Distinct APC Subtypes Drive Spatially Segregated CD4+ and CD8+ T-Cell Effector Activity during Skin Infection with HSV-1

Efficient infection control requires potent T-cell responses at sites of pathogen replication. However, the regulation of T-cell effector function in situ remains poorly understood. Here, we show key differences in the regulation of effector activity between CD4+ and CD8+ T-cells during skin infection with HSV-1. IFN-γ-producing CD4+ T cells disseminated widely throughout the skin and draining lymph nodes (LN), clearly exceeding the epithelial distribution of infectious virus. By contrast, IFN-γ-producing CD8+ T cells were only found within the infected epidermal layer of the skin and associated hair follicles. Mechanistically, while various subsets of lymphoid- and skin-derived dendritic cells (DC) elicited IFN-γ production by CD4+ T cells, CD8+ T cells responded exclusively to infected epidermal cells directly presenting viral antigen. Notably, uninfected cross-presenting DCs from both skin and LNs failed to trigger IFN-γ production by CD8+ T-cells. Thus, we describe a previously unappreciated complexity in the regulation of CD4+ and CD8+ T-cell effector activity that is subset-specific, microanatomically distinct and involves largely non-overlapping types of antigen-presenting cells (APC).


Introduction
Infection results in the priming of pathogen-specific T-cell responses in LNs draining the site of infection. Depending on the nature of the pathogen, this critical step in generating adaptive immunity involves the interaction of naive T cells with various types of migrating and LN-resident DCs [1,2]. During skin infection with herpes simplex virus (HSV)-1, LN-resident CD8a + DCs and skin-derived CD103 + DCs can activate naïve CD8 + Tcells, presumably through the cross-presentation pathway involving the acquisition of noninfectious antigen [1][2][3][4]. By contrast, all subsets of skin-derived migratory DCs, including epidermal Langerhans cells, dermal CD11b + and dermal CD103 + DCs, in addition to LN-resident CD8a + DCs acquire the ability to stimulate naive HSV-specific CD4 + T cells [1,2,4]. Following appropriate activation by DCs, T cells undergo a program of clonal expansion, which is accompanied by the acquisition of effector functions and the induction of migration molecules that facilitate their infiltration of infected tissues.
While CD4 + helper T cells support the generation of antibody and CD8 + T-cell responses in lymphoid tissues, both CD4 + and CD8 + T-cells also contribute directly to pathogen control at sites of infection [5,6]. The latter is achieved through two principle effector functions: the contact-dependent elimination of infected tissue cells and the local production of inflammatory and antimicrobial cytokines [5,6]. The extent to which these T-cell activities contribute to immunity depends on the nature of the infection. For instance, control of non-cytopathic viruses, such as lymphocytic choriomeningitis virus, strictly requires cytolytic Tcell activity [7]. By contrast, immunity against cytolytic viruses, such as vaccinia and vesicular stomatitis virus, does not rely on target cell elimination by T cells [8]. Instead, under circumstances where infection will ultimately result in lytic cell death regardless of T-cell killing, pathogen containment and clearance is dependent on the production of cytokines by effector CD4 + and CD8 + T cells [9][10][11]. Together these diverse effector T-cell (T EFF ) activities are essential for efficient immune protection, however, they may also cause the destruction of uninfected tissues, as seen in the context of immunopathology, autoimmunity or transplant rejection. Therefore, a detailed understanding of T-cell-mediated immunity in peripheral tissues forms an essential basis for therapeutic interventions to modulate T-cell responses against both harmful and innocuous antigens. Nevertheless, the cellular and molecular mechanisms controlling T-cell effector activities in nonlymphoid organs remain poorly defined [2,10].
At its simplest, T-cell effector functions are regulated by T-cell receptor (TCR) stimulation through peptide-MHC complexes on APCs. Importantly in this respect, disengagement of the TCR from antigen-MHC complexes results in the immediate cessation of T-cell cytokine production [10,12]. This ''on-off cycling'' of effector activity provides a sophisticated level of antigen specificity and places important temporal and spatial constraints on T EFF -cell responses [10]. As a consequence, effector T cells circulating through the blood or uninfected tissues are thought to shutdown cytokine production and to regain this effector function only upon reencounter with antigen in infected tissues [10]. In addition, noncognate signals delivered through inflammatory mediators and costimulatory molecules, such as interleukin (IL)-18, IL-12, type I IFNs or CD80 and CD86, may also trigger or further modulate Tcell cytokine production and cytotoxic activity [13][14][15][16]. Thus, the presence of appropriate APCs providing antigen stimulation together with accessory signals is critical in regulating T-cell immunity in situ and targeting effector activities to pathogencontaining tissues [2,17]. Indeed, various types of professional and nonprofessional APCs, including monocyte-derived inflammatory DCs, B cells, neutrophils and parenchymal cells, have been suggested to elicit T-cell effector functions within nonlymphoid tissues [15,[18][19][20][21]. Key aspects in this regulation, however, particularly those pertaining to the infection status of APCs and the role of distinct APC subtypes in driving CD4 + versus CD8 + T EFF -cell responses, remain poorly understood.
Here, we define the cellular interactions that control T EFF -cell activity during the course of skin infection with HSV-1. We focus our analysis on the production of IFN-c, a central component of adaptive immune responses. IFN-cexerts proinflammatory and regulatory effects on a variety of target cells, including the stimulation of antimicrobial activity and the induction of MHC molecules and inflammatory chemokines [5,6]. Protection from HSV infection strictly requires T EFF -cell activities, with both CD4 + and CD8 + T cells contributing to virus control in skin, mucosa and sensory ganglia [22][23][24][25][26]. Moreover, efficient immunity against HSV infection requires IFN-c [27] and, interestingly, it has been proposed that production of this cytokine rather than cytolytic activity is the major CD8 + T-cell mechanism for virus control in neuronal tissues [28,29] as well as during lytic infection in genital mucosa [26]. The shared role of IFN-c as a key effector molecule produced by both CD4 + and CD8 + T cells allowed us to directly compare the regulation of these T EFF -cell subsets side by side. We further took advantage of the tropism of HSV-1 for epithelial tissues [30] to document a distinct anatomical distribution of IFNproducing T EFF -cell subsets in relation to the presence or absence of infectious virus in different microanatomical compartments. Importantly, this unexpected spatial segregation of T EFF -cell effector activity was a direct result of the involvement of largely non-overlapping subsets of professional and nonprofessional APCs in driving CD4 + and CD8 + T EFF -cell responses.

