Chronic UV radiation–induced RORγt+ IL-22–producing lymphoid cells are associated with mutant KC clonal expansion

Significance

Ultraviolet (UV) exposure from the sun is the greatest risk factor for skin cancer. Toward preventive strategies, we elucidated the immune pathways critical to skin cancer development and identified specific cells and signals associated with mutated skin cell growth. UV light and Langerhans cells cooperate to induce lymphoid cells that produce interleukin-22 locally and are associated with mutant skin cell growth. This effect is highly dependent on the differentiation factor RORγt that can be inhibited with a topical agent. Such skin cancer–supporting immune cells were identified with distinctive markers: RORγt+ Sca-1+ CD103+ ICOS+ CD2^+/− CCR6+ intracellular CD3+. Together, these data elucidate key immune components of UV-induced skin cancer that may represent targets for skin cancer prevention.

Abstract

Chronic ultraviolet (UV) radiation exposure is the greatest risk factor for cutaneous squamous cell carcinoma (cSCC) development, and compromised immunity accelerates this risk. Having previously identified that epidermal Langerhans cells (LC) facilitate the expansion of UV-induced mutant keratinocytes (KC), we sought to more fully elucidate the immune pathways critical to cutaneous carcinogenesis and to identify potential targets of intervention. Herein, we reveal that chronic UV induces and LC enhance a local immune shift toward RORγt+ interleukin (IL)-22/IL-17A–producing cells that occurs in the presence or absence of T cells while identifying a distinct RORγt+ Sca-1+ CD103+ ICOS+ CD2^+/− CCR6+ intracellular CD3+ cutaneous innate lymphoid cell type-3 (ILC3) population (uvILC3) that is associated with UV-induced mutant KC growth. We further show that mutant KC clone size is markedly reduced in the absence of RORγt+ lymphocytes or IL-22, both observed in association with expanding KC clones, and find that topical application of a RORγ/γt inhibitor during chronic UV exposure reduces local expression of IL-22 and IL-17A while markedly limiting mutant p53 KC clonal expansion. We implicate upstream Toll-like receptor signaling in driving this immune response to chronic UV exposure, as MyD88/Trif double-deficient mice also show substantially reduced p53 island number and size. These data elucidate key immune components of chronic UV–induced cutaneous carcinogenesis that might represent targets for skin cancer prevention.

The capacity to identify and therapeutically target the major relevant immune pathways driving disease pathogenesis has proven particularly advantageous for controlling inflammatory cutaneous disorders. This has occurred despite the complexity of lymphoid cell subtypes and cytokine receptor networks active within cutaneous lesions and draining lymph nodes. Marked durable responses are now possible with antibody-based therapeutic agents disrupting, for example, the interleukin (IL)-23/IL-17A (Th17) axis in psoriasis vulgaris (1) and the IL-4/IL-13 (Th2) axis in atopic dermatitis (2).

Exposure to ultraviolet (UV) light results in extensive DNA damage to keratinocytes (KC) because of two major phenomena. UV exposure, mostly in the type B UV (UVB) bandwidths, directly targets genomic DNA as a chromophore in which the absorbed energy readily and immediately induces cyclobutane pyrimidine dimers. This is complimented by a more complex indirect (and relatively delayed) series of energy transfers initiated by epidermal molecules, for example, melanin degradation products, that absorb/transfer UV energy before giving rise to triplet state intermediates that may also eventuate into DNA damage (3). Possible fates for such UV-damaged KCs include enzymatic DNA repair, elimination via apoptosis induction, metabolic stress that induces NKG2D-ligand surface expression to mark the damaged cells for elimination, and acquisition of mutations within cell cycle regulators (e.g.TP53, CDKN2A) and stimulators (e.g., RAS, NOTCH-1,2,3) that endow apoptosis resistance and clonal persistence.

Martincorena et al. (4) have definitively shown by genomic deep sequencing that such driver mutations persist within chronically sun-exposed skin and that mutant-associated KC clones accumulate as epidermal collections long before the emergence of clinically apparent skin tumors of premalignant actinic keratoses (AK) and cutaneous squamous cell carcinomas (cSCC). Currently, clinical efforts in preventing the outgrowth of these skin neoplasms have included the topical application of chemotherapeutic (e.g., 5-fluorouracil) or immunostimulatory (e.g., imiquimod) agents or the administration of systemic retinoids in attempt to minimize, eliminate, and/or differentiate the mutant clones (5). In this context, we hypothesize that elucidating the immunopathogenesis of chronic UV exposure might identify other potential targets and strategies for skin cancer prevention.

It is in this setting of epithelial tissue harboring of persistent driver mutant clones that chronic inflammatory states have long been appreciated for their association with malignant development and progression, for example, inflammatory bowel disease and increased risk of colon cancer (6). Indeed, locally secreted IL-17A and IL-22, both known stimulators of epithelial proliferation, have been implicated in epithelial tumor promotion, including within the gut, lung, and skin (7–9). Nonetheless, IL-22–inhibiting antibodies have met with limited clinical success in preventing or treating intestinal tumors, and while anti–IL-22 antibodies injected into established cSCC might prevent progression (10), the role of IL-22–producing cells in mutant KC clonal expansion and photocarcinogenesis prior to tumor emergence has not been fully elucidated.

Recognizing that UVR exposure induces an epidermal damage–driven inflammatory response, we sought to more fully characterize the effects of chronic UV exposure on the local cutaneous immune milieu relevant to the promotion of photocarcinogenesis. A cascade of KC-derived triggers of the local immune response are precipitated by acute UV exposure in the skin, including Toll-like receptor (TLR) ligands (e.g., U1 RNA), antimicrobial peptides (e.g., beta-defensins, S100A7), thymic stromal lymphopoietin (TSLP), and proinflammatory cytokines (e.g., IL-1, IL-36). In addition, chronic UVR has been reported to markedly decrease epidermal Langerhans cells (LC) and increase local and systemic levels of cytokines (e.g., IL-10, TGFβ) that down-regulate cellular antitumor immunity. However, little is known about if and how chronic UV alters the predominant local immune milieu to promote mutant KC clonal expansion so fundamental to cutaneous carcinogenesis. Having previously demonstrated that the absence of epidermal LC markedly decreases mutant KC clone growth during chronic UV exposure, we also considered whether targeting any identifiable associated dominant immune cell population and signals might also limit the expansion of mutant KC and thereby uncouple the tumor-promoting effects of UV-associated inflammation. Herein, we reveal that chronic UV exposure stimulates specific skin immune cell populations, including RORγt-differentiated innate lymphoid cells (ILC) and T cells, that shift the local cytokine milieu to an IL-22/IL-17A–rich state in association with mutant KC growth. Moreover, we identify and implicate a distinct cutaneous (CD4- NKp46- MHCII-) ILC type-3 (ILC3) population (that we term uvILC3) of phenotype RORγt+ Sca-1+ CD103+ ICOS+ CD2^+/− CCR6+ intracellular (ic)CD3+ that drives chronic UV–induced mutant KC clonal expansion in the absence of T cells. We believe that this deciphering of the immunopathogenesis of UV carcinogenesis identifies potential translational targets for cutaneous cancer prevention for patients with chronically photodamaged (mutant clone-harboring) skin and, in particular, potentially for those patients (e.g., organ transplant recipients) with iatrogenically compromised T cell function.

