Tracing the Equilibrium Phase of Cancer Immunoediting in Epidermal Neoplasms via Longitudinal, Intravital Imaging

Abstract

Recognition of transformed cells by the immune system can sometimes generate a rate-limiting “equilibrium phase,” wherein tumor outgrowth is prevented without complete neoplasm elimination. Targeting premalignancies during this immune-controlled bottleneck is a promising strategy for rational cancer prevention. Thus far, immune equilibrium has been difficult to model in a traceable way and most immunoediting systems have been limited to mesenchymal tumor types. Here, we introduce a mouse model for fluorescent tracing of somatic, epithelial transformation. We demonstrate that transplantation can be used to prevent a confounding, artificial tolerance that affects autochthonous, inducible models. Using this system, we observe the expected dichotomy of outcomes: immune-deficient contexts permit rapid tumorigenesis, while initiated clones in immunocompetent mice undergo elimination or equilibrium. Strikingly, the equilibrium phase correlates with localization within hair follicles, which have been previously characterized as relatively immune-privileged sites. Given this, we posit that valleys in the immune surveillance landscape of a normal tissue can provide a cell-extrinsic alternative to the canonical, cell-intrinsic adaptations believed to establish the equilibrium phase. Our model is a prototype for tracing immunoediting in vivo and could serve as a novel screening platform for therapies targeted against immune-controlled premalignancies.

INTRODUCTION

Cancers evolve immune-evasive character when they develop within the context of selective pressure from anti-tumor immunity(Shankaran et al., 2001). This phenomenon, termed “cancer immunoediting,” includes three phases: elimination, equilibrium, and escape(Schreiber et al., 2011). In the “elimination phase,” transformed cells are successfully recognized and eradicated. Since elimination is difficult to observe, it is inferred in most studies via measurement of increased cancer incidences in immunocompromised settings relative to wild type (Shankaran et al., 2001). Sporadic survivors of the elimination phase can enter the “equilibrium phase,” during which the immune system prevents outgrowth of transformed cells without destroying them. This state was demonstrated by murine chemical carcinogenesis experiments in which neoplasms remained latent for over 200 days, only emerging at the initiation of experimental immunosuppression(Koebel et al., 2007). Several studies have identified T cell recognition of tumor-derived neoantigens as the major mechanism of immune control that maintains the equilibrium phase(DuPage et al., 2011, Matsushita et al., 2012, Takeda et al., 2017). However, innate immunity may also be sufficient for an immunoediting response in the absence of adaptive immunity(O’Sullivan et al., 2012). In either case, the conditions or cellular adaptations that allow transformed cells to circumvent elimination and enter the equilibrium phase remain poorly understood and difficult to model. Moreover, since elimination and equilibrium both precede macroscopic outgrowth, termed “the escape phase,” it can be experimentally difficult to distinguish the two states. Finally, the escape phase is believed to require development of sub-clonal, immune-evasive adaptations, such as loss of antigen presentation(DuPage et al., 2012, Matsushita et al., 2012).

The equilibrium phase of immunoediting is interesting since it represents a rate-limiting intermediate in cancer development that might be targetable for cancer prevention. While the true burden of equilibrium-bound clones in the human population is unknown, immune-mediated dormancy appears to be common. Organ transplant recipients (OTRs), who receive immunosuppressive drugs to curb graft rejection, experience drastic increases in incidence across a variety of cancers, indicating a tumor suppressor role (Engels et al., 2011, Euvrard et al., 2003). There have also been cancers transferred from donors to recipients during organ transplantation. Though the donors were disease-free at transplant, malignant cells quickly disseminated from the donated organs in the recipients(Nalesnik et al., 2011). Again, the immunosuppressive drug regimens administered to the OTRs were thought to be responsible for recipient-specific outgrowth. For at least two cases, retrospective analyses determined that primary melanomas had been diagnosed and resected from the donors long prior to the transplant(MacKie et al., 2003, Tsao et al., 1997). This suggests that intact immunity in the donors was able to prevent melanoma progression at secondary sites for long durations, supporting the existence of an equilibrium phase in humans. Generally, the fact that immunosuppression of OTRs causes rapid increases in cancer incidence suggests that occult, immune equilibrium clones are present in healthy, immunocompetent adults(Yeh and Ramaswamy, 2015). We hypothesize that persistent, immune-selective pressure would serve to limit both the size and heterogeneity of equilibrium lesions, rendering them relatively sensitive to existing and emerging, targeted therapies(Stanta and Bonin, 2018). The challenge remains: how do we test the efficacy of such strategies when we cannot directly observe sensitivity of the target cells in vivo?

