Mouse genetics identifies unique and overlapping functions of fibroblast growth factor receptors in keratinocytes

Abstract Fibroblast growth factors (FGFs) are key regulators of tissue development, homeostasis and repair, and abnormal FGF signalling is associated with various human diseases. In human and murine epidermis, FGF receptor 3 (FGFR3) activation causes benign skin tumours, but the consequences of FGFR3 deficiency in this tissue have not been determined. Here, we show that FGFR3 in keratinocytes is dispensable for mouse skin development, homeostasis and wound repair. However, the defect in the epidermal barrier and the resulting inflammatory skin disease that develops in mice lacking FGFR1 and FGFR2 in keratinocytes were further aggravated upon additional loss of FGFR3. This caused fibroblast activation and fibrosis in the FGFR1/FGFR2 double‐knockout mice and even more in mice lacking all three FGFRs, revealing functional redundancy of FGFR3 with FGFR1 and FGFR2 for maintaining the epidermal barrier. Taken together, our study demonstrates that FGFR1, FGFR2 and FGFR3 act together to maintain epidermal integrity and cutaneous homeostasis, with FGFR2 being the dominant receptor.


| INTRODUC TI ON
Fibroblast growth factors (FGFs) are master regulators of development and tissue repair, and abnormal expression of FGFs or their receptors is associated with a wide variety of human diseases. [1][2][3] Most of the 22 members of the mammalian FGF family bind and activate four receptor tyrosine kinases, designated FGFR1-FGFR4, 3 which exert distinct functions in different tissues and organs. In some cases, however, partially overlapping functions of FGFRS have been discovered as revealed by a stronger phenotype of double-compared with single-knockout mice. This was previously demonstrated in our laboratory for the skin through generation and characterization of mice lacking FGFR1 or FGFR2 or both receptors in keratinocytes. 4 Whereas mice lacking only FGFR1 in this cell type are phenotypically normal, mice lacking FGFR2 in keratinocytes develop hair follicle and sebaceous gland abnormalities and mild, although progressive skin inflammation. 4,5 Interestingly, loss of both receptors resulted in a strong phenotype | 1775 MEYER Et al.
characterized by progressive loss of hair follicles and sebaceous glands and development of chronic skin inflammation. 4,6,7 This finding demonstrates that FGFR1 signalling is unmasked in the absence of FGFR2, probably through FGF10/FGF22 signalling to the b splice variant of FGFR1. 2,4 The phenotype of the double mutant mice (designated K5-R1/R2 mice) shows many similarities to the skin inflammation that occurs in patients with the chronic inflammatory skin disease atopic dermatitis (AD). 4,[6][7][8] It results from a defect in the epidermal barrier that is caused at least in part by decreased expression of major tight junction components that are under direct control of FGFR signalling in keratinocytes, 4,[6][7][8] We is also expressed in the murine epidermis, 9 we speculated that loss of this receptor in K5-R1/R2 mice would aggravate the phenotype.
In addition, a phenotype in mice lacking only FGFR3 in keratinocytes was anticipated, as activating mutations in the FGFR3 gene are the cause of the genetic skin disorder acanthosis nigricans and also induce seborrhoeic keratosis and epidermal naevi. [10][11][12][13] Here, we show, however, that loss of FGFR3 in keratinocytes does not obviously affect skin morphogenesis, homeostasis or wound repair in mice. Surprisingly, loss of all FGF receptors in keratinocytes is compatible with life, but the FGFR3 deficiency further aggravated some of the phenotypic abnormalities seen in K5-R1/R2 mice. Overall, these results identify FGFR2 as the major functional FGF receptor in keratinocytes, whereas FGFR1 and FGFR3 have a "back-up" function.