Distribution of IFN-c-producing T EFF cells
To determine population kinetics and cytokine production by T EFF cells in lymphoid and peripheral tissues, we utilized a skin infection with HSV-1 in combination with adoptive transfer of TCR-transgenic T cells specific for determinants derived from the HSV glycoproteins gB (CD8 + gBT-I cells) [31] and gD (CD4 + gDT-II cells) [4], respectively. Consistent with the tropism of HSV for epithelial tissues, immunofluorescence microscopy (IFM) of skin revealed that infection was largely confined to the epidermal layer and hair follicle epithelium (Fig. S1A). Separation of epidermal and dermal tissue (Fig. S1B,C) revealed that HSVspecific T EFF cells began to infiltrate infected skin around 5 days post-infection, albeit with fewer cells in the smaller epidermal compartment (Fig. 1A). T EFF -cell numbers peaked around 8 days after inoculation and declined thereafter (Fig. 1A).
To analyze the production of IFN-c by T EFF cells in situ, we adopted protocols that facilitate intracellular cytokine staining following exposure to the Golgi inhibitor brefeldin A (BFA), either in vivo after intravenous injection [15,32] and/or ex vivo immediately after tissue harvest and during enzymatic digestion [21]. Of note, in order to focus our analysis on IFN-c production in situ, neither of these approaches involved overt restimulation with high concentrations of peptide antigen ex vivo. A considerable portion of gBT-I and gDT-II T EFF cells in the epidermis and dermis of infected skin produced IFN-c 5-6 days post-infection ( Fig. 1B-D). Concomitant with clearance of infectious virus from skin [25], IFN-c production by T EFF cells ceased around day 7, with virtually no IFN-c + T EFF cells present 8 days post-infection (Fig. 1D). Similar kinetics of IFN-c production were also observed for endogenous CD8 + and CD4 + T EFF cells (Fig. S2A,B). While roughly equal portions of gDT-II cells produced IFN-c in the epidermal and dermal layers of skin 5 days after infection, the fraction of IFN-c + gBT-I cells was approximately 3-fold higher in the epidermis as compared to the dermis (Fig. 1C,D). Note, that dermal preparations contained hair follicles of epithelial origin and therefore also harbored some replicating virus ( Fig. S1A-C). The broader distribution of IFN-c + gDT-II cells in the skin also extended to lymphoid tissues, with skin-draining axillary and brachial LNs, but not spleen, containing an appreciable fraction of IFN-c-producing gDT-II cells (Fig. 1E,F). By contrast, IFN-c + gBT-I cells were virtually absent from all lymphoid tissues. Together, these results suggested a distinct anatomical distribution of IFN-c-producing CD4 + and CD8 + T EFF cells in both peripheral and lymphoid tissues.

