Langerhans cells (LCs) reside in the epidermis as a dense network of immune system sentinels, capable of initiating potent immune responses to cutaneous pathogens and neoplastic cells (1–3). As a first line of the cutaneous immune defense system, LCs are uniquely specialized at sensing the environment by extending dendrites through inter-cellular tight junctions to gain access to the outermost part of the skin, the stratum corneum, so that rapid responses can be initiated if a dangerous pathogen is encountered (4). We and others have shown that LCs are capable of priming CD4 T cell responses and can also activate CD8 T cells via antigen cross-presentation, the latter more effectively than CD14+ and CD11c+ dermal dendritic cells (DCs) (5–8). In contrast, during steady-state conditions (non-pathogenic contexts), LCs induce the activation and proliferation of skin-resident regulatory T cells (9, 10) that prevent unwanted immune-mediated pathology. Recent analysis of single cell transcriptional programmes uncovered in situ heterogeneity of human steady-state LCs across different body sites, and suggested the role of PU.1, ID2 and NFkB transcription factors in controlling LC differentiation and activation (11). In the healthy skin LC migration from the epidermis, occurring during non-inflammatory conditions, involves transport of self-antigens, such as those derived from melanin, to skin draining lymph nodes and the priming of tolerogenic regulatory T cells (Tregs) (12–14). In accord with these findings, in a mouse model of autoimmune encephalomyelitis, migratory skin LCs and not lymphoid-resident resident DCs have been shown to induce Foxp3+ Tregs and improve disease outcome (15). While migratory LC appear to play important roles in immunoregulation and cutaneous homeostasis, the nature of the transcriptional states underlying migratory LC immunoregulatory responses remain to be elucidated.

A long standing and widely-held view posits that priming of immunoregulatory responses is a consequence of the immature status of antigen presenting cells (APCs), including LCs and DCs in peripheral tissues (16–20). The immature status of APCs is associated with lower levels of antigen-processing and presentation components as well as co-stimulatory molecules required for efficient T cell priming and activation. According to this view, maturation of APCs that accompanies their migration to draining lymph nodes is necessary for efficient priming of immunogenic but not tolerogenic T cell responses. However, DC maturation has also been shown to occur under steady-state conditions, in particular by activation of the β-catenin pathway and the disruption of E-cadherin mediated adhesion (21). E-cadherin-stimulated DCs undergo maturation with the upregulation of MHC class II, co-stimulatory molecules and chemokine receptors and prime regulatory T cells responses that that suppress immune responses in vivo thereby promoting tolerance. Consistent with these findings, Wnt-β-catenin signalling in intestinal dendritic cells has been shown to be required for expression of anti-inflammatory mediators including retinoic acid-metabolizing enzymes and interleukin-10 and the stimulation of Treg induction (22). Accordingly, loss of β-catenin in DCs results in enhanced inflammatory bowel disease in a mouse model (22).

Although the concept linking the pathogen induced migration and maturation of APCs with their preferential priming of immunogenic responses is consistent with the largely immature status of steady-state LCs (7, 8, 23) and the observations that such LCs can activate skin resident regulatory T (Treg) cells (9, 24), it is challenged by the following. As with DCs, recent findings suggest that the induction and maintenance of cutaneous homeostasis is dependent on LC migration to the lymph nodes (12–14). Notably, such non-inflammatory migratory LCs express significantly higher levels of MHC II and co-stimulatory molecules (7, 8, 25), thereby implicating the importance of LC maturation for the efficient immunoregulation.

Using an in vitro system involving murine bone-marrow derived DCs, we have previously shown that in the absence of pathogen derived signals, the transcription factor interferon regulatory factor 4 (IRF4) promotes the expression of genes required for antigen presentation along with those for T cell tolerance, thus enabling the efficient priming of Treg responses (26). Furthermore, we have recently demonstrated that the migration of human LCs is associated with the upregulation of IRF4 and the enhanced expression of MHC-II complexes and co-stimulatory molecules required for T cell activation (8, 27). The IRF4 dependent gene regulatory network has been proposed to counterbalance the induction of LC activation by pro-inflammatory cytokines coordinated by IRF1 (8, 28). Here we sought to compare the transcriptional programming of steady-state human LCs with their migratory counterparts so as to test our regulatory framework and to gain further insight into transcriptional regulatory circuits that could couple LC maturation with the functional priming of regulatory T cell responses.

