CD4+ Resident Memory T Cells Mediate Long-Term Local Skin Immune Memory of Contact Hypersensitivity in BALB/c Mice

In allergic contact dermatitis (ACD) and contact hypersensitivity (CHS), the healed skin shows greater swelling than the naïve skin in the same individual upon re-exposure to the same hapten. This “local skin memory” (LSM) in healed skin was maintained for a prolonged period of time and mediated by skin CD8+-resident memory T (TRM) cells in C57BL/6 mice. However, the number of CD4+ T cells is elevated in ACD-healed human skin, and the contribution of CD4+ TRM cells to the formation of LSM currently remains unclear. We herein demonstrated that immediately after CHS subsided, the healed skin in BALB/c mice showed an accumulation of hapten-specific CD4+ and CD8+ TRM cells, with a predominance of CD4+ TRM cells. The presence of CD4+ or CD8+ TRM cells in the healed skin was sufficient for the induction of a flare-up reaction upon a re-challenge. The CD4+ and CD8+ TRM cells both produced interferon-γ and tumor necrosis factor early after the re-challenge. Moreover, while CD8+ TRM cells gradually decreased over time and were eventually lost from the healed skin at 40–51 weeks after the resolution of CHS, the CD4+ TRM cell numbers remained elevated during this period. The present results indicate that the long-term maintenance of LSM is mediated by CD4+ TRM cells, and thus CD4+ TRM cells are an important target for the treatment of recurrent human ACD.


INTRODUCTION
Allergic contact dermatitis (ACD) and its animal model, contact hypersensitivity (CHS), are T cell-mediated delayed-type hypersensitivity reactions caused by sensitization and secondary epicutaneous exposure (elicitation or challenge) to contact allergens (e.g., haptens and metallic cations) (1)(2)(3)(4). A flare-up (or retest) reaction of ACD often occurs on previously affected sites (4,5). In guinea pigs, inflammation in a CHS-healed region was found to be more severe than that in the previously unaffected region in the same individual upon re-exposure to the same hapten (6). These findings suggested the existence of local immune memory in healed skin, named local skin memory (LSM), which may be mediated by hapten-specific locally persisting memory T cells. LSM was then shown to be a long-term immune memory that persisted for at least 1 year in BALB/c mice (7,8). LSM is now considered to be mediated by tissue-resident memory T (T RM ) cells based on the findings of studies using C57BL/6 mice (9)(10)(11).
CD4 + and CD8 + T RM cells are a subset of αβ T cell receptor (TCR) + memory T cells that fundamentally differ from their circulating counterparts and are characterized by their long-term residency within virtually all organs and unique transcriptional profiles (12)(13)(14)(15)(16). T RM cells differentiate within tissues from activated T cells (or T RM precursor cells), are maintained by local cytokine signals (17)(18)(19)(20)(21)(22), and thus are concentrated in previously infected or inflamed tissues. The major function of T RM cells is to provide protective immunity against pathogen re-entry into the same tissue; however, when generated in the context of allergy or autoimmune diseases, they are considered to cause disease relapse.
Skin T RM cells have been characterized in various infections and diseases in both human and mouse models. Healthy and affected skin are populated by CD4 + and CD8 + T cells with various expression levels of CD69 and the integrin α E chain (CD103) (17,(23)(24)(25)(26)(27)(28). Sessile skin T RM cells in previously inflamed skin are generally detected in CD69 + CD103 +/− compartments in both CD8 + and CD4 + T cells, and CD69 − compartments represent recirculating populations (25,(29)(30)(31)(32). CD8 + T RM cells are more likely to preferentially localize in the epidermal layer, whereas CD4 + T RM cells favor localization in the dermis, and the T RM cell population that predominantly accumulates in inflammation-resolved skin appears to vary according to pathogens and diseases (11,23,25,31,33,34).
Recent mechanistic analyses of LSM were conducted with C57BL/6 mice, and the findings obtained showed that CD8 + T RM cells predominantly accumulated in CHS-healed skin (9)(10)(11). A previous study demonstrated that C57BL/6 healed skin contained elevated numbers of CD8 + T RM cells, but not their CD4 + counterparts, which induced flare-up reactions in healed skin upon a re-challenge (hereafter referred to as the LSM response) (11). However, this is inconsistent with the finding showing that CD4 + T cells dominantly remained and persisted in ACD-healed human skin (35). Therefore, it currently remains unclear whether CD4 + T cells in healed skin contribute to the formation of LSM. Moreover, the number of CD8 + T RM cells in the CHS-healed skin of C57BL/6 mice was found to progressively and markedly decrease for up to 12 months (11); however, the mechanisms by which LSM is maintained over a prolonged period of time currently remain unclear.
In the present study, we examined LSM in BALB/c mice with long-term LSM. The CHS-healed skin of BALB/c mice accumulated both CD4 + and CD8 + T RM cells, with a predominance of CD4 + T RM cells. We found that the presence of CD4 + or CD8 + T RM cells in healed skin induced the LSM response upon a re-challenge, suggesting that they redundantly mediate LSM. Moreover, while CD8 + T RM cells gradually decreased and disappeared from the skin by 56 weeks after the 1 st challenge, the increase in CD4 + T cell numbers was maintained during this period. Thus, CD4 + T RM cells play a prominent role in the long-term maintenance of LSM.