Modeling and tracing immunoediting in a way that faithfully recapitulates human cancer initiation, without disrupting the microenvironment, introducing artificial sources of immunogenicity, or producing unrealistically excessive mutational burdens, remains a major goal of the immunoediting field(DuPage and Jacks, 2013, Mittal et al., 2014,). Pioneering studies in this field made use of intramuscular mutagen injections, which are effective for initiating sarcomas, but generate high, initial, mutation loads and injection-associated inflammation in both target cells and surrounding stroma(Koebel et al., 2007, Matsushita et al., 2012, Teng et al., 2012). This methodology may obscure the circumstances that distinguish between elimination and equilibrium fates. Moreover, it biases toward neoantigens as the main sources of initial immunogenicity and neoantigen loss as the dominant mechanism for the transition from equilibrium to escape(Schumacher and Schreiber, 2015). Some studies have used genetically-engineered mouse models (GEMMs) to initiate discrete lesions with known driver mutations(DuPage et al., 2011, DuPage and Jacks, 2013, Evans et al., 2016). However, these studies required injections to deliver lentiviral Cre recombinase and co-expression of an exogenous foreign antigen to produce sufficient immunogenicity to establish an immune equilibrium(DuPage et al., 2012). Finally, cell line transplantation methods have also been employed to model immunoediting, for examples in melanoma (Park et al., 2019). These models are intriguing since the cell lines, by virtue of having been derived from primary tumors, have already undergone immunoediting in their original hosts. This leads us to hypothesize that any immune-restricted kinetics upon transplant are artificial relative to what would occur during somatic transformation. At minimum, interpretation of cancer-immune interactions in these transplanted mice is difficult. Overall, carcinoma models of immunoediting have been lacking, with most studies focused on sarcomas. This could reflect the difficulty of initiating immunogenic carcinomas or might suggest that elimination is a more common outcome in epithelia.

We set out to initiate and trace somatic carcinogenesis using specific driver events in only isolated, homeostatic target cells without perturbing any of the surrounding microenvironment. We describe the development of a transplantable GEMM model which allows intravital tracing of skin neoplasms in experimentally variable immune contexts. To our knowledge, this is the first inducible model that allows direct visualization of the entire immunoediting process in vivo.

RESULTS & DISCUSSION

Autochthonous models misrepresent tumor immunogenicity

The skin’s accessibility, immune relevance, and representative epithelial architecture make it an ideal tissue for longitudinal intravital imaging. Since we intended our model to act as an observable surrogate for transformation across a wide range of tissues, we chose to drive cutaneous squamous cell carcinoma (cSCC) formation using Cre-mediated recombination of Kras^LSL-G12D and p53^FL/FL alleles. These alleles have been employed by previous immunoediting studies(DuPage and Jacks, 2013, DuPage et al., 2012), allowing us to compare skin transformation to other tissues, including landmark muscle fibrosarcoma studies. Spatiotemporal recombination is achieved with a keratin-14 promoter-driven rtTA (k14rtTA) and a tet-on Cre (tetO-Cre) allele. An insulator/red/green (IRG) fluorescent reporter allele, which undergoes a Cre-dependent switch from dsRED to EGFP, allows visualization of transformed clones before and after induction (Figure 1a)(De Gasperi et al., 2008).

Figure 1. A non-autochthonous, transplanted GEMM to trace immunoediting in vivo.

(a) Schematic of the alleles used to trace and transform keratinocytes. (b) Cartoon outlining the mixed-donor transplant strategy. “kti-KPP” refers to the combination of all the alleles in A. (c) Images of transplant procedure. (d) 10x Confocal montage image of a healed graft 6 weeks post chamber removal (left) [scale bar = 2 mm] and single clone before (center) and after (right) induction [scale bar = 250 μm]. (e) 20x Confocal image of 2 induced clones surrounded by recipient-derived infiltrates (right) [scale bar = 100 μm] and an expanded view of boxed section (right). (f) Experimental timeline for longitudinal imaging.