Results

LC Facilitate a Shift in Epidermal RORγt+ IL-22+ IL-17+ ILC in Association with UV-Induced p53 Mutant KC Expansion.

Human sun-exposed skin contains clones of KC that are histologically normal but harbor p53 mutations (11, 12). Chronic UVB–exposed mouse epidermis manifests similar p53 mutant KC clonal collections, also termed “p53 islands” (13–15), readily identifiable by immunofluorescent staining of increased p53 protein levels within epidermal sheets. We compared T cell–deficient (FVB.TCRβ^−/−δ^−/−) LC-deficient hu-Langerin DTA (DTA) and T cell–deficient LC-intact normal littermate control (NLC) mice that had been exposed to UVB three times weekly for 9 wk. Consistent with our previous findings (16), the average p53 island size was reduced by ∼30% in the absence of LC (Fig. 1A), and in NLC mice, CD207+ LC were preferentially located within or adjacent to p53 islands (Fig. 1B). In order to assess the effects of chronic UV exposure on the distribution of skin ILC populations in the presence or absence of LC, cell suspensions prepared from unexposed or 9-wk UVB-exposed skin were analyzed by flow cytometry for the expression of characteristic ILC subset-defining transcription factors and their associated cytokines (ILC3: RORγt and IL-22, IL-17A; ILC2: Gata3 and IL-13; ILC1: T-bet and IFN-γ; and NK: Eomes and IFN-γ; Fig. 1C) (17). ILC were identified as CD45+ Thy1+ LIN− cells (gating strategy in SI Appendix, Fig. S1). Chronic UV exposure led to the marked expansion of RORγt+ ILC3 from 5.6% of all ILC in unexposed NLC skin to 30.7% following exposure (Fig. 1C). RORγt+ ILC3 were also significantly increased in LC-deficient DTA skin following chronic UV exposure but to a much lesser extent. In contrast, T-bet–expressing ILC1 predominated in unexposed skin and were maintained at a similar level in LC-deficient DTA skin following UV exposure. However, in LC-intact NLC skin, T-bet+ cells were substantially reduced following UV exposure, while Gata3+ ILC2 were largely unchanged, and Eomes-expressing NK cells (rare in unexposed skin) were barely detectable after UV exposure. This flow cytometry data were used to calculate absolute numbers of cells (SI Appendix, Fig. S2), revealing a 9.65-fold increase in RORγt+ ILC3 in UV-exposed NLC skin compared with a 4.32-fold increase in the absence of LC (P < 0.0001). When cell suspensions were restimulated in vitro with a combination of phorbol 12-myristate 13-acetate (PMA), ionomycin, IL-2, and IL-23 (Fig. 1D), the UV-induced shift toward RORγt+ ILC3 was even further noted by a marked increase in the percentage of ILC ex vivo producing IL-22 and IL-17A in those isolated from UV-exposed skin. IL-22–producing cells increased ∼10×, from an average of 2.78% in unexposed skin to 29.35% in UV-exposed NLC skin and 78.13% of IL-22+ cells costained with IL-17A. In skin from DTA mice, a smaller but proportional increase in IL-22+ cells, from 1.41% in unexposed skin to 16.63% in UV-exposed skin, was observed, again with the majority of these (72.52%) also producing IL-17A. In marked contrast, the percentage of cells producing IFN-γ and IL-13 did not significantly change with UV exposure, nor were they dependent on the presence of LC.

Fig. 1.

LC increase UV-induced mutant keratinocytes in association with RORγt+ IL-22+IL-17+ ILC. (A) Average p53 island area is reduced by 28.98% in LC-deficient (DTA) versus LC-intact (NLC) FVB.TCRβ^−/−δ^−/− T cell–deficient epidermis following 9 wk UVB (400 J/m², 3×/wk) exposure. Each dot represents one mouse; the line is mean. Representative immunofluorescent images of p53 islands (red) and CD207+ LC (green) in epidermal sheets. (Scale bar, 20 μm.) (B) In LC-intact (NLC) mice, CD207+ LC density is greater in association with p53 islands (assoc: in/within 30 μm of island; distant: > 30 μm from island). Flow cytometric analysis of characteristic ILC transcription factors (C) and cytokines (D) in cell suspensions prepared from unexposed (UT) versus chronic UVB–exposed (400 J/m², 3×/wk, 9 wk) LC-deficient (DTA) versus LC-intact (NLC) FVB.TCRβ^−/−δ^−/− mouse skin. Each dot represents one mouse; the line is mean. Representative contour plots are gated on CD45+Thy1+LIN− skin ILC. *P < 0.05, **P < 0.01, ***P < 0.001.

To address whether UV-induced local immune shifts would also be detectable on a, relative to FVB, skin cancer–resistant (and p53 island-resistant) background, we analyzed T cell–deficient C57BL/6 (B6) mice (B6.TCRβ^−/−δ^−/−) for UV-induced ILC. Relative changes in gene expression in unexposed versus chronically UV–exposed skin were first assessed by qRT-PCR (SI Appendix, Fig. S3A). Consistent with findings in T cell–deficient FVB skin, RORγt gene expression increased (mean 8.79-fold) following UV exposure, while no significant changes in expression were observed for Gata3, T-bet, or Eomes. Approximately 10× increased gene expression was also noted for IL-22 (12.07-fold), IL-17A (12.67-fold), and IL-17F (8.86-fold) following chronic UV exposure, while other KC growth factors notably remained unchanged (AREG, IGF1, FGF7) or were decreased (EGF, -2.32-fold). Consistent with a shift away from ILC1/NK, decreases were also seen in expression of IFN-γ (−8.13-fold) and perforin (PRF1, −6.91-fold), while a slight increase in ILC2 cytokine IL-13 (2.52-fold) was observed. Of note, increases in proinflammatory, ILC3-promoting cytokines IL-1β (6.53-fold), IL-6 (4.60-fold), and IL-23 (2.64-fold), along with marked increases in expression of inducible nitric oxide synthase (NOS2, 10.63-fold) and IL-36a (11.67-fold), were also observed. Taken together, these gene expression patterns are consistent with a chronic UV–induced ILC3-supportive cutaneous microenvironment that occurs in the absence of T cells and regardless of genetic (cancer susceptibility) background.