Ethics Statement
All experiments were approved by and performed in strict accordance with the guidelines of the Animal Care and Use Committee of Tottori University (approval numbers 14-Y-6, 16-Y-3, and 19-Y-51).
The sphingosine-1-phosphate (S1P) receptor agonist, fingolimod (hydrochloride) (FTY720) (Cayman Chemical, Ann Arbor, MI, USA), was initially dissolved in ethanol. At the time of administration, 60 µg of FTY720 was dissolved in 300 µl phosphate-buffered saline (PBS) by intense pipetting with a 26G needle and syringe and was then injected i.p. into each mouse (approximately 2 mg/kg). A control injection with the same volume of ethanol (3.0 µl) in 300 µl PBS was performed.

CHS
On day−7 or−6, the right ears of the mice were sensitized with 20 µl of haptens (10 µl, both sides of the ear) under isoflurane anesthesia. In some experiments, sensitization with 100 µl of haptens was performed on shaved back skin. On day 0, the right ears were challenged with 20 µl of haptens and the left ears received 20 µl of vehicle (10 µl, both sides of the ear). After more than 35 days, the naïve and the healed ears were both re-challenged with 20 µl of haptens. In some experiments, both ears were re-re-challenged with 20 µl haptens at 35 days after the re-challenge. The experiments were conducted with 1% (40.4 mM) TNCB unless otherwise noted. Ear thickness was measured with the dial thickness gauge G-1A (Peacock) under isoflurane anesthesia. Hapten applications and measurements were conducted between 9:00 and 16:00 (for 2 days after the 1 st challenge, re-challenge, or re-re-challenge, ear thickness was measured every 22-26 h from the challenge time).

Quantitative RT-PCR
The ears were preserved in RNAlater RNA stabilization solution (Qiagen, Hilden, Germany) at −80 • C until use. According to the manufacturers' instructions, total RNA was extracted with the RNeasy Fibrous Tissue Mini Kit (Qiagen) and BioMasher II homogenizer (Nippi, Tokyo, Japan), and cDNA was synthesized with the PrimeScript RT reagent kit with a gDNA Eraser (Takara Bio, Shiga, Japan). qRT-PCR analyses [the shuttle PCR standard protocol in the Premix Ex Taq protocol (Takara Bio)] were performed using Light Cycler C480 (Roche, Basel, Switzerland) with dedicated 96-well plates. Each well contained cDNA (equivalent to 37.5 ng total RNA) diluted with Easy dilution (Takara Bio), 50% TB Green Premix Ex Taq II (Tli RNaseH Plus) (Takara Bio), 400 nM primers, and distilled water at 15 µl. Cp values were obtained with the Second Derivative Maximum Method using LightCycler software. Each gene expression level relative to Hprt expression in each cDNA sample was calculated with the Ct method. Pre-designed primers [Universal Probe Library Assay Design Center (Roche) or the Perfect Real Time Support System (Takara Bio)] were used, and their sequences were as follows (forward/reverse): Il1b

Immunostaining of Ear Sections
The central region of the ears was cut and snap-frozen in optimal cutting temperature compound (Sakura Finetek Japan, Tokyo, Japan) with liquid nitrogen. Horizontal sections from the base of the ears (thickness of 7 µm) were cut with a cryostat and stored at −20 • C until use. The sections were fixed in cold 4% PFA (3-5 min).
All images were taken using the microscope BX-53 with appropriate mirror units and the digital camera DP73 and then analyzed with cellSens software (Olympus).

Statistical Analysis
Each experiment was repeated more than twice with similar results and representative results were shown unless otherwise noted. Statistical analyses were performed using Microsoft Excel (for the paired or unpaired t-test) and IBM SPSS Statistics version 25 (for one-way ANOVA with post-hoc tests). The significance of differences was established at p < 0.05.