During incrossing, we observed tumor formation without exposure to doxycycline inducer. Sporadic leakiness of the Cre allele caused all mice (n=35) whose genotypes included tetO-Cre, Kras^LSL-G12D, and p53^FL/FL (t-KPP) to develop palpable tumors, indicating leakage of the tetO-Cre allele. Tumor rates were the same in the immunodeficient Rag1^−/− (R-t-KPP) and Rag1^−/−/IL2Rγ^−/− (Rγ-t-KPP) backgrounds (Supplementary Figure S1). This was surprising since intact immunity would be expected to yield slower growth kinetics relative to immunocompromised controls. Previous studies have posited that this may be due to a lack of initial immunogenicity(DuPage et al., 2012, Evans et al., 2016). Alternatively, in autochthonous GEMM contexts, inducible oncogenic alleles may be artificially perceived as self. Since this would not be the case for somatically arising mutations, and because oncogenes are attractive targets for immunotherapy(Chu et al., 2015, Willimsky and Blankenstein, 2007), we wondered if loxP-stop-loxP (LSL) restricted alleles might be promiscuously expressed in vivo. In fact, Kras^G12D-derived RNA was detectable in all analyzed tissues, including the thymus, despite the lack of Cre recombinase (Supplementary Figure S2). This signal was reverse transcriptase-dependent and an LSL-EYFP reporter strain corroborated the result, suggesting a general mechanism of promiscuous LSL readthrough. Despite the presence of these transcripts, proteins of the correct sizes could not be detected by western blot and tissues from LSL-EYFP mice showed no fluorescence (Supplementary Figure S2). Western blots using a polyclonal anti-EYFP antibody produced two unexpected bands for the LSL-EYFP allele, suggesting that promiscuous expression generates peptide variants (Supplementary Figure S2). Crucial to immunoediting experiments, the presence of these peptides in the thymus could influence central tolerance, rendering autochthonously-expressed LSL-restricted alleles less immunogenic than corresponding somatic mutations. This read-through may help to explain the limited immunogenicity of autochthonous models and will be a crucial consideration for studies that attempt to assess the immunogenicity of individual oncogenes. These data provide the rationale for avoiding the use of autochthonous models for immunoediting since they may skew immune activities in unintended ways. Further study will be required to characterize promiscuous expression patterns and their effects on immune surveillance.

Using transplantation to model non-autochthonous, somatic transformation

To avoid any artificial tolerance, we employ a well-characterized skin transplant protocol(Lichti et al., 2008, Yuspa et al., 1970) to exclude inducible alleles from all tissues except a limited number of donor epidermal keratinocytes (Figure 1b). This strategy leaves the host immune system naïve to Kras^G12D and allows direct control over the number and spatial distribution of transformation-competent cells. The low initial burden avoids the possibility of overwhelming or tolerizing the immune system with widespread transformation. We isolate primary, neonatal K14rtTA/tetO-Cre/IRG/Kras^LSL-G12D/p53^FL/FL (kti-KPP) keratinocytes and dilute them with wildtype C57BL/6J keratinocytes and fibroblasts at a 1 to 5 ratio. The resulting primary cell slurry is applied to a temporarily-implanted, silicone chamber in the recipient’s dorsal skin (Figure 1c). The implant allows the cells to settle on the underlying muscle fascia, where they self-organize to generate dermal and epidermal compartments. This reconstitutes fully differentiated skin, complete with hair follicles, sebaceous glands and vasculature. After 7 days, the chamber is removed, allowing the borders of the grafted skin patch to fuse with the surrounding host skin. At 6-weeks after chamber removal, recipient and donor tissues have fully-integrated and grafts contain numerous recipient-derived infiltrates, including dendritic cells and DETCs expressing γδ-TCR. Histology of the graft-recipient boundary indicates normal dermal and epidermal morphology and continuity across the margin of the graft (Supplementary Figure S3).

