Abstract

   The massive infiltration of lymphocytes into the skin is a hallmark of
   numerous human skin disorders. By co-culturing murine keratinocytes
   with splenic T cells we demonstrate here that T cells affect and
   control the synthesis and secretion of chemokines by keratinocytes.
   While pre-activated CD8^+T cells induce the synthesis of CXCL9 and
   CXCL10 in keratinocytes and keep in check the synthesis of CXCL1,
   CXCL5, and CCL20, keratinocytes dampen the synthesis of CCL3 and CCL4
   in pre-activated CD8^+T cells. One key molecule is IFN-γ that is
   synthesized by CD8^+T cells under the control of NFATc1 and NFATc2.
   CD8^+T cells deficient for both NFAT factors are unable to induce CXCL9
   and CXCL10 expression. In addition, CD8^+T cells induced numerous type
   I IFN-inducible “defense genes” in keratinocytes encoding the PD1 and
   CD40 ligands, TNF-α and caspase-1. The enhanced expression of type I
   IFN-inducible genes resembles the gene expression pattern at the
   dermal/epidermal interface in lichen planus, an inflammatory T
   lymphocyte-driven skin disease, in which we detected the expression of
   CXCL10 in keratinocytes in close vicinity to the infiltration front of
   T cells. These data reflect the multifaceted interplay of lymphocytes
   with keratinocytes at the molecular level.

   Keywords: chemokine, keratinocytes, IFN, lichen planus, T cells, Nfatc1

Introduction

   Interactions between keratinocytes and leukocytes play a decisive role
   in the development of numerous skin diseases. Chemotactic chemokines
   attract leukocytes to the skin and direct them to the sites of
   inflammation, infections, and wounds ([45]1, [46]2). Keratinocytes
   synthesize a restricted set of chemokines whose synthesis is amplified
   by numerous stimuli, including type I IFNs, IFN-γ, TNF-α, and others
   that, to a large part, are produced by lymphocytes and other cells. In
   addition, the infiltration of lymphocytes into the skin induces a
   pattern of chemokines that are either synthesized at very low amounts
   or not at all by keratinocytes under homeostatic conditions ([47]3).
   Although there have been numerous attempts to elucidate the role of
   individual chemokines in the generation of skin diseases, our current
   knowledge on the interplay of the immune system with the skin and the
   function of the chemokine network in skin physiology and pathology is
   far from clear.

   There are ~50 chemokines and 20 chemokine receptors ([48]4) that, in
   addition to control leukocyte traffic ([49]5), play an important role
   in the differentiation and function of lymphocytes during immune
   reactions. Both in bone marrow and thymus the development of lymphoid
   cells is guided by the interaction of chemokines with their receptors
   ([50]6, [51]7). In periphery, due to the expression of chemokine
   receptor CCR7, memory T cells have been classified into CCR7^+ central
   memory T[CM] cells that are enriched in lymph nodes and tonsils, and
   CCR7^− effector memory T[EM] cells that are enriched in non-lymphoid
   organs ([52]8). Several chemokine receptors including CCR4, CCR8, and
   CCR10 have been described to direct T cells to the inflamed skin and
   orchestrate T cell-keratinocyte interactions ([53]9, [54]10).

   The differentiation of keratinocytes in the epidermis, the first
   barrier of the organism against the environment, is a fine-tuned
   process that is disturbed under pathological conditions. Keratinocytes
   of the stratum basale proliferate, detach and migrate to the stratum
   spinosum and stratum granulosum and finally differentiate to
   corneocytes of the stratum corneum. Under physiological conditions this
   process takes 1 week in mice and 4 weeks in humans ([55]11). The cells
   of the stratum basale are in close contact with the basement membrane
   and the dermis in which the majority of skin lymphocytes patrol. In
   skin disorders there is often a massive infiltration of lymphocytes
   into the dermis and, less frequently, into the epidermis. Infiltrating
   lymphocytes affect the proliferation of keratinocytes and disturb their
   differentiation to corneocytes. Lichenoid dermatoses such as cutaneous
   lupus erythematosus and lichen planus ([56]12) are prototypic skin
   disorders that are characterized by a massive infiltration of cytotoxic
   lymphocytes and the “liquefaction degeneration” of the basal layer of
   the epidermis. Those infiltrating lymphocytes show a high expression of
   the chemokine receptor CXCR3, the common receptor of the chemokines
   CXCL9, CXCL10, and CXCL11. Due to their massive IFN-γ production,
   infiltrating T cells induce the synthesis of CXCL9 and CXCL10 in
   keratinocytes that, on the other hand, perpetuate and accelerate the
   further infiltration of activated T cells to the skin ([57]13–[58]15).

