Viral MHCI inhibition evades tissue-resident memory T cell formation and responses

CD8⁺ T cells mediate potent immunity against viral infections and respond to foreign antigens presented by major histocompatibility complex class I (MHCI) molecules (Schmitz et al., 1999; Shoukry et al., 2003; Simon et al., 2006). The importance of MHCI antigen presentation is underscored by the fact that viruses have evolved strategies to block MHCI presentation. For instance, cowpox virus (CPXV) inhibits MHCI presentation by two distinct mechanisms. The CPXV203 protein retains MHCI molecules in the ER (Byun et al., 2007), whereas the CPXV012 protein prevents the transporter associated with antigen processing from loading antigen peptides onto MHCI molecules (Alzhanova et al., 2009; Byun et al., 2009). When combined, these mechanisms result in effective evasion of CD8⁺ T cell responses in vivo, and the absence of the CPXV012 and CPXV203 significantly attenuates CPXV in a CD8⁺ T cell–dependent manner (Byun et al., 2009; Gainey et al., 2012; Lauron et al., 2018). Moreover, the ability to inhibit MHCI presentation appears to be an evolutionarily conserved feature, though mechanistically distinct, among CMVs and other viruses (Hansen and Bouvier, 2009). Viral MHCI inhibition evades CD8⁺ T cell responses against murine CMV infection in the salivary glands of naive hosts and is critical in allowing for rhesus CMV superinfection of hosts harboring memory CD8⁺ T cells (Lu et al., 2006; Hansen et al., 2010). However, tissue-resident memory CD8⁺ T cells (T_RMs) are able to protect against local infection when murine CMV is directly introduced into the salivary glands, likely due to an early viral tropism for cells refractory to viral MHCI inhibition (Thom et al., 2015). Therefore, the effects of viral MHCI inhibition on CD8⁺ T_RM responses remain unclear.

CD8⁺ T_RMs typically form in nonlymphoid tissues following viral infection and are a noncirculating subset of memory T cells, whereas the effector memory T cell (T_EM) and central memory T cell (T_CM) subsets continuously recirculate (Carbone, 2015). Because CD8⁺ T_RMs primarily develop and remain at common sites of pathogen entry, they are considered a frontline defense against secondary or recurrent peripheral infections; both CD8⁺ and CD4⁺ T_RMs promote viral control and survival against lethal infection, mediate cross-strain protection, and even provide better protection than the circulating T_EM and T_CM counterparts (Gebhardt et al., 2009; Teijaro et al., 2011; Jiang et al., 2012; Mackay et al., 2012; Wu et al., 2014; Zens et al., 2016).

The factors driving T_RM development have implications for tissue-specific vaccine strategies. For example, the “prime and pull” strategy demonstrates that CD8⁺ T cells can be recruited to the skin or vagina in an antigen-independent manner and drive T_RM formation, resulting in long-term immunity against local challenge (Mackay et al., 2012; Shin and Iwasaki, 2012). Conversely, recruitment or inflammation alone does not generate T_RMs in the lungs unless local cognate antigen is present (Takamura et al., 2016; McMaster et al., 2018), indicating tissue-specific requirements for local cognate antigen during T_RM differentiation. Depots of persisting viral antigens in the lung may also affect the maintenance of memory T cells (Zammit et al., 2006; Lee et al., 2011). However, it is unknown whether persistent antigen presentation occurs in the skin or if MHCI complexes are important for the maintenance of endogenous skin CD8⁺ T_RMs. In the context of viral infections, local cognate antigen recognition promotes the formation of CD8⁺ T_RMs in the skin and is required for CD8⁺ T_RM formation in the central nervous system, peripheral nervous system, and lungs (Wakim et al., 2010; Mackay et al., 2012; Khan et al., 2016; Muschaweckh et al., 2016; Pizzolla et al., 2017).

These findings on the potential role of local antigen during viral infection raise an interesting question: can viral MHCI inhibition affect local antigen recognition and reduce CD8⁺ T_RM formation? To investigate this issue, we compared CD8⁺ T_RM formation and protection following local infection with CPXV and a CPXV mutant lacking the capacity to inhibit MHCI presentation. Surprisingly, viral MHCI inhibition affected CD4⁺ T_RM formation, but not the overall development CD8⁺ T_RM formation, in the skin. We found that local secondary antigenic stimulation promoted CD8⁺ T_RM formation, despite CPXV-mediated MHCI inhibition and CD8⁺ T cell evasion. After T_RM differentiation, local cognate antigen presentation was dispensable for T_RM maintenance but was critical in infected cells to induce protective CD8⁺ T_RM responses.

CPXV-mediated MHCI inhibition could reduce local antigenic stimulation of virus-specific CD8⁺ T cells by infected cells. Because local antigenic stimulation enhances CD8⁺ T_RM formation during vaccinia virus (VACV) skin infection (Khan et al., 2016), we hypothesized that CD8⁺ T_RM formation would be diminished in CPXV-infected skin. To test this, we infected mice by skin scarification (s.s.) with WT CPXV or Δ12Δ203 CPXV, which does not inhibit MHCI antigen presentation, and performed kinetic analyses of CD8⁺ T cells that recognize the immunodominant H-2K^b–restricted epitope B8_19–26. Infection with both viruses resulted in robust expansion and recruitment of B8_19–26-specific CD8⁺ T cells to the skin 1 wk after infection, which were reduced in numbers ∼12-fold by 3 wk after infection (Fig. 1 A). Nonetheless, B8_19–26-specific CD8⁺ T cells were still detectable in the skin and the spleens of previously infected mice long after clearance of the infection (Fig. 1, A–D). CPXV skin infections are highly localized and are cleared within 1–2 wk after infection (Gainey et al., 2012). Mice infected with WT CPXV had a higher relative abundance of B8_19–26-specific CD8⁺ T cells in the skin at 7 d postinfection (dpi) in comparison to Δ12Δ203-infected mice (Fig. 1, C and D), but there were no significant differences in absolute numbers in the skin for all assessed time points.