Doxycycline(dox)-pretreated k14rtTA/tetO-Cre/IRG (kti) mice are used as immunocompetent recipients. Carrying each non-tumorigenic transgene in the recipient ensures that the activator/reporter alleles cannot serve as possible sources of immunogenicity. All lines were backcrossed to the C57BL/6J background and validated via microsatellite sequencing to confirm that the only differences between donor and recipient are at the Kras and P53 loci (Supplementary Figure S4). The dox pre-treatment switches epidermal fluorescence to EGFP in the kti recipient epidermis, allowing discernment of transplanted dsRed⁺ kti-KPP donor keratinocytes at the outset of experiments (Figure 1d). Rag1^−/− (R) mice lacking T and B cells, or Rag1^−/−/IL2Rγ^−/− (Rγ) lacking T, B and NK cells, were used as immunocompromised control recipients.

Longitudinal imaging of neoplasms undergoing immunoediting

Post-grafted skin is allowed 6 weeks for healing and infiltration of the grafts by recipient-derived cells (Figure 1e). To prepare for imaging, mice are anesthetized with isoflurane gas and skin surrounding the grafted area is immobilized over a cover slip to isolate movement from respiration (Supplementary Figure S5). In kti recipient animals, distinct graft boundaries are easily identified by recipient EGFP expression (Figure 1d). Individual dsRED-expressing clones can be imaged within the graft at single-cell resolution. Since epidermal turnover will have occurred repeatedly, fluorescence at 6 weeks indicates that a grafted kti-KPP keratinocyte stem cell produced a homeostatic clonal unit(Gonzales and Fuchs, 2017, Potten et al., 1987, Strachan and Ghadially, 2008). After 8 weeks, the clones are induced with dietary doxycycline and imaged longitudinally to measure EGFP⁺ clone growth (Figure 1f). Clones are manually re-identified across time points via relative positions and morphologies. Because grafting allows for de novo appendage formation, inducible clones can reside in all possible epithelial niches, including the hair follicle.

Epidermal Neoplasms Exhibit Immunoediting’s Three Phases

Immune competence was inversely correlated with tumor incidence. In immunocompromised mice, tumors transitioned from hyperproliferative neoplasia to papillomas, then to large, ulcerated cSCCs (Figure 2a,b). To date, all induced clones have grown rapidly in Rγ mice and developed palpable masses at an average of 94.5 days (Figure 2c,d). EGFP⁺ clones in kti mice do not undergo this immediate expansion, developing tumors in only 25% of animals after long latencies, averaging 273 days (Figure 2c–e). The R background yields intermediate tumor incidence rates, statistically resembling the latency of kti mice more than that of Rγ mice (Figure 2c–e). Like previous studies(O’Sullivan et al., 2012), these data suggest that, in the absence of T and B cells, NK cells can mediate immune equilibrium. Plotting the relative percentages of immunoediting phases across the three backgrounds on a per-graft basis highlights that full elimination is unique to the kti background and that the likelihood of escape correlates with both immune status and time post-induction (Supplementary Figure S6). Although these data demonstrate that in the absence of T and B cells, NK cells can mediate immune equilibrium, they do not eliminate a role for T cells in immunoediting in an intact immune system.

Figure 2. Intact immunity yields an immunoediting response in the epidermis.

(a) Macroscopic images of tumor development stages in Rγ mice with EGFP fluorescence [scale bars = 3 mm]. (b) Histology images of a pre-palpable lesion [scale bar = 100 μm], a papilloma [scale bar = 1 mm], and an SCC [scale bar = 1 mm]. EGFP immunohistochemistry is shown for the pre-palpable lesion [scale bar = 100 mm] and expanded views of boxed regions are shown for the papilloma [scale bar = 250 μm] and SCC [scale bar = 250 μm]. (c) Percent of mice with palpable tumors for each graft recipient genotype (Rγ vs. R: P < 0.0001, R vs. kti: P = 0.0569). (d) Mean times to palpability (P values: * = 0.0001, ** = 0.0017). (e) Longitudinal quantification of EGFP⁺ area normalized to the initial dsRED⁺ area at D0 for individual grafts on different recipient genotypes.

We define the equilibrium phase as any non-palpable clone that remains after 200 days. This aligns with previous studies(Koebel et al., 2007) and is approximately equal to the greatest observed latency in the Rγ cohort (205 days), which is 2.4 standard deviations from the mean. Mechanistically, equilibrium begins as soon as a clone gains immunogenic character and can be recognized by immune effectors. This immunogenicity appears to coincide with initial transformation, since immediate clone growth occurs in Rγ recipients but is prevented in most immunocompetent kti and R mice (Figure 3). We have chosen a conservative definition for the equilibrium phase both to rule out the possibility of a false phase classification and to confirm the long-term nature of the equilibrium phase. Unlike the tumors in Rγ mice, equilibrium bound clones in both kti and R backgrounds are not macroscopically evident in the grafts (Figure 3h,i) but can still be observed via fluorescent imaging (Figure 3a–c).