   To elucidate the molecular events that orchestrate the interplay
   between keratinocytes and skin-infiltrating CD8^+T cells we
   co-cultivated both types of cells and determined their chemokine
   profile. While similar to other authors ([59]3) we detected a strong
   induction of the chemokines CXCL9 and CXCL10 in keratinocytes by IFN-γ,
   we observed that CCL3 and CCL4, which are expressed by cytotoxic T
   cells under the control of NFAT factors, are inhibited. On the other
   hand, CXCL1, CXCL5, and CCL20 expressed by keratinocytes are kept in
   check by T cell-derived IFN-γ. The massive induction of CXCL9 and
   CXCL10 is linked to the expression of NFATc1 and NFATc2 since CD8^+T
   cells deficient for these NFATs were unable to stimulate both
   chemokines. In lichen planus patients who show a massive T cell
   infiltration into the skin we detected the expression of CXCL10 in
   keratinocytes near the invasion front. In addition to chemokines, the
   co-culture of CD8^+T cells with keratinocytes resulted in the
   expression of a number of type I IFN-inducible “survival molecules” by
   keratinocytes such as the PD1 and CD40 ligands, TNF-α and caspase 1
   that affect the fate of keratinocytes.

Materials and Methods

Mice, Isolation and Culture of Cells

   Eight to twelve weeks old C57/B6, sex-, and age-matched mice were used
   in the experiments. Transgenic mice expressing NFATc1-bio protein as
   well as the Nfatc1^flx/flx x Cd4-cre and Nfatc2^−/− mouse lines have
   been described previously ([60]16). NFAT DKO mice were generated by
   crossing of Nfatc1^flx/flx x Cd4-cre and Nfatc2^−/− mice. Primary
   murine keratinocytes were either isolated from tails of adult mice or
   from newborn mice according to Lichti et al. ([61]17) with slight
   modifications. Both mouse tails and newborn mice were disinfected using
   7.5% povidone-iodine solution followed by extensive washes in 70%
   ethanol and distilled water. Upon preparation of skin, skin pieces were
   incubated overnight in dispase solution at 4°C followed by a short
   digestion in trypsin and incubation for 6 h in S-MEM (Gibco) medium
   containing 1.3 mM Ca^++. After 6 h, cells were replenished with
   keratinocyte SFM medium containing 0.06 mM Ca^++ (Gibco) for further
   culture. Since porcine trypsin kills murine keratinocytes, all cell
   culture divisions were performed with TrypsinLE (Gibco; # 12605-010).
   Keratinocyte cultures were treated with murine IL-1α (Preprotech
   211-11A; 30 ng/ml), murine TNF-α (Preprotech 315-01A; 30 ng/ml), or
   murine IL-25 (Biolegend 587302; 2 μg/ml) for 1 and 3 d. To some
   cultures, neutralizing Abs directed against IFN-γ (Biolegend 505702; 5
   μg/ml) or 100 ng/ml CsA were added.

   All co-cultures of keratinocytes with CD8^+T cells were performed at
   least 3 times in duplicate. Splenic T cells were isolated using mouse
   CD8 microbeads kit (Miltenyi Biotech, #130-049-401) according to the
   manufacturer's instructions and cultured in X-VIVO™ 15 medium (Lonza,
   BE02-060F). Cells were pre-activated by αCD3/CD28 (5 μg CD3ε, clone
   73.51, BD Pharmingen) and CD28 (2 μg, clone 145-2C1, BD Pharmingen),
   diluted in 1 ml PBS for coating. The same number or 2-fold more
   pre-activated T cells were co-cultured for 1 or 3 d with freshly
   prepared keratinocytes. Due to the strong differentiation effect of
   Ca^++-containing X-VIVO medium on primary keratinocytes, all
   co-cultures contained 4 parts SFM and 1 part X-VIVO medium.

Transcriptome Analysis by Next Generation Sequencing (NGS)

   Keratinocytes, either cultured alone, or after co-culture with CD8^+ T
   cells, were deep-frozen in liquid nitrogen. Their RNA was extracted
   using Qiagen's RNeasy plus mini kit. The quantity of RNA was assessed
   using a Qubit 2.0 fluorometer, the quality of RNA was determined on a
   Bioanalyzer 2100 using a RNA 6000 nano chip. RNA-Seq libraries for NGS
   were prepared using 20 ng total RNA (RIN>8) and the NEBnext ultra II
   RNA Library Prep kit following the manufacturer's instructions.
   Barcoded RNA-Seq libraries were onboard clustered using HiSeq Rapid SR
   Cluster Kit v2 using 8 p. m. and 59 bps were sequenced on an Illumina
   HiSeq2500 using a HiSeq Rapid SBS kit v2. Sequencing raw data were
   further processed using the software CLC Genomics workbench (v7.5.1
   with CLC's default settings for RNA-Seq analysis). Reads were aligned
   to the GRCm38 genome. For the pathway enrichment analysis (shown in
   [62]Table 1 and [63]Supplementary Table 1) the GORILLA (gene ontology
   enrichment analysis and visualization) tool was used ([64]18).

