Imaging CD8⁺ T cells during diverse viral infections

Abstract

CD8⁺ T cells play a critical role in host defense against pathogens and tumors. Much of our current knowledge of the activation and subsequent effector activities of CD8⁺ T cells has been gained using ex vivo approaches examining the T cell population en masse for surface phenotype, activation status and the production of effector molecules. Thus, the precise behaviors and diversity of individual CD8⁺ T cells responding to virus infection in vivo have not been extensively explored, leaving many unanswered questions relevant to the rational design of antiviral vaccines and therapeutics. Recently, intravital multiphoton microscopy (MPM) has been used to image CD8⁺ T cell priming after infection with disparate viral pathogens ranging from small RNA viruses encoding few proteins to DNA viruses producing hundreds of viral proteins (many immunomodulatory). After priming, effector CD8⁺ T cells have been visualized in virus-infected tissue, both during primary infection and after transitioning to tissue resident memory cells (T_RM). Here, I highlight recent advances in our understanding of antiviral CD8⁺ T cell responses revealed through intravital MPM.

Introduction

In response to a continual onslaught of both friendly and injurious microbial intruders, vertebrate organisms have evolved a complex defense system consisting of both an immediate, innate immune response and a later, antigen (Ag)-specific adaptive phase. The adaptive immune response is primed in the lymph node (LN) draining the site of infection, which produces an army of activated antiviral effector cells, including cytotoxic T cells (CTLs) and antibody-producing B cells. Many of these cells, in particular effector CD8⁺ T cells, rapidly exit the LN and traffic through the blood to the infected tissue, where they can respond and eliminate virus-infected cells. This review will focus on the activities of antiviral CD8⁺ T cells.

Much of our understanding of the behavior of CD8⁺ T cells during priming comes from groundbreaking studies harnessing the power of intravital multiphoton microscopy (MPM) to examine living LNs. Using non-infectious models (eg. ex vivo-activated antigen (Ag)-bearing dendritic cells (DCs)), these seminal analyses provided clear and convincing evidence that naïve T cells encounter these migratory DCs near the site of T cell nodal entry through the high endothelial venules (HEVs) (1–4). The close proximity of immigrant T cells and DCs has been proposed to limit the area that T cells must traverse for priming, thus maximizing the likelihood that a T cell possessing a rare specificity might encounter an antigen presenting cell (APC) bearing its cognate Ag. T cell activation under these conditions proceeds in 3 distinct steps, with different activation “phases” demarcated by changes in both cellular speed and contact stability with DCs (from fast mobility with rapid contacts, to slow movement with stable contacts, then to fast mobility with little DC engagement) (3). Later, the 3-phase model was refined to reflect differences in T cell behavior endowed by the “strength” of Ag presented by the DC during priming; higher affinity Ag can lead to bypass of the first phase of priming and proceed directly to stable T/DC interactions (5). Thus, even small differences in Ag presentation levels or peptide/TCR affinities exert profound influence on CD8⁺ T cell behavior in vivo.

As in the experiments described above, most early immune imaging studies focused on well-described, non-infectious model systems for a number of reasons. Primarily, these experimental systems can be easily manipulated to make immune “players” amenable to imaging (for example, one can transfer higher DC numbers and use different fluorescent labels to aid multiphoton excitation in vivo). In contrast, live pathogens must be manipulated genetically (for fluorescence expression, peptide production, etc.) to produce satisfactory images. Not all pathogens (such as small viruses) can accommodate the large fluorescent proteins in their genomes, and those that can sometimes delete inserted material or express the inserted gene at low levels, rendering imaging viral-driven fluorescence difficult. Finally, imaging most medically important pathogens requires adapting rather expensive MPM systems to biosafety level 2 (BSL2) containment as well as the additional concerns regarding the transport/use of infected animals. All of these factors have hindered viral imaging using MPM.