Figure 3. Longitudinal intravital imaging reveals the growth dynamics of transformed clones growing under different immune contexts.

(a–c) Confocal time courses for a single mouse of each recipient genotype are shown. Each image is a 7×7 confocal montage [scale bars = 1 mm]. (d–f) Quantification of EGFP fold changes for all individually-tracked mice of each recipient genotype. Fold change is calculated by dividing the GFP⁺ area at a given time point by the initial dsRED⁺ area at D0 (see Supplementary Figure S10 for area measurement methods). The colored trace in each graph identifies the animal shown in the imaging time courses from A–C. (g–i) Macroscopic images of 3 different grafts from each recipient genotype. All images were taken at least 100 days after induction.

Longitudinal EGFP area measurements of single clones showed that, while R and kti equilibrium-bound clones fluctuated morphologically over time, total EGFP fluorescent area remained relatively stable compared to Rγ clones (Figure 4a–c). These observations favor a dynamic equilibrium or mass dormancy model rather than cellular quiescence(Yeh and Ramaswamy, 2015). This was not surprising given that the murine epidermis is replaced every 9 to 10 days(Potten et al., 1987) and proliferation should be required to maintain clones as older cells slough away. Both elimination (EGFP⁺ clonal loss) and equilibrium (tumor-free maintenance of EGFP+ clones) could be observed via longitudinal confocal imaging in kti and R backgrounds (Figure 4d and 4e). Transformed clone loss was never observed in the Rγ background. Conversely, EGFP⁺ clones were frequently lost in kti and R mice following induction, sometimes resulting in total clearance.

Figure 4. Individual transformed clones undergo dynamic immune equilibrium in the presence of intact immunity.

(a) Confocal tracing of single-clone growth dynamics as a function of immune status [scale bars = 250 μm]. (b) Average clone growth for all clones in individual mice. (c) Clone growth of the clones shown in A. (d) Clone losses and gains occurring in the immunocompetent kti recipient context across 3 time points. Primary images are 7×7 10x montages [scale bars = 1 mm]. Insets are 20x single images [scale bars = 250 μm]. (e) Quantification of the clones from D. Numbering corresponds to insets.

These data indicate that Kras^G12D/p53^−/− transformation is sufficient for epidermal tumorigenesis in an immune-deficient setting, while also immunogenic enough to yield both elimination and establishment of a long-term dynamic equilibrium in immunocompetent mice. Overall, this supports the hypothesis that the rapid outgrowth of tumors in immune-suppressed contexts, such as OTRs, originates from already transformed cells which had previously been under immune control.

Immune Equilibrium is Correlated with Hair Follicle Localization

The hair follicle is a site of relative immune privilege(Ito et al., 2008, Paus et al., 2005). The hair follicle itself is a complex mini-organ, including both epithelial keratinocytes and mesenchymal fibroblasts. Both cell types collaborate to restrict immune cell access and/or activation through maintenance of a tight sheath of ECM, expression of negative chemotactic receptors, and reduced antigen presentation (Paus et al., 2005). These phenotypes render the proximal 2/3 of the anagen hair follicle relatively exempt from immune surveillance. We hypothesized that this immune-privileged domain might facilitate the transit of transformed, resident cells through immunoediting.

EGFP⁺ lesions that persist beyond the initial elimination phase exhibit or contain circular morphology, indicative of a hair follicle opening at the skin surface (Figure 4a). Retrospective analysis of eliminated clones revealed a lack of hair follicle involvement, suggesting that intrafollicular clones are relatively susceptible to elimination (Supplementary Figure S7). We also noted many instances of lesions emerging later in an imaging time course, suggesting that they had been residing beyond our imaging capability, which degrades with depth (Figure 4d). To confirm hair follicle involvement, we subjected equilibrium phase biopsies to whole-mount imaging (Figure 5). This confirmed that transformed lesions either reside in, or originate from, a hair follicle. EGFP cells were not limited to any specific sub-compartment(s) of the follicle. However, hair follicle morphology was often influenced by transformation, particularly in sebaceous glands, which often expanded drastically (Figure 5a).