To further directly assess for characteristic ILC transcription factors and cytokines (SI Appendix, Fig. S3 B and C) in T cell–deficient B6 mice, cell suspensions prepared from unexposed or chronically UV–exposed B6.TCRβ^−/−δ^−/− skin were also analyzed. As in FVB.TCRβ^−/−δ^−/−.NLC (Fig. 1), UV exposure induced a marked increase in the percentage of ILC-expressing RORγt, from 5.54% in unexposed skin to 30.94% following exposure, while the percentage of cells expressing other transcription factors was essentially unchanged in B6.TCRβ^−/−δ^−/− mice. However, some genetic influences were evident in the absolute number of UV-induced ILC3 (SI Appendix, Fig. S3D), with the B6.TCRβ^−/−δ^−/− mouse skin’s 5.16-fold increase in RORγt+ cells indicative of significantly fewer RORγt+ ILC than the absolute number observed in FVB.TCRβ^−/−δ^−/−.NLC (502 ± 107 B6 versus 1,593 ± 218 FVB, P = 0.0008) (SI Appendix, Fig. S2). When cell suspensions from chronic UV–exposed versus unexposed B6.TCRβ^−/−δ^−/− skin were restimulated in vitro, we again identified a substantial increase in UV-associated IL-22+ cells (SI Appendix, Fig. S3C, mean UV 27.91% versus unexposed 0.25%), the majority (mean 63.76%) of which were also IL-17A+. Under these conditions, we also observed a small but statistically significant increase in IL-13+ cells, and a decrease in IFN-γ+ cells, following UV exposure.

Distinct Phenotypic Characterization of Chronic UV–Induced Skin RORγt+ ILC (uvILC3).

To more fully characterize and contextualize the RORγt+ cells that expand in response to UV exposure and to compare them to previously described ILC3 phenotypic subsets in skin and other organs, we costained cell suspensions prepared from the skin of chronic UVB–exposed (T cell–deficient) B6.TCRβ^−/−δ^−/− mice with RORγt and a panel of relevant markers (Fig. 2A and SI Appendix, Fig. S4A). UV-induced RORγt+ (or uvILC3) cells were observed to uniformly express Sca-1, CD103, ICOS, and CCR6(dim), while ∼50% also expressed CD2. These cells did not coexpress appreciable levels of MHC II, NKp46, cKit, CD4, NRP1, CCR4, CCR10, KLRG1, NK1.1, Foxp3, CD5, or surface CD3ε. Further distinguishing the uvILC3, nearly all UV-induced RORγt+ ILC expressed intracellular CD3ε (icCD3ε), as did nearly all IL-22– and IL-17A–producing cells (Fig. 2B).

Fig. 2.

The phenotype of UVB-induced skin ILC3 in T cell–deficient mice is distinct: RORγt+ Sca-1+ CD103+ ICOS+ CD2± CCR6+ icCD3ε+ CD4− NKp46− MHCII−. Skin cell suspensions prepared from chronic UVB– (400 J/m², 3×/wk, 9 wk) exposed B6.TCRβ^−/−δ^−/− mice were analyzed by flow cytometry for correlation of (A) RORγt expression along with a variety of other markers and (B) intracellular CD3ε expression along with characteristic ILC cytokines. Representative contour plots are gated on CD45+Thy1+LIN− cells. The composite (A) was prepared from five experiments in which RORγt+ ILC ranged from 31.0 to 41.1% of all skin ILC. Additional markers are shown in SI Appendix, Fig. S4. (C) Following chronic UVB exposure, immunofluorescent staining of epidermal sheets from T cell–deficient mice demonstrate intracellular CD3ε+ cells (icCD3; green) and CD207+ LC (blue) in mutant KC p53 islands (red). (D) Immunofluorescent staining of frozen sections from T cell–deficient mouse skin shows increased intracellular CD3ε+ cells (icCD3; red) following chronic UVB exposure. (Scale bar, 20 μm.)

The coexpression of icCD3ε by RORγt+ uvILC3 cells allowed us to identify these cells by immunofluorescent staining of epidermal sheets and skin sections from T cell–deficient (TCRβ^−/−δ^−/−) mice. Following chronic UV exposure, icCD3ε+ uvILC3 cells can be readily detected within p53 islands (Fig. 2C), and when epidermal sheets from UV-exposed, T cell–deficient RORγt-EGFP reporter mice are examined, we find nearly all EGFP+ cells costain with icCD3ε (SI Appendix, Fig. S4B). While such cells are very rare in cross-sections of untreated skin (Fig. 2D), they are readily identifiable in chronically UV–exposed skin and are notably enriched at the dermal–epidermal junction and in association with hair follicles, that is, sites of KC tumor cell origins (18, 19).

Low levels of CD3ε transcripts have been previously reported to be differentially expressed in murine gut ILC3 (20) and in human tonsil ILC3 (21). CD2+ ICOS− epidermal lymphoid cells (22) are reported to have heterogeneous expression of intracellular CD3, although characteristic transcription factor expression was not defined for these cells. Yet, the phenotype of UV-induced RORγt+ ILC is distinct from each of these previously reported populations. To rule out the possibility that coexpression of RORγt and icCD3ε was occurring in a population of T cell precursors in this model (TCRβ^−/−δ^−/−), we assessed Rag1 and Rag2 gene expression (SI Appendix, Fig. S4C). Although readily detectable in control WT.B6 thymus, expression of Rag1 and Rag2 was undetectable in CD45+ cells from both untreated and chronic UVB–exposed FVB.TCRβ^−/−δ^−/− skin, while expression of both CD3ε and RORγt increased with UV exposure. We also investigated the effects of chronic UVB exposure on skin ILC in Rag2^−/− mice (SI Appendix, Fig. S5). As we observed in FVB.TCRβ^−/−δ^−/− mice (Fig. 1) and in B6.TCRβ^−/−δ^−/− mice (SI Appendix, Fig. S3), chronic UVB exposure of Rag2^−/− mice leads to an expanded population of RORγt+ skin ILC3 (untreated 1.23 ± 0.12%, UVB-exposed 19.67 ± 1.56% RORγt+, P = 0.0001), and these UVB-induced RORγt+ ILC3 are surface CD3ε negative but do express intracellular CD3ε.

