Ptpn2 and KLRG1 regulate the generation and function of tissue-resident memory CD8⁺ T cells in skin

Peripheral organs and lymphoid tissues harbor large numbers of nonrecirculating memory T cells, many of which are generated in response to pathogens and commensal microbes (Sathaliyawala et al., 2012; Steinert et al., 2015; Gebhardt et al., 2018). These cells are now commonly referred to as tissue-resident memory T (T_RM) cells and, over the past decade, have emerged as key mediators of tissue immunity. Accordingly, T_RM cells afford rapid and potent protection from viral, bacterial, fungal, and parasitic pathogens and drive local control of latent-reactivating and persisting infections (Park and Kupper, 2015; Gebhardt et al., 2018). Recent preclinical and translational studies have further implicated T_RM cells in tumor control and response to checkpoint blockade immunotherapy, with elevated numbers of T_RM cells linked to better patient survival across a broad spectrum of cancers (Smazynski and Webb, 2018; Park et al., 2019b). Aberrant T_RM responses, on the other hand, are associated with tissue pathology in the context of autoimmune inflammation, allergy, and transplant rejection (Clark, 2015; Park and Kupper, 2015; Turner et al., 2014). Thus, the therapeutic manipulation of T_RM responses carries great potential for the prevention and treatment of a broad spectrum of diseases. Any such attempts, however, will require a profound understanding of the molecular mechanisms that operate during the formation and maintenance of T_RM cells and that determine their function.

The epithelial layers of skin, small intestinal mucosa, and other organs are populated by CD8⁺ T_RM cells that commonly express surface markers such as CD69 and CD103 and share core transcriptional signatures that distinguish them from recirculating T cells (Mackay et al., 2016; Mackay et al., 2013). These T_RM cells can develop from precursors found within the killer cell lectin-like receptor G1 (KLRG1)–negative pool of CD8⁺ T cells expanded in response to acute pathogen encounter (Mackay et al., 2013; Sheridan et al., 2014), although a recent study using a fate-mapping fluorescent reporter system indicated that some effector CD8⁺ T cells may switch off KLRG1 expression and give rise to various subsets of memory cells, including T_RM cells (Herndler-Brandstetter et al., 2018). Regardless, in the case of viral skin infection, CD69⁺ KLRG1⁻ effector cells are highly enriched in the epidermis early after onset of T cell infiltration, while at the same time their KLRG1⁺ counterparts dominate the effector pool in spleen and blood (Mackay et al., 2013). Thus, selective accumulation of CD69⁺ KLRG1⁻ cells in the epidermal layer establishes an early local pool of effector cells with potential to differentiate into CD69⁺ CD103⁺ T_RM cells upon resolution of infection. Interestingly, both KLRG1 and the αEβ7 integrin (the latter incorporating the αE subunit CD103) share one of their major ligands, E-cadherin, a molecule expressed not only by epithelial cells, such as keratinocytes and Langerhans cells (Gründemann et al., 2006), but also by T_RM cells (Hofmann and Pircher, 2011; Mackay et al., 2013). While both CD103 and E-cadherin can promote T_RM cell formation and maintenance (Casey et al., 2012; Hofmann and Pircher, 2011; Mackay et al., 2013; Sheridan et al., 2014), it has remained unclear if simultaneous repression of KLRG1 expression is also necessary for the establishment of T_RM cells.

The extent to which T_RM cells are generated following skin infection is determined by diverse factors and interconnected molecular processes, including the generation of KLRG1⁻ T_RM cell precursors in skin-draining lymph nodes, precursor infiltration of peripheral tissues, local antigen reencounter, repression of tissue exit receptors, chemokine-directed epidermal migration, and, critically, exposure to locally produced cytokines such as TGFβ and IL-15 (Gebhardt et al., 2018). These processes are regulated and fine-tuned at the molecular level as to ensure transition into long-lived memory cells and adjustment of T_RM cell functionality according to the tissue of residence. The protein tyrosine phosphatase nonreceptor type 2 (Ptpn2) is an important negative regulator of CD8⁺ T cell responses in the context of autoimmunity, lymphopenia, cancer, and chronic infection (LaFleur et al., 2019; Wiede et al., 2014a; Wiede et al., 2020; Wiede et al., 2011; Wiede et al., 2014b). Ptpn2 attenuates TCR signaling by dephosphorylating TCR-proximal kinases, such as LCK and FYN, which results in an increased T cell activation threshold and, consequently, reduced responsiveness to low-affinity antigen (Wiede et al., 2011; Wiede et al., 2014a; Wiede et al., 2014b). Furthermore, Ptpn2 dephosphorylates and inactivates components of the JAK-STAT signaling pathways, thereby attenuating cytokine signaling employed by members of the IFN and common ɣ-chain cytokine families, such as type I and II IFN, IL-2, IL-15, and others (Simoncic et al., 2002; Aoki and Matsuda, 2002; Wiede et al., 2011; Wiede et al., 2014a; ten Hoeve et al., 2002; LaFleur et al., 2019). Thus, Ptpn2 appears likely to operate as a rheostat for T_RM cell development and function, although its role in T_RM cell biology has remained largely unexplored.

Here, we investigated the role of Ptpn2 in T_RM cell formation using a model of acute HSV skin infection. We describe that absence of Ptpn2 expression in CD8⁺ T cells results in perturbed effector differentiation with predominant production of short-lived KLRG1⁺ effectors and impaired generation of epidermal T_RM cells. Interestingly, forced expression of KLRG1 alone compromises T_RM development, while KLRG1⁻ effector cells with memory potential develop normally into T_RM cells independently of Ptpn2 expression. Finally, we show that Ptpn2 is expressed by resting T_RM cells and functions to restrict T_RM responses to both virus infection and innocuous model antigens. Thus, our study describes the molecules Ptpn2 and KLRG1 as new regulatory factors for the development and function of T_RM cells and highlights the potential of therapeutic inhibition of Ptpn2 activity as a way to amplify T_RM cell–mediated immunity.

To test whether Ptpn2 impacted CD8⁺ T cell responses to an acute viral infection, we tracked Ova-specific CD8⁺ T (OT-I) cells adoptively transferred into mice infected on the skin with an HSV-1 strain expressing the model antigen Ova (HSV-Ova). In this setting, naive OT-I cells generate strong effector responses that ultimately give rise to recirculating memory T cells as well as long-lived CD69⁺CD103⁺ T_RM cells concentrated at the site of resolved infection (Mackay et al., 2013). To allow for endogenous T cell responses to commence unperturbed and thus provide a level of competition with the transgenic T cells, which we presumed would best reveal any regulatory impact of Ptpn2 on CD8⁺ T cell expansion and activation, we transferred control (Ptpn2^fl/fl) or genetically Ptpn2-deficient (Lck-Cre;Ptpn2^fl/fl) naive OT-I cells 2 d after infection (Wiede et al., 2011). While both populations expanded robustly, numbers of OT-I.Lck-Cre;Ptpn2^fl/fl cells recovered from spleen 1 and 2 wk after infection were increased on average by ∼2.7- and 3.7-fold compared with mice that received control OT-I.Ptpn2^fl/fl cells, respectively (Fig. 1 A). Numbers of OT-I cells recovered from the spleen were very low for both genotypes at later time points. By contrast, fewer OT-I.Lck-Cre;Ptpn2^fl/fl than OT-I.Ptpn2^fl/fl cells were recovered from skin at 2 and 4 wk after inoculation when virus infection had resolved (van Lint et al., 2004; Fig. 1 A). This reduction in numbers was particularly evident for the population of CD69⁺ CD103⁺ T_RM cells that evolved with slower kinetics and overall less efficiency from Ptpn2-deficient OT-I cells compared with control OT-I cells (Fig. 1, A and B). In line with this, histological examination and flow cytometric enumeration of OT-I cells in mechanically separated epidermis and dermis revealed that OT-I.Lck-Cre;Ptpn2^fl/fl cells were sparse in the epithelial compartments where CD69⁺ CD103⁺ skin T_RM cells arise (i.e., epidermis and hair follicle epithelium; Gebhardt et al., 2009; Mackay et al., 2013; Fig. 1, A, C, and D). This apparent defect in epidermal entry and/or retention did not appear to be caused by an inability of Ptpn2-deficient OT-I cells to respond to the T_RM-instructive factor TGFβ, as both OT-I.Lck-Cre;Ptpn2^fl/fl and control OT-I cells activated by peptide-pulsed splenocytes in vitro strongly up-regulated CD103 in response to this cytokine (Fig. S1 A).