Characterization of TNCB-Induced LSM in BALB/c Mice
We examined LSM in more detail using BALB/c mice showing the LSM response (7,8). To induce CHS, the right ears of BALB/c mice were sensitized (day−7) and challenged (day 0) with the hapten 1% TNCB. Ear swelling peaked on day 1 and healed after 5 weeks ( Figure 1A). On day 35, naïve (left) and healed (right) ears were both re-challenged with 1% TNCB. The healed ears showed significantly more swelling than the naïve ears for days after the re-challenge (Figure 1A), suggesting the formation of LSM in CHS-experienced skin. We referred to the greater swelling in rechallenged healed ears than in naïve ears as the LSM response. When we re-re-challenged ex-naïve and ex-healed ears with 1% TNCB at 5 weeks after the re-challenge (on day 70), the extent of swelling was similar in both ears, suggesting that LSM was formed in the naïve ears by the re-challenge ( Figure 1A).
The LSM response occurred when the mice were sensitized on shaved back skin instead of ear skin (Supplemental Figure 1A). Furthermore, ear sensitization alone (day 0) induced weak swelling but not an effective LSM response in healed ears upon (F) Ear swelling upon back skin sensitization + a re-challenge with an irrelevant hapten (Ox) and upon a re-re-challenge with 1% TNCB (n = 6). All experiments were performed with BALB/c mice. Data are means ± SE. *P < 0.05 (two-tailed paired t-test).
the challenge to both ears on day 35 (Supplemental Figure 1B). These results indicate that LSM only forms in the challenged site regardless of the route of sensitization. LSM responses similar to those with 1% TNCB were observed when other haptens (0.5% FITC and 0.88% Ox) were used (Supplemental Figure 1C).
When we re-challenged both ears with 0.1 and 0.01% of TNCB, but not with 0.001%, the CHS (with 1% TNCB)-healed ears still showed a marked LSM response ( Figure 1B). Moreover, the peak swelling of the healed skin upon the re-challenge with 0.1 and 0.01% TNCB (Figure 1B) was markedly stronger than that in the 1 st CHS with the corresponding concentration of TNCB (Supplemental Figure 1D). Thus, TNCB-induced LSM enhanced the local antigen sensitivity and reactivity of the healed skin exposed to the low dose of TNCB (< 0.1%), as shown previously with other haptens (8), but not the high dose of TNCB (1%).
Moreover, we found that LSM responses were still observed by the re-re-challenge with the same dose of TNCB (Figure 1B), in contrast to the re-re-challenged ears with 1% TNCB (Figure 1A), suggesting that the formation of LSM in re-challenged naïve skin requires a higher dose of the antigen challenge of more than 0.1% TNCB.
To confirm the long-term persistence of LSM (8), we rechallenged ears at 15, 30, 44, or 56 weeks after the 1 st challenge. In contrast to previous findings showing that the swelling response of the healed ears upon a re-challenge gradually diminished over time (8), the swelling of both naïve and healed ears upon a rechallenge increased over time in our experiments ( Figure 1C). Nevertheless, the LSM response was observed even after 56 weeks (Figure 1C), confirming the long-term persistence of TNCBinduced LSM.
We found that the dermis of 1% TNCB re-challenged healed ears markedly accumulated Gr-1 + cells (neutrophils and monocyte-derived cells) within 6 h (Figure 1D). At 24 h, Gr-1 + cells accumulated in the hair follicles of the healed ears. Flow cytometry on isolated ear cells at 24 h after the re-challenge showed that the healed ears contained more neutrophils (Ly6G Hi CD11b + ), but not other myeloid cells (Ly6G Lo CD11b + ) ( Figure 1E). Thus, the LSM response was characterized by the extensive accumulation of neutrophils.
To confirm the antigen specificity of TNCB-induced LSM, as shown previously with other haptens (6-8), we re-challenged both naïve and healed ears (harboring 1% TNCB-induced LSM) with Ox ( Figure 1F) or FITC with or without secondary sensitization before the re-challenge (Supplemental Figure 1E). In all cases, the extent of swelling was similar in both ears, indicating that TNCB-induced LSM did not respond to unrelated haptens, and thus this was an antigen-specific response. Moreover, we found that upon the re-re-challenge with TNCB, Ox-or FITC-induced CHS-experienced skin showed a normal LSM response against TNCB ( Figure 1F and Supplemental Figure 1E), indicating that original LSM was maintained after CHS induced by irrelevant antigens.