Utilizing scRNA-Seq we reveal key differences in transcriptional programming of non-inflammatory LCs. Upon migration, LCs induced the expression of an immunoregulatory gene module including the IDO1, LGALS1, LAMTOR1, IL10RA and IL4L1 genes and this was accompanied with their more efficient priming of regulatory T cell responses in co-culture assays. Regulon-focused analyses of transcriptional programmes, including Partial Information Decomposition in Context, identified IRF4 along with KLF6 and RELB as key regulators of the immunoregulatory program. Select predictions were validated by analysis of CRISPR-Cas9 IRF4-edited LCs. Our findings support the concept that efficient priming of immunoregulatory responses by LCs under non-inflammatory conditions requires upregulation of a migration-induced maturation program coupled with an immunoregulation-inducing gene module.

While LC-mediated immunoregulation appears critical for cutaneous and systemic immune homeostasis, analyses of the genomic programs and molecular pathways that enable human LCs to promote immunoregulatory responses has been hampered by the lack of suitable experimental systems and limitations of available technologies. Here, we analyzed primary human LCs both in steady-state and upon migration from the epidermis for their ability to prime Treg responses. Importantly, these functional analyses were coupled with single-cell transcriptional profiling using Drop-seq (29). Multivariate information measures were then used to predict a transcriptional network underpinning LC immunoregulatory function. This approach revealed that migration out of the epidermis strongly induced LC maturation and immunocompetence. Additionally, transcriptomes of steady-state LCs diverged into two clusters, with a gradation of immunocompetence. Importantly, the transcriptional states of in vitro migratory LCs matched recently published transcriptomes of steady-state LCs (11, 38), including expression of EPCAM, a classical in situ LC marker, and indicated that both S1 and S2 clusters represented activated cells, based on the expression of PIM3, MMP9, LMNA and STK17B.

Classically, the ability of APCs to suppress immune responses thereby inducing tolerance has been associated with their immature state characterized by lower expression of the antigen processing and presentation machinery and co-stimulatory molecules (19, 20). Indeed, earlier studies have shown that DCs in an immature state, expressing low levels of antigen presenting and co-stimulatory molecules, can drive tolerogenic responses by inducing anergy of antigen-specific T cells and expanding Tregs (18, 16). In contrast, here we demonstrate that such immature LCs inefficiently induce Tregs, while a specific subset of immunocompetent LCs expressing CD86 are able to do so more effectively. Supporting the linkage between immunocompetence and ability to prime immunoregulatory responses, LC migration out of the epidermis resulted in a further increase in their ability to induce Tregs when compared with the immunocompetent (S2) steady-state LCs. This observation is corroborated by recent findings, suggesting overlap between migratory and immunoregulatory DC transcriptional programmes, that function during homeostasis to keep in check responses to self-antigens in peripheral tissues (41). The observation that maturation is coupled with ability to induce tolerogenic T cells strongly converges with our earlier studies using a model system of murine bone-marrow derived DCs to analyse the transcriptional and epigenomic control of tolerogenic responses (26). It was demonstrated in that system that tolerogenic functions are up-regulated during DC maturation, via the transcription factor IRF4, which functions in part to dampen inflammatory cytokine signalling. In a striking parallel, we have recently shown that IRF4 is up-regulated upon migration of human LCs and it functions to repress oxidative stress and inflammatory cytokine signalling gene modules (8). Thus, in totality these findings strongly suggest that in both murine and human APCs, the transcription factor IRF4, which is upregulated during migration, functions as a regulatory determinant to couple efficient antigen presentation with immunoregulatory programming of T cell responses.

