In the past, cytometry of cellular and molecular mechanisms underlying immune reactions was mainly performed ex vivo. These analyses focused on characterization of the players in the immune system as single entities, without taking tissue context into account. On the other hand, microscopic analyses of intact tissues were restricted to fixed samples. Thus, the dynamic nature of the immune response, with cellular interactions and migration being two of its most prominent features, could not be analyzed until recently.
The development of multiphoton intravital microscopy and its application to address immunological questions has changed this over the past decade. For instance, typical motility patterns of lymphocytes could be tracked for the first time in time-lapse movies taken in living tissue. Lymphocytes in secondary lymphoid organs were shown to move with a median instantaneous three-dimensional (3D) velocity of between 5 and 10 μm/min, and are therefore able to cover distances of several hundred micrometers within a few hours (1, 2). Compared with other imaging techniques (summarized in Table 1), multiphoton microscopy is the only technology able to capture the dynamic nature of these processes in deep tissue, as it allows fast acquisition of 3D image stacks (for example: 300 × 300 × 30 μm3 with a resolution of 512 × 512 × 11 voxels can be recorded every 15 s). Until now, most of the work in the field has focused on the analysis of motility parameters (e.g., the velocity at which cells migrate, their displacement, directed versus random migration) and the duration and type of cellular interactions using fluorescent reporter mice to distinguish between different cell subsets. These studies have already contributed tremendously to our understanding on how the immune system works in vivo, and at the same time have sparked the desire to visualize subcellular events by intravital microscopy. As novel photoactivatable fluorescent proteins and genetically encoded Förster resonant energy transfer (FRET)-biosensors of sufficient brightness for in vivo imaging are becoming increasingly available, the possibility to visualize cellular function on a molecular level is now feasible. Hence, advances in transgenic mouse technology are necessary together with developments of microscopy techniques to better fit the new requirements of dynamic intravital imaging. These requirements are related both to a better optical performance, (e.g., greater imaging depth, higher spatial resolution, lower photobleaching of the fluorophores, and tissue photodamage) and to new possibilities to probe cellular function using the intrinsic molecular specificity of fluorescence. It is becoming increasingly clear that engineering fluorescent mice has to work hand in hand with developments in the field of microscopy. To address this, the first part of the present review deals with the latest achievements in transgenic mouse technology involving fluorescent proteins. In the second part, we review the progress in microscopy design inspired by these achievements, which will provide insights into molecular immunology in vivo in the future.
Table 1. Overview and features of imaging techniques used in biosciences and biomedicine
Intravital /in vivo (cell culture/organ models)
∼ 100 μm (9/17 T)
Spectroscopy possible but worse resolution
∼ 100 μm
Photoacoustic microscopy (121)
Molecular specificity limited
Molecular specificity limited
Molecular specificity limited
Wide-field microscopy (85)
∼ 50 μm
Confocal microscopy (85)
∼ 50 μm
500 μm–1 mm
Confocal or MP micro-endoscopy (81–87)
Similar to MPM; depends on surgery
Only at the surface
Molecular specificity possible but difficult
Only at the surface
No/yes in cell culture
Max. 200 μm thin samples
Typically cell monolayer; up to 130 μm brain slices (121)—resolution comparable to confocal microscopy
Max. 10 μm
Max. cell monolayer
Highlighting Cellular Function: Fluorescent Reporter Mice as Tools for Intravital Microscopy
Second harmonic generation (SHG) as a result of frequency doubling caused by collagen fibers is very useful for orientation within organs (e.g. for identifying the capsule of lymph nodes). Recently, second harmonic generation polarization anisotrophy has been described to be useful in defining the structural state of actomyosin motors (3). In order to draw conclusions about specific cell subsets or tissue structures they have to be rendered fluorescent. Fluorescent high molecular weight dextran (4) can be used for highlighting blood vessels as anatomical landmarks. The use of fluorescent semiconductor nanocrystals (quantum dots) has also been reported for that purpose, with the advantage of less interstitial leakage than dextran (5). Nonhematopoietic stromal cells can be visualized using in vivo labeling with fluorescently conjugated whole antibodies (6). Fab-fragments (2) are used to avoid unwanted Fc-receptor mediated effects. Alexa dyes have been successfully used for this purpose, although they have the disadvantage of limited observation time because of bleaching. ATTO dyes are an interesting alternative to established dyes coupled with antibodies for intravital imaging, as they are less prone to photobleaching and are not cytotoxic (A.E.H. and R.A.N., unpublished observation).