Confinement of IFN-c + CD8 + T EFF cells to skin epithelium
To gain further insight into the microanatomical localization of IFN-c-producing T EFF -cell subsets, we obtained skin tissue for IFM analysis. Staining of skin sections with anti-IFN-c antibodies

Author Summary
HSV-1 is a widely distributed pathogen causing a life-long latent infection associated with periodic bouts of reactivation and severe clinical complications. Adaptive immune responses encompassing CD4 + and CD8 + T-cell activities are key to both the clearance of infectious virus and the control of latent infection. However, precisely how such Tcell responses are regulated, particularly within acutely infected peripheral tissues, remains poorly understood. Using a mouse model of HSV-1 skin infection, we describe a complex regulation of T-cell responses at the site of acute infection. These responses were subset-specific and anatomically distinct, with CD4 + and CD8 + T-cell activities being directed to distinct anatomical compartments within the skin. While IFN-c-producing CD4 + T cells were broadly distributed, including skin regions a considerable distance away from infected cells, CD8 + T-cell activity was strictly confined to directly infected epithelial compartments. This unexpected spatial segregation was a direct consequence of the involvement of largely non-overlapping types of antigen-presenting cells in driving CD4 + and CD8 + T-cell effector activity. Our results provide novel insights into the cellular regulation of T-cell immunity within peripheral tissues and have the potential to guide the development of T-cell subset-specific approaches for therapeutic and prophylactic intervention in antimicrobial immunity and autoimmunity.
confirmed the presence of IFN-c-producing cells during the acute phase of infection ( Fig. 2A,B), with both endogenous CD4 + and CD8 + (Fig. S2C,D), as well as transgenic gBT-I and gDT-II T EFF cells (Fig. 2C,D) contributing to this response. Interestingly, although gBT-I T EFF cells were broadly distributed throughout the skin, IFN-c + gBT-I cells were strictly confined to the epidermis and hair follicle epithelium (Fig. 2C). By contrast, the majority of IFN-c + gDT-II cells localized to the dermal layer, where they were found either in association with hair follicles or in considerable distance to the epithelium (Fig. 2D). Thus, in contrast to the strict confinement of IFN-c + CD8 + T EFF cells to the epithelium, the dermal layer was the predominant site of the CD4 + T-cell IFN-c response.

Antigen requirements for IFN-c production by T EFF cells
The kinetics of IFN-c production by T EFF cells suggested that the presence of infectious virus was likely to play a role in the induction of cytokine production. This was indeed the case, as gBT-I and gDT-II T EFF cells primed by HSV infection did not produce IFN-c in non-specifically inflamed skin after treatment with 1-fluoro-2,4-dinitrobenzene (DNFB) (Figs. 3A and S3A). Likewise, in vitro activated gBT-I T EFF cells transferred into HSV-infected mice lacking H-2K b molecules (H-2K b2/2 ) did not produce significant amounts of IFN-c in the skin (Figs. 3B,C and S3B). Transfer of activated T EFF cells was necessary as H-2K b2/2 mice cannot prime naïve gBT-I cells due to lack of the relevant MHC-I restriction element. IFN-c production by transferred gBT-I T EFF cells was completely restored in similar experiments following HSV-1 infection of bone marrow chimeric mice in which H-2K b molecules were expressed exclusively in radioresistant cells, but were absent from the radiosensitive hematopoietic compartment (Figs. 3D and S3C). By contrast, we observed a significant reduction in the frequency of in vivo primed IFN-c + gBT-I cells in chimeric mice in which H-2K b molecules were selectively missing from radioresistant cells, when compared with fully MHC-I-sufficient control chimeras (Figs. 3E and S3D). Interestingly, the overall frequencies of IFN-c + gBT-I T EFF cells appeared to be increased in this particular experimental set-up, possibly related to altered immune activation thresholds in previously irradiated recipient mice. Regardless, these results indicated that presentation of viral antigens by radioresistant epithelial cells, such as keratinocytes, Langerhans cells [33] and dendritic epidermal T cells (DETC) [34], was necessary and sufficient for optimal IFN-c responses by gBT-I T EFF cells. IFN-c production by gDT-II cells was also a consequence of antigen recognition, as in vitro activated gDT-II T EFF cells transferred into infected MHC-II-deficient mice (I-A/E 2/2 ) failed to produce IFN-c (Figs. 3F,G and S3E). Transfer of activated gDT-II cells was necessary to overcome the inability of I-A/E 2/2 mice to support CD4 + T-cell priming. By contrast, gDT-II T EFF cells primed in vivo produced IFN-c in chimeric mice in which only radiosensitive, but not radioresistant cells expressed MHC-II molecules, although we observed a moderate, yet significant reduction in the frequency of IFN-c + gDT-II cells in this situation (Figs. 3H and S3F). These results implied that bone marrowderived MHC-II + APCs were largely responsible for presentation of viral antigens and eliciting cytokine production by CD4 + T EFF cells.