Recognition of cognate antigen in infected skin promotes the formation of T_RM cells, although the molecular mechanisms involved are only incompletely understood (Gebhardt et al., 2009; Khan et al., 2016; Muschaweckh et al., 2016). Given the established function of Ptpn2 in dampening TCR signaling, we asked whether local antigen recognition in infected skin impacted T_RM cell formation by OT-I.Lck-Cre;Ptpn2^fl/fl cells. To this end, we infected mice with both HSV-Ova and WT HSV-1 (HSV-WT, not expressing Ova) strains on opposite flanks, followed by transfer of naive OT-I.Lck-Cre;Ptpn2^fl/fl or control OT-I.Ptpn2^fl/fl cells (Fig. 1 D). Confirming our previous result from a single infection, OT-I.Lck-Cre;Ptpn2^fl/fl cells generated fewer CD69⁺ CD103⁺ T_RM cells across both flanks when analyzed 30 d later (Fig. 1 E). As expected, antigen presence increased CD69⁺ CD103⁺ T_RM cell formation by control OT-I.Ptpn2^fl/fl cells (Fig. 1 E). By contrast, Ptpn2-deficient OT-I cells generated comparable numbers of T_RM cells in the presence (HSV-Ova) or absence (HSV-WT) of local antigen (Fig. 1 E). Thus, Ptpn2-deficient effector cells infiltrating infected skin failed to respond to important environmental cues such as local antigen stimulation to support the formation of long-lived epidermal memory cells. Interestingly, skin epithelial-infiltrating CD103⁺ OT-I.Lck-Cre;Ptpn2^fl/fl cells were similarly reduced in the presence (HSV-Ova) or absence (HSV-WT) of antigen early after infection (day 9; Fig. S1 B). This defect likely reflected a more general impairment in overall memory potential in the pool of Ptpn2-deficient effector cells infiltrating infected skin, independent of potentially negative effects of excessive local TCR stimulation on memory formation in absence of Ptpn2. Regardless, the results thus far established that OT-I.Lck-Cre;Ptpn2^fl/fl cells were impaired in forming epidermal CD69⁺ CD103⁺ T_RM cells in response to an acute skin infection.

The surface molecules CD127 (IL-7Rα) and KLRG1 are often used as markers to predict the memory potential of activated CD8⁺ T cells, with short-lived effector cells identified as CD127^lo KLRG1⁺ and long-lived memory precursors as CD127^hi KLRG1⁻ cells (Joshi et al., 2007; Sarkar et al., 2008), the latter of which also contain cells that can give rise to CD69⁺ CD103⁺ T_RM cells (Mackay et al., 2013; Sheridan et al., 2014). Given that Ptpn2 has recently been identified as a regulator of CD8⁺ T cell expansion and differentiation during systemic infection (LaFleur et al., 2019; Flosbach et al., 2020), we next compared the phenotype of effector T cells arising in the presence or absence of Ptpn2 expression around the peak of expansion after acute peripheral infection. Flow cytometric analysis of OT-I cells isolated from spleens 9 d after HSV single or double infection revealed a marked reduction in the frequency of KLRG1^– and CD127^hi memory precursors among Lck-Cre;Ptpn2^fl/fl cells when compared with Ptpn2^fl/fl controls (Fig. 1, G–I), although the absolute numbers of KLRG1⁻ effector cells did not differ between the two genotypes (Fig. S1 C). The decrease in the proportion of KLRG1^–Lck-Cre;Ptpn2^fl/fl OT-I cells in spleen was also evident in infected skin. Whereas the vast majority of Ptpn2^fl/fl control cells in skin displayed a KLRG1⁻ phenotype (91.2 ± 1.7%), as previously reported (Mackay et al., 2013), Lck-Cre;Ptpn2^fl/fl OT-I cells comprised a ∼2:1 mixture of KLRG1⁻ and KLRG1⁺ cells (Fig. 1, G and H). This was echoed by an on average 2.4-fold reduction of KLRG1⁻Lck-Cre;Ptpn2^fl/fl OT-I cells compared with control Ptpn2^fl/fl OT-I effector cells in skin (Fig. S1 D). Thus, absence of Ptpn2 expression promoted the expansion and differentiation of KLRG1⁺ CD127^lo effector cells, which caused an increase in the proportion of KLRG1⁺ cells in infected skin and an accompanying reduction in numbers of KLRG1⁻ cells.

In combination with other markers, KLRG1 is often used as a surrogate for terminally differentiated effector T cells with limited proliferative and memory potential (Joshi et al., 2007; Sarkar et al., 2008; Voehringer et al., 2001; Voehringer et al., 2002), although its functional role in the development of recirculating memory cells after systemic infection appears to be minimal (Gründemann et al., 2010). T_RM cells in the epidermal layer of skin develop from KLRG1⁻ precursors, and these dominate the epidermal-infiltrating effector population from the earliest times after infection (Mackay et al., 2013), implying selective entry of KLRG1⁻ cells and/or suppression of KLRG1 expression by locally produced factors such as TGFβ (Schwartzkopff et al., 2015). Of note, KLRG1 and the α_Eβ₇ integrin (incorporating the α_E-chain CD103) both bind to E-cadherin expressed by epithelial cells such as keratinocytes (Cepek et al., 1994; Gründemann et al., 2006; Hofmann and Pircher, 2011), and both CD103 and E-cadherin are required for optimal generation of T_RM cells in a number of organs (Casey et al., 2012; Hofmann and Pircher, 2011; Mackay et al., 2013; Sheridan et al., 2014). Therefore, we speculated that T_RM cell generation might rely on repression of KLRG1 expression, which in part could have explained our findings with Lck-Cre;Ptpn2^fl/fl OT-I cells.