Table 1.

   Compilation of the first top 10 pathways that were changed in murine
   keratinocytes upon co-culture with pre-activated CD8^+ T cells for 1 d.
   GO term Description P-value FOR g-value Enrichment (N, B, n, b)
   G0:0035456 Response to interferon-beta 1.05E-28 1.35E-24 47.44
   (8048,39,87,20)
   G0:0035458 Cellular response to interferon-beta 2.46E-28 1.59E-24 62.01
   (8048,32,73,18)
   G0:0051707 Response to other organism 1.55E-25 6.64E-22 8.37
   (8048,271,149,42)
   G0:0043207 Response to external biotic stimulus 3.63E-23 1.17E-19 6.68
   (8048,364,149,45)
   G0:0006952 Defense response 6.54E-23 1.68E-19 8.33 (8048,411,87,37)
   G0:0009607 Response to biotic stimulus 2.70E-22 5.79E-19 6.38
   (8048,381,149,45)
   G0:0009617 Response to bacterium 1.19E-20 2.19E-17 4.82
   (8048,141,604,51)
   G0:0002376 Immune system process 1.28E-20 2.06E-17 4.45
   (8048,703,144,56)
   G0:0006955 Immune response 7.31E-19 1.05E-15 5.98 (8048,345,160,41)
   G0:0051704 Multi-organism process 2.85E-18 3.67E-15 6.21
   (8048,436,113,38)
   [65]Open in a new tab

   RNAs of keratinocytes were isolated and their transcriptome determined
   in NGS assays. Five hundred three genes whose expression was changed
   2.5 fold between keratinocytes co-cultured with CD8^+ T cells compared
   with keratinocytes alone were classified using the Gorilla (Gene
   Ontology enRIchment anaLysis and visualization) tool ([66]18). GO, gene
   ontology terms. FDR (False Discovery Rate) q-value is the correction of
   the indicated p-value for multiple testing using the Benjamini and
   Hochberg method. The enrichment (N,B, n,b) is defined as follows: N is
   the total number of genes, B is the total number of genes associated
   with a specific GO term, n is the number of genes in the top of the
   user's input list (or in the target set when appropriate), b is the
   number of genes in the intersection. Enrichment = (b/n)/(B/N). The
   individual genes in the 10 intersections are presented in
   [67]Supplementary Table 1.

   The data of NGS assays have been deposited in the GEO database
   ([68]GSE131905).

Immunohistochemistry

   All specimens were obtained from the archive of paraffin-embedded
   diagnostic tissues of the Department of Dermatology at the University
   Hospital of Würzburg, Germany, and were used in accordance to the
   guidelines of the local ethics committee. Anonymized sections were
   prepared from formalin-fixed, paraffin-embedded skin biopsies from
   lichen planus patients. Immunohistochemical stains were performed using
   mAb specific for CD8 (clone SP-16, dilution 1:50, Invitrogen), NFATc1
   (clone 7A6, dilution 1:100, BD Pharmingen), NFATc2 (clone D43B1,
   dilution 1:200, Cell signaling), CXCL9 (MIG, dilution 1:15, R&D
   systems), or CXCL10 (Polyclonal, clone 33036, Dilution 1:15, R&D
   Systems) following the manufacturer's instructions. The stained slides
   were captured with confocal microscope (Zeiss) or a Nikon Ti motorized
   stage fluorescence microscope at the magnifications indicated. The
   absolute numbers of keratinocytes expressing CXCL10 were determined
   using ImageJ.

Enzyme-Linked Immunosorbent Assay (ELISA)

   The concentration of IFN-γ in the culture supernatants was determined
   using commercially available mouse IFN-γ ELISA kits (88-7314-88,
   Invitrogen). Each sample was analyzed in duplicate following the
   manufacturer's instructions.

Chemokine Secretion Assay

   For chemokine measurements, we used the LEGENDplex™ Mouse
   Proinflammatory Chemokine Panel (BioLegend) kit according to the
   manufacturer's instructions. This is a multiple fluorescence-encoded
   bead-based assay allowing the simultaneous detection of 13 different
   chemokines by flow cytometry.

Statistical Analysis

   Statistical analyses were performed using GraphPad (Prism) software,
   version 6.0. Data were presented as mean ± SEM. Unpaired t-test was
   performed to evaluate the statistical significance. Statistical
   significances are indicated (^***p < 0.001, ^**p < 0.005, and ^*p <
   0.05).