DCs trigger IFN-c production by CD4 + T EFF cells
Consistent with an involvement of MHC-II-expressing APCs in driving CD4 + T EFF -cell responses, MHC-II hi cells accumulated in the skin during the first week post-infection [35] (Fig. S4A,B). The majority of these cells had a CD11c int CD11b + phenotype and further expressed CD64 and MAR-1, identifying them as monocyte-derived inflammatory DCs [36,37] (Fig. S4C). IFM of infected skin revealed a broad distribution of MHC-II + cells, with the vast majority localizing to the dermal layer, where they were found in close proximity to IFN-c + gDT-II cells (Fig. 4A). Indeed, partial depletion of CD11c + DCs upon diphtheria toxin (DT) treatment of CD11c.DTR mice resulted in a significant reduction in the frequency of IFN-c + gDT-II cells in infected skin (Figs. 4B,C and S5A). Furthermore, treatment of mice with antibodies blocking the costimulatory molecules CD80 and CD86, typically expressed by professional APCs such as DCs, also abrogated the CD4 + T EFF -cell IFN-c in skin and draining LNs (Figs. 4D and S5B). In stark contrast, costimulation blockade had no bearing on IFN-c production by gBT-I T EFF cells (Fig. 4D). These experiments suggested that MHC-II + DCs were the main drivers of cytokine production by CD4 + T EFF cells. Nevertheless, in Ccr2 2/2 mice, an absence of monocyte-derived DCs [20,38], which numerically dominated the cutaneous DC network during acute infection, rather increased than decreased the frequency of IFN-c + gDT-II and gBT-I cells (Fig. S5C,D), potentially related to impaired virus control in these mice [20]. Furthermore, we observed normal IFN-c production by gDT-II and gBT-I T EFF cells in the absence of Langerhans cells, CD103 + dermal DC and CD8 + LN-resident DCs upon DT treatment of Langerin.DTR mice [39,40], and similarly, also in B-cell-deficient mMT mice [41] (Fig. S5E-G). These results suggested a level of redundancy regarding the involvement of different types of professional APCs in regulating CD4 + T EFF activities in infected skin.

Multiple DC subsets activate CD4 + T EFF cells
To more directly establish a role for DCs in CD4 + T-cell responses, we utilized an ex vivo stimulation assay in which APCs purified from infected mice were cocultured with in vitro generated T EFF cells. Initial experiments revealed that maximal IFN-c production occurred 18 hours after antigen-dependent restimulation for gDT-II and 5 hours for gBT-I T EFF cells (Fig.  S6A,B). We purified CD11c hi DCs from HSV-infected skin and divided them into CD11b hi and CD11b lo subsets (Fig. 5A), with the former expected to contain monocyte-derived and dermal DCs and the latter expected to contain Langerhans cells and CD103 + dermal DCs [4,37,40]. Notably, both subsets induced robust IFNc production by gDT-II T EFF cells, whereas monocytes (CD11b + CD11c 2 Ly6C hi ) and neutrophils (CD11b + CD11c 2 Ly6-C int ) failed to do so (Fig. 5B,C). Unexpectedly, none of these APCs induced IFN-c production by gBT-I T EFF cells (Fig. 5B,C). This was not related to potentially compromised expression of H-2K b molecules after APC isolation, since all subtypes triggered IFN-c production by gBT-I T EFF cells when pulsed with high doses of gB-peptide prior to cell sorting (Fig. S6C,D). In separate experiments, we specifically sorted CD103 + dermal DCs, as this DC subset is capable of cross-presenting viral antigens to CD8 + T cells during skin infection [4]. Once again, while both CD103 + and CD103 2 CD11c hi DCs triggered IFN-c production by gDT-II T EFF cells, neither of the two subsets activated gBT-I T EFF cells (Fig. 5D). Remarkably, this disparate response also extended to DCs isolated from LNs draining the site of infection with CD103 + , CD11b + , CD8a + and Langerhans cell-containing CD103 2 CD11b 2 CD8a 2 subsets inducing IFN-c production by gDT-II, but not gBT-I T EFF cells (Fig. 5E). Of note, gBT-I T EFF cell unresponsiveness towards DC stimulation was observed irrespective of the culture period for 5-18 hours (not shown). Together, these results demonstrated that various skin-derived and LN-resident DC subsets acquired and presented viral antigen for the activation of CD4 + T EFF cells. By contrast, none of these DCs elicited IFN-c-production by CD8 + T EFF cells, even though at least some of them, namely the CD8a + and CD103 + subsets, have the ability to present viral antigens for the activation of naïve CD8 + T cells [3,4].