Counterpart in BALB/c Healed Ears
A previous study reported that CD8 + T RM cells predominantly accumulated in CHS-healed C57BL/6 mouse skin and were responsible for the LSM response (11). IHC on BALB/c mouse ear skin sections showed that the healed ears (day 35) contained higher numbers of CD3 + T cells both in the epidermis (including hair follicles and sebaceous glands) and the dermis than the naïve ears (Figures 2A,B). CD4 + and CD8 + cell numbers were both higher in the healed ears than in the naïve ears, with the CD4 + cells outnumbering the CD8 + cells (Figures 2A,B). The majority of CD4 + cells resided in the dermis of both ears, while CD8 + cells in the healed ears were in the epidermis and the dermis, with approximately 70% being in the epidermis (Figures 2A,B). Only a few dendritic epidermal γδ T cells were detected in BALB/c mice, which is consistent with previous findings (37,38), and γδ T cells did not increase in the healed ears (Figures 2A,B). Flow cytometry on isolated ear skin cells also showed that the numbers of CD3 + CD4 + and CD3 + CD8 + T cells, but not CD3 + CD4 − CD8 − (including γδ T cells), increased in the healed ears on day 36 after the 1 st challenge (Figure 2C). A similar increase in CD4 + and CD8 + cell numbers in the healed ears was also observed in CHS with Ox or FITC (Supplemental Figures 2A,B) and CHS with 1% TNCB with back skin sensitization (Supplemental Figure 2C). The IHC showed that dermal CD4 + cells sometimes outnumbered dermal CD3 + cells. The IF analysis revealed that the CD8 + cells in the ear sections were almost exclusively CD3 + , while CD4 + cells contained 10-20% of CD3 − non-T cells in both ears ( Figure 2D and Supplemental Figure 2D).
Skin T RM cells generally express CD69 and CD103, and sessile T RM cells have been detected in CD69 + compartments in previously inflamed skin (25,31). Moreover, skin T RM cells, but not circulating T cells, were shown to be resistant to in vivo antibody-mediated cell depletion (20,29). On day 36, approximately 70% of CD4 + T cells were CD69 + CD103 + in both ears, while approximately 20% were CD69 + CD103 − , and the majority of CD8 + T cells were CD69 + CD103 + regardless of epidermal and dermal localization ( Figure 2E and Supplemental Figure 2E). Moreover, CD4 + CD69 + and CD8 + CD69 + T cells in the healed ear skin were resistant to in vivo antibody-mediated cell depletion with anti-CD4 or -Thy1.2 mAbs despite the effective depletion of splenic T cells (Supplemental Figure 3A). Furthermore, the CD4 + T cells in the healed skin contained ∼15% of Foxp3 + regulatory T (Treg) cells (Supplemental Figures 3B,C), indicating that the majority of the skin CD4 + T cells were conventional T cells. These results suggest that the increased numbers of CD4 + and CD8 + T cells in the healed ears of BALB/c mice were T RM cells.
The skin and other organs are large and flexible niches for CD8 + T RM cells generated by pathogen infections, that is, repeated infections expand the T RM cell population by inducing their proliferation as well as the formation of new T RM cells from circulating T cells without displacing pre-existing populations (30,39,40). We investigated whether this was the case in CHS and CD4 + T RM cells (Figure 3). After the resolution of 2 nd CHS by the re-challenge (day 70), the ex-naïve ears contained similar numbers of CD4 + and CD8 + T RM cells as the healed ears on day 35. The ex-healed ears contained expanded epidermal, but not dermal CD8 + T RM cells, as well as a higher number of dermal CD4 + T RM cells than the 1 st CHS-healed ears. Since the ex-naïve and the ex-healed ears both showed a similar extent of swelling upon the re-re-challenge (Figure 1A), this result suggested that once T RM cells are formed, the skin shows a constant swelling response upon re-exposure to the same concentration of the antigen regardless of how many T RM cells are retained. After the resolution of the 3 rd CHS by the re-re-challenge (day 105), the ex-ex-healed ears contained higher numbers of dermal CD4 + and unchanged CD8 + T RM cells than the ex-healed ears on day 70. The increased rates of CD4 + and CD8 + T RM cells were the highest in the healed ears at the 1 st CHS (day 35) and gradually decreased in the 2 nd and subsequent CHS. These results demonstrated that both skin T RM cell populations were expanded by repeated CHS, with CD8 + T RM cells expanding exclusively in the epidermis and CD4 + T RM cells in the dermis. Thus, the skin is also a flexible niche for CD4 + T RM cells.  Figure 2B. Bars represent means ± SE. *P < 0.05 between both ears on each day of the analysis (two-tailed paired t-test). The numbers inside and outside the bars indicate the fold change in each value [epidermis (red), dermis (blue), and total number (black)] relative to those in naïve ears on day 35.