Results

Secretion of Chemokines by Primary Murine Keratinocytes ex vivo

   To determine the effect of lymphocytes on the secretion of chemokines
   by keratinocytes, we isolated primary keratinocytes from the tails of
   adult mice, and from newborn mice and cultured them in serum-free SFM
   keratinocyte medium (Gibco) supplemented with 0.06 mM Ca^++. Under
   these conditions, which resemble those of the stratum basale, the
   keratinocytes rapidly formed a lawn of cobblestone-like cells. However,
   upon co-culture with splenic CD8^+ or CD4^+ T cells activated by
   αCD3/CD28 Abs for 1 d (or upon stimulation of keratinocytes with IL-1α,
   TNF-α, or IL-25) the tight layer of keratinocytes showed numerous holes
   in which T cells assembled ([69]Supplementary Figures 1A–C). After 1
   day of co-culture, no measurable differences were observed in cell
   viability and cell proliferation, as compared to individual cell
   cultures ([70]Supplementary Figure 1D and data not shown). After one
   and 3 days of co-culture, the supernatant of cultures was collected to
   determine the secretion of 13 chemokines using a fluorescence-encoded
   bead-based assay (BioLegend). In addition, transcriptional changes of
   keratinocytes induced in the co-cultures were measured in NGS assays.
   Keratinocytes from adult or newborn mice cultured in vitro for 1 or 3 d
   secrete a restricted set of chemokines, mainly CXCL1 and CCL2, and very
   low amounts of CXCL5 ([71]Figure 1, lanes 1, 9, and [72]Supplementary
   Figure 2). Stimulation of adult keratinocytes with IL-1α and TNF-α
   enhanced CXCL1 and CCL2 secretion. Co-cultures of keratinocytes with
   pre-activated T cells resulted in an increase of CXCL5 secretion and
   also in the appearance of newly induced chemokines such as CCL3, CCL4,
   CXCL9 and CXCL10 ([73]Figure 1, lanes 5-6 and 13-14), which remained
   unaffected by adding CsA to the co-cultures ([74]Figure 1, lanes 8/16).

Figure 1.

   [75]Figure 1
   [76]Open in a new tab

   Chemokine secretion of murine keratinocytes ex vivo. 10^5 keratinocytes
   cultured for 7 d in vitro were seeded in 12-well plates. They were
   cultivated alone (columns 1–4 and 9–12) or together with 10^5 splenic
   CD4^+ or CD8^+T cells pre-activated for 1 d by αCD3/CD28 Abs (columns
   5–8 and 13–16) for another day. Keratinocytes remained unstimulated or
   were stimulated by IL-1α and/or TNF-α (30 ng/ml each), CsA (100 ng/ml),
   or IL-25 (2 μg/ml) as indicated. In the assays 7 and 15, 5 μg/ml of an
   Ab neutralizing IFN-γ was added. The supernatants of cultures were
   harvested and chemokine secretion was measured using a
   fluorescence-encoded bead-based assay (BioLegend), ^*p < 0.05, ^**p <
   0.005, ^***p < 0.001.

   The pattern of chemokines secreted by keratinocytes from the skin of
   newborn mice overlapped with that of adult tail keratinocytes but
   differed in detail. Remarkably, the stimuli that enhanced chemokine
   production diverged conspicuously between both types of keratinocytes.
   While IL-1α, TNF-α, and IL-25 affected mildly—or not at all—the
   chemokine production of adult tail keratinocytes, these stimuli
   strongly affected the secretion of CXCL5 and CCL20 by keratinocytes
   from newborn mice. Adding IL-1α and TNF-α to keratinocytes from newborn
   mice resulted in a 100 fold increase of CXCL5 and somewhat weaker of
   CCL20 secretion upon culture for 3 d ([77]Supplementary Figure 2).

CD8^+T Cells Control Chemokine Secretion of Keratinocytes

   In the co-cultures of T cells with keratinocytes, the same number
   (ratio 1:1) or 2-fold more CD8^+T cells (i.e., ratio 2:1) pre-activated
   for 1 d by αCD3/CD28 Abs in X-vivo-medium were added. Since
   keratinocytes were maintained in serum-free keratinocyte medium this
   resulted in a 4:1 mix of keratinocyte: X-vivo media, respectively, and,
   thereby, in a drop of Ca^++ in culture medium and subsequent silencing
   of numerous Ca^++-dependent genes.

   The secretion of CXCL1 and CCL2 chemokines remained either unaffected
   or was slightly decreased while that of CXCL5 and CCL20 was increased
   upon co-culture with CD8^+T cells ([78]Figures 1, [79]2). This is
   especially the case in wild type and NFATc1-ko CD8^+T cells. T cells
   from mice double ko (DKO) for NFATc1 and NFATc2 remained “neutral” with
   respect to CXCL1 and CCL20 secretion and enhanced slightly the
   expression of CXCL5 when compared to wild type and NFATc1-ko CD8^+T
   cells ([80]Figure 2). Since DKO CD8^+T cells are prone to apoptosis one
   may speculate that pro-apoptotic signals emerged from those cells to
   increase the secretion of pro-inflammatory chemokine CXCL5 ([81]19) by
   keratinocytes.

