Scientists have long valued the power of in vivo observation to answer fundamental biological questions. Over the last 20 years, the application and evolution of intravital microscopy (IVM) has vastly increased our ability to directly visualize immune responses as they are occurring in vivo after infection or immunization. Many IVM strategies employ a strong multiphoton laser that penetrates deeply into the tissues of living, anesthetized mice, allowing the precise tracking of the movement of cells as they navigate complex tissue environments. In the realm of viral infections, IVM has been applied to better understand many critical phases of effector T cell responses, from activation in the draining lymph node, to the execution of effector functions, and finally to the development of tissue-resident memory. In this review, we discuss seminal studies incorporating IVM that have advanced our understanding of the biology of antiviral CD8+ T cells.
Viral infections pose an immediate and ongoing global public health challenge, with continued outbreaks of common viruses and the relentless threat of new pandemics resulting from emerging or re-emerging pathogens. Historically, little prior knowledge has been required to generate effective antiviral vaccines. However, for many of the remaining viruses, as well as for advanced preparation for new viruses, a better understanding of the underlying biology of important antiviral immune effectors could inform antiviral vaccine design and maximize effectiveness. Although many different arms of the immune response are important for immunity to viral infections, in this article, we focus on CD8+ T cells.
Once a virus is able to establish infection, CD8+ T cells are critical for the recognition and elimination of infected cells. Because of their importance, immunological studies often use the number and nature (e.g., the ability to produce one or more cytokines and different cytolytic proteins) of CD8+ T cells as a gauge of the effectiveness of antiviral response. Traditionally, these analyses have been performed ex vivo after tissue and/or fluid removal from infected animals at different times postinfection. In recent years, the importance of retaining precise spatial and kinetic information has become increasingly evident for the generation of a detailed understanding of CD8+ T cell activation and antiviral activity (1). Thus, to complement ex vivo approaches, increasing numbers of immunologists are turning to multiphoton intravital microscopy (IVM) to visualize both virus-infected cells and antiviral T cells as the immune response unfolds in infected animals. In this Brief Review, we discuss novel insights into antiviral T cell biology gained uniquely through IVM.
IVM: a brief overview
IVM refers to any microscopic observation of the tissues of a living animal regardless of the methodology used for imaging, including standard confocal microscopy and multiphoton microscopy. However, the application of IVM to immunological studies greatly expanded after the development of off-the-shelf multiphoton lasers (2). Multiphoton lasers deliver powerful pulses of light capable of exciting traditional fluorophores in a restricted tissue space with less scattering of light and deeper tissue penetration than standard lasers. The technical application of multiphoton IVM (hereafter referred to as IVM for simplicity), including the physics behind multiphoton excitation, has been reviewed extensively elsewhere (3–6).
IVM has some specific advantages and disadvantages that should be considered when interpreting studies using this powerful technique. Technically, the term IVM applies only to those studies performed in vivo in living animals (hence, intravitally). Although IVM studies often employ a multiphoton (also referred to as a two-photon) laser, this laser is not required for some exterior tissues (such as the skin) that are amenable to imaging using traditional single-photon confocal lasers (7). For the many interior tissues and organs that do not fall into this category, surgery is required to allow access of the microscope objective (which is not small), and tissue movement must be stabilized. This process of surgery/stabilization can be quite traumatic for some tissues, and how much perturbation and manipulation occurs should be considered when evaluating any IVM study. Even with the routinely imaged lymph node (LN), changes in blood or lymph flow, as well as tissue oxygenation, can impact lymphocyte motility, and care should be taken that consistent results in terms of cellular mobility are achieved between laboratories and even the individuals performing experiments (8, 9).
Some tissues, such as the thymus, simply cannot currently be accessed and stabilized for imaging using IVM without causing damage that would hamper data interpretation (10). Therefore, an alternative approach to imaging infected tissues is the removal of the organ/tissue, followed by thin sectioning using a vibratome, after which the explanted slice is kept in conditions that mimic normal physiology as closely as possible (11, 12). For some studies, the explant approach offers the only viable solution, and it allows for the imaging of tissues from animals other than mice (which will not fit the microscope) or for viruses (or other pathogens) with enhanced biosafety requirements (such as BSL-3 viruses). However, removal from the animal raises concerns about the movement and interactions of cells and limits the amount of time that the imaged specimen accurately reflects what is occurring in vivo, as new cells cannot enter. Although this review focuses on multiphoton IVM studies, we will also highlight important studies that have used explant-based, nonintravital approaches to study viral immunity.
