A novel humanized mouse model to study human antigen-specific cutaneous T cell responses in vivo .

Human skin contains a significant number of T cells that support tissue homeostasis and provide protective immunity. T cell function in human skin is difficult to study due to a lack of adequate in vivo models. In this study we used immunodeficient NOD- scid IL2rγ null (NSG) mice that carried in vivo- generated engineered skin (ES) and received adoptively transferred human peripheral blood mononuclear cells. ES were generated from keratinocytes and fibroblasts only and initially contained no skin-resident immune cells. This reductionist system allowed us to study T cell recruitment and function in non-inflamed and non-infected human skin. We found that functional human T cells specifically infiltrated the human skin tissue and responded to microbial antigen in vivo . Importantly, T cell maintenance and function was supported by the microenvironment of human skin. We have thus generated a novel mouse model with broad 2 utility in studies of human cutaneous antigen-specific T cell responses and the role of the skin microenvironment to skin immunity in vivo . Statistical significance was calculated with Prism 7.0 software (GraphPad) by one- way ANOVA with Tukey’s multiple comparisons test, or by un -paired student’s t-test as indicated. Error bars indicate mean standard deviation.

utility in studies of human cutaneous antigen-specific T cell responses and the role of the skin microenvironment to skin immunity in vivo.

Introduction
As the body's outermost barrier, the skin represents a unique and complex immunological organ. As such, healthy human skin contains around twice the number of T cells found in the blood of which the majority are memory T cells (Bos et al. 1989;Clark et al. 2006) that support tissue homeostasis and ensure adequate response to pathogens (Di Meglio et al. 2011;Klicznik et al. 2018;Nestle et al. 2009). Although significant advances in understanding the role of the skin microenvironment on T cell function and memory development in murine skin have been made (Mackay et al. 2015;Mackay et al. 2013;Pan et al. 2017), the specific contribution of keratinocyte-and fibroblast-derived signals to cutaneous immunity in human skin remain poorly understood. T cell responses are strongly influenced by the surrounding tissue (Hu and Pasare 2013;McCully et al. 2012), and T cells at different barrier sites show distinct functional and metabolic properties (Kumar et al. 2017;Pan et al. 2017), therefore it is crucial to study cutaneous immunity within its physiological compartment in vivo. Due to fundamental structural differences between human and murine skin, as well as a lack of direct correspondence between human and murine immune cell populations (Di Meglio et al. 2011;Gudjonsson et al. 2007;Perlman 2016;Shay et al. 2013), direct translation from the murine cutaneous immune system can be difficult. This emphasizes the need for mouse models that faithfully replicate conditions found in human skin. Humanized mice, in which immunodeficient mice are adoptively transferred with human peripheral blood mononuclear cells (PBMC) and transplanted with human full thickness skin derived from either healthy donors or patients with skin diseases are currently used to study human skin immunology in vivo (Boyman et al. 2004;King et al. 2009;Watanabe et al. 2015). In these models the rejection of skin allografts and xenogenic graft versus host disease (GvHD) development can be readily studied (Racki et al. 2010), but antigen-specific activation of T cells has been much harder to follow. Additionally, if obtained from adult donors full-thickness skin grafts contain resident immune cells (Clark et al. 2006;Racki et al. 2010;Sanchez Rodriguez et al. 2014;Watanabe et al. 2015), making it difficult to functionally analyze and manipulate discrete skin-tropic T cell populations.
We hypothesized that humanized mice reconstituted with simplified human skin consisting only of keratinocytes and fibroblasts would provide a reductionist system allowing us to study human T cell recruitment and function in human skin in absence of acute inflammation. To this end, we used NOD-scid IL2rγ null (NSG) mice that were adoptively transferred with human PBMC and carried in vivo-generated engineered skin (ES). This model serves as a novel platform for studying human cutaneous antigen-specific T cell responses, and the contribution of the skin microenvironment to skin immunity in vivo. Additionally, we provide a method to stimulate robust antigen-specific memory responses of cutaneous T cells in this system.

