Long-lived epithelial immunity by tissue-resident memory T (T_RM) cells in the absence of persisting local antigen presentation

Circulating CD8⁺ Memory T Cells Do Not Control Skin Infection with HSV.

Certain peripheral infections are efficiently eliminated by effector CD8⁺ T cells but are less well controlled by their memory counterparts (3, 4, 9, 20). We wanted to address whether skin infection with herpes simplex virus (HSV) exhibited a similar pattern of circulating memory cell resistance. To this end, we vaccinated C57BL/6 mice with a recombinant influenza virus that contains the immunodominant determinant from the HSV glycoprotein B (gB) molecule (flu.gB) to generate cohorts with effector (day 10 after vaccination) or memory (day 30 after vaccination) T-cell populations in the circulation. When challenged by HSV skin infection, mice recently immunized with the recombinant flu.gB showed marked protection, with strongly reduced viral loads in the inoculation site compared with nonimmunized controls (Fig. 1A). This protection was also associated with the absence of zosteriform skin lesions that developed in nonimmunized mice as a result of viral replication in sensory neurons, followed by viral recrudescence into a large area of flank skin (21) (Fig. 1B). Despite this robust control during the effector phase of the response, CD8⁺ T cell-mediated immunity was very short-lived, as memory mice inoculated 30 d after immunization did not show control of HSV replication (Fig. 1A), and also developed herpetic skin disease (Fig. 1B). This lack of protection by circulating memory T cells was also observed even when we inoculated with 10,000-fold less infectious viral particles for challenge infection (Fig. S1A).

One possible reason for the disparity in protection between the effector and memory time points could have been a decrease in virus-specific T cells, as their numbers are reduced to approximately 10% to 15% of those present at the day-10 peak of the primary response (Fig. 1C), which is consistent with the contraction seen for other antiviral responses (22). We therefore used a prime–boost regimen to increase memory T-cell numbers in the circulation. After transfer of gBT-I T cells, recipient C57BL/6 mice were primed with dendritic cells (DCs) coated with the immunodominant gB peptide and then boosted with the flu.gB recombinant. This regimen promotes rapid expansion of the specific T cells on boosting, which thereafter show only a slow contraction (23) such that, at 6 wk after infection, secondary memory gBT-I T-cell numbers were approximately equivalent to those found during the primary effector response (Fig. 1C). In addition, consistent with previous reports (23), the majority of these secondary-memory T cells were of the T_EM phenotype, expressing low levels of CD62L (Fig. S1B). Despite containing elevated numbers of T_EM cells in the circulation, mice subjected to the prime–boost treatment showed only minimal protection against HSV-1 skin infection (Fig. 1D). Six days after HSV skin challenge, virus loads were comparable between the two cohorts of immunized mice (Flu.gB alone or DC.gB plus Flu.gB), and prime-boosted mice showed a marginal, albeit statistically significant, reduction in viral titers compared with naive controls [i.e., non-pulsed DC (DC-no) plus intranasal mock challenge; Fig. 1D]. Of note, this lack of overt protection was reflected by a similar course of viral skin disease with the development of characteristic skin lesions in all groups of mice. Thus, in the case of HSV skin infection, despite the ability of CD8⁺ effector T cells to mediate control of replicating virus, there was minimal protection by the circulating T cells during the memory phase of the immune response.

Inflammation Enhances T_RM Lodgement in Skin in Absence of Local Antigen Stimulation.