| Mice
Mice lacking FGFR1 and FGFR2 in keratinocytes (K5-R1/R2 mice) were previously described. 4,[6][7][8]14 To generate mice lacking a functional FGFR3 protein in keratinocytes, we mated mice with floxed FGFR3 alleles 15 with K5-Cre mice. 16 Triple mutant mice were obtained by crossing females with floxed FGFR3 alleles with K5-R1/ R2 males ( Figure 1A). The F1 generation was paired inter se until K5-R1/R2/R3 mice were obtained as described in Figure 1A. All K5-Cre mice used for breeding were males, as global deletion occurred with females. 16 Because of the difficult breeding scheme, each experiment included mice from different litters, but at least 1-2 mice from the same litter were used for a direct comparison in all experiments. All mice were in C57BL/6 genetic background. Control mice (Ctrl) were mice with floxed FGFR alleles but without Cre recombinase or occasionally K5-Cre mice. They were housed under specific pathogen-free conditions and received food and water ad libitum.
Mouse maintenance and all animal experiments had been approved by the veterinary authorities of Zurich, Switzerland (Kantonales Veterinäramt Zürich).

| Establishment and culture of primary mouse keratinocytes
Keratinocytes were isolated from single mice as described previously 4 and cultured in defined keratinocyte serum-free medium (Invitrogen) supplemented with 10 ng/mL epidermal growth factor (EGF), 10 −10 mol/L cholera toxin and 100 U/mL penicillin/100 μg/ mL streptomycin (all from Sigma) in keratinocyte medium. 17 Plates were coated with collagen IV (2.5 g/cm 2 ) prior to seeding of the cells.

| 5-Bromo-2′-deoxyuridine (BrdU) incorporation assay
Primary keratinocytes were incubated overnight in keratinocyte serum-free medium without EGF. EGF (Sigma) or FGF1 (Peprotech) was added to a final concentration of 10 ng/mL and incubated for 24 hours. After 20 hours, BrdU (Sigma) was added to the cell culture medium at a final concentration of 100 μmol/L followed by incubation for 4 hours at 37°C and 5% CO 2 . Then, cells were washed with PBS and fixed with 4% paraformaldehyde for 30 minutes at RT. Afterwards, they were permeabilized and DNA was denatured using 0.1% Triton X-100 in 2 mol/L HCl for 30 minutes.

| Histology
Skin samples were fixed overnight with 4% paraformaldehyde (PFA) or 95% ethanol/1% acetic acid prior to paraffin embedding. Sections was used for acquisition of data. Data were analysed using ImageJ software. All analyses were performed blinded with regard to the genotype of the mice.

| Immunohistochemistry and immunofluorescence analysis
Paraffin sections were dewaxed and rehydrated using a xylene/ ethanol gradient followed by antigen retrieval using citrate buffer (10 mmol/L citric acid, pH 6.0) at 95°C for 1 hour and three washes with PBST (PBS, 0.1% Tween). Skin sections were then blocked with PBS containing 12% BSA for 1 hour at RT, followed by incubation with the primary antibodies overnight at 4°C. Cryosections were fixed in 1%-4% paraformaldehyde for 10 minutes at RT, washed for 2 × 5 minutes in PBS and blocked in 5% BSA for 1 hour at room temperature, followed by incubation with the primary antibodies overnight at 4°C. For bright-field microscopy analysis, a biotinconjugated secondary antibody, the Vectastain ABC Kit and the DAB peroxidase substrate kit (both from Vector Laboratories) were used for visualization. For immunofluorescence staining, slides were incubated at room temperature for 1 hour with the Cy2-or Cy3-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories, Inc) and counterstained with Hoechst 33342 (Sigma).
The following antibodies were used for immunohistochemistry or immunofluorescence:

| Analysis of transepidermal water loss (TEWL)
Mouse back skin was shaved one day before TEWL analysis. TEWL was determined using a Tewameter (Courage and Khazaka Electronic GmbH) as described previously. 6 Twenty-five consecutive measurements were taken from four different places on the back on different days.