There are several other technical considerations for IVM studies. Although the number of fluorophores that can be excited by multiphoton lasers is ever expanding, imaging multiple cell populations in the tissue is still limited compared with immunohistochemical approaches. Thus, much of the microscope field of view may appear black, lacking signal; however, it is important to remember that these spaces are filled with both hematopoietic and nonhematopoietic cells that are affecting the movement and interactions of the cells that are visualized. Additionally, although multiphoton lasers afford deeper tissue penetration, there are still limitations to the imaging depth that can be achieved, especially when compared with new cleared tissue techniques that allow visualization of the entire volume of large organs. Finally, although detector sensitivity and laser power have greatly increased over the years, it is still difficult to detect some fluorescent signals, particularly when using Ab injection to label structures or cell populations in vivo.
Even with the aforementioned caveats, IVM and multiphoton microscopy (MPM) are powerful approaches to study antiviral immunity that yield unique spatiotemporal information during the ever-changing in vivo interplay between virus and host (Fig. 1). In this article, we will highlight the many multiphoton IVM-based studies that have advanced our understanding of antiviral CD8+ T cell biology.
Antiviral T cell priming in the LN
Adaptive antiviral immune responses initiate in the LN draining the site of viral infection. The ease of exposing LNs for multiphoton IVM as well as their importance during immune responses have made LNs a frequent target for direct visualization. Indeed, the first application of MPM to virus-infected tissue was in the LN (13).
LNs have a highly specialized structure, and nodal architecture dictates the progression of T cell activation (14). Lymph that drains from infected tissue is routed via afferent lymphatic vessels into the LN, where lymph and any conveyed viral cargo is deposited into the subcapsular sinus (SCS). In the SCS, macrophages sample lymph and remove lymph-borne virions, serving as a barrier to viral infection of the LN and to systemic dissemination (15–18). For this important function, SCS macrophages have been dubbed the “flypaper” of the immune system. Subjacent to the SCS, B cell follicles containing naive B cells, follicular helper CD4+ T cells, and follicular dendritic cells (DCs) receive viral Ag after SCS macrophage capture and handoff (17). In the paracortex of the LN lies the T cell zone, where naive T cells enter through high endothelial venules and scan the DCs that heavily populate this area for cognate viral Ag.
Because Ag availability and form dictate the activation of antiviral CD8+ T cells, how Ag arrives and is distributed in the LN postinfection is the subject of intense investigation. Importantly, many IVM studies use the s.c. injection of virus, which delivers a bolus of virus to the LN within seconds. Non–DC-associated, lymph-borne virions that are not rapidly acquired by SCS macrophages upon deposition into the LN SCS do not have unrestrained access to the LN paracortex. Instead, the lymphoid endothelial cells (LECs) lining the SCS restrict virus entry in two ways: 1) by forming a barrier from the sinus and 2) by enforcing a size-exclusion limit through structures formed by the LEC protein plasmalemma vesicle-associated protein 1 (PLVAP) (19). In contrast to large virions, small proteins (those under 70 kDa) can flow through a series of channels called conduits that run from the SCS to the paracortex of the LN and the high endothelial venules (20, 21). Virions that are not captured in the SCS flow through the sinus network of the LN, eventually entering the medullary sinuses that are also heavily populated with phagocytic macrophages. We have recently demonstrated that, after s.c. injection at the high viral doses often used for IVM studies or vaccination, vaccinia virus (VACV) can enter the conduit system and infect paracortical DCs, which rapidly activate centrally located T cells (22). It is currently unclear what role, if any, the conduit system plays in the transport of virions during natural viral replication.