To address this, we transduced in vitro activated TCR-transgenic CD8⁺ T cells (gBT-I) specific for a HSV-derived epitope in glycoprotein B (gB) with a retroviral vector encoding for KLRG1 and GFP (pMIG.Klrg1-WT; Fig. 2 A). As a control, gBT-I cells were transduced with a control vector encoding GFP, but not KLRG1 (pMIG.Ctrl). Of note, T cell activation with gB peptide–pulsed splenocytes in presence of IL-2 yields activated gBT-I cells with considerable potential to form long-lived memory cells (Mackay et al., 2012) and does not induce KLRG1 expression in the absence of IL-12 supplementation. Accordingly, KLRG1-transduced cells expressed KLRG1 on their surface, whereas cells transduced with the control vector were negative for KLRG1 expression (Fig. 2 B). On day 3 after transduction, GFP⁺ gBT-I cells were sorted and transferred into mice, followed by treatment of flank skin with 1-fluor-2,4-dinitrobenzol (DNFB) to attract cells into the skin for T_RM cell formation, as previously described (Mackay et al., 2012). 4 wk after DNFB treatment, we recovered similar numbers of KLRG1-overexpressing and control gBT-I cells from spleen (Fig. 2 C). By contrast, KLRG1-expressing gBT-I cells in the DNFB-treated skin, and in particular CD69⁺ CD103⁺ T_RM cells, were significantly reduced (Fig. 2 C).

To test whether forced KLRG1 expression also affected T_RM cell formation in response to skin infection, we generated bone marrow chimeric mice as a source of naive pMIG.Klrg1-overexpressing and control gBT-I cells by reconstituting lethally irradiated mice with gBT-I bone marrow cells transduced with pMIG.Klrg1-WT or pMIG control vectors (Fig. 2 D). 8 wk after bone marrow transfer, KLRG1 expression by gBT-I cells in the chimeras was confirmed by surface staining (Fig. 2 E). Naive GFP⁺ gBT-I cells were isolated and sorted from lymph nodes and spleens and transferred into recipient mice 1 d after HSV skin infection (Fig. 2 D). Analysis of GFP⁺ gBT-I cells >4 wk after infection revealed a modest 1.4-fold increase of pMIG.Klrg1-WT transduced gBT-I cells in spleen, whereas CD69⁺ CD103⁺ T_RM cell numbers and frequencies in skin were significantly reduced (Fig. 2, F and G). KLRG1 expression levels in splenic gBT-I cells were comparable to expression levels found on control effector gBT-I cells, largely excluding that nonphysiological KLRG1 expression levels may have caused impaired T_RM cell formation (Fig. 2 E). Furthermore, the vast majority of pMIG.Klrg1-WT transduced gBT-I cells in spleen retained high KLRG1 surface expression (Fig. 2 F). By contrast, their T_RM cell counterparts in skin displayed evidence for partial down-modulation of KLRG1 expression both at the population and single-cell level (Fig. 2 F), implying selection pressure against KLRG1 expression specifically for T_RM cells. Combined, these results established that forced expression of KLRG1 inhibited T_RM cell formation in skin in settings of inflammation and viral infection.

We next set out to address the mechanisms by which KLRG1 can inhibit T_RM cell formation in skin. Previous studies have shown that KLRG1 and CD103 are mutually exclusively expressed in the majority of effector CD8⁺ T cells infiltrating infected skin and intestinal mucosa (Mackay et al., 2013; Sheridan et al., 2014). Therefore, we speculated that forced expression of KLRG1 may impact CD103 expression and tested this by exposing KLRG1-overexpressing cells to TGFβ in vitro. Nevertheless, we found normal CD103 up-regulation, ruling out a general defect in expression of this integrin subunit by KLRG1⁺ cells (Fig. S1 E). Binding of KLRG1 to E-cadherin induces an ITIM–mediated signaling cascade that results in inhibition of TCR signaling and T cell proliferation (Beyersdorf et al., 2001; Tessmer et al., 2007; Henson et al., 2009). To test whether KLRG1-ITIM signaling was involved in the inhibition of T_RM cell formation, we transduced gBT-I cells with vectors encoding for signaling-deficient KLRG1 proteins and tested their ability to form T_RM cells as above (Fig. 2 A). KLRG1-Y7F harbors a point mutation in its ITIM that abrogates recruitment of phosphatases and consequently results in reduced inhibitory capacity (Tessmer et al., 2007; Rosshart et al., 2008). KLRG1-C61-72 harbors four point mutations replacing cysteines in its stalk region crucial for multimerization, which prevents multivalent E-cadherin binding and inhibitory signaling (Hofmann et al., 2012). Flow cytometric analysis confirmed KLRG1 expression on GFP⁺ cells transduced with WT KLRG1, KLRG1-Y7F, and KLRG1-C61-72, although we noted somewhat lower expression of KLRG1-Y7F and KLRG1-C61-72 compared with KLRG1-WT (Fig. 2 B). 3 wk after transfer and skin treatment with DNFB, similar numbers of all types of gBT-I cells were recovered from spleens (Fig. 2, H and I). By contrast, gBT-I cells expressing ITIM mutant or multimerization-deficient KLRG1 molecules displayed a heightened capacity to generate T_RM cells, which manifested in increased numbers and frequencies of CD69⁺ CD103⁺ cells in skin compared with gBT-I cells expressing WT KLRG1 (Fig. 2, H and I). These results implied that signaling via ITIM in the cytoplasmatic domain of KLRG1 was involved in inhibiting T_RM cell formation in the skin, and they inferred that increased KLRG1 expression by Lck-Cre;Ptpn2^fl/fl CD8⁺ T cells in our earlier experiments was one factor explaining reduced T_RM cell numbers in the absence of Ptpn2 expression.

Given the important role of Ptpn2 in regulating T cell differentiation in response to chronic systemic infection (LaFleur et al., 2019), we considered that molecular pathways other than KLRG1–ITIM signaling were also involved in restraining the T_RM cell potential of Ptpn2-deficient CD8⁺ T cells activated in response to skin infection. To gain more detailed insights into effector differentiation in absence of Ptpn2, we FACS-purified Lck-Cre;Ptpn2^fl/fl and Ptpn2^fl/fl OT-I cells from spleens of acutely infected mice (as in Fig. 1) and subjected total populations, as well as KLRG1⁻ and KLRG1⁺ subsets, to mRNA sequencing (Fig. 3 A). As expected, we found a large number of differentially expressed genes (DEGs; |log2 (fold change)| >1; false discovery rate <0.05) between KLRG1⁻ and KLRG1⁺ subsets of both Lck-Cre;Ptpn2^fl/fl (712 DEG) and Ptpn2^fl/fl (528 DEG) OT-I cells. Furthermore, we found 190 DEGs between total populations of Ptpn2-deficient and control OT-I cells (Fig. 3, B and C; and Table S1). Interestingly, genes with differential expression in total Lck-Cre;Ptpn2^fl/fl OT-I cells included those coding for skin-homing and other chemokine receptors, such as Ccr4, Ccr10, Ccr7, and Cxcr5, the memory-associated transcription factors Tcf7 and Id3, and genes that form part of transcriptional signatures for T_RM cells in various tissues, such as Xcl1, Qpct, Itgae (encoding CD103), P2rx7, and S1pr5 (Mackay et al., 2013; Stark et al., 2018). Gene ontology (GO)–term analysis of DEG (|log2 (fold change)| >0.5) revealed enrichment of pathways associated with cell cycle/division and, conversely, underrepresentation of those associated with immune system processes, signal transduction, and cytokine/chemokine receptor activity in Lck-Cre;Ptpn2^fl/fl relative to Ptpn2^fl/fl OT-I cells (Figs. S2, S3, and S4). This was consistent with superior early expansion but impaired skin infiltration and T_RM cell formation observed for Lck-Cre;Ptpn2^fl/fl OT-I cells in our initial experiments (Fig. 1).