Figure 2.

   [82]Figure 2
   [83]Open in a new tab

   Secretion of chemokines CXCL1, CXCL5 and CCL20 by keratinocytes alone
   or upon co-culture with CD8^+T cells. Keratinocytes cultured for 7 d in
   vitro and splenic CD8^+ T cells pre-activated by αCD3/CD28 Ab for 1 d
   were either cultured alone or together in 6-well plates for 1 or 3 d as
   indicated and their chemokine secretion measured.

   The secretion of CCL3 and CCL4 by CD8^+T cells was strongly suppressed
   upon co-culture with keratinocytes or upon culture in SFM medium
   ([84]Figure 3A). This silencing effect was also reflected at the
   transcriptional level. The strong transcription of Ccl3 and Ccl4 genes
   in pre-activated CD8^+T cells was almost abolished upon culture of T
   cells in SFM/X-vivo (4:1) medium, without or with co-culture with
   keratinocytes whereas that of the Ccl5 gene remained unaffected
   ([85]Figure 3B). This is due to the defect of NFAT activity in low
   Ca^++ medium. In activated cytotoxic T cells the promoters of Ccl3 and
   Ccl4 genes – but not of the Ccl5 gene—are bound by NFATc1 ([86]Figure
   3C) and, therefore, are excellent targets for NFATc1 in cytotoxic T
   cells ([87]16).

Figure 3.

   [88]Figure 3
   [89]Open in a new tab

   Silencing of CCL3 and CCL4 secretion in keratinocyte medium. (A)
   Secretion of chemokines CCL3 and CCL4 by keratinocytes, CD8^+T cells,
   or co-cultures of both. Keratinocytes cultured for 7 d in vitro and
   splenic CD8^+ T cells pre-activated by αCD3/CD28 Ab for 1 d were either
   cultured alone or together in 6-well plates for 1 or 3 d as indicated
   and their chemokine secretion was measured. (B) Suppression of Ccl3 and
   Ccl4, but not of Ccl5 transcription upon culture of pre-activated
   CD8^+T cells for 1 d in keratinocyte/X-vivo medium (4:1) alone (column
   2) or in the presence of keratinocytes (column 3). Column 1 shows the
   RNA levels of genes in pre-activated CD8^+T cells cultured in X-vivo
   medium for 1 d. (C) Binding of NFATc1-bio protein to the promoters of
   murine Ccl3 and Ccl4 genes in activated cytotoxic T cells. Results of
   ChIP seq assays ([90]16). Splenic CD8^+T cells from transgenic mice
   expressing NFATc1-bio proteins were stimulated by T+I for 5 h. Their
   chromatin was cross-linked by 4% formaldehyde-treatment of cells in
   vitro, followed by shearing of chromatin and binding to streptavidin
   beads. After the release of DNA from cross-linked chromatin complexes,
   DNA was sequenced in NGS assays ([91]16, [92]20), ^*p < 0.05, ^**p <
   0.005, ^***p < 0.001.

CXCL9 and CXCL10 Are Strongly Induced in Keratinocytes by IFN-γ at the
Transcriptional Level

   CXCL9 and CXCL10 were neither secreted by untreated keratinocytes nor
   by activated WT CD8^+T cells. However, upon co-culture of keratinocytes
   with T cells pre-activated by αCD3/CD28 for 1 d a massive induction of
   both chemokines was observed. This was correlated with the NFAT
   expression of T cells, since NFAT DKO CD8^+T cells were unable to
   induce any CXCL9 and CXCL10 secretion and NFATc2–ko CD8^+T cells showed
   a marked drop in the induction of both chemokines. In contrast, in
   three out of five co-culture assays NFATc1-ko cells did not show any
   restriction in chemokine induction whereas in two assays a decrease was
   detected ([93]Figures 4A,B and [94]Supplementary Figure 3A). A strong
   induction of CXCL9 and CXCL10 chemokines was also observed in
   keratinocytes after culture for 1 d with “T cell-conditioned medium,”
   i.e., with X-VIVO medium in which CD8^+T cells were activated for 1 d
   by αCD3/CD28, whereas a poor induction was observed using conditioned
   medium from NFAT DKO CD8^+T cells ([95]Supplementary Figure 3B). The
   differences in chemokine induction could be explained by marked
   differences in IFN-γ secretion between NFAT DKO, NFATc1-ko, and
   NFATc2-ko CD8^+T cells. While DKO CD8^+T cells did not synthesize any
   IFN-γ, NFATc2-ko showed a strong, and NFATc1-ko cells a more moderate
   decrease in IFN-γ secretion before and after co-culture ([96]Figures
   4A,B). Along with the suppressive effect of Abs directed against IFN-γ
   on CXCL9 and CXCL10 secretion ([97]Figure 1 and [98]Supplementary
   Figure 2), these data indicate that IFN-γ produced by CD8^+T cells
   stimulates keratinocytes to secrete CXCL9 and CXCL10.