T cell Ag recognition in vivo in LNs is uniquely quantifiable using IVM. Because T cell speed, directional movement (velocity), displacement from original location, and contact times with DCs can be imaged and calculated at different times postinfection, changes in any of these parameters that occur upon Ag recognition can be enumerated. Of note, many of the seminal studies of CD8+ T cell priming have used protein or peptide Ag and have not been reconfirmed with bona fide viral infection. However, it is generally assumed that these observations will hold true for antiviral T cell priming. Upon recognition of cognate Ag presented in the context of MHC molecules, naive, motile T cells arrest and transiently and repeatedly contact Ag-bearing APCs (23–25). After these rapid interactions, T cells undergo a period of stable contact and relative immobility, often with many T cells clustered around a single DC, prior to resuming nodal motility. Interestingly, after VACV infection of the LN, we did not observe a period of rapid, transient interactions between T cells and infected DCs; rather, T cells formed stable contacts from the start (16). This is likely due to relatively high numbers of peptide:MHC complexes recognized by TCR transgenic T cells that can be generated on directly infected DCs, which leads to bypass of the initial phase of transient interactions (26). In addition to T cell arrest, a number of other parameters have been analyzed during initial T cell Ag recognition, including T cell calcium flux through calcium-sensitive dyes or via a transgenic calcium reporter animal (27). To more completely understand T cell activation after viral infection, additional viruses will need to be examined using natural routes of infection with a physiological, polyclonal T cell repertoire.
Although only a few viruses have yet been visualized by IVM during T cell activation in the LN, IVM has been used to probe the DC subsets involved in CD8+ T cell activation in the infected LN after s.c. virus injection and after natural viral replication in the tissue. With a relatively low dose of VACV injected s.c., antiviral CD8+ cells are primed at the periphery of the LN by VACV-infected DCs (16, 28). At high doses of the replication-incompetent strain of VACV modified vaccinia Ankara (MVA), T cell priming occurs primarily in the cortical ridge of the LN (29), consistent with our recent results with high-dose VACV infection (22). Kastenmüller and colleagues (30) demonstrated that CD8+ T cell activation after s.c. MVA infection occurs on DC subsets spatially distinct from those needed for CD4+ T cell activation. Later during infection, CD4+ and CD8+ T cells coclustered around Ag-bearing (but uninfected) XCR1 DCs. CD8+ T cells actively recruit these DCs to distinct nodal microenvironments through the secretion of XCL1, thereby optimizing their own priming (30).
In models using a more physiological infection route allowing viral replication in the tissue, nodal Ag delivery can either occur through 1) the direct drainage of free virions, 2) the migration of infected cells to the LN, or 3) the capture of viral Ag at the site of infection by migratory APCs, which then traffic to the LN (31). Currently, only a handful of studies have visualized APCs using IVM during tissue viral replication. After cutaneous infection and skin replication of HSV-1, Hor et al. (32) demonstrated that CD8+ are not activated by early migratory DCs even though CD4+ T cells are. As with s.c. MVA infection, CD8+ T cells are primed by XCR1+ DCs in the LN paracortex, and it is on these DCs that CD8+ T cells receive CD4+ help.
After T cell activation and rapid T cell division within the LN, T cells exit via nodal sinuses and enter the efferent lymphatics (33). After vesicular stomatis virus infection, Benechet et al. (34) visualized the egress of antiviral CD8+ T cells at both cortical and medullary sinuses, which could be impaired through disruption of sphingosine-1-phosphate-receptor-1 signaling. Recently, Lucas et al. showed that alphavirus-induced IFN signaling in the LN coordinates expansion and contraction of the LECs (and, thus, lymphatic sinuses) (35). Thus, although unexplored, viral infection is poised to impact T cell egress from the LN through manipulation of nodal sinus size and/or structure.
Moving past the LN: imaging effector cells at the site of infection
After LN egress, T cells travel in the blood stream to sites of tissue infection, migrating through the blood endothelium into infected tissues. Once there, T cells must find virally infected cells and kill them in processes that are incompletely understood but amenable to analysis via multiphoton IVM. Numerous IVM studies have now visualized various aspects of T cell–mediated control of viral infection; however, these studies vary greatly in the viruses used, tissues examined, and even T cell numbers examined (because of the use of adoptive transfers for imaging).
