Keratin 17 promotes epithelial proliferation and tumor growth by polarizing the immune response in skin.

Basaloid skin tumors, including basal cell carcinoma (BCC) and basaloid follicular hamartoma, are associated with aberrant Hedgehog (Hh) signaling and, in the case of BCC, an expanding set of genetic variants including keratin 5 (encoded by KRT5), an intermediate filament-forming protein. We here show that genetic ablation of keratin 17 (Krt17) protein, which is induced in basaloid skin tumors and co-polymerizes with Krt5 in vivo, delays basaloid follicular hamartoma tumor initiation and growth in mice with constitutive Hh signaling in epidermis. This delay is preceded by a reduced inflammation and a polarization of inflammatory cytokines from a Th1- and Th17-dominated profile to a Th2-dominated profile. Absence of Krt17 also attenuates hyperplasia and inflammation in models of acute dermatitis. Re-expression of Krt17 in Gli2(tg); Krt17(-/-) keratinocytes induces select Th1 chemokines that have established roles in BCC. Our findings establish an immunomodulatory role for Krt17 in Hh driven basaloid skin tumors that could impact additional tumor settings, psoriasis and wound repair.

Main Text

Gli2 mice, in which the bovine K5 promoter drives the constitutive expression of mouse Gli26, develop BCC and BFH6,7, which are both linked to deregulated Hh signaling in humans7,8. Gli2 mice show a reproducible pattern of lesions on the ear that successively involves hyper-keratosis (flaking), thickening and hyperpigmentation (Supplemental Fig. 1a). Mice were scored as positive for the onset of lesions upon the first sign of macroscopic hyperkeratosis in ear tissue. Histologically, the lesions present between P80 and P120 resemble BFH as described 7,8. By P180, larger, nodular, BCC-like tumors frequently occur deeper in the dermis (Supplemental Fig. 1a). Male Gli2 mice consistently develop lesions earlier than females (Supplemental Fig. 1b). Induction of K17, a Gli target gene9, is the main alteration in keratin expression prior to onset of lesions in Gli2 epidermis (Supplemental Fig. 1c).

Gli2 and K17 mice6,10 were interbred so as to assess the impact of K17 loss on genesis of BFH-like tumors. Appearance and progression of hamartoma-like lesions were captured from P30 to P125. At P80, epithelial lesions are clearly less pronounced in Gli2 than in Gli2 ear tissue (Fig. 1a; male data shown). In male Gli2 and Gli2 mice the average onset of lesions is 65±2 days (n=32) and 91±2 days (n=31; p< 0.01), respectively. In females, onset is at 80±5 days (Gli2tg; n=22) vs. 101±2 days (Gli2tg/K17−/−; n=21; p< 0.01)(Fig. 1b; Supplemental Fig. 2a). Gli2 mice lacking K1411 do not display such a delay (Fig. 1c), establishing specificity. Gli2 transgene expression is similar in both genotypes (Fig. 1d). Loss of K17 does not impact Gli2 subcellular localization or hedgehog signaling (Supplemental Fig. 2, b–d). Therefore, the absence of K17 causes a delay in the inception of BFH-like skin tumors in Gli2 mice.

Figure 1

Absence of K17 delays the onset of ear lesions, and epidermal hyperplasia, in Gli2 mice. (a) Age-matched P80 Gli2 and Gli2 male mice. Left, pictures of intact ear. Box highlights lesional tissue in Gli2 mice. Arrows point to blood vessels, prominent in Gli2 mice. Right, hematoxylin-eosin stained ear tissue section, showing expansion of epidermis (epi). (b), Mean age (± s.e.m.) of onset of macroscopic ear lesions in Gli2 and Gli2 mice, stratified by gender. (c) Percentage of mice with ear lesions at P80 in the Gli2, Gli2, and Gli2 strains of mice. (d) RT-PCR assay of levels of Gli2 transgene expression in Gli2 and Gli2 mice (GAPDH: loading control). (e) Immunostaining for BrdU in ear tissue of P80 male Gli2 and Gli2 mice.epi, epidermis; derm, dermis. (f) Quantitation of BrdU-positive keratinocytes/mm of epidermis seen in (e). (g, h) Immunostaining for phospho-Histone H3 (g), marking mitotic activity, and TUNEL staining (h), detecting apoptotic cells, in P80 male Gli2 and Gli2. Scale bars: a (50µm), e,g,h (20µm).