| Isolation of RNA from mouse skin
Mice were killed, shaved, and epidermis from mouse back skin was separated from the dermis after heat shock treatment (30 seconds at

| Real-time RT-PCR
RNA was reverse-transcribed using the iScript™ cDNA Synthesis Kit The following primers were used for qRT-PCR:

| Flow cytometry
Unspecific binding sites were blocked using a CD16/CD32 antibody. Dead cells were stained with Zombie Aqua™ dye (BioLegend).
The antibodies used for flow cytometry analysis are listed below.
Stained cells were quantified using a BD Fortessa machine and BD

| Wound healing experiments
Female K5-R3 and Ctrl mice (9-11 weeks of age) were anaesthetized by intraperitoneal injection of ketamine/xylazine (100 mg ketamine/5-10 mg xylazine per kg bodyweight). After shaving of the back skin, four full-thickness excisional wounds of 5 mm diameter were generated using a biopsy punch, two wounds on each side of the dorsal midline. Wounds were allowed to heal without dressing. Mice were killed by CO 2 inhalation at different time points after wounding, and paraffin sections from the middle of the wounds were used for histological/histomorphometric analysis.

| Statistical analysis
Analysis of mouse skin sections was performed blinded by the investigators. Statistical analysis was performed with PRISM software v5 (GraphPad Software Inc). Mann-Whitney U test was used for comparison between two different groups. *P > .05, **P > .01, *P > .001.

| FGF receptors 1, 2 and 3 are expressed in the epidermis and in cultured keratinocytes
To determine whether the loss of FGFR1 and FGFR2 in keratino-

| Loss of FGFR3 in keratinocytes does not affect epidermal homeostasis and wound healing
To determine the functional relevance of FGFR3 in keratinocytes, we generated mice lacking FGFR3 in this cell type (desig-  Figure S2F).

| Generation of mice lacking all FGF receptors in keratinocytes
We next determined whether loss of FGFR3 aggravates the phenotype of mice lacking FGFR1 and FGFR2 in keratinocytes by generation of triple conditional knockout mice (K5-R1/R2/R3 mice; see breeding scheme in Figure 1A). K5-R1/R2/R3 mice were born with the expected Mendelian ratio. qRT-PCR analysis of RNA from iso- There was no compensatory up-regulation of FGFR3 in K5-R1/R2 mice, but rather a reduced expression of this receptor ( Figure 1B).
Surprisingly, expression levels of FGFR3 were highly variable in keratinocytes with wild-type FGFR3 alleles ( Figure 1B) but this did not correlate with obvious phenotypic differences.
The complete loss of FGFR signalling in cultured keratinocytes from K5-R1/R2/R3 mice was confirmed by proliferation studies with FGF1, which activates all FGFR variants. 22 Whereas cells from control mice incorporated BrdU in response to FGF1 and to epidermal growth factor (EGF)-used here as positive controlkeratinocytes from K5-R1/R2/R3 mice only responded to EGF ( Figure 1D).  Figure 2B).

| Loss of FGFR3 in keratinocytes
Histological analysis of the skin showed the characteristic phenotype of K5-R1/R2 mice with epidermal thickening. 4 Again, this was not obviously aggravated by the additional loss of FGFR3 ( Figure 2C). In addition, keratinocyte proliferation and differentiation as revealed by Ki67 immunohistochemistry ( Figure 2D) or immunofluorescence analyses for differentiation-specific proteins ( Figure S3) were similar in K5-R1/R2/R3 and K5-R1/R2 mice. As previously shown, 4 K5-R1/R2 mice exhibited interfollicular expression of keratin 6 (K6), a sign for abnormal keratinocyte differentiation and enhanced proliferation. 23 This was also not further aggravated by the loss of FGFR3 and also not detected in mice lacking FGFR3 alone, and the other differentiation-specific keratins as well as the late differentiation marker loricrin were normally expressed in K5-R3, K5-R1/R2 and K5-R1/R2/R3 mice ( Figure S3). was equally low in control vs. K5-R3 mice ( Figure 3C, and Figure   S4 for flow cytometry), and the number of toluidine blue-positive mast cells was even mildly reduced by the loss of FGFR3 ( Figure 3D).
With the exception of a significant increase in epidermal γδ T cells in K5-R1/R2/R3 vs K5-R1/R2 mice, the additional loss of FGFR3 did not further affect the number of the other immune cell types ( Figure 3C). Consistent with the similar number of immune cells in K5-R1/R2 and K5-R1/R2/R3 mice, the genes encoding the pro-inflammatory cytokines IL-36β (=IL1-F8) (Il1f8), S100A8 (S100a8) and S100A9 (S100a9) were expressed at similar levels in mice of both genotypes ( Figure 3E).