Figure 4.

   [99]Figure 4
   [100]Open in a new tab

   Induction of CXCL9, CXCL10 and IFN-γ secretion in keratinocytes upon
   co-culture with CD8^+T cells. (A,B) Secretion of chemokines CXCL9 and
   CXCL10 and of IFN-γ by keratinocytes and CD8^+T cells. Keratinocytes
   cultured for 7 d in vitro and splenic CD8^+ T cells pre-activated by
   αCD3/CD28 Ab for 1 d were either cultured alone or together in 6-well
   plates for another 1 or 3 d as indicated and their chemokine and IFN-γ
   secretion measured. In (A) the incubation of NFAT-DKO T cells was
   included, in (B) the incubation of NFATc2-ko instead of DKO T cells.
   (C) Induction of Cxcl9 and Cxcl10 transcription in keratinocytes that
   were cultured alone (column 1) or together with the same number of
   pre-activated CD8^+ T cells for 1 d (column 2). In the co-culture, the
   CD8^+ T cells were washed off, and the RNA levels of keratinocytes were
   determined in NGS assays, ^*p < 0.05, ^**p < 0.005, ^***p < 0.001.

   NGS assays of the transcriptome of keratinocytes from WT mice before
   and after co-culture with T cells for 1 d demonstrated that the
   induction of CXCL9 and CXCL10 secretion occurs at the transcriptional
   level. In cultured murine keratinocytes, 8440 genes were expressed in 5
   and more RPKM, and 503 genes were changed 2.5 fold and more in their
   expression level between keratinocytes without or upon co-culture with
   T cells. When these 503 genes were analyzed using the GORILLA (gene
   ontology enrichment analysis and visualization) tool, among the most
   prominent 10 gene sets, 9 corresponded to genes controlling response
   and defense mechanisms ([101]Table 1). While the top 2 pathways
   comprise genes induced in the interferon-beta response, including the
   genes encoding the transcription factors IRF1 and STAT1, the other
   pathways are characterized by further response genes, including the
   Cxcl9 and Cxcl10 genes ([102]Supplementary Table 1). Upon co-culture
   for 1 d, the RNA levels of the Cxcl9 gene increased more than 100 fold,
   and that of the Cxcl10 gene approximately 10 fold ([103]Figure 4C). A
   2–5 fold increase was detected for the two genes encoding STAT1 and
   STAT2. A similar increase was observed for the Cd274 gene encoding PD1
   ligand and the Tnf and Cd40lg genes ([104]Supplementary Figure 3C) that
   defend keratinocytes against T cell-mediated killing.

   It is worth mentioning that within the gene sets of the top 10 pathways
   induced in keratinocytes by T cells the overwhelming majority of genes
   corresponded to genes induced by type I interferons. In addition to the
   genes encoding the transcription factors STAT1, STAT2, IRF1, IRF7 and
   IRF8, these include the Casp1 and Gsdmd genes that code for caspase-1
   and gasdermin D—prominent inducers of pyroptosis and membrane pore
   formation ([105]21)—and numerous genes coding for guanylate binding
   proteins, Tap transporters and proteasome subunits. All these genes are
   compiled in the [106]Supplementary Table 1.

Local CXCL10 Induction in Keratinocytes From Lichen Planus Patients

   Finally, we wanted to see whether our in vitro/ex vivo findings on the
   interplay between murine CD8^+T cells and keratinocytes are of
   relevance for human skin disorders. To this end, we selected lichen
   planus (LP) as model disease. LP is a T cell mediated inflammatory
   disease that is characterized by the massive infiltration of T
   lymphocytes, the destruction of the basal epidermal layer and the
   induction of numerous genes with a typical IFN signature ([107]12,
   [108]13, [109]22). These include the chemokine Cxcl9 and Cxcl10 genes
   as typical molecular LP markers that distinguish lichenoid tissue
   reactions from atopic dermatitis and psoriasis ([110]12). In our
   co-stains of LP skin sections using Abs directed against CXCL10, CD8,
   CD4, and pan-keratin (or Krt14) we detected a strong expression of
   CXCL10 in keratinocytes located mostly at the edge of elongated
   epidermal ridges closely surrounded by infiltrating T cells. In
   contrast, infiltrating CD8^+ and CD4^+T cells showed rather a very weak
   staining with the CXCL10 Ab ([111]Figure 5 and [112]Supplementary
   Figures 4, 5). For quantification of CXCL10-expressing keratinocytes we
   counted positive cells in each immunofluorescence stain of 10 patients
   and compared results with healthy skin ([113]Figure 5).