Figure 5. Equilibrium phase lesions localize to hair follicles.

(a) Confocal image of a 100μm section of a graft containing multiple transformed clones [scale bar = 1000 μm]. Numbered boxes in top image are enlarged below [scale bar = 250 μm]. (b) Confocal image of a transformed hair follicle with normal morphology [scale bar = 100 μm]. (c) Confocal image of a hair follicle containing a transformed clone that is restricted to the sebaceous gland [scale bar = 100 μm]. (d) Optically cleared whole-mount biopsy of a surface clone and underlying hair follicle [scale bar = 250 μm].

Epidermal equilibrium phase clones are surrounded by recipient cellular infiltrates

The constitutive dsRED expression of k14-negative cells in dox-exposed kti recipients was used to track the infiltration of recipient immune cells during longitudinal imaging (Figure 1e). DsRed⁺ cells near the skin surface exhibited dendritic morphology. Deeper infiltrates were densely distributed, surrounding transformed clones (Figure 6). Across grafts, infiltrates preferentially migrated to the space surrounding EGFP⁺ lesions, with the greatest dsRED densities just outside EGFP^high regions (Supplementary Figure S8).

Figure 6. Transformed clones are surrounded by recipient-derived infiltrates.

(a) Confocal z-stack showing an EGFP+ kti-KPP clone and dsRED+ infiltrates from a kti recipient [scale bar = 100 μm]. (b) Zoomed view of boxed region from the 2^nd z plane [scale bar = 50 μm]. (c) Zoomed view of boxed region from the 5^th z plane [scale bar = 50 μm].

To identify immune infiltrates, mice were euthanized and grafts containing equilibrium phase clones were harvested for immunostaining. We could easily identify Langerhans cells and DETCs among the superficial, dendritic infiltrates (Supplementary Figure S9). F4/80⁺ macrophages, which closely associated with transformed clones, comprised the remainder of CD45⁺ infiltrates. NK cells can inhibit tumor formation in Rag2^−/− mice by inducing M1 macrophages through IFNγ(O’Sullivan et al., 2012). We found that macrophages were tightly associated with equilibrium phase lesions but were mostly localized to the exterior of each clone (Supplementary Figure S9). Interestingly, we observed that some macrophages contained intracellular EGFP⁺ compartments (Supplementary Figure S9), possibly resulting from phagocytic activity. We were only able to detect rare NK cells in grafts (Supplementary Figure S9). This, combined with the fact that macrophage involvement was seen even in Rγ settings, suggests that F4/80⁺ cells did not require T, B, or NK cells for their migration to transformed clones. However, rapid tumor growth in the Rγ background would suggest that macrophage activity is not sufficient, on its own, to restrict tumor formation.

Niche-defined immune equilibrium could provide new therapeutic opportunities

The role of the normal tissue microenvironment during cancer initiation and the early immune response is rarely captured by current modeling approaches. Our transplant method is particularly attractive because it initiates transformation in cells that are properly integrated into a homeostatic tissue microenvironment. Additionally, we demonstrated that autochthonous models may mis-represent the T cell repertoire by generating artificial tolerance to conditional alleles. Surprisingly, we found that the artificial tolerizing effect of autochthonous models may extend beyond T cell antigenicity, since tumors grown in the Rag knockout background were edited nearly as well as the fully immunocompetent kti mice. Further studies will be required to understand the mechanisms driving this innate immunoediting of early transformation.

Our system introduces only two defined oncogenic events. However, since neoantigen load has been correlated with immunoediting in past studies, we were uncertain as to whether our scheme would be immunogenic enough to initiate an immune response. In fact, we did find that minimal transformation consistently generated a robust immunoediting response. Additionally, immunoediting proceeded in the absence of T cells, suggesting that NK cells or other innate lymphoid cells, may do the heavy lifting of early immunoediting preceding the antigenicity required for a T cell response. Together, these data lead us to question the dominance of neoantigens for immunoediting of de novo transformation and suggest a greater role for innate immune mechanisms in targeting early lesions.