Labeling lymphocytes ex vivo with cell-permeable fluorescent dyes like CMAC (blue), CFSE (green), CMTMR (orange), CMPTX (red), or SNARF (far red) and adoptively transferring them to syngeneic recipients allows tracking only over short time periods (i.e., a few days), since the dyes are diluted with every cell division. The use of donor mice expressing fluorescent proteins ubiquitously is an option if highly proliferative cell populations are being investigated in vivo (7–9) although functional information on a single cell level is limited in those systems.
Using reporter mice in which gene expression is linked to the expression of a fluorescent protein provides additional information about actual functionality. This method has been used to label cells based on the expression of certain transcription factors (e.g., Blimp1 (10, 11) and Foxp3 (12, 13), the production of cytokines and chemokines [e.g., IL-2 (14), IL-4 (15), IFN-γ (16) IL-7 L (17), IL-10 (18), CXCL12 (19), CCL17 (20)] chemokine receptor and adhesion molecule expression [CX3CR1 (21), CD11c (22, 23)], and other molecules associated with cellular function, such as production of lysozyme (24) or antigen presentation (25).
Most of these mice have been generated using green fluorescent protein [GFP (26)], limiting the analysis of several reporter genes in the same sample. Cyan fluorescent protein (CFP) is a less desirable candidate for intravital imaging because of its short emission wavelength, which results in a high degree of reflection in the tissue, making it difficult to detect CFP at the depths needed for intravital imaging. Spectral separation between yellow fluorescent protein (YFP) and GFP is a problem in intravital imaging due to their relatively low signals, thus it is necessary to move into the red and even far-red spectrum for a combined use with GFP. Attempts to generate viable transgenic mice using the first cloned red fluorescent protein (RFP) DsRed from the sea anemone Discosoma (27) were unsuccessful (8). Optimizing DsRed resulted in several tetrameric variants that formed green-emitting intermediates and had a tendency to form aggregates in cells (28, 29). The generation of the first monomeric red fluorescent protein (mRFP) overcame these problems but had the disadvantage of decreased brightness (30). Finally, the development of new, bright, and photostable monomeric RFP variants (31), which possess fast maturation times constitute promising tools for transgenic mice (9, 32, 33), some of which have already been reported to be successfully used in multiphoton intravital imaging (34). Bright fluorescent proteins cloned from a different sea anemone, Entacmea quadricolor (35), should enable further shifting into the far-red area of the spectrum, especially when combined with optimized excitation through the application of an optical parametric oscillator (OPO), thereby opening the possibilities for multicolor intravital microscopy using fluorescent proteins.
If appropriate Cre-recombinase expressing mouse strains are available, they can also be used to render specific cell populations fluorescent by crossing these mice on strains in which fluorescent proteins have been targeted into the ubiquitously expressed ROSA26 locus. These reporter constructs consist of a transcriptional stop element flanked by loxP-sites, followed by genes encoding for fluorescent proteins. The stop element becomes irreversibly excised in cells expressing Cre-recombinase, lead to transcription of the fluorescent protein in a cell and all its progenitors. Although this system does not allow conclusions to be drawn about ongoing transcription, it is particularly useful as a lineage marker. Mouse strains using Cre-inducible expression of CFP, YFP (36), eGFP (37), and tdRFP (9) have been generated.