Absence of LSM in T Cell-Deficient or DO11.10 Mice and Development of the LSM Response Without Recruitment of Effector T Cells
To confirm the involvement of T RM cells in the LSM response of BALB/c mice, we initially analyzed C.B-17 SCID mice (lack T and B cells), BALB/c-nude mice (lack thymic T cells), and BALB/c-DO11.10 mice [in which many T cells express transgenic αβ TCR genes specific for the ovalbumin peptide on major histocompatibility complex (MHC) class II (I-A d )]. The SCID and the nude mice exhibited slight swelling with the 1 st CHS, and their healed ears did not show the LSM response upon a re-challenge with 1% TNCB (Figure 4A). The healed ears of SCID and nude mice did not accumulate T cells (Figure 4B and Supplemental Figure 4A) despite the nude mice having extrathymically differentiated T cells (41). The DO11.10 mice showed greater swelling than SCID and nude mice with the 1 st CHS, whereas the healed ears did not show the LSM response upon the re-challenge ( Figure 4A). The healed ears accumulated a large number of CD4 + T cells and a small number of CD8 + T cells (Figure 4B and Supplemental Figure 4A), and the majority of the CD3 + cells in the healed ears were DO11.10 TCR + cells ( Figure 4C). These results suggest that the TNCB challenge in DO11.10 mice induced the recruitment of both DO11.10 TCR + T cells and T cells with endogenous TCR, both of which were unable to mediate LSM. Collectively, the lack of an LSM response in these mice indicates that T cells are required for the initiation of the LSM response upon a re-challenge.
A previous study reported that skin T RM cells alone induced the LSM response without the recruitment of effector T cells from the circulation in C57BL/6 mice (9). To examine whether this was the case in BALB/c mice, CHS-healed mice were treated with the S1P receptor agonist FTY720, which prevents lymphocyte egress from the LNs, before and after the re-challenge in order to block the recruitment of newly activated effector T cells to the inflamed skin. The treatment did not affect the ear swelling response upon the re-challenge but completely inhibited the recruitment of effector T cells in both ears at 24 h after the rechallenge (Supplemental Figures 4B,C). Moreover, the marked accumulation of neutrophils in the re-challenged ears was not inhibited by FTY720 (Supplemental Figure 4D). Therefore, the T RM cells already present in the healed ears induced the LSM response in BALB/c mice.