Analysis of B8_19–26-specific CD8⁺ T cells during the memory phase showed that they were predominantly located in the previously infected skin flank relative to the uninfected contralateral flank (Fig. 1 B). These findings are consistent with previous reports (Gebhardt et al., 2009; Khan et al., 2016; Muschaweckh et al., 2016). We show that these cells expressed higher levels of the T_RM markers CD103 and CD69 than the B8_19–26-specific CD8⁺ T cells in the spleen (Fig. 1, E and F), suggesting that CPXV s.s. infection generates authentic CD8⁺ T_RMs. To verify tissue residency of CD103⁺CD69⁺CD8⁺ cells, we performed parabiosis between CD45.1⁺ naive mice and CD45.2⁺ mice previously infected with CPXV, referred to as naive and memory mice, respectively (Fig. 1 G). Comparison of the CD45.2⁺CD45.1⁻CD103⁺CD69⁺CD8⁺ T_RMs in the skin of naive and memory parabionts revealed the presence of these cells in only the memory parabionts, showing that they are unable to circulate and are indeed a resident population (Fig. 1, H and I). Furthermore, CD45.2⁺CD45.1⁻B8_19–26-tetramer⁺ cells in the spleens equilibrated between memory and naive parabionts, but CD45.2⁺CD45.1⁻B8_19–26-tetramer⁺ T_RMs in the skin were only present in memory parabionts (Fig. 1, J and K). Regarding CD4⁺ T_RM, CPXV s.s. generated CD103⁺CD69⁺CD4⁺ T cells in the skin in comparable numbers to CD8⁺ T_RM, and parabiosis revealed that these cells were bona fide skin-resident CD4⁺ T cells (Fig. 2, A–C). Interestingly, we found that viral MHCI inhibition significantly reduced the abundance of these CD103⁺CD69⁺CD4⁺ T cells in the skin (Fig. 2, A and B). Taken together, these results reveal that viral MHCI inhibition does not affect the overall development of CD8⁺ T_RMs but does impair the formation of CD4⁺ T_RMs in the skin.

We reasoned that the effects of viral MHCI inhibition on local CPXV-specific CD8⁺ T cells might have direct or indirect consequences on the development of CD4⁺ T_RMs. To gain insight into how viral MHCI inhibition affects local antigen-specific CD8⁺ T cells, we performed RNA sequencing (RNA-seq) on B8_19–26-specific CD8⁺ T cells isolated from the skin and spleen of mice coinfected with WT CPXV and Δ12Δ203 on opposite flanks. At 7 dpi, B8_19–26-specific CD8⁺ T cells isolated from WT CPXV and Δ12Δ203-infected skin exhibited similar expression profiles and shared many differentially expressed genes relative to splenic B8_19–26-specific CD8⁺ T cells (Fig. S1, A and B). Gene set enrichment analysis (GSEA) of unique differentially expressed genes revealed significant enrichment of genes associated with activation of lymphocytes in Δ12Δ203-infected skin, while analysis of B8_19–26-specific CD8⁺ T cells from WT CPXV–infected skin demonstrated expression of genes involved in the regulation of cytokine production (Fig. S1 C). Cytokines that were significantly up-regulated in WT CPXV–infected skin were also significantly up-regulated in Δ12Δ203-infected skin (Fig. S1 D). However, we found that IL-10 expression by B8_19–26-specific CD8⁺ T cells was significantly higher in WT-infected skin in comparison to Δ12Δ203-infected skin, both at the transcript and protein levels (Fig. S1, E–G). The increased IL-10 production correlates with the reduced number of CD4⁺ T_RM in WT CPXV–infected skin. IL-10 has been shown to reduce the number of circulating memory CD4⁺ T cells following acute infection with lymphocytic choriomeningitis virus (Brooks et al., 2010). Therefore, these findings could explain why viral MHCI evasion affects CD4⁺ T_RM formation.

CD4⁺ T cell help is critical for the local development of lung CD8⁺ T_RMs (Laidlaw et al., 2014) but dispensable for the formation of CD8⁺ T_RMs in the female reproductive tract (Beura et al., 2018). To test the role of CD4⁺ T cell help in the development of CD8⁺ T_RM in the skin, we depleted CD4⁺ T cells by injecting depleting antibodies 1 d before infection and then 2 and 5 dpi. This approach eliminates CD4⁺ T cells during acute infection but allows for their recovery by 30 dpi. CD4⁺ T cell help was not required for priming of CD8⁺ T cells in the presence or absence of viral MHCI inhibition (Fig. 2, D and E). In accordance with our previous findings (Byun et al., 2009; Gainey et al., 2012), we found that mice infected with WT CPXV by s.s. succumbed to infection in the absence of CD4⁺ T cell help (Fig. 2 F), despite the mounting an anti-CPXV CD8⁺ T cell response (Fig. 2 E). These results underline the capacity of CPXV to efficiently evade CD8⁺ T cells and suggest that CD4⁺ T cell help is required to control localized CPXV infection. In contrast, all mice infected with Δ12Δ203 survived in the absence of CD4⁺ T cell help (Fig. 2 F), which is also consistent with our previous findings that CD8⁺ T cells alone are sufficient to control CPXV infection only in the absence of viral MHCI inhibition (Byun et al., 2009; Gainey et al., 2012). Given that Δ12Δ203 did not cause mortality, this provided a setting to assess the role of CD4⁺ T cells in skin CD8⁺ T_RM development. As expected, depleting CD4⁺ T cells during the primary response reduced the overall number of CD4⁺ T cells in the skin of mice previously infected with Δ12Δ203 but did not affect the overall development of CD4⁺ T_RMs, since the percentage of CD103⁺CD4⁺ T cells in the skin was similar in isotype control–treated mice (Fig. 2, G and H). However, it is possible that these are not tissue-resident cells and instead represent a different subset of CD4⁺ T cells. Surprisingly, the formation of CD8⁺ T_RMs was augmented in the absence of CD4⁺ T cell help (Fig. 2, G–I), indicating that CD4⁺ T cell help is not required for the generation of skin CD8⁺ T_RMs and may even hinder their development.