Secondary lymphoid organs.
As discussed, secondary lymphoid organs (SLOs) are commonly virally infected to image T cell activation. The logical extension of this is visualizing effector function in the same secondary lymphoid organs, as both the number and phenotype of virus-infected cells in these tissues have been previously determined. In a groundbreaking study, Halle et al.(36) used multiphoton microscopy of explanted LNs to visualize CD8+ T cell–mediated killing of murine CMV (MCMV)–infected cells after s.c. infection (in which both nodal hematopoietic cells and nonhematopoietic stromal cells become infected) (37). The authors calculated a per-cell killing rate based on direct visualization; each CD8+ T cell could kill ∼2–16 MCMV-infected cells per day. Interestingly, MVA-infected LN cells were also killed at similar rates, and these were much lower than those estimated previously by in vitro studies. These data suggest that CD8+ T cell killing in vivo is less efficient than previously thought or varies greatly among different tissues (and could be tightly regulated). To our knowledge, this is the sole study to calculate killing effectiveness of antiviral T cells in an organ using direct counting. Many questions thus remain: does this hold true for more viruses in more tissues, and, if so, why are T cells limited in the number of cells they can kill? Likewise, a caveat of systems examining effector function in SLOs, such as the Halle study, is that T cells are often primed first using an unrelated system, and LNs are later infected when effector cells are already present. Therefore, it remains to be determined if kinetics of viral clearance will differ in infected LNs where priming also occurs.
Apart from T cell killing rates, the function of antiviral T cells has been analyzed using IVM in lymphocytic choriomeningitis virus (LCMV)–infected spleens, during which the red and white pulp are both infected. Zinselmeyer et al. (38) IVM imaged chronically infected mice, showing that splenic CD8+ T cell exhaustion in spleens was accompanied by T cell paralysis or immobility, which could be reversed by administration of PD-1 Ab. As acute viral infection can also lead to PD-1–expressing T cells (39), it will be interesting to determine whether this also leads to T cell paralysis or if this is a unique feature of chronic viral infection.
The skin presents a first line of defense to many pathogens, including viruses, providing both an imposing physical barrier and a unique immune environment. As the skin is easily accessible for IVM, numerous groups are currently examining features of antiviral immune responses that unfold in the skin. We have used IVM to image CD8+ T cells responding to epicutaneous VACV infection administered with the same bifurcated needle used for human vaccination (40). After this infection, both keratinocytes and inflammatory monocytes become infected and can be visualized using viral promoter-driven fluorescent protein expression. Intriguingly, CD8+ T cells adeptly killed mobile inflammatory monocytes, as shown by flow cytometry and IVM. Although we were able to visualize T cell–mediated killing of VACV-infected cells in this study, we did not calculate killing rates; however, Halle et al. (36) calculated similarly low rates for the killing of MCMV-infected cells in the skin as in the LN. In our study with VACV infection, we observed that 1) a single T cell appeared capable of killing a virus-infected cell and 2) there were far more T cells than targets in the dermis after viral-induced activation of transferred TCR transgenic T cells. Together, these data suggest that T cells may not be serial killing nor need kill multiple targets in order to control all viral infections in the skin.
Additionally, perhaps not surprisingly, our study demonstrated that not all infected cells are equal in terms of T cell–mediated killing, with infected epidermal keratinocytes refractory to killing (40). Instead of lysis, T cells contributed to infected keratinocyte clearance through the coproduction of IFN-γ and IL-10, which altered the local immune environment to favor clearance (41).
Although the LN represents a vast area for Ag-specific T cells to find cognate Ag–bearing DC, nodal architecture is thought to maximize the likelihood of their encounter. Do similar mechanisms allow activated effector cells to locate virus-infected cells in the large tissue space of the skin? We have previously demonstrated that the chemokines CXCL9/10 are produced in the skin after VACV infection (42). T cell expression of the receptor for these chemokines, CXCR3, increased the ability of effector cells to enter areas heavily populated with infected inflammatory monocytes. Ariotti et al. (43) later showed that during HSV skin infection, CXCR3 expression allowed subtle chemotaxis toward infected cells. Importantly, disturbance of CXCR3 ligands altered T cell distribution, and microinjection of CXCL10 increased the efficiency of CTL accumulation. Thus, with the two viruses so far examined in the skin, CXCR3 is important for directing T cell movement in the tissue for viral clearance. As both studies demonstrated only partial or subtle effects of CXCR3 on finding infected cells, the other factors involved remain to be elucidated.