CD4 + and CD8 + T RM Cells Redundantly Mediate the LSM Response
To identify which T RM cells mediate the LSM response, back skin-sensitized BALB/c mice were injected with anti-CD4 (GK1.5) or -CD8α (53-6.7) mAb before and after the 1 st challenge in order to inhibit the accumulation of each T cell subset during the 1 st CHS and the formation of T RM cells (by cell depletion and/or inhibition of T cell activation upon antigen presentation) (Figure 5A). Anti-CD4, but not anti-CD8α, mAb effectively depleted the circulating T cells, while both mAbs blocked the accumulation of T cells in the inflamed ear skin on day 1 after the 1 st challenge (Supplemental Figures 5A-D). CD4 + cell depletion resulted in more prolonged swelling in the 1 st CHS than in the PBS-treated control mice (Figure 5B). Upon the re-challenge, the extent of swelling in the healed ears was similar to that in the control mice, whereas that in the naïve ears was significantly greater (Figure 5B). Nevertheless, the peak swelling response of the healed ears at 1 day after the re-challenge was significantly greater than that of the naïve ears, suggesting the presence of the LSM response ( Figure 5B). CD4 + cell depletion resulted in the accumulation of a larger number of CD8 + T RM cells not only in the healed ears but also in the naïve ears but prevented the accumulation of CD4 + T RM cells (Figure 5C and Supplemental Figure 5E). In contrast, injecting anti-CD8 mAb had no effect on the swelling responses throughout the experiment (Figure 5B), and the healed ears mainly accumulated CD4 + T RM cells in addition to very few CD8 + T RM cells (Figure 5C and Supplemental Figure 5E).
We then adoptively transferred TNCB-sensitized BALB/c LN cells [whole, CD4 + cell-depleted ( CD4) or CD8] into SCID mice, and the swelling responses upon 1 st challenge and rechallenge were assessed (Figures 6A,B). The transfer of whole LN cells restored the LSM response upon the re-challenge ( Figure 6C) and the formation of CD4 + and CD8 + T RM cells in healed ears (Figure 6D and Supplemental Figure 6A). The transfer of CD8 LN cells also restored the LSM response, and the extent of swelling was similar to that in mice into which whole LN cells were transferred (Figure 6C). The healed ears contained increased numbers of CD4 + T RM cells and no CD8 + T RM cells (Figure 6D and Supplemental Figure 6A). We were unable to accurately assess CD4 LN cell-transferred SCID mice because a small area with abnormally accumulated CD4 + and CD8 + T cells was observed in the healed ear sections of three out of five mice for an unknown reason(s) (Supplemental Figures 6B-D).
We also adoptively transferred enriched splenic T cells from naïve BALB/c mice to DO11.10 mice lacking LSM (Figure 7). While the transfer of CD4 + or CD8 + T cells alone, or both, had no or a weak effect on the swelling response in the 1 st CHS, they restored the LSM response of the healed ears upon a re-challenge in DO11.10 mice.
Collectively, these results indicate that the presence of CD4 + or CD8 + T RM cells in the healed ears was sufficient for the development of the LSM response, and thus these cells redundantly initiate the LSM response.
CD4 + and CD8 + T RM Cells Rapidly Produce IFNγ and TNF Upon a Re-challenge While isolated CD8 + T RM cells in CHS-healed C57BL/6 mouse skin have been shown to secrete IFNγ upon an overnight restimulation (11), the identities of the cytokines secreted by CD8 + and CD4 + T RM cells early after the re-challenge that initiated the LSM response remain unknown. We attempted to clarify this using comparisons of the early phase (0, 1.5, 3, and 6 h) mRNA expression of several cytokines in whole ear extracts after the rechallenge ( Figure 8A and Supplemental Figure 7A). The levels of Ifng, Tnf, Il4, and Il13 were higher than those in the naïve ears within 3 h of the re-challenge, and thus these cytokines were the candidate early responsible genes that initiate the LSM response. The levels of Il1b, Il9, Il22, and the neutrophil chemoattractants Cxcl1 and Cxcl2 were higher in the healed ears than in the naïve ears at more than 3 h after the re-challenge, suggesting that these are downstream effectors. The levels of Il6, Il17a/f, Il18, Il33, and Tslp were similar between the naïve and the healed ears (Supplemental Figure 7A).
We compared the expression of the above genes at 6 h after the re-challenge between WT BALB/c and DO11.10 mice lacking LSM. All genes in both ears of DO11.10 mice showed similar expression levels, with most being similar to or lower than the levels present in WT naive ears (Figure 8B). This result supports the notion that T cells accumulating in the healed ears of DO11.10 mice cannot respond to TNCB, and the expression of downstream effector genes and the LSM response are not induced without the secretion of the early responsible genes.