Given that circulating memory CD8⁺ T cells proved ineffective in protecting against peripheral HSV infection, we wanted to test whether T_RM cells would be superior in this respect. Although primed T cells intrinsically infiltrate a wide variety of extralymphoid tissues in the absence of local infection (10, 24) and thereafter leave behind a residual resident population (17), we reasoned that targeted skin inflammation would enhance memory T-cell lodgement. To demonstrate such inflammation-enhanced T_RM formation, we transferred in vitro activated gBT-I T cells into C57BL/6 mice and then treated one flank (left) with the contact-sensitizing agent 2,4-dinitrofluorobenzene (DNFB), while leaving the other (right) as an untreated control. In this approach, DNFB acted as a nonspecific inflammatory stimulus to recruit the virus-specific T cells into the skin in the absence of transient or ongoing local antigen presentation. As shown in Fig. 2 A and B, memory T cells could be found in far greater numbers within the left flank of mice that had undergone DNFB treatment compared with the nonsensitized right flanks. The lack of T_RM cells in the nontreated flanks was not a consequence of a wider absence of memory cells, as they were easily detected in the spleen (Fig. 2B). Intravital microscopy showed that the T cells localized to the outermost epithelial layers of the skin, similar to what is seen after natural HSV infection (18) (Fig. S2). Overall, the results demonstrate that nonspecific inflammation alone resulted in the efficient recruitment and lodgement of T_RM cells in flank skin, which then persisted for at least 1 y after lodgement in the absence of ongoing antigen stimulation.

CD103 Is Selectively Up-Regulated on CD8⁺ T Cells in Skin in Absence of Local Antigen Stimulation.

CD103, part of the α_E(CD103)β₇ integrin, is expressed by CD8⁺ T_RM cells in different peripheral tissues (19, 25, 26), and has been shown to be important for survival of T cells in these compartments (26, 27). Accordingly, T cells embedded in skin by DNFB treatment expressed high levels of CD103 (Fig. 3A), and the proportion of CD103⁺ CD8⁺ T cells in DNFB-treated skin increased with time in a similar manner to that seen after HSV skin infection (Fig. S3). The expression of CD103 by the T cells recruited by using a nonspecific inflammatory signal was surprising given that Wakim et al. (26) showed that up-regulation of this molecule in brain appeared to require local antigen recognition. It was possible that epithelial tissues, such as skin and mucosa (described later), uniquely permitted CD103 up-regulation in the absence of local antigen recognition. To show this, we infected left and right flanks of mice with WT HSV or a recombinant virus [K.L8A (28)], respectively, with the latter having a mutated immunodominant gB determinant. HSV enters free nerve endings in the skin and then travels to the distinct sensory ganglia (the dorsal root ganglia) that innervate the regions of skin regions involved in initial infection (29). Thus, in doubly infected mice, WT virus primes cytotoxic T cells specific for the natural gB determinant. These cytotoxic T lymphocytes are recruited into inflamed skin and ganglia infected with either the WT virus (left side) or mutant K.L8A virus (right side). Examination of the recruited T cells showed that, whereas antigen recognition was critical to CD103 up-regulation in the sensory ganglia, it was dispensable in the case of those memory cells resident in the skin (Fig. 3B). Note that CD103 is up-regulated after T cells enter the ganglia, as the early recruits do not express this marker (Fig. S4). The antigen-independent CD103 up-regulation in skin, but not sensory ganglia, was also seen when activated OT-I T cells of irrelevant (i.e., ovalbumin) specificity were transferred before flank infection (Fig. 3B). The CD103 up-regulation on the OT-I cells and their survival in skin also argued that these events were not caused by aberrant cross-reactivity between the gBT-I transgenic cells and DNFB. Overall, the results suggest that skin has an inherent capability of supporting the formation of CD103⁺ T_RM cells.

Skin T_RM Cells Provide Local Protection Against Infection in Absence of Ongoing T-Cell Stimulation.

We next wanted to determine whether the virus-specific T_RM cells lodged by DNFB treatment could control local HSV infection. To this end, we transferred activated T cells to C57BL/6 mice that were treated on the left flank with DNFB and then infected on the left and right flanks with HSV. Fig. 4A shows that, although herpetic zosteriform lesions appeared on control flanks deficient in T_RM cells, skin containing this population was protected from disease. Consistent with this protection, virus replication was severely suppressed in skin previously treated with DNFB (Fig. 4B). Virus control was dependent on the presence of CD8⁺ memory T cells, as mice that had undergone DNFB treatment but had not received transferred gBT-I cells showed little signs of protection in terms of alleviated disease (Fig. 4A) or reduced virus load (Fig. 4B). Such protection was also evident for at least 100 d after T_RM lodgement (Fig. S5). Protection was antigen-specific, as skin containing ovalbumin-specific memory OT-I CD8⁺ T cells showed a lack of protection (Fig. 4B). As surface replication feeds virus into the sensory ganglia that persists thereafter (30), we reasoned that inhibition of skin infection should also limit the extent of latent infection. The latency set-point was indeed reduced, as shown by using a recombinant virus (KOS.LβA) that expresses the reporter β-gal gene under the latency promoter (31) (Fig. 4C) and, separately, by estimation of the average genome copy number persisting after active virus replication had subsided (32) (Fig. 4D).