Visualizing differentiation processes via in the combination of several fluorescent proteins indicating developmental stages in one cell requires special characteristics of the fluorescent proteins, such as the combination of fast maturation with a short half-life. Sakaue-Sawano et al. (38) succeeded in visualizing cell cycle progression in transgenic mice by fusing monomeric forms of Azami Green and Kusabira Orange with Geminin and Cdt-1, respectively, two proteins known to oscillate inversely during cell cycle progression (39). Although the authors used single-photon excitation to generate their data in developing brain, these mice provide the first proof that temporal cellular processes can be visualized with a dual-color fluorescent protein system in vivo, and similar systems are potentially useful for multiphoton imaging (40). Most of the cellular differentiation processes exceed the intravital imaging acquisition periods that are currently feasible, thus it remains a challenge to improve the anesthesia technique for small laboratory animals to extend the duration of intravital imaging recording periods to observe differentiation processes in multiphoton movies.
Cellular interactions are a prerequisite in the initiation and maintenance of every immune response, and multiphoton imaging has made it possible to analyze these contacts in terms of duration and frequency, between T cells and dendritic cells (DCs) (4, 23, 41), B cells and T cells (42, 43), B cells and follicular dendritic cells (2, 6, 44, 45), and B cells and DCs (46) in vivo. These studies have provided important insights in where and how these contacts happen but, at the same, time have raised new questions: What are the consequences of these interactions on a molecular level in each cell? The duration of T cell–DC interaction has been shown to determine the outcome in terms of an immunogenic versus tolerogenic response (41), but which signaling pathways are differentially activated during these two types of interaction? Approaches to visualize the outcome of interactions by multiphoton intravital microscopy include the analysis of adoptively transferred cells that had been loaded with a Ca2+-sensitive dye to demonstrate the effect of B cell–DC interactions (46). A dye-based method has also been described to visualize cytotoxic effects on cells in vivo by double-labeling them with a membrane-permeant dye (CellTracker Orange) in combination with the blue nuclear dye Hoechst 33342 (47). The loss of labeled cytoplasmic proteins results in a decrease of orange fluorescence, while blue fluorescence at the same time increases; this might be either an effect of unquenching or release of Hoechst 33342 as the nucleus disintegrates.
Analyzing Molecular Interactions in Intravital Multiphoton Imaging: Towards Quantitative FRET
Genetically encoded FRET-based probes are an elegant option to visualize protein–protein interactions, protein folding, protein cleavage, and syntheses which constitute a central corner stone of cellular function and communication.
Fluorescence techniques, in general, and fluorescence microscopy, in particular, use FRET to probe these phenomena on a molecular basis (48–54). The development of transgenic mice, which selectively express FRET-based constructs indicating Ca2+ levels, membrane voltage, or apoptosis, open new perspectives for probing cellular function dynamically in living organisms.
Successful use of FRET by intravital microscopy has been reported for imaging of apoptosis events during intratumoral CD8+ T cell cytotoxic activity (55), for proving that contacts between pathogenic CD4+Th17 cells and neurons induce fluctuations of the neuronal intracellular Ca2+ concentration as an early stage of tissue damage in experimental autoimmune encephalomyelitis, a mouse model of multiple sclerosis (Fig. 1) (34) and for imaging protein–protein interactions within tumors (54).
Currently, most of the FRET pairs used in biosciences are based on fluorescing proteins like CFP and YFP or derivatives (49, 56, 57). Especially with the development of red fluorescent proteins, pairs like GFP-mCherry, GFP-tdRFP, or even RFP-farRFP, which are characterized by more efficient FRET, lower crosstalk, or lower photobleaching are increasingly being used (50, 52, 58).
As far as the microscopy techniques to probe FRET are concerned, a great deal of experience was accumulated over the last decades (48, 51, 59, 60). They range from qualitative techniques measuring only the acceptor fluorescence signal, to ratiometric two- or four-wavelength techniques (34, 51, 57, 61), photobleaching-based techniques (51), fluorescence polarization-based techniques (53), and time-resolved techniques (51, 56, 60). Photobleaching-based techniques, which necessitate the acquisition of upto seven images of the same area within the sample at one time point, are currently accepted to be the most accurate (51). Fluorescence lifetime imaging (FLIM) is similarly reliable and only necessitates the time-resolved detection of the donor fluorescence in the presence and absence of the acceptor (56). However, these techniques have been developed and applied mostly in extracellular conditions or in cell culture, and not in an intravital setting.