UV Exposure Induces a LC-Promoting RORγt/IL-22/IL-17 Pathway and an IFN-γ Response in T cell–Intact Mice.

Because ILC populations may be increased in T cell–deficient mice and a more complex response to UV might be expected in the presence of T cells, we next analyzed skin from T cell–intact, LC-deficient (DTA) versus LC-intact (NLC) mice to assess for changes in the expression of key genes by qRT-PCR in association with chronic UV exposure (Fig. 3A). In LC-intact mice following UV exposure, increased expression of RORγt, IL-22, IL-17A, IL-1β, and IL-6 were noted, consistent with the cytokine milieu observed in UV-exposed, LC-intact T cell–deficient mice (SI Appendix, Fig. S3). However, in contrast to UV-exposed, T cell–deficient mice that showed an 8.13-fold decrease in expression of IFN-γ, T cell–intact mouse skin showed a 13.84-fold increase in IFN-γ mRNA. Notably, in the absence of LC (DTA), UV exposure did not significantly increase expression of RORγt, IL-22, IL-17A, or IL-1β, while induction of IFN-γ and IL-6 were blunted. Furthermore, average p53 island area was decreased 44.59% in LC-deficient DTA mice (Fig. 3B). Cell suspensions prepared from the skin of LC-intact mice that were untreated (UT) or chronic UVB–exposed (UVB) were also assessed (Fig. 3 C–E). UV exposure can lead to an increase in Treg and subsequent immune suppression (23, 24), and a subset of RORγt+ Treg have been described (25, 26). While our chronic UVB protocol did lead to an increase in the percentage of both RORγt+ and Foxp3+ cells, these were independent subsets without coexpression (Fig. 3C). A shift in T-bet fluorescence was also noted but difficult to quantitate because a distinct population was not evident. Costaining with TCRβ and TCRδ revealed that in T cell–intact mice, the RORγt+ population is comprised of ∼53% αβ T cells, 39% γδ T cells, and 8% ILC (Fig. 3D). Further examination of these ILC revealed that nearly 100% of the RORγt+ surface CD3−TCR− cells expressed intracellular CD3ε (SI Appendix, Fig. S6) as seen in the T cell–deficient models. Following in vitro restimulation, cell suspensions from chronically UV–exposed T cell–intact skin again showed increased IL-22+ and IL-17+ cells, with 90.29% of IL-22+ cells also expressing IL-17 (Fig. 3E). Consistent with the observed pattern of gene expression, IFN-γ+ cells were also increased following chronic UV exposure. Together, this data suggests that LC potentiate UV-induced pathways, which lead to increased IL-22– and IL-17–producing T cells and uvILC3 but also IFN-γ–producing T cells, the net effect of which is nonetheless still associated with the expansion of mutant p53 KC clones.

Fig. 3.

Chronic UVB induces distinct populations of IL-22+ IL-17A+ ILC3 and Th cells. T cell–intact, LC-deficient (DTA) versus LC-intact (NLC) B6 mice remained untreated (UT) or were exposed to chronic UVB (400 J/m², 3×/wk, 9 wk) before skin was harvested for analysis of gene expression (A) or quantification of p53 islands (B). P53 islands in DTA mice are 44.59% smaller than those found in NLC mice. A flow cytometric analysis of CD45+Thy1+LIN− T cells + ILC from untreated (UT) or chronic UVB–exposed (UVB) WT.B6 mice (C) shows increased RORγt and T-bet expression. The UVB-induced RORγt+ cells are Foxp3− and composed of γδ T cells, αβ T cells, and CD45+Thy1+LIN−TCR− ILC (D). Increased RORγt and T-bet expression correlates with increased IL-22, IL-17A, and IFN-γ (E) production following chronic UVB. *P < 0.05, **P < 0.01, ***P < 0.001. This figure is representative of 12 experiments, with each panel repeated two to three times, and utilized 10 UT DTA, 19 UVB DTA, 10 UT NLC, 20 UVB NLC, 27 UT WT.B6, and 32 UVB WT.B6 mice. NLC and DTA mice were analyzed individually; WT.B6 were analyzed in pools of two to four mice/pool.

RORγt Deficiency Is Associated with Reduced p53 Island Size.

We hypothesized that LC facilitation of p53 island expansion was due at least in part to the promotion of elevated levels of IL-22 and IL-17 produced by RORγt+ T cells and uvILC3. To test this, we compared p53 island development in B6 wild-type (WT) and RORγt^−/− mice (Fig. 4 A and B). While the number of p53 islands detected was significantly, albeit modestly (10.81%), reduced in the absence of RORγt+ cells, the mean area of the p53 islands was markedly reduced (67.36%). The epidermis of untreated mice contained equivalent numbers of CD3+ cells (SI Appendix, Fig. S7); however, following UV exposure, far fewer CD3+ cells were observed in the epidermis of RORγt-deficient mice (14.29 ± 4.21/mm²) relative to WT mice, where CD3+ cells (144.54 ± 24.6/mm²) were found preferentially associated with the p53 islands (Fig. 4A). To consider the full potential effect of uvILC3 on UV-induced mutant KC clonal expansion, we examined T cell–deficient (B6.TCRβ^−/−δ^−/−) RORγt-intact versus -deficient mice (Fig. 4C) and found RORγt deficiency was associated with a 42.58% decrease in p53 island incidence and 64.50% decrease in p53 island size. Intracellular CD3ε+ cells were barely detectable in UV-exposed, T cell–deficient RORγt-deficient epidermis (2.97 ± 1.58/mm²) but were found associated with p53 islands in RORγt-intact epidermis (18.78 ± 2.58/mm²), albeit at lower levels than are seen for CD3+ cells in T cell–intact mice. These data strongly suggest that uvILC3 are the principal immune cell drivers of UV-induced KC clonal expansion under conditions of T cell compromise.

