Skin-resident memory CD8+ T cells trigger a state of tissue-wide pathogen alert

See allHide authors and affiliations

Science  03 Oct 2014:
Vol. 346, Issue 6205, pp. 101-105
DOI: 10.1126/science.1254803

Resident memory T cells sound the alarm

Immunological memory protects against reinfection. Resident memory T cells (TRM) are long-lived and remain in the tissues where they first encountered a pathogen (see the Perspective by Carbone and Gebhardt). Schenkel et al. and Ariotti et al. found that CD8+ TRM cells act like first responders in the female reproductive tissue or the skin of mice upon antigen reencounter. By secreting inflammatory proteins, TRM cells rapidly activated local immune cells to respond, so much so that they protected against infection with an unrelated pathogen. Iijima and Iwasaki found that CD4+ TRM cells protected mice against reinfection with intravaginal herpes simplex virus 2.

Science, this issue p. 98, p. 101, p. 93; see also p. 40


After an infection, pathogen-specific tissue-resident memory T cells (TRM cells) persist in nonlymphoid tissues to provide rapid control upon reinfection, and vaccination strategies that create TRM cell pools at sites of pathogen entry are therefore attractive. However, it is not well understood how TRM cells provide such pathogen protection. Here, we demonstrate that activated TRM cells in mouse skin profoundly alter the local tissue environment by inducing a number of broadly active antiviral and antibacterial genes. This “pathogen alert” allows skin TRM cells to protect against an antigenically unrelated virus. These data describe a mechanism by which tissue-resident memory CD8+ T cells protect previously infected sites that is rapid, amplifies the activation of a small number of cells into an organ-wide response, and has the capacity to control escape variants.

Tissue-resident memory CD8+ T cells (TRM cells) are a subtype of memory lymphocytes (1) that permanently reside in nonlymphoid tissues in mice and humans (211). Analysis of herpes simplex virus (HSV)–1 and HSV-2 shedding episodes in infected human mucosa has shown that emerging lesions are often controlled within 6 to 12 hours (12). Furthermore, the severity of viral lesions during reactivation is likely determined by local immune control, and data in mouse models suggest that such tissue protection can be mediated by locally residing memory CD8+ T cells (1315).

The mechanisms by which a small number of local memory cells can protect a peripheral tissue have not been established and, given the low numbers of TRM cells in tissues, unlikely to solely involve the direct killing of target cells (16). In addition, although TRM cells are able to recruit circulating memory CD8+ T cells to the peripheral tissue within 48 hours of activation (17), such recruitment is unlikely to achieve early pathogen control (12).

To investigate how small numbers of tissue-resident memory T cells confer rapid protection of local tissue, we created a pool of TRM cells by intra-epidermal DNA vaccination of mice that had received small numbers of green fluorescent protein CD8 T cells specific for the HSV-1–derived glycoprotein B peptide (gB498-505 (gBT-GFP hereafter)) (fig. S1) (9, 18). Weeks later, skin areas harboring TRM cells were challenged with HSV-1 or gB498-505 peptide. Immunohistochemical analysis of HSV-1–infected or gB498-505 peptide–challenged skin tissue of gBT-GFP TRM cell mice and naive mice at 9 hours after infection/peptide administration did not reveal any difference in the infiltration of macrophages, gBT-GFP memory T cells, or CD3 cells. A moderate increase in the number of neutrophils was only observed upon peptide administration (fig. S2, A to F). To evaluate other possible effects of TRM cell activation on the surrounding tissue, we obtained transcriptional profiles from the entire skin tissue at the same early time point after in situ triggering of TRM cells. Comparison of the transcriptional profiles in skin exposed to control [ovalbumin (OVA257-264)] peptide or cognate (gB498-505) peptide revealed differential expression of a large number of genes [cut-offs: false discover rate (FDR) < 0.05; log2 fold change > ± 1.2] (Fig. 1A and fig. S3). To distinguish between noise caused by variation in tissue composition and signal due to TRM cell triggering, transcriptional profiling was performed on a second cohort of mice (Fig. 1B and fig. S3). Genes the expression of which was altered at a comparable magnitude in both data sets (difference in magnitude of induction <1.5; blue in fig. S3B) were retained for further analysis.

Fig. 1 TRM cell triggering alters tissue-wide gene expression profiles.