T_RM Cells Offer Superior Protection Against Peripheral Infection Compared with Circulating Memory T Cells.

Although the preferential protection on DNFB-treated flanks suggested that local T_RM cells offered superior immunity compared with the circulating memory populations, one could have argued that the effect actually involved some DNFB-induced perturbation of the skin architecture that then enhanced recruitment of the circulating memory population. It has been shown that T_RM precursors are only present in the circulation for a short time after antigen stimulation (17), so we progressively delayed DNFB treatment after transfer of activated T cells. Fig. 5A shows that a delay of 15 d resulted in dramatic reduction of T_RM lodgement in the skin despite equal numbers of virus-specific memory T cells present in the spleen (Fig. 5B). As a consequence, this approach allowed the generation of two cohorts with equivalent numbers of circulating memory T cells, but different densities of T_RM cells in the skin. After flank infection by scarification, only mice with T_RM cells were protected against skin disease and showed reduced levels of local virus replication (Fig. 5C). These data formally demonstrate that T_RM cells afford superior protection against localized skin infection with HSV compared with their circulating counterparts.

T_RM Cells Can Protect Against HSV Challenge in Different Tissues and After Different Lodgement Modalities.

Although the transferred activated T cells were useful in demonstrating that T_RM cells could protect against a localized infection, we wanted to show that these results could be translated to a variety of different settings. To this end, we repeated the DNFB-mediated lodgement and protection experiments, this time vaccinating mice with the recombinant influenza virus, flu.gB, used in the experiments shown in Fig. 1. Fig. 6A shows that the in vivo-primed virus-specific CD8⁺ memory T cells lodged and survived in skin treated with DNFB but were undetectable in untreated control regions of skin. Areas treated in this fashion showed superior protection after HSV challenge compared with control flank skin in the same animals (Fig. 6B). Thus, effective peripheral resident memory could be generated with a recombinant vaccine that otherwise offered poor control of peripheral infection.

Finally, to show wider application, we extended this approach to a different barrier tissue and used a different inflammatory stimulus to achieve lodgement of the T_RM cells. For this, we used the surfactant nonoxynol-9 (N9), the active ingredient in a number of commercially available spermicides, to generate nonspecific inflammation within the female reproductive tract (33). Mice received activated gBT-I T cells and were then treated with N9 for 6 d by intravaginal application. Fig. 6C shows that this treatment resulted in efficient recruitment and retention of virus-specific T cells in the vagina compared with nontreated controls. These T cells expressed high levels of CD103 (Fig. S6), consistent with their localization to the epithelium (27). Note that vaginal T cells in control mice also had some CD103 expression. However, when combined with the results in Fig. 6C, we estimate that N9 increased CD103⁺ T-cell numbers approximately 10-fold. Critically, the combination of N9 treatment and transfer of activated gBT-I cells gave superior protection against HSV challenge compared with N9 or transferred gBT-I cells alone (Fig. 6D). Importantly, this suggests that regional protection by embedded T_RM cells can be extended to whole organs, in this case the lower female reproductive tract.

Mice.

C57BL/6, gBT × B6.CD45.1 (gBT-I.CD45.1), gBT-I.GFP, and OTI × B6.CD45.1 (OTI.CD45.1) were bred in the Department of Microbiology and Immunology at the University of Melbourne. The gBT-I and OT-I mice are CD8⁺ TCR transgenic mice that recognize the H-2K^b-restricted HSV-1 gB epitope of aa 498 to 505 (gB_498–505) and the ovalbumin-derived epitope of aa 257 to 264 (OVA_257–264), respectively. Animal experiments were approved by the University of Melbourne Animal Ethics Committee.

Virus Infection and DNFB and N9 Treatment.