While some of the “rules” governing T cell behavior in both lymphoid and peripheral tissues can be extrapolated from non-infectious models, viral infection adds additional layers of complexity (route of infection, host cell tropism, immune evasion) to an already multifarious system. For example, some viruses do not infect DCs and must be primed with Ag gathered from an exogenous source (termed cross-priming). Thus, adoptive transfer systems will fail to recapitulate priming under these conditions. Many viruses that can infect DCs modify the host cell, with some shut-off host-protein translation all together (for a few examples, see: (6–8)). Differences in viral dose as well as the precise viral strain can alter results, particularly if immunomodulatory proteins are deleted. For instance, the priming site of naïve CD8⁺ T cells differs in LNs of mice infected with 10⁶ plaque forming units (pfu) of recombinant vaccinia virus (VACV) versus those infected with 10⁸ pfu of a greatly attenuated, replication incompetent strain of VACV known as modified vaccinia Ankara (MVA) (9, 10).

The robust application of MPM imaging to a substantial number of diverse viruses (over a range of doses and infectious routes) is needed to gain a comprehensive understanding of the behaviors common to all antiviral CD8⁺ T cells rather than those that are virus specific. Likewise, that adoption of some common viral strains and doses (when prudent) by multiple labs should increase the coherence of results obtained using MPM. With the caveat that there is likely great variability in antiviral CD8⁺ T cell behavior during different viral infections, this review will highlight some of the many important advances made using intravital MPM imaging of virus-infected tissues.

Viral Trafficking to and Infection of the Lymph Node

Antiviral CD8⁺ T cell responses initiate in the LNs that drain the sites of tissue infection. Even before T cells are spurred into action, however, viral Ags must reach nodal APCs, either as free viral proteins or along with cells trafficking into the reactive node. Precious little is currently known about the precise mechanisms that enable viral Ag (or infectious virions) to reach the node, and this probably varies considerably among viruses. To date, the nodal delivery of only a few model viruses has been studied using MPM, and only then using subcutaneous injection of relatively high (questionably physiologic) doses of virus.

Despite incomplete understanding of virus drainage, one can trace the general pathway of cells and interstitial molecules (including hypodermic-introduced virions) from peripheral tissue to the LN. Fluid collected from the tissue is routed into and through a complex series of lymphatics (fluid-collecting vessels that maintain the proper fluid volume in the interstitium) that flow into a regional draining LN (DLN). Lymph vessels transport not only fluids, but also proteins and peptides (Ags), viruses and unicellular pathogens, and cells (11). Incoming lymph cargo (including cells) is first deposited into the LN in a large cavity underneath the collagenous capsule, aptly named the subcapsular sinus (SCS). Intriguingly, diaphragms formed by lymph endothelial cells selectively regulate access of lymph-borne proteins, etc. into the conduits that permeate the cortex of the node (12). Very small particulates, soluble proteins and lymph fluid deposited in the SCS can enter the node's parenchyma through these conduits. Larger molecules are excluded from the parenchyma, instead flowing through transitional and cortical sinuses into the even more extensive sinuses of the medulla before exiting through the efferent lymphatics. The filtered post-node contents contained by the efferent lymphatics are eventually deposited into the bloodstream.

Importantly, migratory dendritic cells (DCs) mobilized by infection also utilized the lymphatics to traffic to the DLN. At steady-state, approximately 5% of LN DCs arrive at skin-draining LNs every 24 hours; DC migration is greatly elevated by infection or inflammation (13) (for a review, see (14)). DCs appear to be the primary targets of infection for some viruses, and may be infected in the tissue, traffic into the LN, and potentially shed progeny virions leading to more infection (15). For other viruses, migratory DCs acquire exogenous Ag liberated from infected cells and thus serve as an important source of non-infectious Ag in the node (16). Migratory DCs entering the SCS must navigate and penetrate the sinus' floor (either transcellularly or junctionally) before subsequent intranodal migration for situation in the paracortex (17). LEC diaphragms appear to also permit SCS floor penetration (12).

Lymph node sinuses are not hollow vessels devoid of cells, but are instead densely populated by macrophages. SCS macrophages express CD169 and are found just underneath the floor of the SCS with processes extending upward into the sinus, as well as in transitional interfollicular sinuses (18). Medullary macrophages, heavily distributed in the medullary sinuses, are distinguished by the expression of F4/80 in addition to CD169. Because of their strategic locations in and adjacent to LN sinuses, these macrophages have direct access to incoming lymph fluid that cannot reach the B cell follicles or T cell zone due to size exclusion (12, 19).