Human T cells specifically infiltrate human engineered skin in a xenograft mouse model.
To follow human skin infiltration by autologous human immune cells, we generated engineered skin (ES) from human keratinocytes and fibroblasts isolated from healthy human skin and immortalized (Merkley et al. 2009). ES was generated as described before (Wang et al. 2000) and allowed to heal and differentiate for a minimum of 30 days (Fig.1a). H&E staining revealed that the ES displayed histological features of human and not murine skin.
Additionally, organotypic proteins such as human type VII collagen at the basement membrane of the ES and human cytokeratin 5/6 within the correct epidermal layers of the ES confirmed correct differentiation and human origin (Fig.1b). In parallel, human PBMC from the skin donor were isolated and stored in liquid nitrogen until use (Fig.1a). After complete wound healing of the ES PBMC were adoptively transferred, thus creating a mouse model designated huPBMC-ES-NSG.
Engraftment of human CD45 + cells was detectable after 14 days in the spleen and after 21 days in the ES (Fig. 1c-d). After a period of 18 -34 days mean levels of human CD45 + cells in spleen and ES were at >18% (Fig. 1e). The majority of human cells (>94%) in spleen and ES were CD3 + T cells (Fig. 1f) and human CD3 + cells preferentially infiltrated human ES not the adjacent murine skin (Fig. 1g). CD4 + and CD8 + as well as TCRγδ + T cells engrafted within the spleen and ES at levels comparable to human PBMC and skin ( Fig. 1h-i). The ratios of CD4 + and CD8 + T cells engrafting in spleen and ES reflected the physiological levels found in human PBMC and skin, respectively (Table in Fig. 1j).
In previous studies development of xenogeneic graft versus host disease (xeno-GvHD) occurred around 5 weeks after adoptive transfer of 10 7 human PBMC into NSG mice (Ali et al. 2012;King et al. 2008). To delay the development of GvHD we reduced cell numbers to 1.8 -3x10 6 /mouse. Additionally, we limited all experiments to 35 days, which was before onset of GvHD thus avoiding convoluting effects on our studies.
Taken together, these data illustrate that the human T cell compartment can be reconstituted in spleen and ES of the huPBMC-ES-NSG mouse. Next, we sought to determine whether the model was suitable to study human T cell function within human skin in vivo.

Cutaneous and splenic T cells from huPBMC-ES-NSG maintain the functional profile of T cells found in human skin and blood
We assessed the function of splenic and skin T cells following ex vivo stimulation and intracellular cytokine staining. Production of the Th2, Th17 and Th22 cytokines IL-13, IL-17 and IL-22, respectively, were preserved in CD4 + T cells isolated from the huPBMC-ES-NSG mouse when compared to T cells from human blood and skin( Fig. 2a-b; e-g). By contrast, . CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted December 7, 2018. . https://doi.org/10.1101/490060 doi: bioRxiv preprint increased percentages of CD4 + T cells isolated from the spleen and ES produced GM-CSF (Fig. 2c;. Interestingly while IFNγ + CD4 + cells were increased in the spleen when compared to PBMC, the proportion of IFNγ producing CD4 + cells within the ES was comparable to skin from healthy donors (Fig. 2i). In addition to conventional T cells, CD4 + CD25 + Foxp3 + regulatory T cells engrafted within the spleen and ES ( Fig. 2d; j).
Analogous to the cytokine profiles of splenic and cutaneous CD4 + T cells, we assessed the cytokine secretion of CD8 + T cells isolated from spleen and ES (Fig. 3). The cytokine profiles of CD8 + T cells in ES and spleen were comparable to healthy human skin and PBMC with the exception of GM-CSF, which was increased within the ES, similar to the CD4 + T cell population.

Engrafted T cells share a skin-homing memory-like phenotype
Since CD4 + T cells represent the majority of skin-homing and -resident T cells (Clark et al. 2006;Watanabe et al. 2015), we focused on the function of cutaneous CD4 + T cells.
Confirming previous studies of PBMC engraftment in NSG mice, we found that human CD4 + T cells isolated from mice did not express markers of naïve T cells such as CCR7 and CD45RA despite being present in the ingoing PBMC population (Ali et al. 2012) (Fig. 4a).
Additionally, around half of them expressed cutaneous leukocyte antigen (CLA), a glycan moiety that promotes skin-homing (Clark et al. 2006) (Fig. 4b). Taken together these data indicate that engrafted CD4 + T cells show a skin-tropic, memory-like phenotype.