Fig. 4.

Chronic UV–induced epidermal KC p53 mutational burden is dependent on RORγt+ cells. T cell–intact (A and B) and T cell–deficient (C) RORγt-intact versus RORγt-deficient mice were exposed to chronic UVB (400 J/m², 3×/wk, 10 wk), and epidermal sheets prepared for quantification of p53 islands and CD3+ T cells or, in T cell–deficient mice, intracellular CD3+ ILC. CD3+ cells are more frequently seen in RORγt-intact epidermis and associated with p53 islands (assoc: in/within 30 μm of island; distant: > 30 μm from island). Representative immunofluorescent images (B) of T cell–intact RORγt-intact (WT) versus deficient (RORγt^−/−) show CD207+ LC (green), CD3+ cells (blue), and p53 island (red). (Scale bar, 20 μm.) T cell–deficient mice (D and E) remained unexposed or were exposed to chronic UVB (400 J/m², 3×/wk, 7 wk). During the final 2 wk of exposure, mice were treated topically with vehicle or 1% RORC inhibitor GSK2981278, and then skin was harvested for analysis of gene expression (D) (RQ to unexposed) and quantification of p53 islands (E). *P < 0.05, **P < 0.01, ***P < 0.001.

Topical Application of RORγt Inhibitor Reduces UV-Induced Cytokine Expression and Mutant KC Clonal Expansion.

To begin to consider the translational potential of uvILC3 inhibition, we exposed T cell–deficient mice to chronic UVB and, during the final 2 wk of exposure, percutaneously applied the RORγ/γt inhibitor GSK2981278 (27). RORγ/γt inhibitor application reduced expression of IL-22 by 78.32%, IL-17A by 90.65%, IL-17F by 89.12%, and IL-36α by 56.61% (Fig. 4D), and there was an observed trend toward reduction of IL-1β, IL-6, and IL-20. To determine whether these changes in cytokine expression were sufficient to affect mutant KC clonal expansion, we analyzed p53 island size in these same mice (Fig. 4E) and found a 36.53% decrease in p53 island size in mice treated with the RORγ/γt inhibitor. These results were largely recapitulated in T cell–intact mice (SI Appendix, Fig. S8) treated with 1% RORγ/γt inhibitor during their final 3 wk of UV exposure, with a decreased expression of IL-22 by 79.12%, IL-17A by 76.30%, IL-17F by 88.08%, IL-17C by 77.47%, and IL-36α by 59.02% (SI Appendix, Fig. S8A). We also observed increased expression of T-bet (up 50.87%) and NKp46 (up 57.42%), suggesting a shift in treated skin toward a cytokine milieu known to be more conducive to cell-mediated immunity. However, IFN-γ expression, while increased, did not reach statistical significance, and we also observed an increase in Treg-associated Foxp3 expression (up 59.65%). These changes could reflect antitumor effects, such as increased NK or Th1 activity, but could also reflect increased Treg activity that may contribute to immune suppression. In control- (vehicle only) treated mice, total mutant p53 island area was found to significantly correlate with cytokine gene expression levels of IL-22 (r = 0.6213, P = 0.0276), IL-20 (r = 0.8429, P = 0.001), and IL-36a (r = 0.6075, P = 0.0312), consistent again with a major role for these cytokines in driving clonal proliferation (SI Appendix, Fig. S8B).

Mutant p53 KC Clonal Expansion Is Dependent on IL-22.

To determine the degree to which IL-22 alone contributes to the promotion of mutant p53 KC clones, we assessed p53 island incidence and size in IL-22^−/− mice relative to WT (Fig. 5 A and B). Following chronic UVB exposure, a significant, albeit modest, decrease (13.91%; P = 0.0451) in the number of p53 islands was detected, while a substantially larger decrease (60.10%; P < 0.00001) in the size of the p53 islands was seen in IL-22–deficient mice. CD3+ cells were found preferentially associated with p53 islands in both WT and IL-22^−/− mice (Fig. 5A); however, the number of p53 island–associated CD3+ cells was much greater in the relatively p53 island–rich WT (430.98 ± 30.78/mm²) than in IL-22^−/− (164.74 ± 19.90/mm²) epidermis. We also assessed minimal epidermal thickness and number of proliferating (Ki67+) epidermal cells in these mice (SI Appendix, Fig. S9). No baseline difference in either epidermal thickness or Ki67+ cell density was detected in WT versus IL-22^−/− skin; however, both metrics were reduced in IL-22^−/− mice following chronic UVB exposure. Further examination of the Ki67+ cell density, specifically within p53 islands, revealed a 4.11-fold increase in proliferating mutant p53 KC (Fig. 5D) in WT epidermis when compared with IL-22^−/−. The substantial dependence of mutant KC clonal expansion on IL-22 occurs even in the absence of T cells (i.e., in B6.TCRβ^−/−δ^−/−, Fig. 5C), as IL-22 deficiency limited p53 island number (12.31% decrease; P = 0.0261) and, much more substantially, p53 island size (60.84% decrease; P = 0.0006). Epidermal icCD3ε+ cells were also enumerated in these mice, and although the number of such cells is fewer relative to CD3+ cells in T cell–intact mice, we again found icCD3+ cells preferentially associated with p53 islands in IL-22–intact mice and at greater than twice the density of that seen in IL-22^−/− mice (13.87 ± 3.91 versus 6.20 ± 1.27/mm², P = 0.0359), in which no preferential localization was detectable. These results confirm that IL-22 production by T cells (e.g., Th17/22 cells), or under a state of T cell compromise by ILC (e.g., our implicated uvILC3), impart a major promotion effect on the clonal expansion of p53 mutant KC within chronically UV–exposed skin.

Fig. 5.

Mutant p53 keratinocyte clonal expansion is dependent on IL-22. T cell–intact (A, B, and D) and T cell–deficient (C) IL-22–intact versus IL-22–deficient mice were exposed to chronic UVB (400 J/m², 3×/wk, 12 wk), and epidermal sheets prepared for quantification of p53 islands and CD3+ T cells or intracellular CD3+ ILC (assoc: in/within 30 μm of island; distant: greater than 30 μm away from island). Representative immunofluorescent images (B) of T cell–intact IL-22–intact (WT) versus deficient (IL-22^−/−) show CD207+ LC (green), CD3+ cells (blue), and p53 island (red). Costaining with Ki67 (D, blue) shows decreased p53 island proliferation in the absence of IL-22. (Scale bar: 20 μm.) *P < 0.05, **P < 0.01, ***P < 0.001.