(A to D) Transcriptome analysis of full-thickness skin from mice harboring gBT-GFP TRM cells (A and B) or OTI-GFP TRM cells (C and D) upon local administration of either gB498-505 or OVA257-264 peptide. Relative abundance is plotted for averaged normalized read counts. All detected nondifferentially expressed genes are depicted in gray. Genes that are differentially expressed in all four comparisons (log fold change (logFC) > ±1.2; FDR < 0.05; difference in magnitude between replicate experiments < 1.5) (fig. S3) are depicted in red. In these and further plots, horizontal green lines represent logFC limits for significance ±1.2. (E) Average logFC for gBT TRM cells and OTI TRM cells harboring skin upon triggering with cognate peptide. Genes listed in table S1 (differentially expressed upon TRM cell triggering) are shown in red; genes that were only up-regulated in one of the TRM cell groups in black. (F) Average logFC for skin harboring gBT TRM cells upon triggering with either HSV-1 or cognate peptide. Genes listed in table S1 are shown in red; genes specifically up-regulated upon HSV-1 infection are depicted in black. Group 1, correlated behavior between both triggers, enriched in interferon-responsive genes; group 2, preferentially or only induced by HSV, enriched in secreted molecules; group 3, reduced by HSV, too small for pathway analysis. (G) Comparison of normalized transcript counts of the differentially expressed gene set in table S1 (induced transcripts) with normalized transcript counts of a set of T cell–specific genes (T cell transcripts). Among induced transcripts, IFITM3 is depicted in blue; the T cell transcripts gene set (IFN-γ, CD2, zap70, CD5, CD69, CD8a, and CD8b1) is depicted in red. Values are representative of eight comparisons.

To determine whether the observed changes in gene expression were due to T cell receptor (TCR) recognition of antigen, cohorts of mice harboring TRM cells specific for the OVA257-264 epitope (OTI hereafter) were challenged with either gB498-505 or OVA257-264 antigen. In this setup, activation of the OTI-GFP TRM cells by cognate OVA257-264 resulted in a reproducible change of the skin transcriptional profile (Fig. 1, C and D). Furthermore, changes in gene expression in gBT-GFP skin TRM cells challenged with cognate gB498-505 peptide and OTI-GFP TRM cells challenged with cognate OVA257-264 peptide were highly correlated (Fig. 1E). Thus, triggering of TRM cells harboring skin with peptide antigen leads to a rapid alteration in the transcriptome that is visible at the level of the entire tissue before substantial influx of immune cells is seen.

Combination of all four data sets resulted in a list of 89 genes that are differentially expressed (all increased) upon specific triggering of TRM cells (table S1 and Fig. 1E). Induction of part of this gene set was already observed 3 hours after antigen administration, and induction was essentially complete after 6 hours (fig. S4). Supporting the immunohistochemical results, T cell–specific genes did not show any significant difference between TRM cells harboring skin treated with specific or control peptide (fig. S5). Independent full transcriptome gene ontology analyses (19) of the four data sets indicated inflammation and immunity as dominant signatures of all data sets (table S2).

Transcriptome analysis of two independent experiments in which gBT-GFP TRM cell skin was challenged with HSV-1 or control showed a similar pattern of gene induction. For most genes within the gene set (table S1), the magnitude of induction was larger upon peptide triggering, possibly because a greater number of TRM cells can encounter antigen early after peptide administration (group 1 in Fig. 1F). In addition, a second group of genes, including a large number of chemokines and cytokines involved in innate immune cell movement, was more strongly or only up-regulated upon virus infection (group 2 in Fig. 1F). Together, these data show that antigen-specific activation of TRM cells is sufficient to initiate an early response that is visible at the level of the entire tissue.

Strikingly, many of the genes that were induced upon peptide administration were expressed at levels >10 to >100 times the level of T cell–specific genes (Fig. 1G). Furthermore, analysis of the identified gene set revealed the induction of a broad-spectrum antipathogen response. To dissect whether this rapid tissue response depends upon systemic antigen-specific memory T cells, or only requires the TRM cell population, OTI-GFP cells from male donors were transferred into syngeneic female recipients and activated by vaccination. In this setting, the systemic memory T cell pool (central memory + effector memory cells; TCM + TEM)—but not the tissue-resident memory pool—is cleared (5) (fig. S1). Comparison of the transcriptional profile in skin of recipients harboring either TRM cells or both TRM and TCM + TEM cells indicates that activation of the TRM cell pool is sufficient to induce expression of the large series of genes within skin (Fig. 2A).

Fig. 2 TRM cells mediate the induction of an antiviral state through IFN-γ.