Epithelial APCs activate CD8 + T EFF cells
Given that IFN-c + gBT-I cells were found exclusively in skin epithelium (Figs. 1B-D and 2B), we reasoned that this compartment contained APCs capable of stimulating CD8 + T EFF cells. Therefore, we purified CD45.2 + hematopoetic and CD45.2 2 parenchymal and stromal cells from epidermal sheets of infected mice by cell sorting and tested their ability to activate CD8 + T EFF cells. Note that here we used expression of CD45.2 to distinguish between hematopoeitc and non-hematopoeitc epidermal cells, whereas in other analyses we used this molecule as a marker for CD45.1 + CD45.2 + gBT-I and gDT-II cells. Both fractions induced IFN-c production by gBT-I T EFF cells, although the keratinocytecontaining CD45.2 2 subset appeared to be slightly more potent in this regard (Fig. 6A,B). Induction of IFN-c production was antigen-specific, since co-cultured OT-I T EFF cells of an irrelevant specificity did not respond to either of the APC subsets (Fig.  S7A,B). Furthermore, stimulation of IFN-c production by CD45.2 2 epidermal APCs was specific for CD8 + T EFF cells since these APCs failed to activate gDT-II T EFF cells (Figs. 6B and S7C). In line with this, the vast majority of CD45.2 2 epidermal cells from infected skin lacked expression of MHC-II molecules (Fig. S7D). To better define the nature of epidermal APCs capable of stimulating CD8 + T EFF cells, we sorted these cells into keratinocytes (CD45.2 2 EpCAM + ), DCs (CD45.2 + CD11c hi ), DETCs (CD45.2 + Vc3 + ), as well as residual CD45.2 + CD11c 2 cells. As expected, keratinocytes induced moderate levels of IFN-c production by gBT-I T EFF cells, and so did epidermal DCs (Fig. 6C,D), in contrast to their counterparts isolated from total skin preparations and LNs (Fig. 5C-E). While residual CD45.2 + epithelial cells had only a weak stimulatory capacity, remarkably, DETCs were by far the most potent APCs triggering IFN-c production by gBT-I T EFF cells (Fig. 6C,D).

Only infected APCs activate CD8 + T EFF cells
Given that the epidermis was the predominant site of viral replication in vivo, we hypothesized that the stimulatory capacity of epidermal APCs resulted from their direct infection. In agreement, intravital two-photon microscopy of skin infected with a cyan fluorescent protein (CFP)-expressing HSV-1 strain revealed that slow-moving gBT-I T EFF cells were swarming around virally infected epidermal cells (Movie S1). By contrast, gBT-I T EFF cells more distal to infection foci displayed significantly higher mean velocities (Fig. 7A,B). Importantly, using IFM, we observed that IFN-c + gBT-I cells co-localized with HSV-infected cells in the epidermis and hair follicle epithelium (Fig. 7C). To further identify infected cells, we inoculated mice with a recombinant strain of HSV-1 expressing green fluorescent protein (HSV.GFP) and analyzed epidermal cells 5 days post-infection using flow cytometry. We observed small numbers of GFP-expressing cells amongst various populations of epidermal cells, including keratinocytes (CD45.2 2 EpCAM + ), DETCs (CD45.2 + Vc3 + ), Langerhans cells (CD45.2 + EpCAM + MHC-II hi ), other DCs (CD45.2 + Ep-CAM 2 MHC-II hi ), as well as undefined CD45.2 + MHC-II 2 cells (Fig. 7D). As expected, GFP + cells were absent after infection with the wild-type HSV-1 KOS strain. Next, we sorted GFP + and GFP 2 DETCs, MHC-II hi DCs and keratinocytes from epidermal sheets (Fig. 7E) and tested their ability to stimulate T EFF cells ex vivo. Strikingly, all GFP + , presumably infected, APCs were able to trigger IFN-c production by gBT-I T EFF cells (Fig. 7F,H). In contrast, IFN-c production was not elicited by their GFP 2 counterparts. Finally, DCs, but not DETCs, elicited IFN-c production by gDT-II T EFF cells, irrespective of their GFPexpression status (Fig. 7G,H). Together, these results indicated that various types of epidermal APCs induced cytokine production by CD8 + T EFF cells. Importantly, direct viral infection was a strict requirement for their stimulatory capacity. Overall, our data highlight a previously unappreciated complexity in the regulation of T-cell effector activity that was subset-specific, microanatomically distinct and involved largely non-overlapping subsets of professional and nonprofessional APCs for CD4 + and CD8 + T-cell responses (Fig. S8).