Historically, the brain has been considered an immune-privileged organ, containing a paucity of peripheral immune cells (44). However, it is becoming increasingly apparent that the CNS is immune specialized rather than privileged. Peripheral immune cells can cross the blood-brain barrier and exert effector functions within the parenchyma that are crucial for control of neurotropic viruses (45). Importantly, immune responses in the brain must be carefully managed to avoid collateral damage to healthy brain tissue, which lacks the regenerative capacity of other peripheral tissues.
Many viruses used for IVM imaging in other tissues can also infect the CNS. Intracranial LCMV infection results in fatal CD8+ T cell–mediated immunopathology, characterized by seizures and vascular leakage (46). Instead of providing protection against LCMV infection, CD8+ T cells produced chemokines that recruited myelomonocytic cells into the CNS and thus indirectly caused fatal blood-brain barrier disruption (47). Similarly, chemokine production following HSV infection and immune cell migration contributed to immunopathology in the CNS (48). Although it remains to be determined if this is true for all viruses, it is clear that CD8+ T cells in the CNS can potentiate pathology.
What is the impact of Ag presentation in the CNS on subsequent effector T cell expansion and function? Although initial T cell priming occurs in the draining LN, effector CD8+ T cells can undergo additional rounds of cell division within the brain after LCMV infection (49). In contrast to intracranially inoculated mice, adoptively transferred therapeutic T cells do not recruit myelomonocytic cells into the CNS of LCMV-infected carrier mice (50). In these carrier animals, microglia (the resident APCs of the CNS) upregulate a transcriptional profile promoting Ag presentation to effector CD8+ T cells, which can purge viral infection. Thus, the impact of effector T cells can vary profoundly in the same tissue infected by different routes.
A number of medically significant viruses infect the respiratory tract, including influenza virus (IAV), respiratory syncytial virus, and rhinovirus. Although the lung represents an important site of human viral infection, it poses several challenges for IVM imaging (51). Respiration-induced movement of the tissue necessitates extensive stabilization (thus, manipulation) for imaging, whereas the air-filled alveoli increase lung volume and thereby decrease tissue imaging depth. Nonetheless, researchers have developed sophisticated methods to image virus-infected lungs in breathing animals. A thoracic window, originally developed to image the pulmonary circulation of dogs, has been successfully applied to the mouse lung (52). The conducting airways of the lungs, however, are still generally too deep to image, leading to the development of methods for tracheal imaging, a model of human conducting airways (53).
Because of the inherent difficulty in imaging the lung, CD8+ T cell responses in this tissue have not been robustly visualized. However, IVM has revealed a number of key features of antiviral T cell biology in the lung. In IAV-infected lungs, CD8+ T cells form stable contacts with lung DCs, although T cell division was not observed at the timepoint examined in the lung as in other infected tissues (54). T cells moved more slowly in the lung prior to viral clearance, suggesting T cell interaction with infected cells. Similarly, after MVA infection, Ag-specific CD8+ T cells moved more slowly than nonspecific T cells in the bronchus-associated lymphoid tissue of lung explants (55). As with IAV infection, MVA-specific T cells formed stable interactions with lung DCs. It will be interesting to systematically characterize the kinetics and phases of T cell contacts with lung DCs during different viral infections to understand if there are common “rules” that guide these interactions and lead to division in the tissue. Likewise, as unrestrained effector activation in the lung can lead to pathology (56), the precise role of CD8+ T cells during various viral infections, including noncytopathic viral clearance, will be important to directly visualize.
After the primary response: memory cells in action
After viral clearance, most of the expanded antiviral effector T cells undergo marked attrition, yet some memory T cells persist in both the SLO and peripheral tissues. In this section, we review IVM studies visualizing antiviral memory T cells in both locales.