Given that RORγt+ uvILC3 and UV-induced Th17/22 cells may produce IL-17 in addition to IL-22 (Figs. 1 and 3), we compared chronic UVB–exposed T cell–intact IL-17A/F^−/− mice with WT.B6 (SI Appendix, Fig. S10). Although similar numbers of p53 islands were observed in both mouse strains (WT 8.17 ± 0.46, IL-17A/F^−/− 7.45 ± 0.52/cm², P = 0.1637), the p53 islands in IL-17A/F^−/− mice were 26.14% reduced in size (WT 12390 ± 735, IL-17A/F^−/− 9151 ± 573 mm², P = 0.0023). Indeed, IL-17A does appear to have a modest effect on mutant KC clonal expansion, albeit not at the level of impact seen in IL-22 deficiency. Taken together, these findings further strengthen the translational potential of targeting RORγt differentiation to limit mutant KC clonal expansion induced by chronic UV exposure.

TLR Signaling Initiates the Immunopathogenesis of UV-induced Mutant KC Clonal Expansion.

We have indicated that LC facilitate p53 island growth by producing ILC3 and/or Th17/22-stimulating cytokines IL-23 and IL-6 in response to UV-induced signals. KC are known to produce a variety of damage-associated molecular patterns (DAMPs) in response to UV exposure, including HMGB1 (28) and U1 RNA (29), and several of these DAMPs act by signaling through TLR including those present on epidermal LC (30, 31). To determine whether TLR signal disruption could limit the critical downstream cytokine signaling cascade we found associated with mutant KC clonal expansion, we studied mice deficient in the two major TLR adaptor proteins, MyD88 and Trif, to examine the effects of TLR signal disruption on the development of p53 islands. When cells were quantified in epidermal sheets at baseline, no difference was found in the number of CD207+ LC, Vγ5/Vδ1+ (17D1+) dendritic epidermal T cells (DETC), or other CD3+ cells in untreated WT.B6 compared with MyD88^−/−Trif^−/− double-deficient mice (SI Appendix, Fig. S11). Following chronic UV exposure (Fig. 6A), we observed that disruption of TLR signaling led to a marked decrease (57.20%; P = 0.0002) in the number of p53 islands detected as well as in the mean island size (43.21%; P < 0.0001).

Fig. 6.

TLR signaling initiates the immunopathogenesis of UV-induced mutant KC clonal expansion. (A) P53 island density (Left) and average island size (Right) are reduced by 57.20 and 43.21%, respectively, in MyD88^−/−Trif^−/− (KO) mice compared with C57BL/6 (WT) mice following chronic UVB exposure (400 J/m², 3×/wk, 9 wk). Each dot represents one mouse. ***P < 0.001. (B) LC line XS106 expresses a variety of DAMP receptors. Gene expression following 40 cycles of qRT-PCR, expressed as 40-Ct (threshold cycle), and shown relative to β-actin (ACTB). Production of ILC3-activating cytokines by XS106 (C) or isolated LC (D) following exposure to DAMP ligands. Cells were cultured in the presence of predetermined optimal concentrations of DAMP ligands for 48 h, and then supernatants were collected for analysis of IL-23p19 or IL-6 by Bio-Plex assay. Concentrations used: 1 μg/mL LPS, ODN1668 and IMQ; 20 μg/mL HMGB1, 25 μg/mL Poly I:C, 1 mM ATP, 10 μg/mL LTA and BD2. nd = not detected, nt = not tested.

Since we postulate that LC could respond to UV-induced KC DAMPs and in turn stimulate T cells or ILC, we next assessed pattern recognition receptor (PRR) expression by LC. When qRT-PCR was used to assess expression of a panel of PRRs in the LC cell line, XS106 (32, 33), we observed relatively high expression of TLR2, TLR9, and AHR; moderate expression of TLR3, TLR4, TLR5, purinergic receptor P2X7 (P2Rx7), and the receptor for advanced glycation end products (RAGE); and very low to undetectable expression of TLR7 (Fig. 6B). To determine if LC produce the ILC3/Th17-stimulating cytokines IL-23 or IL-6 in response to PRR signaling, we exposed the XS106 LC line to a panel of PRR ligands and assayed culture supernatants using Bio-Plex. Of PRR ligands tested, we found that TLR4 ligand lipopolysaccharide (LPS), TLR9 ligand ODN1668 (a class-B CpG oligonucleotide), and high mobility group box-1 (HMGB1), a ligand for multiple receptors including TLR2, 4, 5, 9, and RAGE, stimulated production of high levels of IL-23p19 (Fig. 6C), while significantly less IL-23p19 was produced in response to TLR3 ligand polyinosinic-polycytidylic acid (poly I:C), P2Rx7 ligand ATP, and TLR2 ligand lipoteichoic acid (LTA). XS106 cells also produced IL-6 in response to LPS and ODN1668. LC isolated from WT.B6 skin were also tested for responsiveness to ATP and HMGB1 (Fig. 6D) and produced significant levels of IL23p19 in response to HMGB1.

To directly test the effects of increased IL-23 in vivo, we exposed FVB.TCRβ^−/−δ^−/− mice to chronic UVB for 5 wk and then gave one cohort subcutaneous (s.c.) injections of recombinant IL-23, 3×/wk for an additional 2 wk, while the other group received vehicle (phosphate-buffered saline) alone. An analysis of p53 islands in epidermal sheets prepared from these mice revealed a 2.29-fold increase in the number of detectable p53 islands and a 2.33-fold increase in average p53 island area in the mice receiving IL-23 (SI Appendix, Fig. S12). Furthermore, icCD3+ cell density was increased 9.49-fold, while CD207+ LC density was only marginally increased (1.18-fold).

Human Sun-Damaged Skin and Premalignant AK Have Increased Expression of IL-17 and IL-22.