Discussion
Our results highlight a stringent and complex regulation of T EFF -cell responses that targets effector activities strictly to the site of infection and related lymphoid tissues. Thus, IFN-c production was limited to the time of acute infection and occurred in an antigen-dependent fashion, requiring in situ restimulation via peptide-MHC complexes on bone marrow-derived professional APCs for CD4 + T EFF cells and on directly infected tissue cells for CD8 + T EFF cells. Consistent with a critical involvement of DCs in peripheral CD4 + T EFF -cell responses, we observed a pronounced accumulation of monocyte-derived inflammatory DCs in infected skin. Such DCs are thought to exert multiple functions, including the local production of inflammatory mediators [42], trafficking of antigen to lymph nodes [43] and replenishment of peripheral DC populations following the resolution of infection [35,44]. In addition, our experiments employing genetic approaches, ex vivo stimulation assays and costimulation blockade provide compelling evidence that DCs in inflamed skin play a key role in stimulating CD4 + T-cell effector activity. These results reinforce the concept that DCs regulate various aspects of peripheral T-cell responses [2,15,20,21,45,46]. In fact, the presence of DCs appeared to be essential for CD4 + T EFF -cell responses as non-professional APCs, such as keratinocytes and DETCs, largely lacked MHC-II expression and failed to elicit IFN-c production in ex vivo assays. It is possible that certain APC subsets may dominate the regulation of peripheral CD4 + T EFF -cell activities, as suggested for CD11c hi CD11b hi dermal DCs after skin injection of model antigens [21] or CCR2-depedent monocyte-derived inflammatory DCs during mucosal HSV-2 infection [20]. Nevertheless, our study revealed a considerable degree of redundancy in this regard, with various DC populations from skin and LNs displaying strong stimulatory capacities for CD4 + T EFF cells. These results imply that all DC subsets, relative to their abundance in infected skin, contribute to CD4 + T EFF activation in vivo. Consistent with this, we observed normal IFN-c + production in mice deficient in specific APC subsets, such as monocyte-derived DCs, Langerhans cells, CD103 + dermal DCs or B cells. According to our analysis, monocyte-derived inflammatory DCs are by far the most abundant DC subtype in HSV-infected skin [35] and therefore, may be the major drivers of peripheral CD4 + T EFF -cell responses during HSV-1 skin infection. Nevertheless, our results demonstrate that they may not be essential in this regard as other DC subsets may compensate for their absence.
The IFN-c response by CD4 + T cells occurred in both draining LNs and the epithelial and dermal layers of infected skin, including regions a considerable distance away from infection foci in the epithelium. This remarkably broad distribution echoes the diverse functions of CD4 + T EFF cells in infection control, ranging from the initiation of antibody class-switching in LNs to the regulation of inflammatory cell infiltration and activity as well as direct antimicrobial effects within infected tissues [5]. Interestingly in this respect, IFN-c can exert long-range effects on target cells located as far as 80 mm from CD4 + T EFF -cell-APC conjugates, as recently shown for skin infection with Leishmania major [47]. Thus, DC-mediated CD4 + T EFF -cell activation in the dermis may be an essential component of the host defense that restricts infection to the skin epithelium and limits its spread after virus reemergence in sensory nerve endings. The importance of the CD4 + T EFF -cell response is further illustrated by the lack of CD8 + T EFF -cell IFN-c production in the dermis, as shown here. Supporting this notion, CD4 + T EFF -cell responses are thought to dominate the clearance of HSV-1 from the skin [22,23], most likely via antibody-independent functions such as direct inflammatory and antiviral activities [5,48,49].
In striking contrast to the stimulation requirements for CD4 + T EFF cells, we identified nonprofessional APCs, such as keratinocytes and DETCs, as the main drivers of IFN-c production by CD8 + T EFF cells. According to our histological analysis, keratinocytes are the most abundant cell type in infected epidermis suggesting that they may be largely responsible for activating CD8 + T EFF cells. In addition, various types of inflammatory cells that infiltrate the epithelial layer during infection may contribute to this response. Importantly, both keratinocytes and DETCs are highly susceptible to direct infection by HSV-1 in vivo [30,50]. Indeed, their stimulatory capacity, and surprisingly also that of epidermal DCs, was strictly dependent on direct infection. DETCs are invariant cd-T cells that form a dense network in the epidermis of mice and have been implicated in both innate and adaptive immune responses [51]. Interestingly, it has been speculated that human cd-T cells may act as professional APCs capable of stimulating naïve CD4 + T cells [52], although we did not find evidence supporting a similar role for DETCs, as they failed to activate CD4 + T EFF cells. Nevertheless, their contribution to CD8 + T-cell responses may be particularly relevant at lesion borders where high numbers of infected DETCs [50] could elicit strong IFN-c responses required to curb the lateral spread of infection. In addition to DETCs, other epidermal-infiltrating T cells may also act as potent APCs for local CD8 + T EFF cells upon infection with virus [53]. A role for infected T cells and DCs in triggering local CD8 + T EFF -cell activity is further supported by our observation that chimeric mice, in which only radiosensitive APCs could activate CD8 + T cells, had a residual IFN-c response.