The application of FRET techniques to dynamic intravital imaging is related to the uncertainty introduced by refractive index mismatches within tissue. Until now, especially ratiometric techniques based on two-wavelength detection have been used (34, 55, 57). For more accuracy, a four-wavelength detection method (two for the donor and two for the acceptor) has been developed (61).
Still a true quantitative FRET measurement is needed for intravital imaging, as all ratiometric techniques are constrained by the limitations of different scattering and photobleaching characteristics of donor and acceptor molecules in tissue. As photobleaching techniques require high laser illumination, they are rather invasive. Moreover, they are quite slow because of complicated calibration procedures at each of the regions to be imaged, so their application to dynamic intravital imaging is inadequate. The development of dynamic intravital FRET-FLIM is highly desirable in this context. Repeated but still rather slow intravital FRET-FLIM has been applied in a cancer research application (54).
FLIM can be performed either in the frequency domain or in the time domain. The benefits and pitfalls of the different acquisition and evaluation techniques have been already extensively reviewed (48, 59, 60, 62–66). As far as performance and accuracy for multiexponential decays are concerned, time-domain techniques proved to be superior to frequency-domain techniques (62). In the context of multiphoton microscopy, it is accepted that techniques based on time-correlated single-photon counting (TCSPC) are more reliable (58, 62, 65). Although time-gated camera-based techniques (65) and streak-camera techniques (67, 68) allow for fast acquisition, they are better suited for wide-field microscopy than for deep-tissue standard multiphoton laser scanning microscopy (MPLSM). They are mostly characterized by low signal-to-noise ratio and signal efficiency (only ∼10% of the photons impinging the detector are considered for the fluorescence decay), and coarse lateral resolution. In contrast, TCSPC techniques theoretically allow for imaging with diffraction-limited resolution and are more appropriate for the low signal typical for deep-tissue imaging. However, due to the dead-time introduced by the TCSPC electronics (time-to-amplitude converter [TAC] or time-to-digital converter [TDC]), standard TCSPC techniques can acquire ∼ 10% of the emitted photons, that is, only the photon emitted after every eighth laser pulse can be evaluated (56, 62). This means that standard TCSPC techniques can be easily applied to acquire single 3D snapshots of slowly or nonchanging structures deep within tissue but do not allow for monitoring dynamics, for example, of the immune system. Synchronized multichannel TCSPC techniques are very promising in this context, because they dramatically increase the photon efficiency and therefore acquisition speed, for example, in combination with multibeam scanning (56).
Apart from photobleaching-based FRET techniques, the rather neglected phenomenon of chromophore photobleaching is used by techniques like fluorescence recovery after photobleaching (FRAP) or fluorescence loss in photobleaching (FLIP) to obtain information about slow biomolecular kinetics or transport processes within cells (69, 70). Until now, these techniques have been applied only in extracellular conditions or in cell culture. We expect that the application of FLIP and FRAP to dynamic intravital microscopy is rather difficult due to the sustained need of high laser powers and thus stress for the fluorophores (71, 72) and tissue (73) over time.