So how do viruses reach the DLN? Rapidly. Minutes after subcutaneous (sc.) infection of VACV, we recovered viral DNA (from transported virions) in the DLN (9). Because DC migration occurs on a much slower time frame, these data strongly suggested that viral transport occurred via lymphatic drainage independent from cellular trafficking. By MPM, fluorescently conjugated VACV virions rapidly concentrated on SCS macrophages, and the delivery of viable, fluorescent-protein-encoding VACV or vesicular stomatitis virus (VSV) led to rampant infection of this cell layer. Remarkably, virus-driven fluorescent protein expression was visualized within a few hours of infection (with similar kinetics to in vitro-infected tissue culture cells).

Junt et al. sc. injected vesicular stomatitis virus (VSV), and likewise found marked accumulation of labeled virions on SCS macrophages (20). Based on the proclivity of SCS macrophages to gather virions and filter them from the lymphatics, the Von Andrian laboratory coined the phrase “cellular flypaper” to describe these cells. A testament to the filtering power of SCS macrophages, conditional ablation of macrophages allows VSV (which is not normally lethal to mice) to escape the LN, traffic along the nerves, and eventually kill infected animals (21). Examining a third (and actual murine!) virus, lymphocytic choriomeningitis virus (LCMV), Sung et al. visualized infected cells in both the LN SCS and medulla, but not in the node's interior, consistent with the exclusion of molecules larger than 70 kDa (22). Viral vaccines (eg. non-infectious, killed viruses) also have similar distribution patterns in the node post-injection; however for influenza virus (IAV), medullary DCs preferentially acquired inactivated virus (23, 24). Thus, the lymphatics can efficiently and rapidly deliver both large (VACV) and small (VSV, LCMV) virions into nodal sinuses for acquisition by resident macrophages and DCs.

Many important questions remain regarding virion transfer from the peripheral infection site. First, how does this occur after more physiologically relevant routes of viral delivery, such as via arthropod bite or during replication of the virus in the tissue (rather than injection)? Where and in what form do DCs encounter virus and become infected, and can they make it to the LN once hijacked by a cytolytic virus? Since some viruses (such as VSV) do not require a specific cellular receptor for infection, how are virions able to pass by myriad receptive lymphatic endothelial cells to actually arrive at the LN? A better understanding of these questions would shed much light on early viral dissemination and strategies to prevent it.

Antiviral T Cell Priming

Once viral Ag arrives in secondary lymphoid organs, APCs process and present proteins as short peptides bound by MHC class I (for activating CD8⁺ T cells) and class II molecules (CD4⁺). Non-cell associate Ag can be accessed by DCs extending probing dendrites into the SCS or the conduits of the node (25, 26). Alternatively, DCs can acquire Ag in peripheral tissue and migrate into the LN to present Ag (or even donate that Ag to other APCs for presentation). Most of the MPM studies of naïve T cell priming mimic the last set of conditions, as in-vitro grown DCs are activated using LPS, pulsed with Ag, and then adoptively transferred into a recipient animal where the mature DCs home to the regional LN. Under these conditions, T cell priming occurs in the node's paracortex adjacent to the sites of naïve T cell entry through the HEVs (1, 3). After subcutaneous (sc.) injection of live virus, however, it was soon discovered that naïve antiviral CD8⁺ T cells did not follow the same roadmap for activation.

Node

The first foray into MPM visualization of antiviral CD8⁺ T cells utilized VACV-infected, explanted popliteal nodes (27). In a pioneering study, Norbury et al. imaged nodes after infection with recombinant VACV expressing a model Ag (to activate T cell receptor transgenic, Ag-specific T cells) and GFP (to visualize virus-infected cells). Approximately 6 hours post-infection with VACV, Ag-specific T cells clustered around virus-infected DCs in the popliteal node. Although more macrophages than DCs were infected with VACV, T cells almost exclusively engaged infected DCs, suggesting these cells serve as the primary APC during antiviral responses. Norbury et al. further demonstrated (using ex vivo approaches) that DCs isolated from the node activated CD8⁺ T cells, which was a contentious topic at the time (2002) as both macrophages and DCs possess the cellular machinery necessary to serve as competent APCs. T cell clustering around nodal APCs would later be adopted as a proxy for T cell activation in numerous static and MPM imaging studies.