Since CD8⁺ T_RM development occurred despite viral MHCI inhibition, we tested if CD8⁺ T_RM formation is independent of local antigen. To this end, we performed coinfections on opposite flanks with CPXV viruses that express either a WT (WT and Δ12Δ203-CPXV) or a variant of the immunodominant B8_19–26 epitope (B8Y3AF5A and Δ12Δ203B8Y3AF5A) that is not presented on MHCI (Lauron et al., 2018; Fig. 3 A). During the acute phase of infection, B8_19–26-specific CD8⁺ T cells had modestly higher frequencies of CD69 expression in the skin where B8_19–26 was expressed in comparison to the skin where the B8_19–26 epitope was absent, whereas CD103 expression was unaffected (Fig. S2, A and B). Nonetheless, a similar percentage and number of B8_19–26-specific CD8⁺ T cells were recruited to both infected skin flanks in a local antigen-independent manner (Fig. 3 B). This was also the case when coinfections were performed with Δ12Δ203 and Δ12Δ203B8Y3AF5A. At 35 dpi, however, B8_19–26-specific T_RMs were significantly enriched in the skin where B8_19–26 was locally expressed as compared with infection where B8_19–26 was locally absent (Fig. 3 C), even in the presence of CPXV-mediated MHCI inhibition. Furthermore, the abundance of total CD8⁺ T_RM was equal on both flanks of coinfected mice (Fig. 3 D), suggesting that T_RM formation was only affected for B8_19–26-specific CD8⁺ T cells. Taken together, these data indicate that local antigen does promote CD8⁺ T_RM formation during CPXV infection, regardless of viral MHCI inhibition.

To identify the cells that could display local cognate antigen, we assessed CPXV tropism in the skin of mice infected by CPXV s.s. There was significant recruitment of leukocytes in the skin at 6 dpi (Fig. 4 A). The majority of infected cells were MHCII/CD45 double positive and composed of mainly CD11b⁺ dendritic cells (DCs) and monocyte-derived DCs (moDCs; Fig. 4, B and C). Moreover, CPXV⁺ and MHCII⁺ cells were found in close proximity to CD8⁺ cells (Fig. 4 D), suggesting that CD8⁺ T cells may engage infected CD11b⁺ DCs or moDCs at the site of infection, though viral MHCI inhibition should prevent virus-specific T cell stimulation. Consistent with this notion, DC line DC2.4 cells infected with WT CPXV do not stimulate IFN-γ production by CPXV-specific CD8⁺ T cells in vitro (Fig. 4 E). This effect is dependent on CPXV012 and CPXV203, as CPXV-specific CD8⁺ T cells co-cultured with Δ12Δ203-infected DC2.4 cells produced significantly higher levels of IFN-γ and up-regulated expression of Nur77 (Fig. 4 E), a transcription factor that is specifically induced upon TCR engagement (Moran et al., 2011). However, our results suggest that CPXV-specific CD8⁺ T cells can receive antigenic stimulation at the site of infection in vivo (Fig. 3 C).

We therefore tested if our in vitro findings extend to in vivo infection by coinfecting IFN-γ.Thy1.1 knockin (KI) reporter mice with WT CPXV and Δ12Δ203 on opposite flanks. Similar to our in vitro results, CPXV s.s. resulted in significantly lower frequencies of IFN-γ–producing CD8⁺ T cells in WT CPXV–infected skin in comparison to Δ12Δ203-infected skin (Fig. 4 F). Conversely, the expression of Nur77 and the frequency of Nur77⁺CD8⁺ T cells were not significantly different (Fig. S2 C and Fig. 4 F). We also found similar results following coinfection of Nur77 GFP reporter mice (Fig. S2 D). In accordance with our findings that local cognate antigen promotes T_RM formation (Fig. 3 C), these findings suggest that CD8⁺ T cells recruited to CPXV-infected skin receive local cognate antigen stimulation in the context of viral MHCI inhibition. However, it still remained unclear whether CPXV-specific CD8⁺ T cells could recognize infected target cells in vivo, so we performed an in vivo cytotoxicity assay. After transfer of infected targets to day 7–infected mice, ∼72% of Δ12Δ203-infected targets were rapidly killed, whereas only ∼3% of WT-infected targets were killed (Fig. 3, G and H). These data suggest that, as a consequence of viral MHCI inhibition, WT CPXV–infected targets provide little to no antigenic stimulation and may only contribute minimally to local antigen-driven T_RM differentiation. Strikingly, CD103⁺ DCs, which are essential for cross-priming of naive CPXV-specific CD8⁺ T cell precursors (Gainey et al., 2012; Lauron et al., 2018), remain largely uninfected following CPXV skin infection (Fig. 4 C) but were also found in close proximity to CD8⁺ cells at the site of infection (Fig. 4 D). These data strongly imply that CD8⁺ T cells in the skin are engaging cognate antigen on local cross-presenting cells.

Infection with WT CPXV is prolonged in comparison to Δ12Δ203 infection (Gainey et al., 2012), and it is therefore possible that during WT CPXV infection continuous recruitment of CD8⁺ T cells could compensate for decreased secondary antigenic stimulation on skin T_RM precursors. To test if continuous recruitment of CD8⁺ T cells to the skin augments CD8⁺ T_RM formation in the presence of viral MHCI inhibition, we used the sphingosine 1-phosphate receptor-1 (S1PR) agonist FTY720, which prevents lymphocyte egress from lymphoid and infected tissues. FTY720 treatment early during infection (1, 3, and 5 dpi) significantly reduced the presence of circulating CD8⁺ T cells in the blood and the recruitment of CD8⁺ T cells to infected skin, but not the priming of B8_19–26-specific CD8⁺ T cells in the inguinal lymph node (Fig. S3, A–C). FTY720 treatment also prolonged skin infection with both WT CPXV and Δ12Δ203, as indicated by lesion scab loss (Fig. S3, D and E), but Δ12Δ203-infected lesions still cleared slightly faster than WT CPXV–infected lesions.