We next asked whether the subpopulations of KLRG1⁻ and KLRG1⁺ cells differed depending on their expression of Ptpn2. Compared with the 190 DEGs for the total effector populations, considerably fewer genes were differentially expressed between Lck-Cre;Ptpn2^fl/fl and Ptpn2^fl/fl OT-I cells among the KLRG1⁻ (93 DEGs) and KLRG1⁺ (59 DEGs) subsets (Fig. 3, B and C; Table S2; and Table S3). Hierarchical clustering analysis further revealed clustering of KLRG1⁻ cells of both genotypes with total Ptpn2^fl/fl OT-I cells, whereas KLRG1⁺ cells of both genotypes clustered with total Lck-Cre;Ptpn2^fl/fl OT-I cells (Fig. 3 D), indicating that in vivo–activated Ptpn2-deficient OT-I cells displayed a transcriptional bias toward KLRG1⁺ effectors. Gene set enrichment analysis further showed that differential gene profiles between Lck-Cre;Ptpn2^fl/fl and Ptpn2^fl/fl OT-I cells for all subfractions (total, KLRG1⁺, and KLRG1⁻) were enriched significantly for previously identified effector versus memory CD8⁺ T cell signature genes (Kaech et al., 2002; Luckey et al., 2006; Fig. 3 E). Taken together, these findings further support the above suggestion that after in vivo priming following HSV infection, Ptpn2 deficiency resulted in the preferential generation of OT-I cells with a transcriptional profile biased toward more terminally differentiated KLRG1⁺ effector cells with restricted memory potential.

We next compared the capacities of purified KLRG1⁺ and KLRG1⁻ effector cells of both genotypes to form T_RM cells. To this end, we isolated and sorted total, KLRG1⁺, and KLRG1⁻ OT-I cells from spleens 8 d after infection and transferred these into new recipient mice infected with HSV-Ova 4 d earlier, followed by analysis of T cell numbers in the skin 30 d after infection (Fig. 4 A). Upon transfer of total effector cells, we recovered 2.2-fold more Ptpn2^fl/fl than Lck-Cre;Ptpn2^fl/fl OT-I cells from the skin of recipient mice (Fig. 4 B), echoing the differences after transfer of naive OT-I cells (Fig. 1 A). By contrast, purified KLRG1⁻ effector cells generated comparable numbers of total and CD69⁺ CD103⁺ T_RMPtpn2^fl/fl, and Lck-Cre;Ptpn2^fl/fl OT-I cells in skin, whereas purified KLRG1⁺ cells of both genotypes failed to give rise to T_RM cells, consistent with previous reports (data not shown; Mackay et al., 2013). Confirming these results, we also found that OT-I cells of both genotypes activated in vitro under conditions that do not induce KLRG1 expression generated comparable numbers of T_RM cells following direct transfer into the skin by intradermal (i.d.) injection (Fig. 4, C and D). Interestingly, numbers of Lck-Cre;Ptpn2^fl/fl OT-I in spleen were moderately increased compared with Ptpn2^fl/fl OT-I cells in this setting (Fig. 4 D). Combined, these results revealed that Ptpn2-deficient KLRG1⁻ effector cells have a normal capacity to form T_RM cells and further suggested that the defect in T_RM cell formation seen in our previous experiments was primarily caused by the relative reduction of KLRG1⁻ cells among the effector pool in spleen and skin (Fig. 1, G–I).

Finally, we asked whether Ptpn2 impacted the function of T_RM cells in skin. Flow cytometric analysis confirmed sustained Ptpn2 expression by T_RM cells generated following skin infection with HSV or by i.d. deposition of in vitro–activated CD8⁺ T cells (Fig. 5 A). The latter approach offered an opportunity to examine the role of Ptpn2 in regulating T_RM cell–mediated local immunity. Thus, in a first set of experiments, we generated Lck-Cre;Ptpn2^fl/fl or Ptpn2^fl/fl OT-I T_RM cells in skin of K5-rTA;TetO.mOva mice in which expression of membrane-bound Ova (mOva) by skin keratinocytes can be induced following treatment with doxycycline (Dox; Fig. 5 B). Upon induction of Ova, T_RM cells in skin are the first to recognize local antigen and precipitate an inflammatory response with appearance of skin lesions focused at sites of highest T_RM cell density (Fig. S5 A and data not shown). While the time of disease onset in the first week after Ova induction was comparable between the cohorts harboring Lck-Cre;Ptpn2^fl/fl or Ptpn2^fl/fl T_RM cells, disease progression and overall severity were more pronounced in mice with Ptpn2-deficient OT-I cells (Fig. 5 C). In line with this, Lck-Cre;Ptpn2^fl/fl CD69⁺ CD103⁺ T_RM cells produced more granzyme B than their Ptpn2^fl/fl counterparts at the onset of disease (Figs. 5 D and S5 B). In support of the notion that Ptpn2 fine-tunes T_RM cell responses to antigen, separate experiments with WT mice revealed strongly elevated production of granzyme B and IFNɣ by Lck-Cre;Ptpn2^fl/fl CD69⁺ CD103⁺ T_RM cells in response to low-dose i.d. peptide challenge (Fig. S5 C).

To further refine the role of T_RM cells in dermatitis, we generated Lck-Cre;Ptpn2^fl/fl or Ptpn2^fl/fl OT-I T_RM cells in skin of K5-rTA;TetO.mOva mice and depleted recirculating CD8⁺ T cells using anti-CD8 depleting antibody, leaving the skin T_RM cell population largely intact (Hobbs and Nolz, 2019; data not shown). 9 d later, Ova expression was induced by Dox treatment (Fig. 5 E). As shown in Fig. 5 F, dermatitis developed in the absence of recirculating OT-I cells, confirming that T_RM cells were key drivers of disease. Furthermore, dermatitis progression and severity were significantly more pronounced in mice harboring Ptpn2-deficient T_RM cells, again emphasizing an important role of Ptpn2 in regulating T_RM cell function. Thus, consistent with the established role of Ptpn2 in preventing CD8⁺ T cell–mediated autoimmune disease in models of type I diabetes (Wiede et al., 2019; Wiede et al., 2014b), lack of Ptpn2 expression specifically in T_RM cells was sufficient to cause aggravated skin inflammation in a dermatitis model.

In a final set of experiments, we generated Lck-Cre;Ptpn2^fl/fl or Ptpn2^fl/fl OT-I T_RM cells in WT mice by i.d. injection of in vitro activated OT-I T cells and challenged them with HSV-Ova skin infection at the site of initial T_RM cell deposition (Fig. 5 G). We have previously shown in related experiments that T_RM cells afford a level of rapid protection from HSV challenge infection potent enough to prevent the development of characteristic zosteriform skin lesions (Gebhardt et al., 2009; Mackay et al., 2012). When we measured viral titers in skin harboring Ptpn2-deficient or control T_RM cells 4 d after infection, we observed a 100-fold reduction in viral load in skin with Ptpn2-deficient OT-I cells (Fig. 5 H). Combined, these results demonstrate an important role of Ptpn2 beyond the regulation of T cell effector differentiation; that is, restraining T_RM cell–mediated immune responses in the context of skin inflammation and infection.

T_RM cells are key determinants of effective tissue immunity and as such are promising targets for immunotherapeutic intervention and vaccination in a broad spectrum of diseases, including infection, chronic inflammation, cancer, autoimmunity, and organ transplantation. A detailed understanding of the molecular pathways and environmental factors that regulate their generation, persistence, and function is therefore critical to the rational development of novel immunotherapies aimed at modulating tissue immunity. In the present study, we identified two T cell–intrinsic molecules as new regulators of T_RM cell biology in skin; namely, the protein tyrosine phosphatase Ptpn2 and the putatively immunoregulatory receptor, KLRG1.