The Yewdell group later MPM imaged CD8⁺ T cell priming in the inguinal LN after sc. VACV injection in the mouse flank. Within hours of sc. injection of 10⁶ plaque forming units (pfu) of VACV expressing GFP, we visualized numerous GFP⁺ CD169⁺ macrophages just underneath the SCS of the node, with a minority of GFP⁺ DCs admixed (9). Although virus-specific naïve CD8⁺ T cells entered the node through centrally located HEVs, almost all virus-infected cells were located distally from the HEVs near the node's periphery. Unexpectedly, antiviral CD8⁺ T cells moved centripetally to interact with virus-infected cells in an area between B cell follicles that we termed the peripheral interfollicular region (PIR). CD8⁺ T cells clustered with virus-infected cells in the PIR and expressed high levels of the early activation marker CD69 (determined by staining live tissue slices after infection). Blockade of naïve T cell entry into the node showed that the first 24 hours post-infection were essential to prime the anti-vaccinia CD8⁺ T cell response. These data suggested that direct priming of antiviral CD8⁺ T cells in the PIR generates most anti-VACV effectors.

Why do CD8⁺ T cells selectively interact with DCs rather than macrophages in the LN after VACV infection? To address this, we utilized a transgenic mouse model that allowed ablation of LN DCs through administration of diphtheria toxin (CD11c-DTR mice (28)). Although CD8⁺ T cells normally interacted with GFP⁺ CD11c⁺ DCs in these mice, depletion of DCs resulted in long-lived T cell clustering around infected macrophages in the PIR (29). Importantly, DC-depleted LNs could not sustain full T cell activation; although CD8⁺ T cells proliferated, they did not upregulate activation markers and could not perform full effector functions. Together, these data revealed that T cells can interact with peripheral macrophages for priming, but DCs are needed to generate full-fledged antiviral effectors.

Myriad questions regarding CD8⁺ T cell priming after viral infection can potentially be resolved through the application of intravital microscopy. In future studies, it will be important to image DLNs after vector-mediated delivery of live virus or after drainage of tissue-replicating virus. Likewise, visualizing rare interactions between DCs presenting actual viral Ag (not model Ags) and naïve, endogenous T cells (rather than adoptively transferred cells enriched for particular specificities), should shed needed light on the precise events that lead to a robust antiviral T cell priming under physiological circumstances. No doubt more surprises will be uncovered as technologies advance.

Antiviral Effector T Cells in Tissues

Despite the many remaining questions concerning the precise nature of antiviral CD8⁺ T cell priming in the DLN, it is clear that a hallmark feature of activated cells is their ability to leave the LN, traffic to the periphery and attempt to eliminate virus-infected cells. To better understand this critical phase of antiviral CD8⁺ T cell responses, T cell behavior in a number of different virally infected tissues has been visualized by MPM.

Lung

Infections with respiratory viruses, such as influenza A virus (IAV), respiratory syncytial virus (RSV) and even the common cold (rhinovirus), cause significant human morbidity and mortality. Despite much interest, MPM imaging respiratory infections has proven difficult, as the lung presents a moving target located deep within the body (30). Regardless of technical hurdles, several groups have successfully imaged lungs using either specialized stages stabilizing lung tissue or perfusion chambers holding live, explanted lung slices. To date, however, these methods have not been largely applied to respiratory virus infection.

The Förster laboratory has examined lung infection with several different viruses (though no major human respiratory pathogens). Halle and colleagues infected mice intranasally (in.) with the replication-deficient poxvirus modified vaccinia Ankara (MVA), an often-employed vaccine vector (31). MVA infection resulted in the formation of bronchus-associated lymphoid tissue (BALT) that was maintained by lung DCs. Intriguingly, when the authors injected naïve CD8⁺ T cells into animals already possessing MVA-induced BALT, CD8⁺ T cells immigrated into these areas of lymphoid tissue and formed clusters with BALT-resident Ag-bearing DCs. T cell mobility within the BALT was strikingly similar to that of naïve, Ag-specific CD8⁺ T cells moving in LNs that possessed their cognate antigen (3). Using a combination of imaging and ex vivo analyses, Halle et al. demonstrated that T cells present in BALT were dividing, suggesting that BALT is an important locale for T cell priming outside of secondary lymphoid organs.