Figure 5.

   Figure 5
   [114]Open in a new tab

   Enumeration of selectively accentuated CXCL10+ keratinocytes in lichen
   planus patients. Representative immunofluorescence stainings for
   CXCL10, CD8, and pan-cytokeratin (Pan-Kera) in (A) lichen planus (LP)
   and (B) healthy skin (HS). Rectangle marks position of blow-up (right
   side). (C) CXCL10^+ cells were counted in at least 3 high power fields
   (HPF)/patient (n = 10) and compared to HS. Confocal microscopy, 40x
   magnification.

   The immunofluorescence stains with two Abs directed against CXCL9 did
   not show a distinct expression of CXCL9 in keratinocyte at the
   keratinocyte/T cell border but a broad, more diffuse expression in the
   invading cells, including CD4^+T lymphocytes and other cells
   ([115]Figure 6).

Figure 6.

   Figure 6
   [116]Open in a new tab

   Expression of CXCL9 in the skin of a lichen planus patient. Skin
   sections were stained with Abs directed against CXCL9, NFATc1, and CD4
   and co-stained with DAPI as indicated. Immunofluorescence microscopy,
   40x magnification.

   The data on the role of NFAT factors in CXCL9 and CXCL10 induction
   prompted us to speculate that in lichenoid tissue reactions the nuclear
   expression of NFATs might be a further molecular sign of infiltrating T
   cells. However, in all immunofluorescence stains of such skin sections
   we were unable to detect any clear-cut nuclear NFATc1 or NFATc2
   localization in infiltrating CD8^+ and CD4^+ T cells, but a prominent
   cytosolic expression of both NFAT factors ([117]Supplementary Figures
   6A,B).

Discussion

   The experimental data presented here extend a large body of published
   results on the interplay between epidermal keratinocytes and T
   lymphocytes ([118]23, [119]24). Employing co-culture experiments and
   analyzing supernatants (“conditioned medium”) of activated CD8^+T cells
   we demonstrated that T lymphocytes modulate the chemokine synthesis of
   keratinocytes, including the IFN-γ-mediated induction of CXCL9 and
   CXCL10 chemokines. In addition, we detected the induction of a large
   set of type I IFN-inducible genes in keratinocytes. Similar changes in
   gene expression have been described in the skin of lichen planus
   patients ([120]13) in which we detected the distinct expression of
   CXCL10 in keratinocytes near the border to invading T cells.

   Compared to the situation in vivo, there are limitations of our
   co-culture system. The infiltration of lymphocytes into the skin and
   the attack of the stratum basale are a feature of many skin diseases.
   In our co-cultures, keratinocytes of stratum basale were exclusively
   propagated and studied. Therefore, the attack of pre-activated T cells
   and the “liquefaction degeneration” of the basal layer of the epidermis
   by cytotoxic lymphocytes might reflect certain pathological situations
   in vivo, such as in the skin of patients suffering from lichen planus.

   The massive changes in keratinocyte gene expression induced by T cells
   suggest the existence of a fine-tuned balance between keratinocytes and
   T lymphocytes that is disturbed in numerous skin diseases. This might
   be due to metabolic changes in primary keratinocytes which, like
   activated CD8^+T cells ([121]16), exhibit a high glycolysis rate (own
   unpublished data). Our data indicate that CXCL9 and CXCL10 are induced
   by IFN-γ secreted by T cells. Both chemokines attract (autoreactive) T
   cells expressing CXCR3, the common receptor for CXCL9, CXCL10 and
   CXCL11 ([122]4), to the skin ([123]25, [124]26). Albeit the co-culture
   system does not reflect the complex cellular interplay between
   keratinocytes, T cells, macrophages, dendritic cells, and others in
   human skin diseases, it contributes to the understanding of T
   cells/keratinocytes interactions. In LP patients, this “feed-back”
   mechanism leads to a “flood” of invading T lymphocytes into the dermis,
   the liquefaction degeneration of the basal epidermal layer and the
   interference with keratinocyte differentiation ([125]27). Surprisingly,
   unlike CXCL10 we were unable to detect a distinct expression of CXCL9
   in keratinocytes of LP patients, in spite of the comparable induction
   of CXCL9 and CXCL10 secretion by murine keratinocytes upon co-culture
   with CD8^+T cells. However, a similar dichotomy in the expression of
   CXCL9 and CXCL10 (and CXCL11) has been described before for different
   human skin disorders. In those studies CXCL10 and CXCL11 were found to
   be predominantly expressed in basal keratinocytes whereas CXCL9
   expression was predominantly localized in dermal infiltrates ([126]15,
   [127]28).