The function of Ptpn2 in regulating intracellular signaling is critical to the establishment and maintenance of immune tolerance. This is most evident from the observation that genetic deletion of Ptpn2 in immune cells, T cells, or adoptively transferred CD8⁺ T cells promotes systemic inflammation and autoimmunity, respectively (Wiede et al., 2011; Wiede et al., 2019; Wiede et al., 2014b; Heinonen et al., 2004). Furthermore, single-nucleotide polymorphisms in the human PTPN2 gene have been linked to development of autoimmunity and chronic inflammatory disorders in patients (Todd et al., 2007; Smyth et al., 2008). Consistent with its molecular function in attenuating TCR and cytokine receptor signaling (Aoki and Matsuda, 2002; Simoncic et al., 2002; ten Hoeve et al., 2002; Wiede et al., 2014a; Wiede et al., 2011), Ptpn2 has previously been shown to restrain CD8⁺ T cell responses to self-antigen, tumors, and acute and chronic systemic infections, although recirculating memory CD8⁺ T cells can from normally after acute bacterial infection in absence of Ptpn2 (LaFleur et al., 2019; Wiede et al., 2014a; Wiede et al., 2020; Wiede et al., 2011; Wiede et al., 2014b; Flosbach et al., 2020). We describe a similar function of Ptpn2 in regulating effector CD8⁺ T cell differentiation during an acute virus infection of skin. We show that absence of Ptpn2 expression in CD8⁺ T cells results in transient amplification of T cell expansion early after infection, with a strong numerical increase seen predominately for KLRG1⁺ effector cells and an overall transcriptional bias reflecting increased proliferation and decreased memory potential relative to Ptpn2-competent cells. Of note, our results reveal an important consequence of perturbed effector T cell differentiation in the absence of Ptpn2 that is strongly reduced seeding of skin with epidermal-infiltrating KLRG1⁻ memory precursor cells and, consequently, diminished formation of T_RM cells at the site of resolved infection. Interestingly, these defects echo transcriptional changes observed already in splenic Ptpn2-deficient T cells before they infiltrate the skin, including down-regulation of genes encoding for skin-targeting chemokine receptors, such as Ccr4 and Ccr10, and perturbed expression of genes associated with T_RM cell–specific transcriptional signatures, such as Xcl1, Qpct, Itgae, P2rx7, and S1pr5 (Mackay et al., 2013; Fernandez-Ruiz et al., 2016). Thus, a large fraction of Ptpn2-deficient effector cells in infected skin may be transcriptionally poised in such a way that prohibits commitment to the T_RM cell differentiation program. This could also explain their failure to respond appropriately to key environmental cues such as the presence of cognate antigen presented by infected tissue cells (Gebhardt et al., 2009; Khan et al., 2016; Muschaweckh et al., 2016). Furthermore, Ptpn2 deficiency results in a proportional increase in KLRG1⁺ effector cells in skin, and we show that forced expression of KLRG1 alone is sufficient to interfere with T_RM cell differentiation.

Expression of KLRG1 identifies terminally differentiated effector CD8⁺ T cells with reduced proliferative capacity and memory potential (Joshi et al., 2007; Sarkar et al., 2008; Voehringer et al., 2001; Voehringer et al., 2002). Nevertheless, its use as memory marker is of limited value in some infections (Croom et al., 2011) and may further be complicated by the more recent demonstration that KLRG1 expression is more dynamic than initially anticipated such that some KLRG1⁺ effector T cells may lose expression of KLRG1 and become long-lived memory cells (Schwartzkopff et al., 2015; Herndler-Brandstetter et al., 2018). Ligation of KLRG1 in vitro has been shown to dampen proliferation and induction of cytolytic function in CD8⁺ T cells (Gründemann et al., 2006). Somewhat surprisingly, however, its genetic deletion has no bearing on splenic memory CD8⁺ T cells generated in response to systemic virus infection (Gründemann et al., 2010). In line with this, we show that forced expression of KLRG1 is inconsequential for the generation of memory T cells in the spleen. Importantly, however, we demonstrate that KLRG1 selectively interferes with the generation of T_RM cells in the epidermis, where one of its major ligands, E-cadherin, is expressed constitutively by keratinocytes, Langerhans cells, and T_RM cells (Gründemann et al., 2006; Mackay et al., 2013; Hofmann and Pircher, 2011).

KLRG1 shares its ligand E-cadherin with the CD103-incorporating integrin α_Eβ₇ (Cepek et al., 1994; Gründemann et al., 2006). Interestingly, expression of all three molecules is regulated by the same T_RM cell–instructive cytokine, TGFβ, which induces CD103 and E-cadherin expression in activated CD8⁺ T cells and represses or down-regulates that of KLRG1 (Kilshaw and Murant, 1990; Schwartzkopff et al., 2015; Nath et al., 2019). Of note, T cell–intrinsic CD103 and E-cadherin both support optimal generation and/or maintenance of T_RM cells by mediating tethering to neighboring cells and promoting survival and functionality via outside-in signaling (Gebhardt et al., 2018). Binding sites for KLRG1 and α_Eβ₇ on E-cadherin are distinct, making sterical hindrance unlikely (Li et al., 2009). However, KLRG1 may exert its inhibitory activity, at least in part, by interfering with signals mediated by CD103 binding to E-cadherin, which may occur either in cis, within the T cells, or in trans, involving interactions of T cells with neighboring epithelial cells. Since KLRG1 expression levels after retroviral transduction were comparable in magnitude with those seen in normal cells responding to virus infection in vivo, it appears unlikely that nonphysiological expression levels caused the observed phenomena. Furthermore, we provide evidence that KLRG1 signaling via its cytoplasmic ITIM contributes to the inhibition of T_RM cell formation, which is consistent with previous studies demonstrating that ITIM-mediated signaling can inhibit TCR signaling and T cell proliferation (Beyersdorf et al., 2001; Tessmer et al., 2007; Henson et al., 2009). The signaling-deficient KLRG1-Y7F construct used in our experiments harbors a single point mutation that specifically abrogates ITIM-mediated recruitment of phosphatases (Tessmer et al., 2007; Rosshart et al., 2008), whereas KLRG1–C61-72 has four point mutations that are located to the stalk region crucial for multimerization and therefore prevent multivalent E-cadherin binding and inhibitory signaling (Hofmann et al., 2012). Importantly, both KLRG1 mutants exert weaker inhibitory activity on T_RM cell formation, inferring that KLRG1–ITIM signaling hinders T_RM cell formation. It should be noted, however, that expression levels of mutant KLRG1 proteins were slightly lower compared with WT KLRG1 and that monomeric KLRG1–C61-72–E-cadherin interactions are of lower affinity (Hofmann et al., 2012). Thus, binding to E-cadherin and putative interference with CD103 signaling may have also been weaker for mutant KLRG1 molecules. Regardless, the combined evidence suggests that repression of KLRG1 is important for normal differentiation of CD69⁺ CD103⁺ T_RM cells in skin and further implies that increased expression of KLRG1 seen in Ptpn2-deficient effector cells contributes to their impairment in T_RM cell formation.