Despite the fact that all DC subsets could activate CD4 + T EFFcells, only HSV-infected epidermal DCs were capable of activating CD8 + T EFF cell to produce IFN-c. Strikingly, uninfected DCs from the same location failed to stimulate CD8 + T EFF cells, highlighting the importance of direct infection in determining the outcome of CD8 + T EFF -cell-DC interactions. Previous studies have suggested that various types of professional and nonprofessional APCs, including inflammatory DCs and neutrophils, can trigger IFN-c production by CD8 + T EFF cells during pulmonary infection with influenza virus [15,18]. Given the ability of influenza virus to infect a broad range of target cells in addition to its primary tropism for lung epithelial cells [17], it is tempting to speculate that in this situation, direct infection may also be required for CD8 + Tcell activation. Supporting this notion, a large number of inflammatory DCs and neutrophils from influenza virus-infected lungs express viral antigens, most likely as a consequence of direct infection [15,18], and infected neutrophils display a far superior ability to elicit CD8 + T EFF -cell cytokine production than their uninfected counterparts [18].
One of the more surprising findings from our study was that uninfected DCs failed to elicit IFN-c production by CD8 + T EFF cells, even though they had access to viral antigen and efficiently activated CD4 + T EFF cells. We have previously shown that LN DCs are able to activate naïve CD8 + T cells at various stages during skin infection, unequivocally demonstrating that they present viral antigens in the context of MHC-I molecules [3,4,54]. The CD8a + and CD103 + DC subsets are of particular interest in this regard, because those DCs cross-present viral antigens in draining LNs 5 days after HSV-1 infection [4], corresponding to the time point analyzed here. Despite this, CD103 + DCs from skin and LNs failed to trigger IFN-c production by CD8 + T EFF cells. This finding parallels observations with CD8 + T-cell responses to migratory DCs after influenza virus infection, where naïve but not memory T cells proliferate in response to antigen presented on migratory DCs [55].
Although not directly addressed in our study, it is tempting to speculate that the inability of uninfected DCs to activate CD8 + T EFF cells may be a means to prevent their elimination by triggering cytotoxic effector functions. While T-cell killing of DCs has been proposed to represent a negative feedback regulation on T-cell priming [56,57], it should be noted that prolonged antigen presentation is a common feature in a variety of infections [58][59][60][61]. Therefore, CD8 + T-cell-mediated elimination of DCs in vivo may be inefficient at best, which is also consistent with our preliminary data demonstrating the failure of gBT-I T EFF cells to lyse CD103 + DCs from LN of HSV infected mice during shortterm co-culture. The DC-dependent production of IFN-c by CD4 + T EFF cells observed in our study further supports this assumption. Of note in this respect, minute amounts of surface antigen can trigger T-cell cytotoxicity [62,63]. Furthermore, crosspresenting nonprofessional APC, such as liver sinusoidal endothelial cells, can drive CD8 + T-cell TNFa production during viral hepatitis [64]. Thus, it appears unlikely that quantitative differences in antigen presentation between directly infected and cross-presenting DCs alone can explain the CD8 + T EFF -cell unresponsiveness described here. In the light of this, and given that  infected epidermal DCs were indeed able to trigger CD8 + T EFFcell IFN-c production, we speculate that direct infection may alter the functional status of DCs, for instance through interference with putative inhibitory pathways [65], to allow for the activation of CD8 + T EFF cells. Such modulation of DC stimulation thresholds may be particularly relevant when low levels of peptide-MHC-I complexes are presented to CD8 + T EFF cells. Further studies will be required to elucidate the precise molecular mechanisms operating in DCs and/or T cells to prevent the activation of CD8 + T EFF cells by uninfected cross-presenting DCs.
Overall, the stringent temporal, cellular and molecular constraints on T EFF -cell responses identified in our study are likely in place to prevent collateral damage and autoimmune inflammation initiated by T EFF -cell activation in tissues not involved in infection. Our results are compatible with a scenario where CD8 + T-cell responses are strictly focused on infected tissue compartments, whereas CD4 + responses may induce a more regional state of antimicrobial protection in tissues surrounding infection foci. Having identified dramatically distinct requirements for CD4 + and CD8 + T-cell effector activity, our study has provided novel insights into the regulation of cellular immune responses in nonlymphoid tissues. Such knowledge has the potential to guide the development of T-cell subset-specific approaches for therapeutic and prophylactic intervention in antimicrobial immunity and autoimmunity.