The Förster laboratory next examined murine cytomegalovirus (MCMV)-infected lungs. Human cytomegalovirus (HCMV, a closely related virus) is a common human pathogen with infection frequently manifest in neonates as pneumonitis (inflammation of the lung tissue) (32). To better understand HCMV disease progression and viral clearance, Stahl and colleagues MPM imaged the lungs of neonatal mice after laryngopharynx (lp.) infection with MCMV. While MCMV disseminated in neonatal mice resulting in distribution of infected cells throughout the body, non-hematopoeitic cells of the lung were highly infected. Intriguingly, Stahl et al. visualized novel accumulations of uninfected cells near infected lung cells, which they termed nodular inflammatory foci. Although virus-specific CD8⁺ T cells could eliminate these nodules in adult mice, this process was markedly less efficient in neonates. Using MPM, the authors showed that virus-specific CD8⁺ T cells contacted myeloid cells in the nodules, priming these T cells independently from LNs. Together, the works of Stahl and Halle raise important questions concerning T cell priming in non-lymphoid tissues, particularly in the lung.

Preliminary imaging of the lung during IAV infection has been performed, although the limited size of the IAV genome has thus far prevented inclusion of fluorescent reporters suitable for marking infected cells (though new methods have recently been developed). Mattheu and colleagues examined lung-resident CD8⁺ T cell behavior after live IAV infection, finding effector T cells frequently associated with lung tissue-resident DCs (33). Though not the focus of this review, CD4⁺ T cell behavior in IAV-infected lung slices has also been examined (34).

Skin

The skin poses a formidable barrier to pathogen infection, yet after an epithelial breach many human viruses replicate in the dermis or epidermis. Recently, we examined the primary CD8⁺ T cell response to acute poxvirus infection of the skin using MPM (35). After epicutaneous (ec.) inoculation of recombinant VACV with the same bifurcated needle as currently employed for human smallpox vaccination, two major populations of skin cells became infected: mobile inflammatory monocytes and sessile, non-haematopoietic keratinocytes. Though effector CD8⁺ T cells easily located both groups of infected cells, they did not effectively kill infected keratinocytes for a number of reasons, including lack of access to compact nests of viral infection. Conversely, virus-infected monocytes were easily targeted and killed by Ag-specific T cells. At relatively defined times post-infection, we observed effector T cells engaging and killing VACV-infected cells.

In this study, we did not directly assess the mode or kinetics of killing employed by antiviral CTL, largely because the entry of T cells into the skin was asynchronous after endogenous priming and migration into the skin. Additionally, the skin, like the lung, represents a sight for T cell priming outside of secondary lymphoid organs, though this has not been shown during viral infection (36). Thus, much remains to be uncovered during viral skin infections.

Central nervous system

Viral infection of the central nervous system (CNS) poses an interesting challenge to the immune response, where viral clearance must be carefully balanced to avoid immune pathology. To understand how this is accomplished in vivo, Kang et al. examined intracerebral (ic.) infection of mice with LCMV (37). After LCMV-induced viral meningitis, CD8⁺ T cells migrate into the CNS, resulting in seizures and mortality in infected animals. Kang et al. demonstrated that the migration of effector T cell occurs in cell cycle, with T cells undergoing an arrest in cellular mobility preceding division in the CNS. These data again suggest that CD8⁺ T cell division is not hard-wired and can occur outside of secondary lymphoid organs.

Although not a viral infection, MPM imaging of protozoan infection may also provide clues to antiviral T cell behavior in the CNS. Harris et al. examined murine brains during Toxoplasma-induced encephalitis, delineating a role for chemokines in CD8⁺ T cell guidance over long distances for interactions with protozoan-infected cells (38). Interestingly, they described the movement of CD8⁺ T cells as modified Levy walks (a behavior previously described in macroscopic ecosystems such as the ocean). This movement, simplistically consisting of long steps followed by shorter, more rapid motion, was shown to maximize T cell localization of Toxoplasma-infected cells. Whether this holds true for viral infection (where localized replication often leads to foci with high numbers of infected cells) or only dispersed protozoan prey remains to be determined.