We next performed coinfections of mice with WT CPXV and Δ12Δ203 on opposite flanks and then blocked CD8⁺ T cell recruitment to and egress from infected skin with FTY720 treatment starting at 7 dpi (Fig. S3 F). Starting FTY720 treatment at this later time point also prolonged the infection on both flanks of coinfected mice, yet lesion scab loss still occurred by 15 dpi (data not shown). While FTY720 significantly reduced the presence of CD4⁺ and CD8⁺ T cells in the blood of coinfected mice, the relative abundance of B8_19–26-specific CD8⁺ T cells at 9 dpi and CD8⁺ T_RMs at 40 dpi was not significantly different in WT CPXV– and Δ12Δ203-infected skin, with or without FTY720 treatment (Fig. S3, G and H). These results indicate that CD8⁺ T_RMs develop without continuous recruitment of CD8⁺ T cells, which is consistent with the findings that recruitment or inflammation alone is sufficient to induce T_RM formation (Mackay et al., 2012; Shin and Iwasaki, 2012). However, differences in viral replication between WT CPXV and Δ12Δ203 may also influence T_RM development, making it difficult to assess the effects of viral MHCI inhibition on T_RM formation.

We therefore sought to uncouple the potential effects of prolonged viral replication and the effects of viral MHCI inhibition. To do so, we treated infected mice with the antiviral compound cidofovir. Topical treatment with cidofovir resulted in a faster rate of viral clearance in comparison to untreated control mice (Fig. 5, A and B). Moreover, WT CPXV– and Δ12Δ203-infected skin lesions were cleared at a similar rate with cidofovir treatment, whereas WT CPXV–infected skin lesions were cleared at a slower rate than Δ12Δ203-infected skin lesions in untreated control mice. Interestingly, cidofovir-treated mice infected with WT CPXV or Δ12Δ203 showed no differences in the abundance of B8_19–26-specific T_RMs, despite simultaneous viral clearance and equal acute recruitment of B8_19–26-specific CD8⁺ T cells in infected skin (Fig. 5 C).

We postulated that the effects of viral MHCI inhibition on the local microenvironment might also be a confounding variable. We therefore devised a coinfection approach in which the environment and viral replication could be better controlled in a setting where B8 antigen presentation is or is not inhibited by CPXV012 and CPXV203 (Fig. 5 D). For inhibited B8 antigen presentation, we infected one flank with a 1:1 mixture of WT CPXV and Δ12Δ203B8Y3AF5A. In this mix, WT CPXV provides a setting with inhibited B8 antigen presentation, while Δ12Δ203B8Y3AF5A serves to control for the local infectious environment and the total amount of virus per infection site but does not contribute to B8 antigen load. For uninhibited B8 antigen presentation, we infected the opposite flank with a 1:1 mixture of Δ12Δ203 and B8Y3AF5A. Infection with Δ12Δ203 provides a source of uninhibited B8 Ag presentation, because it lacks the CPXV MHCI inhibitors. The B8Y3AF5A virus contains both MHCI inhibitors CPXV012 and CPXV203. It also serves as a control for the local infectious environment and has comparable viral clearance to WT CPXV, even though it does not contribute to B8 antigen load. Therefore, the effects of viral MHCI inhibition on the local microenvironment and the duration of infection for each mixture should be comparable. As expected, each mixture resulted in comparable total levels of viral replication and clearance, as determined by plaque assays and lesion scab loss (Fig. 5 E). At 35 dpi, the abundance of CD8⁺ T_RMs was comparable under both conditions (Fig. 5 F). Remarkably, the abundance of B8_19–26-specific CD8⁺ T_RMs was significantly reduced when B8 Ag presentation was inhibited by CPXV012 and CPXV203 (Fig. 5 G), indicating that viral MHCI inhibition does indeed affect local antigen-driven T_RM formation.

We next assessed the effects of CPXV-mediated MHCI inhibition on lung CD8⁺ T_RMs, since the developmental requirements can differ substantially between certain tissue microenvironments (Pizzolla et al., 2017). To distinguish circulating memory CD8⁺ T cells from T_RMs, we performed in vivo intravascular (IV) staining (Anderson et al., 2012). This method exclusively labels circulating and IV lymphocytes (IV⁺ cells), but not tissue-associated lymphocytes (IV⁻). Within the lung tissue, the overall proportions of IV⁻CD8⁺ T cells were equal in mice i.n. infected with WT CPXV or Δ12Δ203 at 12 and 30 dpi, suggesting that viral MHCI inhibition also does not affect the overall development of lung CD8⁺ T_RMs (Fig. 6, A and B). In contrast, higher proportions of IV⁻ B8_19–26-specific T_RMs were found in the lungs following Δ12Δ203 infection compared with WT CPXV infection (Fig. 6 C). This difference was also reflected in the spleen and blood at 30 dpi, despite equal proportions of B8_19–26-specific CD8⁺ T cells in the spleen and blood at 12 dpi (Fig. 6, D and E). Viral MHCI inhibition also significantly reduced the expression of CD103 on IV⁻ B8_19–26-specific CD8⁺ T cells at 12 dpi, but not CD69 expression (Fig. 6, F and G). The reduced expression of CD103 did not appear to affect CD8⁺ T_RM development, since IV⁻ B8_19–26-specific CD8⁺ T cells in the lungs of WT CPXV– and Δ12Δ203-infected mice expressed comparable levels of CD69 and CD103 by 30 dpi (Fig. 6, F and G). These data indicate that viral MHCI inhibition also reduces the formation of antigen-specific CD8⁺ T_RMs in the lungs.