   It should be noted here that CXCR3 is not a skin-specific chemokine
   receptor but attracts immune cells to sites of IFN-γ-mediated
   inflammation ([128]3). CXCR3 is not only expressed on activated T cells
   but also on dendritic cells, natural killer cells, fibroblasts, smooth
   muscle, and epithelial cells ([129]25, [130]29). Accordingly, the
   expression of CXCR3 and its activation by chemokine ligands has been
   associated not only with numerous skin diseases but also with other
   disorders, such as arthritis, type 1 diabetes ([131]30, [132]31), and
   allograft rejections including the rejection of heterotopic heart
   allografts ([133]32, [134]33). These findings underline the importance
   of CXCL9/CXCL10 signals in the control of (auto-) immunity and have
   been exploited in immune therapies against tumors ([135]3, [136]34) in
   which CXCR3 is strongly expressed ([137]35, [138]36).

   Apart from CXCL9 and CXCL10, the induction of genes encoding caspase-1
   and gasdermin D are additional steps in the T cell-mediated “killing
   program” of keratinocytes. Caspase-1 is an inflammatory response
   initiator that cleaves other proteins, including the pyroptosis inducer
   gasdermin D ([139]21). However, keratinocytes are also able to fight
   back, as it is shown by the induction of the Cd274 gene coding for the
   PD1 ligand, which—through its binding to PD-1 on activated T
   cells—dampens T cell activity ([140]37, [141]38). Most of these genes
   are induced via type I IFN pathways. This is demonstrated by the
   induction of Irf genes Irf1, Irf7, and Irf8, and of Stat factors STAT1
   and STAT2 that are induced by type I IFNs ([142]Supplementary Table 1).

   The decrease in expression of Ccl3 and Ccl4 genes in pre-activated
   CD8^+T cells that we observed upon culture of T cells in Ca^++-poor
   SFM/X-vivo medium might be of relevance for the fate of T cells
   infiltrating the skin. In the basal layer of the epidermis
   extracellular Ca^++ concentrations are very low and facilitate the
   proliferation of keratinocytes whereas increasing Ca^++ supports
   keratinocyte differentiation ([143]39, [144]40). Therefore, one may
   speculate that under physiological conditions invading T cells that
   need Ca^++ for full activity are inhibited. This is reflected in the
   suppression of numerous Ca^++-dependent genes in our assays, including
   Ccl3, Ccl4, Ifn-γ, and further NFAT-dependent genes.

   The need for T cell activation and the inability of NFAT DKO T cells to
   induce CXCL9 and CXCL10 indicate that the IFN-γ-mediated induction of
   both chemokines is linked to NFAT activity in T cells. In activated
   CD8^+ T cells and, in vivo, in cytotoxic T cells the synthesis of IFN-γ
   is under the control of the NFAT factors NFATc1 and NFATc2. In CTLs,
   inactivation of NFATc1 led to a complete loss of IFN-γ induction upon
   TPA+ionomycin treatment whereas NFATc2 inactivation exerted a milder
   defect, leading to a reduction of ~60% ([145]16). Therefore, we
   expected to see a strong defect in CXCL9 and CXCL10 induction by
   NFATc1-ko, but not NFATc2-ko T cells. However, the opposite was
   observed. The strong decrease in IFN-γ production by NFATc2-ko CD8^+T
   cells suggests that under the chosen experimental conditions, i.e., 1
   day activation of splenic CD8^+T cells by αCD3/CD28 (followed by
   co-culture with keratinocytes for 1 d), NFATc2—and not NFATc1—is the
   most prominent NFAT factor to stimulate IFN-γ expression in murine
   CD8^+T cells. It remains to be shown whether this is also true for
   skin-infiltating T cells in human skin diseases.

Data Availability

   The datasets generated for this study can be found in the data of NGS
   assays have been deposited in the GEO database ([146]GSE131905).

Ethics Statement

   Animal experiments were performed according to project licenses (Nr.
   55.2-2531.01-80/10), which are approved and controlled by the Regierung
   von Unterfranken, Würzburg. All anonymized specimens from LP patients
   were obtained from the archive of paraffin-embedded diagnostic tissues
   of the Department of Dermatology at the University Hospital of
   Würzburg, Germany, and were used in accordance with the local ethics
   committee.

Author Contributions

   TR, VS, MA, SK-H, and KMuh: Performed experiments. KMur, FG, and AK:
   Immunofluorescence stains. KMuh, MK, and TB: NGS assays. MG and AK:
   Analyzed data and supported the preparation of the manuscript. KMuh and
   ES: Led the investigation and wrote the manuscript.

Conflict of Interest Statement

   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.

Acknowledgments