Spleen

The studies highlighted thus far have all examined T cell behavior during acute virus infection. However, some viruses establish a state of chronic infection characterized by CD8⁺ T cell exhaustion (a functional decline of T cells during which they can no longer exert full effector function (39)). During exhaustion, T cells upregulate the cell surface molecule PD-1, and signaling through this molecule (via its ligands PD-L1 or PD-L2) contributes to maintenance of the exhausted state. Zinselmeyer et al. studied T cell exhaustion and PD-I expression during persistent viral infection in the spleen by intravenously (iv.) infecting mice with LCMV strain Armstrong (acute) or clone 13 (chronic) (40). Using MPM, the authors defined T cell motility paralysis as a previously unknown consequence of exhaustion. Blockade of the PD-1/PD-L1 pathway during LCMV clone 13 infection increased T cell motility and antiviral cytokine production.

Memory CD8⁺ T Cell Responses

After the primary CD8⁺ T cell response wanes, a small number of memory cells survive and possess the capacity to rapidly respond to viral reinfection. A number of laboratories have begun to use MPM to better characterize and understand these important mediators of protection.

Lymph node

Although the LNs have been intensively scrutinized during the priming phase of T cell responses, a few groups have also delved into memory responses occurring in the node. During the anamnestic T cell response to LCMV infection, Sung et al. found LN-resident CD8⁺ memory T cells were first located within the LN T cell zone. Within 8 hours of infection, however, 90% of memory CD8⁺ T cells relocated to the periphery of the LN, where they facilely eliminated virus-infected cells (22). This peripheral positioning was dependent on expression of the chemokine receptor CXCR3. In a later set of experiments, Kastenmuller et al. similarly demonstrated that sc. delivery of the replication-incompetent VACV strain MVA resulted in CXCR3-dependent movement of central memory T cells to the LN's peripheral regions (10). Together, these studies show that antiviral central memory CD8⁺ T cells are poised to quickly respond and relocate to areas of acute virus infection within LNs.

Skin

Within non-lymphoid tissue such as the skin, memory CD8⁺ T cells comprise a scattered, slowly patrolling pool known as tissue-resident memory cells (T_RM) (reviewed in (41, 42)). Perhaps not surprisingly, T_RM occupy areas of the tissue (skin, mucosa) that are often the first site of pathogen re-exposure (42). In the skin, CD8⁺ T_RM elicited by herpesvirus infection and visualized using MPM are restricted to the epidermis, including the shafts of hair follicles (43, 44). Mueller and colleagues demonstrated that this is a specialized feature of memory CD8⁺ T cells, as CD4⁺ T cells responding to the same virus are preferentially positioned in the dermis (43).

Epidermal CD8⁺ T cells moved more slowly than their dermal CD4⁺ counterparts (likely due to their location in the tightly packed epidermis). Likewise, epidermal-positioning resulted in the adoption of a DC-like morphology by T cells, with numerous motile, probing processes. Ariotti et al. showed that the migratory behavior of T_RM allowed contact with numerous surrounding cells and the rapid recognition of Ag-expressing cells in vivo, thus potentially maximizing the detection of rare targets (44). Intriguingly, T_RM can provide cross-pathogen protection through the broad activation innate and non-antigen specific adaptive immunity in the tissue (42, 45). Future MPM studies are needed to determine the precise localization and behavior of antiviral T_RM during this process.

Concluding Remarks

Both old foes and newly emerging viral pathogens present a continuing threat to man. While we have substantially progressed in our understanding of CD8⁺ T cell-based control of virus infection, much remains to be discovered. Imaging viral infection and the ensuing immune response as they occur in vivo will provide critical insight into the fundamental processes that shape antiviral immunity. At the most basic level, understanding how viruses manipulate the host and subvert immunity is fascinating. At the applied level, correctly understanding both the priming and effector phases of CD8⁺ T cell responses could greatly inform rational vaccine design and the development of antiviral therapeutics.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Funding

This was supported by the Intramural Research Program of the National Institute of Allergy and Infectious Diseases, National Institutes of Health.