Central memory T cells.
Central memory T cells (TCM) circulate between lymphoid organs and can enter the tissue and provide rapid protection after secondary infection. Within the LN, TCM reside in similar locales as naive CD8+ T cells, the central paracortex primarily in the paracortex (57). After LCMV or vesicular stomatis virus infection, TCM rapidly redistribute to LN periphery in a CXCR3-dependent manner. Kastenmüller et al. (29) later showed that this peripheral, CXCR3-mediated TCM redistribution also occurred after MVA vaccination. This peripheral redistribution is likely crucial for the rapid encounter of viruses entering via the lymphatics, although neither study directly visualized infected cell killing by memory CD8+ T cells.
Tissue-resident memory T cells.
Tissue-resident memory T cells (TRM) have only recently been fully appreciated as a crucial defense at barrier sites and are particularly important for the control of viral infection (58). Currently, antiviral TRM have been imaged in both the skin and female reproductive tract (FRT). After cutaneous HSV-1 infection, CD8+ and CD4+ TRM exhibit distinctive migration patterns in the skin; whereas CD4+ TRM circulated between the dermis and the epidermis, CD8+ TRM persisted in the epidermis, largely around the initial site of infection (59–61). Interestingly, epidermal CD8+ TRM exhibited a unique dendritic morphology thought to be imposed by the tight spaces between epidermal keratinocytes (62). Ariotti and colleagues (63) further showed that secondary HSV-1 infection was more effectively controlled if infection occurred at the same site as initial infection, which correlated with the increased concentration of CD8+ TRM that continuously patrol that area. No rate of killing has been visualized for TRM, as these cells likely prevent the establishment of large numbers of infected cells in the tissue they survey.
A unique feature of TRM biology is their ability to regulate their own expansion. Upon secondary challenge of the FRT with LCMV, virus-specific TRM paused and proliferated (64). Whereas TCM can enter the tissue and convert to TRM, daughter cells of the original TRM dominated the antiviral response. Park et al. (65) reported a similar occurrence in HSV-infected skin, with local division of TRM. The preexisting TRM population remained even after new TRM were generated from circulatory precursors. These data explain how the human skin can be populated with such vast numbers of TRM (66) and raise important questions about purging pathogenic T cells during T cell–driven diseases.
Challenges and future questions
Even with the advances made in recent years, a number of challenges remain to the development of a complete understanding of antiviral effector cell biology as it occurs in vivo. Multiphoton microscopic imaging is still only as good as the model systems of viral infection, and some of these currently cannot accurately recapitulate human disease because of tissue amenability for imaging. For instance, many viruses infect the lower FRT and upper airways, and these areas are not easily visualized. Even in the skin, events occurring deep within the dermis are out of range for IVM. Thus, much of our current knowledge is pieced together from the events that can be imaged. Additionally, the use of many different viruses and infection routes relegates most of these elegant studies to unrelated stories, making it difficult to ascertain a consensus behavior for T cells (or lack thereof if T cell function is virus specific). Adding to the complexity, we have only scratched the surface of issues related to viral tropism and dose, viral pathogenesis, and immunomodulation using IVM. We still have much to learn.
As in other biologic fields, IVM has greatly advanced our understanding of antiviral T cell biology in vivo through direct visualization of the key events of the immune response. IVM studies have yielded important insight on viral Ag presentation leading to the activation of CD8+ T cells as well as T cell LN egress, local tissue migration, and killing capabilities. As noted above, many more studies will be needed to comprehensively understand unique and common features of the CD8+ T cell response to viral infection. Unquestionably, however, the combination of IVM and traditional immunologic assays will continue to broaden our understanding of antiviral CD8+ T cell–based immunity.
This work was supported by the Intramural Research Program of the National Institute of Allergy and Infectious Diseases, National Institutes of Health (to C.S.M. and H.D.H.).
Abbreviations used in this article:
female reproductive tract
lymphocytic choriomeningitis virus
lymphoid endothelial cell
modified vaccinia Ankara
secondary lymphoid organ
central memory T cell
tissue-resident memory T cell
The authors have no financial conflicts of interest.