The naïve and the healed ears that were not treated (no stimulation) or re-challenged at 3.5 h before (1% TNCB) were subjected to enzymatic digestion with brefeldin A for 3.5 h for intracellular cytokine analyses (Figures 8C,D and Supplemental Figure 7B). Similar to T RM cells forcibly activated with ionomycin + PMA during enzymatic digestion, a cell population strongly producing IFNγ and/or TNF emerged in both T RM cells in the re-challenged healed ears, with more effective production being observed in CD8 + T RM cells. The CD4 + T cells in naïve ears did not produce large amounts of IFNγ or TNF when re-challenged. In contrast, CD4 + and CD8 + T RM cells in the healed ears did not produce IL-4 ( Figure 8D). These results suggest that CD4 + and CD8 + T RM cells in the healed ears both produced IFNγ and TNF early after the re-challenge.
To clarify whether these cytokines were required for the initiation of the LSM response, we simultaneously injected CHShealed mice with anti-IFNγ and TNF-neutralizing mAbs with the re-challenge. However, the treatment had no or only a small inhibitory effect on the swelling response in both ears at 24 h after the re-challenge ( Figure 8E). An injection with an irrelevant mAb (ACK2) showed that mAb reached all dermal tissues, but not some epidermal regions in the re-challenged healed ears (Supplemental Figure 7C), suggesting that mAbmediated neutralization is ineffective for epidermal CD8 + T RM cells. Therefore, we used Rux (a JAK1/JAK2 inhibitor, the downstream effectors of many cytokine receptors including IFNγ) and BAY (a NF-κB inhibitor, the downstream effectors of the TNF superfamily, including TNF and IL-1). The injection of BAY induced diarrhea at 10 mg/kg and death at 20 mg/kg on the next day, and 5 mg/kg BAY + 10 mg/kg Rux also induced diarrhea. Thus, we injected mice with 6 mg/kg BAY + 8 mg/kg Rux (a dose that is not harmful to mouse health) in addition to the topical application of 0.3% BAY + 0.5% Rux to the ears at 2 h before the re-challenge. This treatment weakly suppressed ear swelling in both the naïve and the healed ears on the next day; however, the LSM response persisted (Figure 8F). Therefore, we examined whether massive neutrophil accumulation in the healed ears in the early phase of the LSM response (at 4 h, the mice are alive) may be inhibited by injecting a lethal dose (20 mg/kg) of inhibitors ( Figure 8G and Supplemental Figure 7D). Rux alone had no effect and BAY alone significantly inhibited the accumulation of neutrophils in the healed ears. The combined injection of Rux and BAY more effectively suppressed neutrophil accumulation in both ears. These results suggest that JAKs and NF-κB signaling cooperatively function in the LSM response.
CD4 + T RM Cells Are Maintained, While CD8 + T RM Cells Disappear Over Time Since LSM is a long-term memory (Figure 1C), we investigated whether T RM cells were maintained in the healed ears over a prolonged period. The epidermal and the dermal CD8 + T RM cells in the healed ears both gradually disappeared over time and reached the same levels as those in the naïve ears between 45 and 56 weeks (Figure 9 and Supplemental Figures 8A,B), as reported in C57BL/6 mice (11). In contrast, the healed ears contained more CD4 + T RM cells than the naïve ears at every time point by 56 weeks after the 1 st challenge, and their numbers in both ears persisted and even increased throughout that period (Figure 9 and Supplemental Figures 8A,B). The CD4 + T RM cells always exclusively localized in the dermis and outnumbered the CD8 + T RM cells. The ratio of Treg cells in the naïve and the healed ears did not markedly change even after 30 weeks (Supplemental Figure 3C). The IF analysis of the naïve and the healed ear skin sections on day 394 (56 weeks) showed that, similar to that of day 35 (Figure 2D), the CD8 + cells were exclusively CD3 + , while 90% of CD4 + cells were CD3 + (Supplemental Figure 8C). These results indicated that the LSM response in BALB/c mice was initially generated by both CD4 + and CD8 + T RM cells after the resolution of the 1 st CHS (at 5 weeks) and was then mediated by CD4 + T RM cells after more than 45 weeks because of the gradual loss of CD8 + T RM cells over time.
We then investigated the expression of genes required for the survival of T RM cells to explain the gradual loss of CD8 + T RM cells. Transforming growth factor β (TGFβ) and IL-15 signaling are involved in CD8 + T RM survival, and IL-7 signaling is involved in CD4 + T RM survival (20,21,42). The expression of IL-2/15 receptor β (IL-2/15Rβ) and IL-7Rα on CD4 + and CD8 + T cells in the naïve and the healed skin remained unchanged between 5 and 30 weeks after the 1 st challenge (Supplemental Figure 8D). The mRNA levels of Tgfb1/3, Il15, and Il7 in the whole ear extract were also unchanged between the naïve and the healed ears and between 5 and 34 weeks after the 1 st challenge (Figure 9B). The expression of Tgfb2 in the healed ears was slightly lower at 34 weeks than at 5 weeks but was still prominent (Figure 9B). These results suggest that the gradual loss of CD8 + T RM cells may not be solely explained by the loss of these molecules from T RM cells or skin.