Having established the importance of local cognate antigen in CD8⁺ T_RM formation against CPXV, we questioned whether local cognate antigen plays a role after CD8⁺ T_RM populations are established. To address this, we generated B2m^fl/flRS26-Cre-ER^T2 mice in which tamoxifen treatment ablates MHCI expression. B2m^fl/flRS26-Cre-ER^T2 mice infected by s.s. were treated with tamoxifen at 49 dpi, and CD8⁺ T_RMs were analyzed in the skin ∼1 mo after treatment (Fig. 7 A). Tamoxifen treatment eliminated the expression of MHCI on CD45⁺ cells from the skin and spleen of B2m^fl/flRS26-Cre-ER^T2 mice (Fig. 7 B), but not from tamoxifen-treated B2m^fl/fl littermate controls. We detected B8_19–26-specific T_RMs at comparable levels in the skin of tamoxifen-treated B2m^fl/flRS26-Cre-ER^T2 and B2m^fl/fl mice (Fig. 7 C), indicating that persistent antigen presentation is not required for CD8⁺ T_RM maintenance. This was also observed for total CD8⁺ T_RMs in the skin (Fig. 7 D). Furthermore, we assessed TCR engagement on CD8⁺ T_RMs using the Nur77 GFP reporter mice and found that CD8⁺ T_RMs are Nur77 negative (Fig. 7 E). Therefore, CD8⁺ T_RMs do not receive cognate antigen stimulation in the skin and are maintained in an antigen-independent manner.

Since viral MHCI inhibition evades CD8⁺ T cell responses during primary CPXV infection, we next investigated the effects of viral MHCI inhibition on CD8⁺ T_RM effector function in the absence of other anti-CPXV immune responses. To do so, we generated a recombinant influenza virus that expresses the CPXV B8_19–26 epitope (Flu-B8) and infected mice by the i.n. route. This resulted in the formation of B8_19–26-specific T_RMs in the lungs of infected mice at 30 dpi (Fig. S4). We next challenged the mice with CPXV at 30 dpi (Fig. 8 A). To prevent circulating memory CD8⁺ T cells from contributing to viral control upon CPXV challenge, we treated Flu-B8–infected mice with FTY720 throughout the course of the CPXV challenge. Viral titers in the lungs of mice challenged with Δ12Δ203 were significantly lower than in mice challenged with WT CPXV (Fig. 8 B), demonstrating that viral MHCI inhibition abrogates protective CD8⁺ T_RM responses in the lungs. Furthermore, CD8⁺ T_RM–mediated protection was antigen dependent, since challenge with B8-deficient viruses resulted in titers comparable to challenge with WT CPXV.

To test if these findings also hold true during CPXV skin infection, we used μMT mice in order to avoid humoral responses against CPXV challenge, as these mice lack mature B cells. μMT mice previously infected by s.s. with Δ12Δ203 harbored few B8_19–26-specific T_RMs in the skin (Fig. S5, A and B), which is consistent with the findings that infection of μMT mice generates significantly lower numbers of memory CD8⁺ T cells in comparison to infection of WT C57BL/6 mice (Asano and Ahmed, 1996; Shen et al., 2003). These findings are unlikely due to differences in viral replication or dissemination, since Δ12Δ203 replicates and is cleared in the μMT mice at comparable rates to C57BL/6 mice (Fig. S5, C and D). We next tested whether depleting CD4⁺ T cells during acute infection of μMT mice would increase CD8⁺ T_RM formation. Similar to what we observed in WT C57BL/6 mice (Fig. 2, G–I), depleting CD4⁺ T cells significantly increased CD8⁺ T_RM formation in the skin of μMT mice infected by s.s., but not i.n., infection (Fig. S5, A and B). In contrast, depleting CD4⁺ T cells did not significantly increase the number of B8_19–26-specific CD8⁺ T cells in the spleen of μMT mice previously infected by s.s.

Based on these findings, we induced CD8⁺ T_RM formation in CD4⁺ T cell–depleted μMT mice by s.s. with Δ12Δ203 and subsequently challenged the skin with WT CPXV and Δ12Δ203 at 30 dpi (Fig. 8 C). Viral titers in the skin of naive mice were not significantly different in WT CPXV– and Δ12Δ203-infected lesions 4 d after challenge (Fig. 8 D). Similarly, viral titers were not significantly different in previously infected μMT mice lacking skin CD8⁺ T_RMs (i.n. infected). Conversely, μMT mice harboring skin CD8⁺ T_RMs (s.s. infected) presented viral titers that were significantly reduced by several logs in the Δ12Δ203 lesions (Fig. 8 D), whereas WT CPXV titers were not significantly reduced in comparison to naive and i.n.-infected mice. The differences were also apparent by lesion formation (Fig. 8 D). These data demonstrate that CPXV effectively evades CD8⁺ T_RM responses, but CD8⁺ T_RMs can provide protection against CPXV in the absence of viral MHCI inhibition. Intriguingly, similar results were obtained in μMT mice that harbored fewer skin CD8⁺ T_RMs (isotype-treated controls), indicating that CD4⁺ T cell help is not required for the generation of functional skin CD8⁺ T_RMs (Fig. 8 E).

In this study, we assessed the effects of viral MHCI inhibition on the generation and function of T_RMs using CPXV and the Δ12Δ203 mutant. Consistent with what was previously reported for VACV skin infection (Khan et al., 2016; Muschaweckh et al., 2016), local antigenic stimulation promoted CD8⁺ T_RM formation in CPXV-infected skin, indicating that the viral MHCI inhibitors CPXV012 and CPXV203 do not completely obstruct local antigenic stimulation of CD8⁺ T cells or subsequent CD8⁺ T_RM development. We note that the duration of local antigen exposure may be shorter throughout Δ12Δ203 infection than WT CPXV infection as a consequence of unimpeded clearance by CPXV-specific CD8⁺ T cells. Despite the potentially reduced window for antigen stimulation, Δ12Δ203 infection in both the skin and the lung significantly enhanced local antigen-driven T_RM formation compared with infection with WT CPXV, suggesting that early encounters with local antigen have greater effects on T_RM differentiation than the length or persistence of antigen presentation.

