shows time-lapse confocal microscopy of the endocytosis of systemically injected 10 kD Texas red–dextran (red) into the kidney-proximal tubuli of a transgenic mouse expressing the membrane marker m-GFP (green).
This research was supported by the Intramural Research Program of the National Institutes of Health, National Institute of Dental and Craniofacial Research.