DISCUSSION
ACD/CHS-healed skin in humans and animals forms LSM, which increases local antigen sensitivity and causes disease relapse. ACD-healed human skin dominantly has persisting CD4 + T cells in the dermis (35); however, CHS-healed C57BL/6 mouse skin predominantly has increased numbers of CD8 + T RM cells in the epidermis that are responsible for the formation . The mice were injected with Rux and/or BAY (20 mg/kg each) at 1 h before the re-challenge (n = 4-5 each group). All graphs represent means ± SE. *P < 0.05 (two-tailed paired or unpaired t-test). The asterisks in/on each bar indicate significant differences from each ear of the wild-type (B) or control treatment (E-G).
of LSM (10,11). It currently remains unclear whether CD4 + T RM cells contribute to the formation of LSM and flare-up reactions upon antigen re-exposure. We herein demonstrated that the CHS-healed skin of BALB/c mice contained increased numbers of CD4 + and CD8 + T RM cells, with a predominance of CD4 + T RM cells. Moreover, we showed not only that CD4 + or CD8 + T RM cells redundantly mediated the LSM response (the flare-up) upon the re-challenge but also that long-term LSM was subsequently mediated by CD4 + T RM cells over time because of the gradual loss of CD8 + T RM cells. Thus, the present results provide evidence to support the targeting of CD4 + T RM cells, in addition to CD8 + T RM cells, in order to prevent flare-up reactions of human ACD. In BALB/c CHS-healed skin, increased numbers of CD4 + T cells almost exclusively resided in the dermis, as shown in ACDhealed human skin (35), while CD8 + T cells largely resided in the epidermis, as shown in CHS-healed C57BL/6 mouse skin (11), suggesting that BALB/c mice have a combination of the traits of the C57BL/6 mouse and human skin in the context of LSM. Our result showing that the skin CD4 + or CD8 + T RM cells mediate the flare-up of CHS upon antigen re-exposure indicates that the flare-up is initiated within the epidermis by MHC class I-regulated LSM or within the dermis by MHC class II-regulated LSM. Thus, there is mechanistic redundancy and robustness in the formation of LSM. Moreover, our results suggest that both systems use the same molecular cascade to induce the flare-up reaction, namely, the rapid production of IFNγ and TNF from each T RM cell, which induced downstream molecules (such as IL-1, IL-9, and Cxcl1/2) produced by many other cell types in the skin (2,(43)(44)(45).
Similar to that in C57BL/6 mice (11), the CD8 + T RM cells in the CHS-healed skin of BALB/c mice gradually decreased in number over time. This decline has also been shown in skin CD8 + T RM cells generated from antigen-specific TCRtransgenic T cells by a viral infection or unspecific stimulation (34,46,47); however, the decrease observed in these cases was very sharp (many cells were lost within several weeks) and a small population stabilized and persisted. The non-recirculating property of skin CD8 + CD69 + CD103 + T RM cells in previously inflamed skin has repeatedly been demonstrated (25,30,33,34,48), suggesting that loss occurs within the skin. The present results showed that the mRNA expression of the T RM survival factor TGFβ2 slightly decreased in the healed ear skin over time; however, the mechanisms regulating the gradual loss of CD8 + T RM cells from the healed skin have not yet been elucidated in detail.
In contrast to their CD8 + counterparts, increased CD4 + T RM cell numbers in BALB/c CHS-healed skin persisted over time. However, it currently remains unclear whether they are a truly resident, non-migrating population. In the naïve skin, the vast majority of CD4 + T cells are recirculating and in equilibrium with the circulation despite their surface expression of CD69 and CD103 (29). In previously inflamed skin, heterogeneity has been reported in the increased number of CD4 + T cells, with both recirculating and resident populations being detected (29,31), and the sustained increase in CD4 + T cell numbers was attributed to the enhanced recruitment of memory cells from the circulation as well as the formation of the long-persisting T RM population (29). Nevertheless, the non-recirculating property of the increased number of skin CD4 + T cells with the T RM phenotype (CD69 + ) in the short term (∼2 weeks) generated after skin pathogen infections has been demonstrated (25,31). Moreover, residence is a dominant feature of CD4 + CD69 + memory T cells generated by viral infection in various nonlymphoid tissues (49). Since the majority of CD4 + T cells in the BALB/c CHS-healed skin were CD69 + cells that were resistant to antibody-mediated cell depletion, their sustained increase may result from their long lifespan and persistence within the skin. However, we cannot rule out the possibility that, during the longterm resting state after the resolution of CHS, some CD4 + T RM cells in the BALB/c healed skin down-regulate the expression of T RM markers, exit the skin, reenter the circulation, and eventually migrate to the distal skin and reassume the T RM phenotype, as reported previously for human skin (24).
The swelling response of both the naïve and the healed ears upon a re-challenge increased over time after the resolution of the 1 st CHS (Figure 1C). The increases in CD4 + T RM cell numbers in the ear skin over time (Figure 9) may be related to heightened responsiveness. However, if this is the case, the mechanisms by which TNCB-specific skin CD4 + T RM cells increase in number during the resting state, particularly in the naïve skin, and how this leads to stronger swelling intensities in both ears upon the re-challenge at 56 weeks than in the healed ears at 5 weeks after the 1 st challenge currently remain unclear. Further studies that consider the possibility of T RM cell relocation (24) and the enhanced recruitment of memory cells (29) are needed to obtain a more detailed understanding of this phenomenon.
The mechanisms responsible for the differential formation of CD4 + and CD8 + T RM cells in the CHS-healed skin of C57BL/6 and BALB/c mice have yet to be clarified. Moreover, the formation of CD4 + T RM cells in the C57BL/6 healed skin differs with CHS protocols and appears to require repeated hapten applications or intense inflammation for effective formation, with the CD8 + T RM cells dominating in many cases (9)(10)(11). Therefore, the C57BL/6 mice appear to produce more skin CD8 + T RM cells than CD4 + T RM cells in CHS, while the BALB/c mice are the opposite. However, the inflammation-healed skin of C57BL/6 mice was found to produce similar or higher numbers of CD4 + T RM cells than CD8 + T RM cells following a skin infection with Candida albicans (31) or the cowpox virus (25). These findings suggest that C57BL/6 mice retain the ability to form large numbers of skin CD4 + T RM cells under different settings. Thus, different regulatory mechanisms function in C57BL/6 and BALB/c mice to establish the balance between skin CD4 + and CD8 + T RM cell formation in the context of CHS.
In addition to ACD, growing evidence points to the involvement of T RM cells in inflammatory skin disorders frequently relapsing in the same locations, such as atopic dermatitis, fixed drug eruption, psoriasis, and vitiligo (13,15,16,50,51). Pathogenic CD4 + and/or CD8 + T cells with the T RM phenotype have been shown to remain in the remitting skin after treatments for these disorders (23,(52)(53)(54). In some settings, remitting skin after years of treatment still contains elevated numbers of CD4 + and CD8 + T cells (52,55), suggesting that the human skin CD4 + T RM cells also have a long lifespan and that CD8 + T RM cells do not always decrease in number after resolution. Based on the present results, the CD8 + and CD4 + T RM cells in the remitting skin may form LSM for each disease, create an antigen-sensitive state, and induce local flareup reactions upon re-exposure to antigens.
The present results highlight the important contribution of skin CD4 + T RM cells to the formation of LSM and the initiation of flare-up reactions in ACD/CHS. Thus, the removal of pathogenic CD4 + T RM cells, in addition to CD8 + T RM cells, has the potential to break the cycle of relapsing-remitting skin diseases.

DATA AVAILABILITY STATEMENT
The datasets generated for this study are available on request to the corresponding author.

ETHICS STATEMENT
The animal study was reviewed and approved by Animal Care and Use Committee of Tottori University.

AUTHOR CONTRIBUTIONS
AM designed and performed the research. AM and S-IH analyzed the data and wrote the manuscript.

FUNDING
This research was supported by JSPS KAKENHI Grant Numbers JP18K08297 and JP15K19076 (to AM) and JP19K07481 (to S-IH) and a research grant from the Takeda Science Foundation (to AM).