Our results also suggest that local antigenic stimulation of CD8⁺ T cells may promote CD8⁺ T_RM development without eliciting T cell–mediated cytotoxicity, raising the possibility that T_RM precursors receive local antigenic stimulation from cross-presenting cells in situ (Boissonnas et al., 2013; Broz et al., 2014). This possibility would explain why local antigen stimulation enhances T_RM formation that is not significantly affected by the presence of viral MHCI inhibition. Since cross-presentation by BATF3⁺ DCs imprints T_RM precursors in the lymph nodes (Iborra et al., 2016) and is required for CPXV-specific T cell priming (Gainey et al., 2012; Lauron et al., 2018), these cells are candidates for involvement in local cross-presentation; further investigation will be needed to determine their relative contribution to in situ cross-presentation during T_RM development. Intriguingly, our results also raise the possibility that local cross-presentation may be involved in activating skin CD8⁺ T_RM cytokine production.

Local antigen presentation by APCs is critical for the activation of CD8⁺ T_RMs in the female genital tract and central nervous system (Wakim et al., 2010; Shin et al., 2016), but direct contact with infected target cells is not necessarily required for CD8⁺ T_RM–mediated protection. For instance, in situ peptide stimulation of CD8⁺ T_RMs is sufficient to induce a tissue-wide antiviral state and protects against antigenically distinct viruses (Ariotti et al., 2014; Schenkel et al., 2014). Therefore, CD8⁺ T_RM activation and protection can ensue independently of direct cell–cell contact with infected target cells. We found that CD8⁺ T_RM protection during viral infection is abrogated by viral MHCI inhibition, indicating that TCR/peptide-MHCI (pMHCI) stimulation via engagement of infected target cells is needed for optimal CD8⁺ T_RM functions. We also found that CD4⁺ T cell help was not required for the generation of functional skin CD8⁺ T_RMs and that the absence of CD4⁺ T cell help increased skin CD8⁺ T_RM formation. As T_RMs have been shown to compete within epidermal niches (Zaid et al., 2014), it is possible that the absence of CD4⁺ T cells during acute infection reduces competition for local survival signals and space, lessening the constraints on the number of CD8⁺ T_RMs.

Additionally, we show CD8⁺ T_RM development in the lungs of CPXV-infected mice remained largely unaffected by viral MHCI inhibition, but viral MHCI inhibition reduced the abundance of antigen-specific lung T_RMs. The induction of CPXV-specific CD8⁺ T cell responses is not affected by viral MHCI inhibition during primary CPXV infection (Gainey et al., 2012; Lauron et al., 2018), but the breadth and the levels of viral epitopes presented in this context can differ from those presented in the absence of viral MHCI inhibition (Basta et al., 2002; Shen et al., 2002). As a consequence, increased antigen presentation and antigen abundance during respiratory infection with Δ12Δ203 potentially promotes CD8⁺ T cell cross-competition for pMHCI complexes, which can affect the T_RM repertoire (Muschaweckh et al., 2016; Pizzolla et al., 2017). These findings were not found following CPXV skin infection, suggesting that differences in the microenvironment also play a role in the development of antigen-specific T_RMs; we did, however, observe an effect in the skin when we experimentally controlled for environmental conditions. Moreover, respiratory viral infections result in persistent antigen presentation (Zammit et al., 2006; Jelley-Gibbs et al., 2007; Turner et al., 2007) and can contribute to the locality of memory T cells (Kim et al., 2010). Persistent antigen presentation was therefore an appealing explanation for the generation and long-term maintenance of T_RMs in peripheral tissues.

Here, we used inducible deletion of B2m to test if persistent antigen stimulation is required for maintaining skin T_RMs. We found that deletion of B2m after T_RM formation did not affect the number of T_RMs, suggesting that local cognate antigen or even low-level stimulation from cross-reactive pMHCI complexes is not required for T_RM maintenance. However, since our studies focused on the CD45⁺ compartment, it is possible that low-level expression of MHCI on nonhematopoietic cells could contribute to T_RM maintenance. Nonetheless, our results add to previous reports (Casey et al., 2012; Mackay et al., 2012; Shin and Iwasaki, 2012; Davies et al., 2017; Slütter et al., 2017) by suggesting that T_RM maintenance is also apparently independent of cross-reactive pMHCI complexes.

Collectively, our findings shed light on the maintenance, generation, and activation of CD8⁺ T_RMs and highlight the effects of viral MHCI inhibition on T_RM biology. The effects of viral MHCI inhibition on CD8⁺ T cells can be difficult to demonstrate, as most of the identified viral MHCI inhibitors exhibit varying degrees of species specificity and efficacy across MHCI allomorphs (Cox et al., 1994; Le Gall et al., 1998; Rehm et al., 2002; Wagner et al., 2002). Remarkably, CPXV012 and CPXV203 potently act on both murine and human MHCI molecules, and the in vivo effects on CD8⁺ T cell responses are evident. The CPXV infection model thus provides a unique opportunity to investigate the effects of viral immune evasion on antiviral CD8⁺ T cell development and effector functions. Although we demonstrate that CPXV-mediated MHCI inhibition evades CD8⁺ T_RM responses, the relevance of skin T_RM responses following CPXV reinfection of its natural wild rodent hosts or humans is unclear. For instance, skin T_RMs would only be activated if CPXV reinfection occurred at or by sites of previous infections where T_RMs have formed. In contrast, T_RMs may be relevant for natural CPXV reinfections acquired through respiratory routes, where lung T_RMs are more likely to reencounter viral antigens. In the case of the later, the role CPXV012 and CPXV203 would certainly be advantageous for the virus. Given the efficacy of CPXV012 and CPXV203, it is likely that CPXV-mediated MHCI inhibition also contributes to the evasion of T_CMs and T_EMs, but we did not directly test this. Nonetheless, our study demonstrates a proof of concept that viral inhibition of MHCI antigen presentation can affect memory CD8⁺ T cell development and responses, which should be considered in regards to vaccine design against viruses that target the MHCI pathway.