In chronic infection with lymphocytic choriomeningitis virus, Ptpn2 limits the generation of Tim3⁺ terminally exhausted T cells by preventing excessive type I IFN signaling (LaFleur et al., 2019). It is therefore reasonable to assume that Ptpn2 acts in a similar manner in acute HSV skin infection to restrict JAK-STAT signaling induced by proinflammatory cytokines known to drive terminal differentiation of KLRG1⁺ effector cells, including type I and II IFN, IL-12, and IL-2 (Kaech and Cui, 2012). Conversely, this implies that under suboptimal or less inflammatory priming conditions, such as during tumor challenge, Ptpn2 deficiency may actually favor the development of memory precursors and T_RM cells. In line with this, we and others have recently shown that genetic deletion of Ptpn2 in total T cells, CD8⁺ T cells or chimeric antigen receptor T cells augments cancer immunity (LaFleur et al., 2019; Wiede et al., 2020). In case of cutaneous melanoma, Ptpn2 deficiency indeed results in enhanced generation of tumor-specific CD69⁺ CD103⁺ T_RM cells (Wiede et al., 2020), and we have previously shown that these cells can drive long-term melanoma control (Edwards et al., 2018; Hochheiser et al., 2019; Park et al., 2019a). Thus, while Ptpn2 modulates the generation of memory precursor cells in a context-dependent manner, it does not appear to influence the process of T_RM cell differentiation in peripheral tissues per se. Our demonstration that in vivo– or in vitro–primed Ptpn2-deficient KLRG1⁻ effector cells display normal capacities to generate T_RM cells upon transfer into new recipients further supports this notion. In these experiments, Ptpn2-deficient KLRG1⁻ effector cells actually generated more memory T cells in spleen than their WT counterparts, which is in line with a recent report showing superior engraftment after i.v. transfer of briefly in vitro–stimulated CD8⁺ T cells in absence of Ptpn2 (Flosbach et al., 2020). Both observations may be explained by increased proliferative momentum of activated Ptpn2-deficient cells before transfer, which is also echoed by our combined transcriptional and GO analyses, as well as to their heightened responsiveness to cytokines known to promote the survival and homeostatic turnover of recirculating memory T cells, such as IL-7 and IL-15.

Importantly, we further show that T_RM cells maintain expression of Ptpn2 and that its genetic deletion amplifies T_RM cell–mediated skin immunity in the context of virus infection and autoantigen-induced inflammation. We speculate that the decreased threshold for productive TCR activation in the absence of Ptpn2 regulation may be the major factor promoting accelerated and possibly augmented effector activity in T_RM cells. As such, our results lend strong support to the concept that therapeutic Ptpn2 inhibition may prove beneficial in bolstering protective immunity to fight infection and cancer. At the same time, they corroborate earlier studies in illustrating potential side effects of therapeutic Ptpn2 inhibition; that is, triggering autoimmune inflammation including type I diabetes and dermatitis (Heinonen et al., 2004; Wiede et al., 2019; Wiede et al., 2011; Wiede et al., 2014b). Of note in this respect, our study further emphasizes that T_RM cells could be major targets of Ptpn2 inhibition in therapeutic settings. Thus, transient topical application of Ptpn2 inhibitors may prove sufficient to achieve clinical benefit, at least in accessible organs such as skin, which has the added benefit of avoiding potential side effects from systemic dosing. In summary, our study provides fresh insights into the molecular underpinnings of T_RM cell generation and function in skin. Future studies will have to explore in more detail the therapeutic implications of our findings for immunomodulatory strategies aimed at modulating T_RM cell function in infection and tissue pathology.

Female C57BL/6, B6.SJL-PtprcaPep3b/BoyJ (B6.CD45.1), C57BL/6-Tg(K5-rTA)04Cbn Tg(TetOmOva)13Cbn/Apb (K5-rTA;TetO.mOva), and gBT-I × B6.CD45.1 (gBT-I.CD45.1) mice were bred in the Department of Microbiology and Immunology of The University of Melbourne. Ptpn2^fl/fl;OT-I and Lck-Cre;Ptpn2^fl/fl;OT-I are Ptpn2-competent or -deficient mice, respectively, with CD8⁺ T cells expressing a transgenic T cell receptor specific for the Ova peptide SIINFEKL, as previously described (Wiede et al., 2011), and were bred at the Victorian Comprehensive Cancer Centre. K5-rTA;TetO.mOva mice (available from the Australian Phenome Bank) express a Dox-inducible (Tetracycline-on [TetO]) mOva under the control of the keratin 5 (K5) promoter in keratinocytes in the epidermis and hair follicles. In these mice, following oral Dox treatment, rTA expression activates the TetO promoter for mOva expression in the skin. In gBT-I mice, CD8⁺ T cells express a transgenic T cell receptor recognizing the HSV gB–derived epitope gB_498–505. All mice were aged between 6 and 12 wk at the start of experimentation, and all experiments were approved by the relevant Animal Ethics Committee of The University of Melbourne.

gBT-I or OT-I splenocytes were activated in vitro by incubation for 4–5 d with gB (SSIEFARL; 0.2 µg/ml; Genescript) or Ova (SIINFEKL; 0.2 µg/ml; Genescript) peptide–pulsed splenocytes in the presence of recombinant human IL-2 (25 U/ml; PeproTech) and lipopolysaccharide (75 ng/ml; Sigma-Aldrich), as previously described (Gebhardt et al., 2009). In some experiments, recombinant human TGFβ1 (10 ng/ml; R&D Systems) was added during the last 2 or 3 d of culture.

Naive Ptpn2^fl/fl or Lck-Cre;Ptpn2^fl/fl OT-I cells (5 × 10⁴) were transferred i.v. into congenic WT mice infected with HSV-Ova 2 d earlier. In some experiments, total, KLRG1⁺ or KLRG1⁻ effector populations (2.5–10 × 10⁴) were isolated from spleens of HSV-infected mice, sorted, and transferred i.v. into infected congenic recipients. In vitro–activated OT-I cells were injected i.d. into B6.CD45.1 or K5-rTA;TetO.mOva mice (1–5 × 10⁶). In other experiments, in vitro–activated and transduced gBT-I cells were i.v. injected into recipient mice (1–2.5 × 10⁶) treated on the skin with DNFB, as described previously (Mackay et al., 2012). Genetically modified gBT-I cells derived from bone marrow chimeric mice were transferred i.v. (2–10 × 10⁴) into congenic mice infected with HSV-1.

HSV-1 skin infection was performed by light skin abrasion using 10⁶ PFUs of the KOS strain of HSV-1 or Ova-expressing HSV-1, as described previously (van Lint et al., 2004). Mice were anaesthetized with a mixture of ketamine (100 mg/kg; Parnell Laboratories) and Xylazil (15 mg/kg; Troy Laboratories) by i.p. injection, and lubricating eye gel (Allergen Australia) was applied to the eyes to prevent drying. Mice were shaved with a razor (Wahl) and depilated using Veet cream (Reckitt Benckiser) before skin was lightly abraded with a power tool (Dremel) and 10 µl virus solution applied to abraded skin. An adhesive waterproof film (Opsite) was used to cover the virus, and mice were bandaged with micropore and macropore tape for 2 d. For HSV challenge experiments, virus titers in skin were determined by removing and homogenizing a 1-cm² area of skin that was assayed for infectious virus by plaque assay on Vero cells (CSL Australia), as described previously (van Lint et al., 2004).