Breaking Spatiotemporal Limits: Photoactivation in Intravital Microscopy
The way the immune system is organized, with cells migrating between primary and secondary lymphoid tissues and also to sites of inflammation, requires that we cannot consider only one organ (or part of an organ) to understand the complex dynamics underlying immune responses. With the 20× objective typically used in intravital microscopy, the field of view is limited to about 500 × 500 μm2 in x/y and 200–300μm in z, covering only relatively small regions of an organ (even if it is a small organ such as a mouse lymph node), and information about cells migrating out of these boundaries is lost. Whole body-imaging techniques do not achieve the resolution required to answer questions about interorgan communication on a single cell level. Moreover, it is sometimes desirable to mark single cells within a certain tissue location to retrieve them after imaging to perform further analyses. Mice expressing photoactivatable fluorescent proteins in certain cell subsets constitute a promising tool to overcome these limits. The fluorescence of these proteins can be manipulated by exposing them to light of a specific wavelength and intensity: some become activated from a nonfluorescent into a fluorescent state, such as paGFP (74) pa mRFP (75), and kindling fluorescent protein-1 (KFP-1) (76) (with red fluorescence of KFP-1 being potentially reversible, depending on the intensity of the activating light). Others convert from one emission maximum to another, therefore, changing their color [e.g., Kaede (77) or EosFP (78) KikGR (79), all switching from green to red]. A transgenic mouse ubiquitously expressing Kaede has been generated and was used to track the turnover of different immune subsets in secondary lymphoid organs (80); in this case, photoactivation was performed using UV radiation. Most of the photoactivatable fluorescent proteins require wavelengths in the UV-range, which are potentially harmful for cells. The use of nonlinear optics for photoactivation circumvents this limitation and in addition makes it possible to activate more defined areas, such as single cells within the tissue. Technically the selective photoactivation of cells necessitates exact control of the scanning mirrors, for example, repeated illumination of the same region within the field of view with accuracy on the scale of nanometers. Fast control of the photoactivation laser is also required, and this can be achieved by means of electroptical modulator-based shutters. Until now, only one study in which photoactivation was used in combination with multiphoton microscopy has been published, using a mouse expressing paGFP (81).
Technical Advances IN Dynamic Intravital Multiphoton Imaging: Improving Depth AND Resolution
With respect to deeper penetration in living tissue, two new techniques are revolutionizing the field of multiphoton microscopy: multiphoton endoscopy and middle infrared (IR) excitation.
The implementation of OPOs in TPLSM allows excitation in the IR range, starting at the limits of Titanium–Sapphire (Ti:Sa) lasers at ∼1,050 nm and extending as far as 1,600 nm. Using IR instead of standard near infrared (NIR) excitation leads to increased penetration depth, as predicted by theory (scattering laws). Thus, IR excitation makes regions accessible, which until now could not be dynamically visualized in living organisms. An increase of up to 80% in depth was shown in different organs of the central nervous and immune system as well as in the skin (82, 83).
The idea to apply an endoscopy technique in multiphoton microscopy took shape after the first successful deep-tissue intravital experiments and attracted a great deal of attention as it theoretically allows the investigation of all possible organs within a living organism without being limited by scattering. Initial setups used a bundle of optical fibers to conduct the excitation light focused by an objective lens to the region of interest within the organ of interest. However, this technique is limited by the heterogeneous illumination of the sample and by substantial spherical aberration, resulting in low spatial resolution, and a deformed image. A true solution to counteract these shortcomings was achieved by the development of gradient refractive index (GRIN) optics (84–87). These are thin (∼500 μm) microstabs with an anisotropic refractive index (gradient of refractive index) to partially counteract the spherical aberration and ensure homogenous illumination of the sample. Moreover, its combination with adaptive optics dramatically improves optical performance. Applications in neurology that showed growth of dendrites over weeks, angiogenesis in glioblastoma (88) or light-based modulation of ion channels, that is, optogenetics (89), as well as in cancer research (90–92) have already demonstrated the power of this technique. The best numerical aperture (NA) reached by multiphoton endoscopy amounts to 0.48, which results in a lateral resolution of 779 nm and an axial resolution of 5,533 nm at an excitation wavelength of 850 nm. These values hold true at the contact region between the GRIN-optics and tissue but are expected to increase deeper within the tissue due to scattering.
By combining two deciding technical developments for intravital microscopy of small animals, fluorescence lifetime endoscopy (FLE) considerably extends the capacity to monitor cellular function on a molecular basis to organs and tissue areas, which until now have been inaccessible (93). Still, technical efforts are required to perform fast high-resolution FLE.