LC-specific immunoregulatory programme comprised 30 genes, including those encoding classic markers of immunotolerance (such as IDO1, IDO2, CD274) and genes encoding LC maturation (CD80, CD83, CD86). This programme was shared with other skin-resident DC populations, such as CD14/CD141- expressing cells and skin-resident MoDC populations [ Supplementary Figures 2H–K (5, 38)], highlighting its generality, and the importance of functional conservation of transcriptional programmes between LCs and other skin DC populations (5, 8). Interestingly, the programme was expressed strongest in LCs, indicating their immunoregulatory competence. The increased expression of T cell co-stimulatory genes, such as CD86 in the steady-state (immunocompetent) and migrated LCs, are suggestive of their importance in Treg induction. Indeed, CD86 activity has been implicated in DC-mediated tolerance induction through interaction with CTLA4 on T cells (42, 43). However, since LC ability to induce immunoregulatory responses is substantially enhanced upon their migration out of the epidermis, it is likely to be governed by additional factors beyond the capacity to process and present antigens and interact more effectively with T cells via co-stimulatory molecules. Consistent with this possibility, we observed inducible expression of the immunoregulatory gene module, including IDO1, LAMTOR1, IL4L1 and LGALS1 across LC populations. While specific genes were present in select LCs in the steady state, both investigated by us and reported by others (38) ( Supplementary Figures 2I–K ), their expression greatly increased on migration. IDO1 is a classical immunoregulatory mediator, which catabolizes tryptophan leading to skewing of T cell differentiation towards Tregs (44). Galectin-1 encoded by the LGALS1 gene has been shown to promote the generation of tolerogenic DCs and to enable Tr1 type Tregs to suppress Th1- and Th17-mediated inflammation (45, 45). Thus, Galectin-1 secreted by LCs could function in an autocrine as well as paracrine manner to promote Treg responses. The enzymes IL4I1, a mediator of H₂O₂ production and HMOX1, which degrades Heme, have been shown to be expressed by DC and are implicated in the suppression of effector T cell activation and the induction of Tregs (46–49). Additionally, LAMTOR1 is implicated in macrophage polarization towards an immunoregulatory M2 phenotype (50).

Interestingly, PIDC analyses suggests a combinatorial set of TFs that may orchestrate the immunoregulatory transcriptional programme in LCs. This set includes IRF4 (8, 26), KLF6 (51), RELB (52, 53) and ELK1 (54) that have been previously implicated in regulation of immune functions in multiple contexts. In contrast, HMGN3 binds to nucleosomes and regulates chromatin organization (55). Thus, its predicted function in LCs is intriguing, and warrants further investigation.

Importantly, analyses of CRISPR-Cas9 mediated knock-down of IRF4 in LCs confirmed the dependence of several immunoregulation-related genes, including IL4L1 and LGALS1 and LAMTOR1 on IRF4. Interestingly, IRF4 did not seem to directly regulate expression of IDO1. However, a recent study demonstrated that IRF4 can form a multipartite transcriptional complex with AHR, a well-established inducer of IDO1 (56, 57), that binds to promoter elements of immunoregulation associated genes (58). High levels of IRF4 expression could thus potentially promote more efficient AHR action, and an increase in IDO1 expression. Induction of AHR, IRF4 and IDO1 axis upon migration provides a mechanism for inducible IDO1 expression upon activation and suggests the existence of a positive feedback loop for promoting immunoregulatory responses. This has been previously observed in other DCs, where kynurenine metabolites, produced during IDO-mediated catabolism of tryptophan, feedback to AHRs to sustain IDO expression (59, 56)