Mice were anaesthetized with ketamine (100 mg/kg; Parnell Laboratories) and Xylazil (15 mg/kg; Troy Laboratories) by i.p. injection. Skin was shaved and depilated before 15 µl of DNFB (0.25%) in acetone and oil (4:1 vol/vol) was applied to a 1.5cm² area of flank skin.

In vitro–activated OT-I cells were transferred i.d. into K5-rTA;TetO.mOva mice. More than 4 wk later, circulating CD8-expressing cells were depleted in cohorts of mice by i.p. administration of 300 µg and 100 µg CD8-depleting antibody (clone 2.43; Walter and Eliza Hall Institute), respectively, on two consecutive days. Absence of circulating CD8⁺ T cells in the blood was confirmed 5 d later by flow cytometry. Starting 9 d after CD8 depletion, mice were continuously treated with Dox in their drinking water (1 mg/ml). To assess skin inflammation, mice were shaved and depilated using Veet cream. Dermatitis was quantified using the following scoring system (from 0 to 4): 0, no signs of disease; 1, marginally dry or red skin; 2, early developing lesions covering <10% of flanks; 3, developed lesions, covering >10% and <50% of flanks; 4, pronounced lesions covering >50% of the flank. Mice were euthanized when reaching a score of 4 but were kept in the analysis with score 4 until the end of the experiment to calculate the area under the curve for statistical evaluation.

In vitro–activated OT-I cells were transferred i.d. into WT mice. More than 4 wk later, mice were shaved and depilated using Veet cream, and 0.1 µg Ova peptide SIINFEKL (Genescript) was i.d. injected in a volume of 100 µl PBS. Skin was harvested 6 h later, and cells were isolated in the presence of 10 µg/ml Brefeldin A (Sigma-Aldrich).

Mice were perfused with 10 ml PBS (Media Preparation Unit, University of Melbourne) before tissues were harvested. Lymphocytes were isolated from the spleen by grinding organs through a metal mesh to create single-cell suspensions. T cells were isolated from the skin as described previously (Gebhardt et al., 2009; Mackay et al., 2013). Briefly, skin was incubated in dispase solution (2.5 mg/ml; Roche) for 90 min and the epidermis separated from the dermis. The dermis was finely chopped and incubated in collagenase type III (3 mg/ml; Worthington) and the epidermis placed in trypsin (0.25%) and EDTA (0.1%, in PBS; Sigma-Aldrich) and incubated for 30 min at 37°C. Samples were filtered through a 75-µm mesh, and single-cell suspensions were stained with antibodies for flow cytometry. For cell sorting, CD8⁺ T cells were enriched from the spleen by incubation with anti-CD4 (GK1.5), anti-Mac-1α (clone M1/70), anti-F4/80 (clone F4/80), anti-erythrocyte (clone TER-119), and anti-I-A/E (clone M5114) monoclonal antibodies, followed by incubation with goat anti-rat IgG-coupled magnetic beads and removal of bead-bound cells. Isolated cells were surface-stained with the following antibodies: anti-mouse CD8a-FITC/APC/PE (53–6.7; BD Biosciences), CD8β-Alexa Fluor 700 (YTS156.77; BioLegend), CD45.2-AF780/BV786 (104; eBioscience or BD Biosciences), Vα2-PeCy7/BV421 (B20.1; BD Biosciences), Vβ5.1/5.2-PE (9MR9-4, BioLegend), CD69-PE/APC (H1.2F3; eBioscience), CD16/CD32 (2.4G2; BD Biosciences), CD45.1-AF700/AF780 (A20; BD Biosciences or eBioscience), CD103-FITC (2E7; eBioscience), CD103-APC/PerCPCy5.5 (2E7; BioLegend), KLRG1-FITC/APC/BV605 (2F1; eBioscience or BioLegend), and CD127-APC (A7R34; eBioscience). Cell viability was determined using a fixable LIVE/DEAD near-infrared cell-staining kit (BD Biosciences), propidium iodide (Sigma-Aldrich), or DAPI (BioLegend). SPHERO calibration particles (BD PharMingen) were used for cell enumeration. For intracellular cytokine staining, surface-stained cells were fixed and permeabilized with BD Cytofix/Cytoperm (BD Biosciences) for 20 min on ice, and subsequent washing and staining steps were performed using Perm/Wash Buffer (BD Bioscience). Intracellular staining was performed overnight at 4°C using the antibodies IFNɣ-FITC/PECy7 (XMG1.2; BD PharMingen) and granzyme B-PECy7 (NGZB; eBioscience). For staining of Ptpn2, surface-stained cells were fixed in 2% paraformaldehyde for 15 min at room temperature, washed with PBS, and permeabilized with ice-cold methanol/acetone (50:50 vol/vol) for 30 min on ice. Cells were washed three times with 5% FCS in PBS and stained with mouse anti-Ptpn2 (6F3; Medimabs) for 1 h at room temperature. Subsequently, cells were washed three times with 5% FCS in PBS and stained with anti-mouse IgG H+L F(ab)2 fragment–Alexa Fluor 647 (Molecular Probes, Invitrogen) for 30 min at room temperature and washed three times with 5% FCS in PBS before analysis. Cells were sorted using a BD FACS Aria III or analyzed using a BD FACS Canto II or BD Fortessa. Data were analyzed using FlowJo version 9 or 10.

Skin was harvested into 20% sucrose (Sigma-Aldrich) in PBS for 30 min on ice and snap frozen in tissue-tek OCT compound (Sakura Finetek) with liquid nitrogen. Samples were cut into 20-µm thin slices using a microtome (Leica) and placed on charged glass slides. Samples were air-dried for at least 2 h and then rehydrated with PBS for 5 min. Protein block (DAKO) was applied for 30 min at room temperature in a semihumidified chamber. Sections were stained with Alexa Fluor 647–conjugated CD45.2 (clone A20; BioLegend) for 45 min and then washed four times for 3 min with PBS. Hoechst 33342 (Bio-Rad) was added for 3 min to visualize cell nuclei and then washed off with PBS before mounting. Tiled images were acquired with a Zeiss LSM780 confocal microscope at 20× zoom and processed with Imaris8 (Bitplane) software. Cells in skin epithelium (epidermis and hair follicle epithelium) and dermis were manually counted in two sections per mouse, and the ratio of cells was calculated for each individual mouse.

RNA was extracted from sorted in vivo primed effector OT-I using TRIzol reagent (Invitrogen) and Direct-zol RNA MiniPrep columns (Zymo) as described previously (Russ et al., 2014). RNA quality was confirmed using a Bioanalyser, and paired-end libraries were prepared for mRNA-seq using an Illumina TruSeq RNA v2 kit and sequenced using an Illumina HiSeq2500. Reads where aligned to mm10 genome using TopHat, reads were assigned to genes using Subread R package, and differential expression was performed using edgeR R package. Genes were deemed differentially expressed if the adjusted P value was < 0.05 and |log2 (fold change)| > 1. For GO-term analysis, genes with false discovery rate < 0.05 and |log2 (fold change)| > 0.5 were taken into account. Genes with read counts of <5 were removed from the analysis. Data were visualized using RStudio (integrated development for R; RStudio). The GEO accession number for submitted RNA-sequencing data is GSE166133. Gene set enrichment analysis (Mootha et al., 2003, Subramanian et al., 2005) was performed with the following gene sets from the molecular signatures database (MSigDB) 3.0 (Liberzon et al., 2011): KAECH_DAY8_EFF_VS_MEMORY_CD8_TCELL_DN/UP (Kaech et al., 2002) and GOLDRATH_EFF_VS_MEMORY_CD8_TCELL_DN/UP (Luckey et al., 2006).