(A) Female recipients of either sex-matched or mismatched OTI-GFP+ cells were vaccinated to induce skin TRM cells. Skin areas harboring TRM cells were treated locally with OVA257-264 peptide and sacrificed 9 hours later. The absence of systemic memory T cells (see fig. S1) does not significantly reduce tissue conditioning by peptide triggering. On average, induction of the identified gene set was slightly more pronounced in sex-matched recipients (by a factor of 1.15, not significant), which could either reflect a limited contribution of the circulating memory T cell pool or the slight reduction in TRM cell numbers in skin of mismatched recipients (fig. S1F). (B) Recipients of OTI-GFP+ TRM cells derived from either wild-type or IFN-γ–deficient donors were vaccinated to induce skin TRM cells, and the effect of TRM cell triggering was then analyzed as in (A). (C) IFN-γreceptor–proficient or -deficient recipients of OTI-GFP+ TRM cells were vaccinated to induce skin TRM cells, and the effect of TRM cell triggering was then analyzed as in (A).

Upstream regulator analysis of the induced gene signatures by Ingenuity Pathway Analysis indicated that the cytokine interferon-γ (IFN-γ) is the most likely factor controlling the transcriptional alterations seen in TRM cell–conditioned skin (table S3), and previous work has shown that CD8+ TRM cells rapidly re-express IFN-γ after local antigen rechallenge (17). Analysis of full-thickness skin 9 hours after triggering of a population of wild-type or Ifng−/− TRM cells revealed that a large part of the transcriptional alterations seen upon TRM cell triggering are dependent on TRM cell–derived IFN-γ (Fig. 2B). Furthermore, this IFN-γ acts on skin cells other than TRM cells themselves, as the tissue response is also largely lost in Ifngr1−/− recipient mice in which only TRM cells express IFN-γ receptor 1 (Fig. 2C). These data suggest that shortly after TCR triggering, activated TRM cells express IFN-γ to enhance expression of proteins involved in pathogen control within the surrounding tissue. To test this hypothesis, we analyzed the expression pattern of IFITM3 (interferon-induced transmembrane protein 3; blue in Fig. 1G), a protein with broad-spectrum antiviral activity (20), and one of the transcripts induced by TRM cell triggering (Fig. 1 and table S1). Within 6 hours of TRM cell activation by cognate antigen, most epidermal and dermal cells expressed IFITM3, with maximal levels at 9 to 18 hours after conditioning (Fig. 3). By 36 hours, IFITM3 expression was largely restricted to the outer layers of the epidermis, indicating that local TRM cell activation leads to a transient change in the skin transcriptome that is still visible in aging keratinocytes by the time newly formed keratinocytes have returned to steady state (Fig. 3B).

Fig. 3 Tissue conditioning by TRM cells results in the induction of an antiviral state in large numbers of surrounding cells.

(A) Immunohistochemical detection of IFITM3 in the skin of mice harboring OTI-GFP+ TRM cells, analyzed at steady state or 9 hours after treatment with either cognate OVA257-264 or control gB498-505 peptide (representative of three mice per group). (B) Immunohistochemical detection of IFITM3 in the skin of mice harboring OTI-GFP+ TRM cells at the indicated time points after treatment with OVA257-264 peptide (representative of three mice per time point). The boundary between epidermis (top in all images) and dermis is highlighted by a black line. Scale bar, 20 μm. (C and D) Naive gBT-GFP+ cells were transferred into recipients that were subsequently tattoo-vaccinated with DNA encoding TTFC-gBpep to create a population of resident TRM cells. Several weeks after tattooing, skin harboring TRM cells was injected with either gBpep or OVApep and processed for transcriptome analysis 9 hours later. (C) Ratio between IFITM3 counts and the median counts of 20 housekeeping genes. (D) The same analysis is depicted for samples in which skin harboring OTI TRM cells was injected with OVApep and analyzed at the indicated time points.