Cutaneous CD4 + T cells are locally activated by microbial antigen
Skin CD4 + T cells play a crucial role in controlling cutaneous microbes (Belkaid and Tamoutounour 2016). Their specific role in responses against the commensal fungus Candida albicans (Lagunes and Rello 2016) is underlined by the fact that primary and acquired immunodeficiencies that lead to the impairment of CD4 + T cell immunity can cause . CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted December 7, 2018. . https://doi.org/10.1101/490060 doi: bioRxiv preprint pathogenic C.albicans infections (Gosselin et al. 2010;Klein et al. 1984;Lagunes and Rello 2016;Ling et al. 2015;Puel et al. 2011). Consistent with that, the human circulating T cell pool contains C.albicans-specific memory T cells (Acosta-Rodriguez et al. 2007;Hernández-Santos and Gaffen 2012).
We hypothesized that memory T cells specific for C.albicans, engraft in the ES and can mount a local antigen-specific T cell response upon encounter of microbial antigen. Therefore, we applied heat killed C.albicans (HKCA) to the ES in vivo. Injection of free HKCA failed to elicit a detectable expansion or proliferation in the spleen or ES of treated mice (Supp. Fig.1a . CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted December 7, 2018. . https://doi.org/10.1101/490060 doi: bioRxiv preprint Together these data show that the huPBMC-ES-NSG model provides a tool to monitor antigen-specific T cell responses within human skin in vivo.

Antigen-specific T cell responses remain detectable in donor-mismatched skin tissue
So far, we used a completely matched system where ES, PBMC and moDC were from the same donor. However, access to skin that is matched to the PBMC, presents a limiting factor in studies of human immune responses. To broaden the model's applicability, we sought to determine whether antigen-specific T cell responses could still be detected in the skin when using donor-mismatched tissues. Therefore, we compared cutaneous CD4 + T cell responses to HKCA in donor-matched and -mismatched ES. ES were generated from two different donors Together, these data suggest that T cells were not activated by allogeneic keratinocytes or . CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted December 7, 2018. . https://doi.org/10.1101/490060 doi: bioRxiv preprint fibroblasts and that mismatched tissues still allow the study human antigen-specific cutaneous CD4 + T cell responses in the huPBMC-ES-NSG model.

Human skin tissue promotes T cell maintenance and function
Based on the finding that human T cells preferentially infiltrated the human over the murine skin and T cells within the ES closely resembled T cells found in human skin ( Fig. 1-3), we hypothesized, that human skin and ES provide similar signals to infiltrating T cells.
Interestingly skin-derived chemokines, such as CCL2, CCL5, CXCL10 and CXCL12 (Fig. 6 a-d) and cytokines involved in T cell survival and maintenance, such as IL7 and IL15 were detectable at comparable levels in ES and normal human skin ( Fig. 6 e,f). Based on these data, we hypothesized that the antigen-specific T cell response we detected was dependent on skin tissue-derived signals. To test this, we injected either the ES of huPBMC-ES-NSG or a defined area of murine skin on the back of huPBMC-NSG mice with autologous HKCA/moDC or LPS/moDC (Fig. 6g). Compared to ES treated with HKCA/moDC, significantly lower numbers of CD3 + cells infiltrated the murine skin. Importantly their frequency remained unaltered upon injection of HKCA/moDC, indicating a lack of antigenspecific activation (Fig. 6h). This suggests that the C.albicans specific T cell response detected in the huPBMC-ES-NSG model is not only a memory re-call response that is elicited by T cell-APC interaction, but requires tissue derived signals to support full memory T cell function.