Vectors encoding murine Klrg1.WT (NCBI reference sequence NM_016970.1), Klrg1.Y7F (TAT → TTT) and Klrg1.C61-72 (C61S TGC → AGC, C62Q TGC → CAA, C69S TGT → TCT, C72S TGC → AGC) cDNA flanked by Bgl2 and Mfe1 restriction sites were ordered from Integrated DNA Technologies IDT. Gene fragments were removed from IDT vectors using Bgl2 (New England Biolabs) and MfeI (New England Biolabs) digestion and inserted into the multiple cloning site of the murine stem cell virus/internal ribosome entry site GFP vector (pMIG, plasmid 9044; Addgene), linearized with Bgl2 (New England Biolabs) and EcoR1 (New England Biolabs).

Viral particles were produced in 293T human embryonic kidney cells with 10 µg vector DNA along with the packaging retroviral constructs GAG (4.8 µg) and pEco (2.4 µg) using standard calcium phosphate precipitation. Supernatant was collected at 48 h, passed through a 0.45-µm filter, and spun onto non–tissue culture–treated 12-well plates (In Vitro Technologies) coated overnight with 32 µg/ml retronectin (Takara) at 3,000 r.p.m and 32°C for 90 min. Supernatant was removed, and 10⁶ CD8⁺ T cells were incubated overnight on virus-coated plates in the presence of recombinant human IL-2 (25 U/ml; PeproTech), starting 2 d after in vitro activation with peptide-pulsed splenocytes. Alternatively, bone marrow cells from gBT-I.CD45.1 mice (isolated 1 d prior) were cultured overnight on virus-coated plates in the presence of stem cell factor (100 ng/ml), IL-6 (10 ng/ml), thrombopoietin (50 ng/ml; Shenandoah Biotechnology), and FLt3L (10 ng/ml; BioXCell). Transduction efficiency was determined by GFP expression.

C57BL/6 mice were lethally irradiated (2 × 550 rad, 3 h apart), i.v. transferred with retrovirally transduced bone marrow cells (1–4 × 10⁶), and allowed to reconstitute for at least 8 wk.

Data were analyzed and plotted using Prism 8 (GraphPad Software). Comparisons between two groups were performed using a two-tailed Mann–Whitney test, unless indicated otherwise. Differences between groups were considered significant for P values of <0.05.

Fig. S1 shows CD103 expression by in vitro–activated T cells in response to TGFβ (Ptpn2^fl/fl versus Lck-Cre;Ptpn2^fl/fl and pMIG.Klrg1.WT versus pMIG.Ctrl), numbers of CD103⁺ OT-I cells in epidermis and dermis in both flanks of mice infected with HSV-Ova and HSV-WT, and mean numbers of KLRG1⁻Ptpn2^fl/fl and Lck-Cre;Ptpn2^fl/fl cells in spleen and skin. Fig. S2 lists significantly (P < 0.05) regulated biological processes (GO-term analysis) comparing transcriptional profiles of total, KLRG1⁺, and KLRG1⁻ populations from Ptpn2^fl/fl and Lck-Cre;Ptpn2^fl/fl OT-I cells, as in Fig. 3. Fig. S3 lists significantly (P < 0.05) regulated cellular components (GO-term analysis) comparing transcriptional profiles of total, KLRG1⁺, and KLRG1⁻ populations from Ptpn2^fl/fl and Lck-Cre;Ptpn2^fl/fl OT-I cells, as in Fig. 3. Fig. S4 lists significantly (P < 0.05) regulated molecular functions (GO-term analysis) comparing transcriptional profiles of total, KLRG1⁺, and KLRG1⁻ populations from Ptpn2^fl/fl and Lck-Cre;Ptpn2^fl/fl OT-I cells, as in Fig. 3. Fig. S5 shows representative images of Dox-treated K5-rTA;TetO.mOva mice with Ptpn2^fl/fl and Lck-Cre;Ptpn2^fl/fl T_RM cells and granzyme B and cytokine production by Ptpn2^fl/fl and Lck-Cre;Ptpn2^fl/fl T_RM cell after Dox treatment or specific peptide challenge, respectively. Table S1, Table S2, and Table S3 list DEGs between Ptpn2^fl/fl and Lck-Cre;Ptpn2^fl/fl OT-I cells for total, KLRG1⁺, and KLRG1⁻ populations, respectively.

We thank the Flow Cytometry Core Facility, the Melbourne Bioresource Facility, and the Biological Optical Microscopy Platform at the University of Melbourne for technical support.

This study was supported by the National Health and Medical Research Council of Australia (grant APP1084805 to T. Gebhardt; grant APP1059514 to T. Gebhardt, D. Tscharke, and S. Bedoui; and grant APP1184615 to T. Tiganis). K. Hochheiser was supported by the German Research Council (grant HO 5417/1-1) and is a Rhian and Paul Brazis Fellow in Translational Melanoma Immunology administered by the Peter MacCallum Cancer Foundation. T. Gebhardt is a Senior Biomedical Research Fellow supported by the Sylvia and Charles Viertel Charitable Foundation. T. Tiganis is a National Health and Medical Research Council of Australia Principal Research Fellow (APP1103037). S.L. Park was supported by an Elizabeth & Vernon Puzey postgraduate scholarship and a Cancer Council Victoria postdoctoral fellowship. M.J. Herold is supported by the National Health and Medical Research Council of Australia (grants APP1186575, APP1159658, and APP1156095) and the Leukemia and Lymphoma Society (grant LLS SCOR 7001-13).

Author contributions: Conceptualization: K. Hochheiser, F. Wiede, T. Tiganis, S. Bedoui, and T. Gebhardt; Data curation: K. Hochheiser, B. Russ, C.M. Jones, and T. Gebhardt; Formal analysis: K. Hochheiser, M.H. Enders, M. Olshansky, B. Russ, T. Tiganis, and T. Gebhardt; Funding acquisition: K. Hochheiser, T. Tiganis, S. Bedoui, and T. Gebhardt; Investigation: K. Hochheiser, T. Wagner, D. Freestone, M.H. Enders, B. Russ, S. Nüssing, E. Bawden, Asolina Braun, Annabell Bachem, E. Gressier, R. McConville, S.L. Park, C.M. Jones, and S. Turner; Methodology: F. Wiede, C.M. Jones, G.M. Davey, I.A. Parish, and M.J. Herold; Project administration: K. Hochheiser, S. Bedoui, and T. Gebhardt; Resources: F. Wiede, C.M. Jones, G.M. Davey, D. Tscharke, S. Turner, and T. Tiganis; Software: M. Olshansky; Supervision: D.E. Gyorki, I.A. Parish, T. Tiganis, S. Bedoui, and T. Gebhardt; Validation: K. Hochheiser, T. Wagner, and T. Gebhardt; Visualization: K. Hochheiser, M.H. Enders, D.E. Gyorki, and T. Gebhardt; Writing – original draft: K. Hochheiser, C.M. Jones, and T. Gebhardt; Writing – review & editing: K. Hochheiser, F. Wiede, S.L. Park, G.M. Davey, D.E. Gyorki, I.A. Parish, S. Turner, T. Tiganis, S. Bedoui, and T. Gebhardt.