Having definitively shown that expression of the RORγt-related cytokines IL-17 and IL-22 are associated with chronic UVB–induced expansion of p53 mutant KC in mice and that inhibition of this pathway results in marked reduction in both the expression of these cytokines and in p53 island size, we next assessed the extent to which this pathway might be operative in human normal (non–sun-damaged) skin as well as skin biopsy lesions that developed under several states of chronic UV exposure: photodamaged skin, AK, and cSCC (Fig. 7). Relative to non–sun-damaged skin, both IL-22 and IL-17A gene expression were found to be markedly increased in sun-damaged skin and AK, both premalignant states of known mutant KC clonal expansion. Expression of IFN-γ was unchanged among all groups, while IL-13 was decreased in cSCC, and AREG was modestly increased in both AK and cSCC. These findings in human skin are further consistent with the overall hypothesis that RORγt-dependent IL-22–producing cells provide critical signals for the promotion of UV-induced carcinogenesis by stimulating mutant KC clonal expansion. Collectively, our findings further suggest the potential for translatable intervention for the prevention of UV-associated skin cancers, for example, via the topical application of inhibitors of an immunologic cell/signal pathway that includes RORγt and IL-22 among others.

Fig. 7.

IL-17 and IL-22 expression are increased in human sun-damaged skin and AK. RNA was isolated from FFPE tissue obtained from normal non–sun-damaged (NNS) or sun-damaged (SD) skin, AK, and squamous cell carcinomas (SCC) for analysis of gene expression (RQ to NNS). *P < 0.05.

Discussion

KC-derived skin cancers, predominantly basal cell carcinoma and cSCC, are collectively the most common malignancies worldwide (34), with estimates exceeding five million cases annually in the United States (35). UV radiation exposure from the sun is considered the major environmental inducer of KC DNA mutations, including within the tumor-suppressor gene p53, and chronic exposure is associated with cutaneous AK and cSCC formation. For years prior to this neoplastic outgrowth, hundreds to thousands of subclinical mutant KC clones present throughout areas of chronically UV–exposed skin comprise areas of “field cancerization” that represent an opportunity to prevent malignant outgrowth in photodamaged patients. Toward that goal, we have herein elucidated fundamental immune alterations induced by chronic UV exposure that are associated with keratinocyte clonal expansion and identified targets for potential translational preventive intervention.

A network of skin-resident immune cells—including LC, ILC, and T cells—are interspersed among epidermal KC, while multiple dendritic cell types and other ILC and T cell populations are adjacently situated in the dermis and subcutis. Key to the understanding of UV-induced immune pathogenesis relevance to cutaneous carcinogenesis are our recent findings that epidermal LC, previously implicated as primary protectors against skin cancer via their capacity to function as tumor antigen–presenting cells, function as major mediators of tumor promotion activities regardless of carcinogenesis induction (e.g., chemical and UV) protocols (16, 36, 37). Through coordinated flow cytometric and confocal immune cell imaging, isolation, and characterization in normal and immune mutant skin, we have now more fully clarified the immune pathway induced by chronic UV exposure that includes TLR-mediated damage recognition, activation of RORγt-differentiated lymphoid cells, and local production of IL-22 that are each observed in association with the expansion of p53+ keratinocyte clones.

The critical contributions of LC, RORγt+ cells, and IL-22 to UV-induced keratinocyte clonal expansion are each observed in the absence of αβ and γδ T cells, thereby implicating a major role for locally resident ILCs driving this phenomenon. Indeed, identified uvILC3 sit adjacent to keratinocyte clones, where they produce the potent keratinocyte growth factor IL-22 and which in turn has the capacity to further locally augment epidermal proliferation via positive feedback loops, for example, via antimicrobial peptide (AMP) and IL-20. Under normal homeostatic epidermal damage response, including to UV exposure, the IL-22 immune axis stimulates epidermal repair by adjacent proliferating KC that can replace defunct/apoptotic KC. However, chronic UV exposure may also give rise to apoptosis-resistant mutant KCs, and IL-22 (even at such physiologic levels) may chronically and locally accelerate keratinocyte clonal expansion. While we have not shown that larger p53+ keratinocyte clone size is correlated with increased mutational burden, others have shown that such clones contain p53 mutant cells (11–15); thus, we suggest that the larger the pool of these early mutant cells, the more such cells are available and/or susceptible to further mutations that ultimately drive transformation and tumor outgrowth.

Dermal IL-17– and IL-22–producing γδ T cells, also called γδ17 cells, are protective in the context of skin bacterial or fungal infections (38, 39) in which they promote neutrophil recruitment. In addition, IL-22 released by αβ T cells (40), dermal γδ T cells, or RORγt+ ILC3 (41) promotes epidermal repair via induction of antimicrobial peptides, promotion of KC proliferation, and inhibition of KC differentiation. The absence of γδ T cells results in spontaneous inflammation that can be controlled by anti-inflammatory Vγ5Vδ1+ DETC (42), and DETC also function in immune surveillance, limiting carcinogenesis (43) and locally producing KC growth factors that promote wound healing and KC homeostasis (44). Thus, T cell–deficient mice are critical to understanding the role of ILC in regulating local immune responses. Yet, studies utilizing T cell–deficient mice, including TCRβ^−/−δ^−/− double knockouts, are known to manifest exaggerated numbers and activities of ILC populations and thus may not be fully indicative of normal human cutaneous immune states. Such T cell–deficient states nonetheless do share immune deviations present in pharmacologically T cell–compromised organ transplant recipients (OTR) who are pharmacologically immune suppressed (e.g., by taking calcineurin inhibitors) and known to be at increased risk for development of cutaneous SCCs. Thus, understanding the role of ILC in photocarcinogenesis may have relevance to human KC-derived malignancy risk under such clinical settings of immune compromise.

Furthermore, there is increasing recognition that ILC, and other locally resident immune populations (e.g., epithelial γδ T cells), may exert greater local influence on adjacent cells during normal physiologic (i.e., typically lower intensity) chronic inflammatory states (45, 46). Indeed, we show chronic UV exposure provokes the emergence of a phenotypically distinct population of RORγt+ Sca1+ CD103+ ICOS+ CCR6+ icCD3+ uvILC that associates with p53+ KC islands and produces cytokines such as IL-22 that appear to play a critical role in promoting expansion of such mutant KC clones. While conventional αβ and recirculating γδ T cells require TCR-dependent antigen presentation for activation of effector function, ILC (and tissue-resident innate-like γδ T cells) respond directly to locally released factors to rapidly modulate surrounding epithelial and stromal elements. Nonetheless, our studies in T cell–intact mice support that procarcinogenic signals in the skin may also be provided by conventional T cells, specifically of the well-characterized Th17/22 subtype (47).