The transplant model allowed us to identify a possible role for variations in the niche in promoting the long-term survival of transformed clones. We observed that hair follicle residence seems to correlate with equilibrium phase rather than canonical cell-intrinsic adaptations of the transformed cells. This supports previous assertions that hair follicles are relatively immune-privileged sites in the skin. It is not well understood how NK cells or ILCs may be regulated in the hair follicle, as most previous work has focused on adaptive immune mechanisms.

The therapeutic potential of the immune system in treating cancer is unquestionable considering the successes of immunotherapies like CAR T cells and checkpoint inhibitors. However, we posit that such treatments could be even more effective if applied prophylactically. The existence of a niche-maintained equilibrium phase suggests that a unique therapeutic window may exist wherein neoplasms are sensitive to immune clearance. Unlike most cancers targeted by immunotherapies, these neoplasms may not yet have evolved immune-evasive traits because the niche has sequestered them from immune selective pressures present elsewhere in the epithelium. Moreover, they may be susceptible to innate immune mechanisms that have not yet been the subject of therapeutic modulation.

Intriguingly, this class of neoplasm has never been directly assayed for sensitivity to targeted agents since it has remained refractory to modeling efforts. In this study, we introduced a prototype for modeling immunoediting using non-autochthonous GEMMs. This model provides a prototype screening system to directly measure effects of targeted therapies or other cancer-preventative measures on a well-defined class of neoplastic disease.

MATERIALS & METHODS

Skin transplants and induction

Donor keratinocytes (1×10⁶) and fibroblasts (1×10⁶) were harvested from newborn (P2–P4) pups and dissociated via floating overnight on 0.25% Trypsin as described previously (Yuspa et al., 1970). Once dissociated, immune cells were removed by magnetic separation for CD45 (Miltenyi mouse α-CD45 MicroBeads). The resulting keratinocyte/fibroblast cell slurry (in DMEM with 10% chelexed FBS) was grafted onto recipient animals. Grafting was conducted according to previously described protocols(Lichti et al., 2008, Yuspa et al., 1970) using a custom-made silicone chamber (inner diameter: 8mm) that was implanted in the dorsal skin between the shoulder blades. Isolated, inducible keratinocytes (kti-KPP) were diluted with C57BL/6J keratinocytes in a 1:5 ratio. Fibroblasts were entirely C57BL/6J-derived. Graft chambers were removed 8–10 days after placement and the grafted skin was covered with a breathable Tegaderm bandage (3M) which was allowed to fall off naturally during the next several days. Animals were given Ibuprofen (0.2mg/mL) in their drinking water at the time of grafting until fully healed. Grafted mice were monitored daily during healing (~2 weeks).

Induction was achieved via feeding the mice dietary doxycycline for 21 days (6gm/kg Doxycycline in grain-based pellets, BioServ cat# S4096–1). Induction of the IRG allele was evident by confocal imaging as early as 48 hours. After 3 weeks, the mice were placed back on normal diet.

In vivo confocal imaging and tumor tracking

Mice were anesthetized with vaporized isoflurane. The skin was depilated via waxing or shaving, cleaned, and immobilized over a coverslip on a custom-made microscope stage (Supplementary Figure S5) attached to a Nikon C2 laser-scanning microscope. Filter sets for 488nm and 594nm fluorescence were used to image the EGFP and dsRED proteins expressed by the IRG allele. A 408nm channel was sometimes used to image skin and hair autofluorescence, which was useful in determining the graft border in non-fluorescent recipient strains. The entire grafted area was scanned as a montaged z-stack at 10x and stitched together using Nikon NIS software. Z-stacks spanned 60μm from the skin surface and included 5 planes spaced 15 μm apart. Areas of interest were re-scanned at higher magnification and resolution. After imaging, mice were allowed to recover from anesthesia and rehoused until the next imaging session. Mice were checked daily for development of palpable tumors. After becoming palpable, tumors were imaged using a Leica M165 FC dissecting fluorescent microscope.

DATA AVAILABILITY STATEMENT

There are no datasets for this manuscript.

Abbreviations:

GEMM

genetically-engineered mouse model

DOX

doxycycline

LSL

LoxP-Stop-LoxP

LC

Langerhans Cell

DETC

dendritic epithelial T cell

OTR

Organ transplant recipient

cSCC

cutaneous squamous cell carcinoma