C57BL/6 mice were purchased from the National Cancer Institute. Transgenic and knockout mice were purchased from The Jackson Laboratory, with the exception of B2m^fl/fl and IFNγ KI Thy1.1 mice. IFNγ KI Thy1.1 mice were provided by Dr. Casey Weaver (University of Alabama School of Medicine, Birmingham, AL). B2m^fl/fl mice were generated as described by Bern et al. (2018). DC2.4 cells were cultured in RPMI supplemented with 10% FBS (Mediatech), 100 U/ml penicillin, 100 g/ml streptomycin, 1 mM sodium pyruvate, L-glutamine, and nonessential amino acids (Gibco). Madin–Darby Canine Kidney (MDCK) and Vero cells were maintained in MEM with 5% FBS, MEM vitamins (Gibco), L-glutamine, 100 U/ml penicillin, and 100 g/ml streptomycin. Human embryonic kidney cells (293T) were maintained in Opti-MEM (Life Technologies) with 10% FBS, L-glutamine, 100 U/ml penicillin, and 100 g/ml streptomycin. DC2.4 cells were a gift from Dr. Kenneth Rock (University of Massachusetts Medical School, Worchester, MA). MDCK and 293T cells were a gift from Dr. Richard Webby (St. Jude Children’s Research Hospital, Memphis, TN). Vero cells were obtained from the American Type Culture Collection.

For the generation of IAV-B8_19–26 (Flu-B8), the CPXV B8_19–26 epitope was inserted into the stalk region of neuraminidase, in-frame, between nucleotides 148 and 170 using inverse PCR (Phusion-HF; Thermo Fisher) and the restriction enzyme AarI (Thermo Fisher) to ligate the mutant plasmid. Co-cultures of 293T and MDCK cells were transfected with eight bidirectional pHW2000 plasmids containing cDNA for A/Puerto Rico/08/1934 (H1N1; 1 µg per plasmid) with polyethylenimine (8 µg total). Rescue of the reverse genetic virus was performed as previously described (Williams et al., 2018).

Bacterial artificial chromosome–derived CPXV viruses were generated as previously described (Lauron et al., 2018). To determine viral titers, tissues were processed as previously described (Gainey et al., 2012). In brief, harvested tissues were placed in 1 ml MEM, homogenized using a MiniBeadbeater-8, and stored at −80°C. Samples were then serially diluted for plaque assays using Vero cells.

Mice 8–10 wk of age were sex and age matched for each experiment. i.n. and s.s. infections were performed as previously described (Gainey et al., 2012). In brief, mice were anesthetized, and 30 µl of virus inoculum diluted in PBS was administered i.n. For s.s., fur from the flank of mice was depilated using Nair. The following day, mice were anesthetized and the flanks were infected by s.s. at three adjacent sites with 10⁵ PFU per site. Parabiosis surgery was performed as previously described (Peng et al., 2013; Sojka et al., 2014). Parabiosed mice were rested for 2 wk before sacrificing and harvesting tissues. Mouse studies were approved by the Animal Studies Committee at Washington University (protocol #A-3381-01) and adhered to the Institutional Animal Care and Use Committee guidelines.

A 2 × 3 cm² piece of skin was harvested from infected mice and prepared as previously described (Hsu et al., 2009). Sections were blocked using 10% goat serum in PBS for 1 h before incubating with primary antibodies for 1 h. Cells were then washed with 0.5% Triton in PBS, incubated with Streptavidin-AF555 (Thermo Fisher), washed with 0.5% Triton in PBS, and mounted with DAPI (Vectashield). Stained sections were analyzed with a laser scanning confocal microscope (LSM880; Zeiss). Primary antibodies were obtained from BD PharMingen and include CD103-Biotin (M290), IA/IE-AF647 (M5/114.15.2), and CD8α-AF488 (53–6.7).

To induce TCR ligation in vivo, 50 µg anti-CD3e monoclonal antibody (145-2C11; Thermo Fisher) was administered into mice by i.v. injection 16 h before harvesting tissues. FTY720 (Cayman Chemical) was administered by i.p. injection (10 mg/kg) in aqueous solution. A 1% cidofovir solution (Sigma-Aldrich) was prepared in DMSO, and 16 µl per lesion was applied topically on 3 dpi and twice daily for the remaining time points indicated. To deplete CD4⁺ T cells, 200 µg anti-CD4 (GK1.5) or isotype control (anti-rat IgG2b; Bio X Cell) was injected i.p. To deplete natural killer (NK) cells, 100 µg anti-NK1.1 (PK136) was injected i.p. weekly. For IV staining, fluorochrome-conjugated anti-CD45.2 (104; BioLegend) was administered by i.v. injection. 3 min after injection, mice were sacrificed, and tissues were harvested. Mice were treated topically with 1 mg 4-hydroxytamoxifen (Sigma-Aldrich) dissolved in ethanol and were fed a tamoxifen diet (Envigo) during the indicated times shown.