Our data also suggests that inhibition of RORγt+ uvILC has the potential to limit the immune contributions to UV-induced cutaneous carcinogenesis by decreasing KC clonal expansion. Even if specific pharmacologic targeting of RORγt+ cells and/or IL-22/IL-22Ra signaling were possible in an attempt to limit UV-induced skin cancer, however, the variety of RORγt+ subsets present in skin and the diversity of activities attributable to IL-22 may make such clinical intervention more challenging. Indeed, the inhibition of the IL-17A and IL-22 axis may adversely affect responses to bacterial (48, 49) and fungal (50) infections and may alter the cutaneous microbiome (e.g., by decreasing epidermal AMP levels) and decrease wound healing. Another subset of RORγt+ ILC3 is reported to regulate sebaceous gland function, which in turn also tunes the skin microbiome (51). However, these concerns are not necessarily prohibitive to treatment targeting these cells/pathways. For example, in mouse models of psoriasis, γδ17 and RORγt+ ILC3 account for the production of IL-17A and IL-22 (52). These cell populations are also increased in human psoriatic skin (53, 54). Yet, the systemic administration of therapeutic inhibitory antibodies, including those decreasing IL-17A– and IL-22–producing cells, in psoriasis is well tolerated, providing support for the safety of a local immune-based skin cancer prevention strategy. A remaining concern is the complex association of systemic inhibition of IL-22 with carcinogenesis (7), as it provides protection in some experimental models but promotes tumor development in others. Given that IL-22 is fundamental to the stimulation of keratinocyte production of antimicrobial peptides, it is possible that alterations in skin microbiota, that is, induced by chronic UV exposure, may also influence KC clonal expansion. Indeed, microbiota may produce bioactive metabolites that may substantially modulate host physiology (55). Furthermore, the pleiomorphic functions of IL-22 are further evident in its role to limit genotoxic effects within gut epithelium (56). Thus, whether the potential anticancer benefits, especially in high-risk clinical settings of chronic photodamage and/or suppression of T cell immunity as seen in OTR patients, of specific targeting of uvILC3, RORγt+ cells, and/or the IL-22 pathway outweigh the potential perturbations to normal cutaneous physiology would need to be studied clinically.

Materials and Methods

Animals.

All of the in vivo studies were approved by the Yale Institutional Animal Care and Use Committee. Mice were bred and maintained under specific pathogen-free conditions with food and water provided ad libitum. The Yale animal facility is accredited by the Association for Assessment of Laboratory Animal Care. Mouse strains utilized and corresponding stock numbers are provided in SI Appendix.

UVB Exposure.

Animals were exposed to a bank of four FS20T12 broadband UVB bulbs (National Biological Corp.) with emitted light filtered (Kodacel, Eastman Kodak Co.) to remove wavelengths <290 nm. Exposure was monitored using a calibrated meter (Intensity Meter 200, G&R Labs). For chronic exposures, mice were treated three times weekly on nonconsecutive days for 9 to 12 wk (as indicated) and initiated at 160 J/m² and increasing by 80 J/m² to reach a maximum of 400 J/m² for the duration of the experiment. Hair was removed from the dorsal body wall with an electric clipper prior as needed.

Preparation of Skin Cell Suspensions.

Epidermal and dermal cell suspensions were prepared as described in SI Appendix. For experiments assessing cytokine production, CD45+ cells were enriched using magnetic beads (Miltenyi) and restimulated for 6 h in vitro with 20 ng/mL PMA (Sigma) plus 0.5 μg/mL ionomycin (Sigma), 10 ng/mL IL-2 (BioLegend), 50 ng/mL IL-23 (BioLegend), and 2 μM Monensin (BD GolgiStop) in complete RPMI (36).

Flow Cytometry.

Cells were blocked, stained for surface antigens, fixed, and then stained for intracellular proteins. A table of antibodies used (SI Appendix, Table S1), methods, and the gating strategy for ILC identification (SI Appendix, Fig. S1) are provided.

Immunofluorescence and Confocal Microscopy.

Epidermal sheets were prepared and stained as previously described (16). A table of antibodies is provided (SI Appendix, Table S2) along with additional methods (SI Appendix, Supplemental Methods). Z-stacked images were captured via a Leica SP5 confocal microscope and fluorescence evaluated relative to isotype controls. P53 islands, CD207+ LC, and CD3ε+ cells were quantified as previously described (16) using Volocity 6.2 (PerkinElmer) and Fiji [ version 2.0.0-rc-59 (57)]. Individuals collecting and analyzing images were blinded to the experimental groups.

Gene Expression Analysis.

RNA was isolated and reverse transcribed and gene expression assessed using TaqMan reagents. Additional methods are provided in SI Appendix. The Ct values obtained were normalized to β-actin (mouse) or HPRT1 (human), and expression differences relative to control were calculated using RQ = 2^-ΔΔCt.

RORC Inhibition.

A 1% solution of GSK2981278 (BioVision) in 60% EtOH/40% H₂O (or vehicle alone) was applied topically (100 μL) to dorsal skin daily for the final 2 or 3 wk (as indicated) of UVB exposure plus 3 d after the final UVB exposure. Skin was harvested 4 d after the final UVB exposure. On days that mice were exposed to UVB, the inhibitor or vehicle was applied immediately after the exposure.

LC Response to TLR Ligands.

The LC cell line XS106 was obtained from A. Takashima, The University of Toledo, Toledo, OH, and cultured as previously described (36). LC were isolated from the ears of FVB/NJ mice using a “crawl-out” procedure (16). Cells emigrating from the epidermal sheet were purified using MHCII magnetic beads (Miltenyi). TLR ligands were titrated on XS106 cells and results using the optimal concentration are reported. A table of ligands and optimal concentrations is provided (SI Appendix, Table S3). XS106 cells or isolated LC (500,000/mL) were cultured with indicated TLR ligands for 48 h and then supernatants tested using Bio-Rad Bio-Plex assays for IL-23p19 and IL-6.

Statistics.

Differences in relative gene expression in Fig. 7 (human skin) were assessed using an unpaired one-tailed Student’s t test with Welch’s correction. All other experimental groups were compared using an unpaired one-tailed Student’s t test with significance established at P < 0.05. The Benjamini–Hochberg procedure with Q = 0.05 was used to correct for multiple comparisons of relative gene expression. All statistical analyses were performed using GraphPad Prism 8.2 software.

Data Availability

All study data are included in the article and/or SI Appendix.