In most models of infection control by CD8+ T cells, both the initial T cell activation and the final output signal (e.g., cytolysis) are dependent on recognition of cognate antigen. In contrast, the above-described tissue conditioning by TRM cells requires antigen as input signal but generates an output signal—the up-regulation of genes involved in broad-spectrum defense—that does not rely on antigen recognition, a mechanism reminiscent of that of effector CD4+ T cells (21). To test the potential relevance of this state of TRM cell–induced “pathogen alert,” we analyzed whether TRM cell activation could lead to control of an antigenically unrelated pathogen in vivo. Skin-resident OTI-GFP TRM cells were activated by local injection of cognate peptide, and 9 hours later the same area was infected with antigenically unrelated HSV-1. At two time points after virus administration, progression of HSV-1 infections was scored microscopically (day 1) and macroscopically (day 3). As expected, disease progression in naïve mice was not influenced by OVA257-264 peptide administration (Fig. 4A). In contrast, in mice harboring skin OTI-GFP TRM cells, application of cognate OVA257-264 peptide resulted in a strong reduction in HSV-1 disease severity relative to control conditions (Fig. 4A). Analysis of HSV-1–challenged skin tissue by staining with antibody to HSV showed that OTI-GFP TRM cell activation resulted in a substantial reduction of both tissue necrosis and lateral spreading of herpetic lesions (Fig. 4B). In line with this, viral DNA levels were reduced in skin of mice harboring activated OTI-GFP TRM cells at the time of infection (Fig. 4C, P < 0.0001). Taken together, these data demonstrate that the tissue conditioning that is induced by TRM cell activation leads to enhanced pathogen control that is independent on the antigenic identity of this pathogen.

Fig. 4 CD8+ TRM cell triggering provides cross-protection against an antigenically unrelated pathogen.

(A) Naïve mice or mice harboring OTI-GFP+ TRM cells (both hind legs) were injected locally with phosphate-buffered saline (PBS), cognate OVA257-264 peptide, or control gB498-505 peptide and locally infected with HSV-1 9 hours later. Sixty hours after infection, the extent of each infection was scored by visual inspection by an observer blinded to experimental group (all naïve groups: n = 2; OTI-TRM + PBS: n = 5; TRM + gB498-505: n = 10; TRM + OVA257-264: n = 13; both legs analyzed separately). (B) Mice harboring OTI-GFP+ TRM cells (both hind legs) were injected locally with cognate OVA257-264 peptide or control gB498-505 peptide and were locally infected with HSV-1 9 hours later. After 60 hours, the amount of viral DNA in infected skin was measured. Data are representative of four independent experiments with at least five mice per group. To allow comparison between experiments, the amount of viral DNA in the OTI-TRM+ gB498-505 group was set to 100% for each experiment. (C) Immunohistochemical detection of HSV-1 infection in naïve mice that received a local injection with OVA257-264 peptide (B) or mice harboring OTI-GFP TRM cells triggered with cognate OVA257-264 peptide, control gB498-505 peptide, or PBS (C). For each condition, two different magnifications of the same sample are shown. Data are representative of 2 (naïve group), 5 (TRM + PBS), 10 (TRM + gB498-505), and 13 (TRM + OVA257-264) mice per group.

Three aspects of TRM cell–mediated tissue conditioning are noteworthy. First, tissue conditioning is almost immediate. This property is likely to be of major relevance as, at least in case of HSV-2, early immune control is the major determinant of episode severity (14). Second, tissue conditioning forms an effective amplification system, in which activation of a rare cell type leads to a tissue-wide response. Third, TRM cell–mediated tissue conditioning results in protection that is ultimately antigen independent: Although initial TRM cell activation requires recognition of antigen, the genes that are up-regulated in response display activity toward a wide array of pathogens.

From a conceptual point of view, these data place TRM cells as a bridge between the adaptive and innate immune system, in which the TCR in TRM cells has a function similar to that of Toll-like receptors in innate immune cells. From a practical point of view, the fact that TRM cell triggering leads to an output signal that no longer requires antigen recognition may also help counteract viral escape. Recently, strategies have been put forward to create TRM cell populations at sites of potential pathogen entry (10, 22, 23). The current data not only help to provide a mechanistic explanation for the effects of such vaccines but also suggest that in case of pathogens that exist as quasispecies, protection may conceivably be provided not only against the vaccine-encoded sequence but also against viral variants that are transferred in parallel.

Supplementary Materials

Materials and Methods

Figs. S1 to S5

Tables S1 to S3

Reference (24)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: We thank members of the Netherlands Cancer Institute Flow Cytometry, Digital Microscopy, Deep Sequencing Core, and Animal Pathology Facilities for technical support; R. van Mierlo and M. Toebes for assistance; and members of the Schumacher laboratory for discussion. The data presented in this manuscript are tabulated in the main paper and in the supplementary materials. Expression data were deposited under Gene Expression Omnibus accession number GSE60599. This work was supported by The Netherlands Organization for Scientific Research grant 912.10.066 and European Research Council grant Life-his-T to T.N.S., and Dutch Cancer Society grant NKI-2008-4112 and The Netherlands Organization for Health Research and Development TOP grant 91213018 to H.J.
View Abstract

Stay Connected to Science

Navigate This Article