The ability of mature LCs to induce immunoregulatory T cell responses appears to be seemingly be at odds with their superiority to induce efficient cytotoxic CD8 T cell responses, through antigen cross-presentation and IL15 release, reported by us and others (6–8, 25), as well as with their ability to prime Th2 and Th17 helper T cells (6, 60). When considering regulation of T cell responses by dendritic cells, two mutually non-exclusive models have been proposed: one model postulates specialized DC subsets that are dedicated for the induction of specific T cell responses e.g. induction of CD8T cells by IRF8+ DCs vs enhanced priming of CD4 T cell responses by IRF4+ DCs (26). In a contrasting model, particular DCs can have sufficient flexibility to induce divergent T cell responses that are dictated by the inflammatory context and cytokine milieu (61–63). Research by us and others indicates that LC indeed display considerable degree of plasticity, likely due to their unique functional roles in the epidermis and transcriptional programming that in the steady state is dependent on IRF4 but not IRF8 expression (8, 28). Importantly, the differential outcome of LC function can be induced by signals emanating from the microenvironment. By analyzing molecular regulatory circuits across LCs and other DC types we proposed that immunoregulatory and immunogenic responses of LCs are directed by signaling from the epidermis and involve counter-acting gene circuits that are coupled to a core maturation gene module regulated by NFkB, and two members of IRF family: IRF4 and IRF1 (28). Our model proposes, that while epidermal signaling in the steady-state promotes LC immunoregulatory function, the disruption of cell-cell contacts coupled with inflammatory signaling induces LC immunogenic programming, via IRF1 and IRF1-dependent transcriptional programmes (27, 64). During inflammation, with exposure to proinflammatory cytokines, such as TNF, LCs ability to stimulate CD8 T cells increases significantly. LC-derived IL15 has been shown to specifically promote activation and expansion of CD8 T cells (65–67). Given these findings it is highly plausible that LCs promote both regulatory and cytotoxic T cell responses, with the timing and balancing of these counteracting functions contributing to the potency of immunogenic responses and their eventual resolution in the epidermis.

It is important to highlight, that our experiments were conducted ex vivo, using LCs isolated from the skin microenvironment. While this system allows investigations of human LCs, it presents its own limitations. As the ex vivo manipulation of LCs (digestion, in vitro culture, etc.) might have led to changes not reflective of what happens in vivo, further research in an appropriate in vivo model is required to confirm the identified immunoregulatory transcriptional networks in LCs in the steady-state. Our analysis of LC functional ability to regulate T cell activation via generation of Tregs was limited to migrated LCs, as the strongest expressors of immunoregulatory programme. These LCs partially inhibited CD4 T cell proliferation, suggesting the need for more complex signalling inputs, or the limitations of our experimental conditions i.e., the mixture of cells used and/or short assay time.

Our analyses document that efficient priming of immunoregulatory responses by LCs critically requires upregulation of a migration-coupled maturation program superimposed with a immunoregulation-inducing gene module. While the induction of this immunoregulatory programme in LCs is complex, IRF4 is likely to act as a pivotal switch regulating LC immune function and orchestrating complementary modules in LC transcriptional programming. The enhancement of LC immunoregulatory abilities on maturation could be explored therapeutically to reinstate tolerance in the skin during inflammatory conditions.

The RNA-sequencing data are deposited at GEO under accession number GSE142298, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE142298.

The studies involving human participants were reviewed and approved by South East Coast - Brighton & Sussex Research Ethics Committee in adherence to Helsinki Guidelines [ethical approvals: REC approval: 16/LO/0999). The ethics committee waived the requirement of written informed consent for participation.

MP, SS, JD, HS: intellectually conceived and wrote the manuscript, planned the experiments and analysed the results. SS, JD, KC, GD, AV: run functional experiments, flow cytometry, single-cell sequencing. AV, JD, PSK, JW: developed and optimized scRNA-sequencing. JD, BMA, AV, PSK, MP: analysis and meta-analysis of scRNA-seq data. JD, GD, BMA, MP reconstruction and modelling of gene regulatory networks. MP, MA-J, HS: discussions, data analysis, reviewing of the manuscript. All authors contributed to the article and approved the submitted version.

The study was funded by a Sir Hendy Dale Fellowship from Wellcome Trust, 109377/Z/15/Z. Development of single cell Drop-Seq technology was funded by MRC grant MC_PC_15078. This project in the Pitt Center for Systems Immunology was supported by UPMC-ITTC funds.

We are grateful to the subjects who participated in this study. We would like to thank Prof Peter Friedmann for in-depth review of the manuscript. We acknowledge the use of the IRIDIS High Performance Computing Facility and Flow Cytometry Core Facilities, together with support services at the University of Southampton.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.