Viruses used were the KOS strain of HSV, K.L8A (28), and KOS.LβΑ, as well as WSN/NA/gB (i.e., flu.gB) (46). KOS.LβA was generated by recombining SC16LβA (31) with the KOS strain by transfecting vero cells with a mixture of SC16.LβA and KOS DNA and selecting plaque-purified nonneurovirulent recombinants. Epicutaneous infection by scarification was carried out by using 1 × 10⁶ pfu HSV (KOS, K.L8A, or KOS.LβΑ) as described (32). Intravaginal infection was performed on progesterone-treated mice (Depo Provera, 2 mg per mouse). For infection, mice were intravaginally swabbed with calcium alginate swabs and inoculated with 1 × 10⁶ pfu HSV. For flu.gB infections, 50 pfu was administered intranasally and mice were bled at peak of response to measure gBT-I T-cell responses. Mice that showed high gBT-I T-cell responses were used for HSV rechallenge experiments (>15% gBT-I of total CD8⁺ T cells). For DNFB treatment, mice were shaved and depilated before the application of 15 μL of 0.5% DNFB in acetone/oil (4:1) to a 1 cm² area of skin. For N9 treatment, Gynol II vaginal contraceptive jelly [3% (wt/vol) N9; Ortho Options] was diluted 3:1 in PBS solution and administered intravaginally in a 60 μL volume by using a blunt-ended pipette, following swabbing the vaginal vault with a moist calcium alginate swab.

Determination of Viral Titer, Copy Number, and β-Gal Detection.

The level of infectious virus was determined within homogenized skin or in vaginal fluids by pfu assays as described (32). Vaginal fluids were collected by using calcium alginate swabs. For determination of the number of latently infected ganglia cells, X-gal staining for β-gal expression was performed on whole mounted ganglia as described (43). Average viral copy number was determined by real-time PCR as previously described (43).

In Vitro Activation and Adoptive Transfer of Transgenic CD8⁺ T Cells.

All adoptive transfers of gBT-I and OT-I cells were carried out intravenously with lymph node suspensions (5 × 10⁴) or in vitro-generated effector splenocytes (5 × 10⁶ gBT-I cells, 1 × 10⁷ OT-I cells), which were activated by peptide-pulsed targets as described (19).

Generation and Transfer of Bone Marrow-Derived DCs.

Bone marrow-derived DCs were generated according to a standard protocol described previously (5). Briefly, bone marrow cells were cultured for 7 to 10 d in the presence of 20 ng/mL GM-CSF and IL-4 to allow for DC differentiation. DCs were then matured overnight in the presence of 150 ng/mL LPS, and half these cells were further pulsed with gB_498–505 peptide (1 μg/mL, 45 min) or left untreated. A total of 1 to 2.5 × 10⁵ DCs were transferred intravenously into recipients.

Flow Cytometry and mAbs.

Skin tissue was incubated for 90 min at 37 °C in Dispase (2.5 mg/mL) followed by the separation of epidermis and dermis. Epidermal sheets were subsequently incubated for 30 min in trypsin/EDTA (0.25%/0.1%), and the remaining tissue was chopped into fragments and incubated for 30 min in collagenase type 3 (3 mg/mL) and DNase. Ganglia were digested for 90 min in collagenase type 3 (3 mg/mL) as described (19). Vaginal tissues were chopped into fragments and incubated in 1.3 mM EDTA for 30 min at 37 °C, followed by digestion for 90 min in collagenase type 3 (1 mg/mL) as described (18). Skin, vagina, or ganglia cell suspensions were stained with antibodies for flow cytometry. The following antibodies were purchased from BD Pharmingen: anti-CD45.1, anti-Vα2, anti-CD8α, and anti-Vβ8. Anti-CD45.2 and anti-CD103 were purchased from eBioscience. A FACSCanto II system and FlowJo software (TreeStar) were used for analysis.

Intravital Two-Photon Microscopy.

Mice were anesthetized and depilated, and the skin was separated from the peritoneum and adhered to a stable raised platform, attached to a imaging platform maintained at 35 °C. Images were acquired by using an upright LSM710 NLO multiphoton microscope as previously described (18).

Statistics.

Comparison of data sets was performed by one-way ANOVA followed by Tukey posttest comparison.