Ethics statement
All experiments were done according to Australian NHMRC guidelines contained within the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes and under approvals ID1112038 and ID1112345 from the University of Melbourne Animal Ethics Committee. Viruses HSV-1 KOS, KOS pCMV/EGFP Cre (HSV.GFP) and 17 HSVgDU L 47DYFP (HSV.CFP) were grown and titrated as previously described [46]. HSV.GFP expresses an EGFP/Cre fusion gene under the control of the CMV-IE promoter from the intergenic space between UL3 and UL4 resulting in EGFP expression by infected cells. HSV.CFP was derived from HSV-1 gDU L 47 (strain 17 expressing YFP from U L 47 and CFP-tagged gD protein) [66], from which the YFP was removed and sequences from U L 47 restored. HSV.GFP and HSV.CFP were made by homologous recombination between parent viral genomes with appropriate transfer plasmids in 293A cells after which green/ yellow fluorescence were selected for and against, respectively. Final recombinants were verified by PCR and sequencing of relevant parts of the genome after at least three rounds of plaque purification.

Bone marrow chimeric mice
Mice were irradiated with two of doses 550 cGy 3 hours apart followed by reconstitution with 5610 6 T-cell-depleted donor bone marrow cells and treatment with purified anti-Thy1 antibody (clone T24/31.7 from hybridoma supernatants) 1 day later. Chimeric mice were allowed to reconstitute for at least 8 weeks before experiments. WTRI-A/E 2/2 chimeras were not treated with anti-Thy1 antibody and received 5610 6 enriched splenic CD4 + T cells from wild-type mice 1 day and 4 weeks after irradiation.

Adoptive transfer of transgenic T cells
gBT-I and gDT-II cells were isolated from lymphoid tissues and gDT-II cells were further enriched by positive and negative selection using magnetic beads, as described previously [4]. 5610 4 gBT-I or 1610 4 gDT-II cells were transferred into naïve mice intravenously via the tail vein, respectively. For transfer of in vitro activated T cells, 1.5610 6 transgenic cells were injected.

Isolation of skin cells and ex vivo analysis of IFN-c production
As indicated, skin tissue was chopped and digested with 3 mg/ ml collagenase type 3 (Worthington Biochemicals, USA) and 5 mg DNase (Roche, Germany) for 90 min at 37uC. Alternatively, skin was digested in 2.5 mg/mL dispase II (Roche) diluted in PBS for 90 min at 37uC. Then, the epidermis and dermis were separated mechanically and epidermal sheets were incubated in trypsin/ EDTA (0.25%/0.1%) (SAFC Biosciences), while the dermis was chopped and incubated in collagenase type 3 and DNase, as previously described [49]. When analyzing ex vivo IFN-c production, 10 mg/mL Brefeldin A (Sigma) was included during each enzymatic digestion step. For some experiments, mice were additionally injected with 0.25 mg BFA intravenously 6 hours prior to sacrifice.

Enrichment of DCs
Axillary lymph nodes of HSV-1-infected mice were desiccated with a scalpel blade and digested with continual mixing in RPMI 1640 containing 1 mg/mL collagenase type 3 and 2 mg/mL DNase for 20 min, prior to the addition of 600 mL 0.1 M EDTA and continual mixing for 5 minutes further. DCs were subsequently enriched for by magnetic beads, as previously described [4].

Intravital two-photon microscopy
Mice were anesthetized and HSV-1-infected flank skin was mounted on an imaging platform and acquired with an upright LSM710 NLO multiphoton microscope as described previously [48]. Imaging data was processed and automatic cell tracking aided by manual corrections was performed with Imaris 7.1 software. For movies, image sequences were composed in Adobe After Effects CS5.

Statistical analysis
Graphs were plotted using Prism 5 (Graphpad) and comparison of data sets was performed by one-way analysis of variance followed by Tukey post-test, or Mann-Whitney or student t tests, as indicated. All graphs depict means 6 s.e.m.. By contrast, IFN-c + CD8 + T EFF cells are strictly confined to epithelial skin regions harboring infectious virus, including the epidermis and hair follicles, and are absent from lymphoid tissues. (B) This distinct anatomical distribution of IFN-c + CD4 + and CD8 + T EFF cells results from their different responsiveness towards stimulation by APCs. Irrespective of their infection status, MHC-II + professional APCs, such as DCs, activate CD4 + T EFF cells in skin epithelium, dermis and LNs, whereas nonprofessional APCs, such as keratinocytes or DETCs, fail to do so. By contrast, IFN-c production by CD8 + T EFF cells is triggered only by directly infected cells, the majority of which are nonprofessional epithelial APCs, such as keratinocytes and DETCs. Noninfected DCs in the dermis or draining LNs fail to elicit IFN-c production by CD8 + T EFF cells, even though they activate CD4 + T EFF cells and initiate division and effector differentiation of naïve CD8 + T cells.

(TIF)
Movie S1 Reduced velocity of CD8 + T EFF cells in proximity to HSV-infected cells. Wild-type mice were subjected to skin infection with HSV.CFP and received in vitro