| FGFR deficiency in the epidermis causes dermal fibrosis
An additional hallmark of the phenotype of K5-R1/R2 mice, which we had previously not characterized, is the development of skin fibrosis. This is reflected by the enhanced dermal thickness and the presence of a dense connective tissue characterized by high levels of collagen, which had replaced the adipose tissue ( Figure 4A). The  Figure 4E).
Taken together, these results demonstrate that loss of FGFR1/ R2 in keratinocytes causes skin fibrosis, which is further aggravated upon loss of FGFR3.

| D ISCUSS I ON
Epidermal FGFR signalling plays a key role in skin homeostasis, repair and disease, but the contribution of the individual FGF receptors to different skin functions has remained largely unknown. Here, we show that FGFR2 is the major FGFR in the murine epidermis, whereas FGFR3 has only a minor supporting role. FGFR3 in keratinocytes is even dispensable for wound healing, in spite of its up-regulation in wounded mouse skin. 24 This is surprising, as activating mutations in this type of receptor cause acanthosis nigricans, seborrhoeic keratosis and epidermal naevi. [10][11][12][13] However, our findings are consistent with data from human keratinocytes demonstrating that knock-down of FGFR3 does not affect normal keratinocyte growth in vitro, 25 even though this receptor is strongly expressed in the human epidermis. 26 These results point to a minor role of  In spite of the keratinocyte-specific deletion of FGF receptors, dermal fibrosis was seen in K5-R1/R2 mice and was more severe in K5-R1/R2/R3 mice. This may well be a consequence of the increase in immune cells that occurs in K5-R1/R2 and also in K5-R1/R2/R3 mice. However, additional loss of either mast cells or γδ T cells, two immune cell populations that are strongly increased in K5-R1/R2 and in K5-R1/R2/R3 mice, did not affect the dermal or epidermal phenotype in K5-R1/R2 mice. 6,7 These findings suggest that fibroblasts are activated as a consequence of the epidermal abnormalities.
Consistent with this assumption, it has been shown that a defect in the epidermal barrier results in up-regulation of cytokines, such as S100A8, S100A9 and S100A12, [28][29][30] which in turn cause fibroblast activation and fibrosis. S100A8 and S100A9 are also up-regulated in K5-R1/R2 4,8 and to a similar extent in K5-R1/R2/R3 mice, and these cytokines likely contribute to the dermal fibrosis. Furthermore, we show here that expression of follistatin is strongly down-regulated in the epidermis of K5-R1/R2 and K5-R1/R2/R3 vs control mice, which is likely to result in higher levels of bioactive activin, a potent pro-fibrotic factor. 31,32 Therefore, the epidermal abnormalities seen in our FGFR mutant mice and also in patients with atopic dermatitis are likely to contribute to the development of skin fibrosis. Thus, amelioration of the epidermal alterations and the resulting inflammation is of crucial importance for the prevention of fibroblast activation, and activation of FGFR signalling in keratinocytes may be a promising strategy to improve the epidermal barrier and even to prevent/ ameliorate the resulting dermal fibrosis.

ACK N OWLED G EM ENTS
We thank Drs.

CO N FLI C T O F I NTE R E S T
The authors confirm that there are no conflicts of interest.