Single-cell suspensions from the spleens, lymph nodes, and lungs and an ∼2 × 3 cm² piece of skin were prepared at the indicated days after infection as previously described (Gainey et al., 2012; Jiang et al., 2012). Skin and lung tissues were minced and incubated in with HBSS containing 1 mg/ml collagenase A (Roche) and 22.4 g/ml DNase I (Roche) at 37°C for 30 min. Samples were then filtered through a 70-µm mesh strainer before staining. For IL-10 staining, cells were incubated in RPMI containing GolgiPlug (BD Biosciences) at 37°C for 2 h before staining. Samples for flow cytometric analyses were stained on ice with Fixable Viability Dye eFlour 506 (eBioscience) before tetramer and cell surface staining. For tetramer staining, cells were incubated with tetramers for 45 min at room temperature before cell surface staining of the indicated surface markers. Tetramers were produced in the Immunomonitoring Laboratory within the Center for Human Immunology and Immunotherapy Programs (Washington University). Ex vivo restimulation and intracellular cytokine staining for IFN-γ was performed as previously described (Lauron et al., 2018). For staining of Nur77, the Foxp3/Transcription Factor Staining Buffer Set was used according to the manufacturer’s protocol (eBioscience). The following antibodies were purchased from Abcam, BD Biosciences, eBioscience, BioLegend, or Thermo Fisher: CD3 (145-2C11), NK1.1 (PK136), CD19 (eBio1D3), CD64 (X54-5/7.1), CD8α (53–6.7), CD4 (RM4-5), VACV (Ab19970), CD4 (RM4-4), CD44 (IM7), CD103 (2E7), CD69 (H1.2F3), H-2K^b (AF-88.5), CD11c (N418), CD45.1 (A20), CD45.2 (104), CD24 (M1/69), IA/IE (M5/114.15.2), SIRPα (P84), CD11b (M1/70), Ly6C (HK1.4), Nur77 (12.14), and IL-10 (JES5-16E3). Cell sorting was performed on a BD FACSAria Fusion sorter.

B8_19–26-specific CD8⁺ T cells (LIVE/DEAD⁻CD4⁻CD45.2⁺CD3⁺CD8⁺CD44⁺B8_19–26-tetramer⁺) were sorted from the spleen and skin of mice coinfected on opposite flanks with WT CPXV and Δ12Δ203 at 7 dpi. Cells were pooled from five mice per experiment to reach at least 20,000 cells per sample. RNA was extracted using the Arcturus PicoPure RNA Isolation kit according to the manufacturer’s protocol (Thermo Fisher). mRNA-seq libraries were created using SMARTer cDNA synthesis technology (Clontech), indexed, pooled, and sequenced at the Genome Technology Access Center (GTAC) at Washington University School of Medicine (St. Louis, MO). Sequencing was run as paired-end reads of 100 bp on the Illumina HiSeq 2500 platform. Raw reads were deposited in the National Center for Biotechnology Information Sequence Read Archive (BioProject ID: PRJNA498691; BioSample accession nos. SAMN10330286, SAMN10330287, SAMN10330288, SAMN10330289, SAMN10330290, and SAMN10330291). Basecalls and demultiplexing were performed with Illumina’s bcl2fastq software and a custom python demultiplexing program with a maximum of one mismatch in the indexing read. RNA-seq reads were then aligned using STAR aligner (Dobin et al., 2013). Gene counts were derived from the number of uniquely aligned unambiguous reads by Subread:featureCount (Liao et al., 2014) and subsequently analyzed using DESeq2 (Love et al., 2014). GSEAs were performed using the GenePattern module GSEAPreranked.

Mice were infected by s.s. with WT CPXV. At 6 dpi, infected mice and naive controls were injected i.p. with 200 µg/mouse of α-NK1.1. At 7 dpi, DC2.4 cells were mock-infected or infected with CPXV (multiplicity of infection of 2) and incubated at 37°C for 2.5 h. Cells were then washed with PBS and costained with CellTrace Far Red dye and either 20 µM, 4 µM, or 0.8 µM CellTrace Violet dye (Thermo Fisher). Labeled cells were transferred i.v. (10⁶ cells per group) into naive mice or day 7–infected mice. After 4 h, the lungs were harvested as described above and cells were examined by flow cytometry. Far Red⁺ cells were gated, and the percent killing was calculated using the following formula: 100 − ([(% infected cells in infected mice/% mock-infected in infected mice)/(% infected cells in naive mice/% mock-infected in naive mice)] × 100).

The data were analyzed with an unpaired Student’s t test or one-way ANOVA followed by Tukey’s post-test comparison using Prism GraphPad software.

Fig. S1 shows that skin CD8⁺ T cells up-regulate local IL-10 production during acute CPXV infection. Fig. S2 shows that local antigen recognition up-regulates CD69 and Nur77 expression by CD8⁺ T cells. Fig. S3 shows that CD8⁺ T_RMs develop without continuous recruitment of CD8⁺ T cells during acute CPXV infection. Fig. S4 shows that Flu-B8 respiratory infection generates B8_19–26-specific CD8⁺ T_RMs in the lungs. Fig. S5 shows that the number of CD8⁺ T_RMs is reduced in B cell–deficient mice.

We thank Sathi Wijeyesinghe and Dr. David Masopust for their technical advice regarding IV staining. We also gratefully acknowledge Dr. Herbert Virgin, Dr. Brian Edelson, Dr. David Sibley, and Dr. Haina Shin for their helpful suggestions on this study. We thank the GTAC, Department of Genetics, Washington University School of Medicine for help with RNA-seq analysis.

This work was supported by the National Institutes of Health (NIH; grant U19-AI109948 and the Initiative for Maximizing Student Development grant R25GM103757), the Chancellor’s Graduate Fellowship, and the American Society for Microbiology Robert D. Watkins Graduate Research Fellowship. The GTAC is partially supported by the National Cancer Institute (Cancer Center Support grant P30 CA91842 to the Siteman Cancer Center) and by the National Center for Research Resources, a component of the NIH (ICTS/CTSA grant ULTR002345), and NIH Roadmap for Medical Research. This publication is solely the responsibility of the authors and does not necessarily represent the official view of the National Center for Research Resources or NIH.

The authors declare no competing financial interests.

Author contributions: E.J. Lauron, L. Yang, I.B. Harvey, D.K. Sojka, G.D. Williams, M.A. Paley, M.D. Bern, E. Park, F. Victorino, A.C.M. Boon, and W.M. Yokoyama designed, analyzed, and performed experiments. E.J. Lauron and W.M. Yokoyama wrote the manuscript.