Discussion
The majority of T cells in healthy human skin are CD45RO memory T cells that ensure adequate protective immunity on this peripheral barrier site (Bos et al. 1989;Klicznik et al. 2018;Watanabe et al. 2015). Most of our understanding of the development and function of protective T cell immunity in the skin is based on research using animal models with limited . CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted December 7, 2018. . https://doi.org/10.1101/490060 doi: bioRxiv preprint direct translation to humans (Di Meglio et al. 2011;Gudjonsson et al. 2007;Schreiner and King 2018). Therefore skin-humanized mouse models present a powerful platform to study cutaneous immune processes in vivo. Existing models in which PBMC and skin are grafted have been most useful for studying inflammatory processes of the skin (Boyman et al. 2004;Issa et al. 2010;Racki et al. 2010;Watanabe et al. 2015) but studies of human cutaneous immunity in steady-state in vivo have been difficult to realize. Additionally, healthy human skin contains a variety of adaptive and innate immune cells that impact the outcome of any intervention and study (Di Meglio et al. 2011;Klicznik et al. 2018;Pasparakis et al. 2014).
Studying distinct immune processes in the skin requires simple traceable models, in which specific components can be independently manipulated. In that respect, neonatal foreskin grafts that contain 45-fold less T cells than adult human skin have been used to reduce the impact of resident cells (Watanabe et al. 2015), but even this tissue already contains T cells (Schuster et al. 2012). Here we combined engineered skin (ES) (Wang et al. 2000) devoid of any resident leukocytes, and adoptive transfer of human PBMC into immunodeficient NSG mice to generate a humanized mouse model (huPBMC-ES-NSG) that allows for specific study of individual immune cell populations, as well as the impact of tissue-derived signals on immunological processes in the skin.
Importantly, the use of ES permits precise control over the cell populations that partake in a specific immune response and targeted manipulation of keratinocytes and fibroblasts using CRISPR/Cas9 will facilitate to study the influence of selected pathways and the effects of these alterations on cutaneous immunity in vivo.
Poor engraftment of APC can limit antigen-specific T cell responses in vivo (King et al. 2009), which can be partly overcome by injection of DC loaded with antigen (Harui et al. 2011). However, we have found that antigen-specific recall responses require the support of the micromilieu, such as tissue-derived IL15 and IL7 (Belarif et al. 2018;Wang et al. 2011) and CD4 + T cells failed to respond to antigen presented in murine skin. Importantly, we found . CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted December 7, 2018. . https://doi.org/10.1101/490060 doi: bioRxiv preprint that healthy human skin-and ES-derived signals were remarkably similar, especially chemokines and cytokines involved in T cell recruitment and activation, such as CCL2 (Carr et al. 1994), CCL5 (Kawai et al. 1999), CXCL10 (Fukui et al. 2013), CXCL12 (Nanki and Lipsky 2001) and T cell function and maintenance, like IL7 (Adachi et al. 2015;Belarif et al. 2018) and IL15 (Adachi et al. 2015;Wang et al. 2011). This indicates that the huPBMC-NSG-ES model enables us to study the contribution of tissue-derived cues on cutaneous T cell maintenance and function.
Interestingly, the antigen-specific response was detectable in both, an autologous and an allogeneic setting. Thus, access to matched tissue samples is not limiting the study of cutaneous T cell responses in these humanized mice.
By engineering human skin tissue, we created an environment that allows the study of cutaneous T cell responses in absence of inflammation or infection in vivo. Additionally, the model may serve as a platform to test novel therapeutic interventions to treat cutaneous inflammation, tumors or autoimmune diseases. Taken together, the huPBMC-ES-NSG model provides a highly versatile tool to study cutaneous T cell responses and study and manipulate tissue-derived signals that impact skin immunity.

Material and Methods
Mice. Animal studies were approved by the Austrian Federal Ministry of Science, Research and Economy. NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice were obtained from The Jackson Laboratory and bred and maintained in a specific pathogen-free facility in accordance with the guidelines of the Central Animal Facility of the University of Salzburg.
. CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted December 7, 2018. . https://doi.org/10.1101/490060 doi: bioRxiv preprint Human specimens. Normal human skin was obtained from patients undergoing elective surgery, in which skin was discarded as a routine procedure. Blood and/or discarded healthy skin was donated upon written informed consent at the University Hospital Salzburg, Austria.
. CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted December 7, 2018. . https://doi.org/10.1101/490060 doi: bioRxiv preprint T cell isolation from skin tissues for flow cytometry. Healthy human skin and ES were digested as previously described (Sanchez Rodriguez et al. 2014). Approximately 1cm 2 of skin was digested overnight in 5%CO2 at 37°C with 3ml of digestion mix containing 0.8mg/ml Collagenase Type 4 (Worthington; #LS004186) and 0.02mg/ml DNase (Sigma-Aldrich; DN25) in RPMIc. ES were digested in 1ml of digestion mix.
Samples were filtered, washed and stained for flow cytometry or stimulated for intracellular cytokine staining. Approx. 3 cm 2 of shaved dorsal mouse skin were harvested and single cell suspensions prepared as described (Gratz et al. 2014) and stained for flow cytometry.

ProcartaPlex™ immunoassays from human skin and engineered skin
Human skin or ES from huPBMC-ES-NSG mice were stored at -70°C until use. Skin was taken up in PBS with Protease Inhibitor Cocktail (1:100) (Sigma-Aldrich; P8340) at 1ml/50mg skin, homogenized and filtered through 0.22µm SpinX columns. Suspensions were stored at -70°C until use. ProcartaPlex immunoassay was performed according to the manufacturer's protocol and measured using Luminex Magpix® system. . CC-BY-NC-ND 4.0 International license was not certified by peer review) is the author/funder. It is made available under a The copyright holder for this preprint (which this version posted December 7, 2018. . https://doi.org/10.1101/490060 doi: bioRxiv preprint Statistical analysis. Statistical significance was calculated with Prism 7.0 software (GraphPad) by one-way ANOVA with Tukey's multiple comparisons test, or by un-paired student's t-test as indicated. Error bars indicate mean + /-standard deviation.

Conflict of Interest.
The authors declare no conflict of interest.

Acknowledgements.
We especially thank all human subjects for blood and skin donation and the nurses at the