Other promising techniques that provide insight into previously inaccessible tissue regions by intravital imaging are photoacoustic microscopy and optical coherence tomography (OCT). OCT-based techniques have been already reviewed elsewhere (94). Photoacoustic microscopy allows whole body imaging but it is characterized by relatively poor spatial resolution. Notably, light-sheet microscopy allows for very rapid, almost photobleaching- and photodamage-free imaging of whole organisms (95). This technique is adequate for transparent samples, for example, zebra fish embryos or fixed organs mounted in 2,2′-thio(diethanol) (96) but not for intravital imaging of small animals like the mouse. Its resolution is limited by a rather low NA of the objective lenses but was improved by combining it with structured illumination (97). These techniques have been extensively discussed elsewhere.
A central limitation in intravital multiphoton microscopy results from the depth-dependent deterioration of the effective point-spread function due to short-path scattering of the excitation light (83, 98). Although the surface of lymph nodes the spatial resolution corresponds to the values calculated by the paraxial resolution, that is, 350 nm lateral and 1,300 nm axial for a 20× objective lens (NA 0.95) at λexc = 800 nm, at 60 μm depth we measured values of 1,400 nm lateral and 4,300 nm axial resolution (98). In this context, the requirements for better spatial resolution deep within tissue become evident.
While at the surface of organs/tissue, the spatial resolution decreases with increasing excitation wavelength, the scattering (short-path scattering) related to refractive index mismatches decreases nonlinearly (either λ−2 or λ−4) with increasing excitation wavelength (99). Therefore, although the spatial resolution at the surface of brain slices under IR excitation of the OPO (l = 1,110 nm) is slightly worse than under NIR excitation of conventional Ti:Sa lasers (λ = 850 nm), in 70 μm depth the advantage of IR excitation is striking (Fig. 2).
In recent years, we experienced a breakthrough in the improvement of spatial resolution for far-field fluorescence microscopy by powerful techniques like STED and RESOLFT (100–108), PALM/STORM (109–112) or linear and nonlinear pattern-illumination techniques (113–117).
Thus, with 7.5 nm lateral and 70 nm axial resolution (by STED), far-field fluorescence microscopy restores detailed morphological information which until now has been only accessible by electron microscopy. As already shown for STED (107), these techniques can be extended to multiphoton approaches, and thus applied to live tissue imaging. For instance, two-photon excitation STED in brain slices was reported (118). Although the lateral resolution in brain tissue was improved twice in these experiments, it did not reach the values expected from theory. Moreover, a relatively high laser power of up to 34.8 mW at 736 nm (continuous wave) must be used to induce the stimulated emission and to achieve this improvement in resolution. Such stress is feasible for short exposure of the tissue to the excitation light, that is, if the region of interest is imaged only once or a few times, but it is incompatible to repeated imaging of the same area as required by dynamic intravital and live tissue imaging.
Because of the technical difficulties in live tissue and since the spatial resolution in deep-tissue imaging is limited by scattering rather than by diffraction while the current super-resolution techniques are developed to break the diffraction rather than the scattering limit, the customized design of appropriate intravital super-resolution techniques remains a challenge for microscopy engineering.
In summary, new multiphoton microscopy techniques for dynamic intravital imaging will enlighten our understanding of, areas within the living organism which until now have been inaccessible, and with better optical quality. Especially those techniques which allow for a more complex, dynamic monitoring of cellular function on a molecular level will be at the forefront of further developments in intravital imaging. In this context, we foresee great potential for multiphoton endoscopy (with considerably improved spatial resolution), and for time-resolved techniques, especially TCSPC-based FLIM, as it intrinsically allows for quantitative monitoring of cellular function. Along with novel achievements in transgenic mouse technology, such as genetically encoded biosensors for different signaling pathways, it will be possible to gain information about molecular events taking place inside cells of the immune system, within their spatiotemporal context.
The authors thank Dr. Jason Millward for critical reading of the manuscript.