Histological anomalies common to Hh pathway-activated mouse skin7 were scored in Gli2 and Gli2 ear tissue (Supplementary Fig. 3a). Such anomalies, absent in wildtype and K17 mouse ears (Supplementary Fig. 3b), are prominent in Gli2 ear but markedly reduced in Gli2 ear (Supplementary Fig. 3c). Overall tissue thickness and penetration of epithelial downgrowths are also reduced in Gli2 ear tissue (Supplementary Fig. 3d,e). K17, K5, and K14 are uniformly distributed in the lesional epithelium (Supplemental Fig. 3f). Co-assembly of K5 and K17 in Gli2 lesional epithelium is conveyed by their co-localization and co-immunoprecipitation (Supplemental Fig. 3g,h). The wound-inducible K6α, K6β and K16, absent in intact epidermis, are induced in the upper layers of thickened Gli2 epidermis, preferentially, but are markedly reduced in Gli2 skin (Supplemental Fig. 3f,i).

Reduced proliferation, rather than increased cell death, is a key contributor to delayed tumor onset in Gli2 skin. Relative to Gli2, indeed, the frequency of mitotically-active cells is depressed by > 3-fold in Gli2 ear epithelium (Fig. 1e–g). In contrast, TUNEL-positive, apoptotic cells are restricted to the upper epidermis of lesional skin and show similar density in both genotypes (Fig. 1h).

Inflammation has emerged as a driver of angiogenesis and tumor growth12 and coincides with K17 induction and loss of barrier function in several skin diseases13,14. Immunoreactivity for markers of innate immune cells (CD11b), T cells (Thy-1), and phagocytes (iNOS) are enhanced in Gli2 compared to Gli2 ear skin of p80 male mice (Fig. 2a). PECAM staining is also decreased in Gli2 ear skin (Fig. 2a), reflecting decreased angiogenesis. Myeloperoxidase (MPO) enzymatic activity, inherent to neutrophils15, is increased 17.4 ± 0.5 fold in P80 male Gli2 ear tissue but only 5.8 ± 0.1 fold in Gli2 males (data normalized to P80 female Gli2 ear; Fig. 2b). Female Gli2 mice also show a reduced level of MPO activity at P80, being at 0.75 fold of that seen in Gli2 controls (Fig. 2b). Skin barrier integrity, assessed via a whole-mount dye penetration assay16, is intact as expected in P70 wildtype ear skin (Fig. 2c). In contrast, a sizable portion of the ear is dye-permeable in P70 Gli2 mice (Fig. 2c); again, this readout is markedly decreased in Gli2 mice (Fig. 2c).

Figure 2

Role of inflammation in the onset of ear lesions. (a) Immunodetection of infiltrating immune cells and vasculature in Gli2 and Gli2 male mice at P80 using antibodies to CD11b, Thy-1, iNOS, and PECAM-1 (see arrows). Labeling key provided in lower left corner. (b) Quantification of myeloperoxidase activity (MPO; mean ± s.e.m.) in ear tissue of mice at P80, normalized to female Gli2 mice. (c) In situ beta-galactosidase staining in P80 male ear tissue of various genotypes. Blue staining reflects loss of barrier integrity. (d) Hematoxylin-eosin stained ear tissue of male mice at P40. (e) Myeloperoxidase activity in ear tissue of P40 male mice, normalized to female Gli2 (mean ± s.e.m.). (f) Quantification of BrdU labeled cells/um of epidermis in P40 male mice. (g) Immunostaining for IL-1β in the epidermis (epi) of P80 male ear tissue. Scale bars: a (20µm), c (25µm), d (50µm).

At P40, i.e., prior to onset of histological anomalies (Fig. 2d), MPO activity is 5.9±1.9 fold greater in male Gli2 mice relative to females, substantiating the gender bias in this model. MPO activity is lower in P40 male Gli2 mice (0.55 ± 0.20 relative to female Gli2 mice; Fig. 2e). While epidermal thickness is the same (Fig. 2d), mitotic activity is higher in Gli2 vs. Gli2 epidermis at P40 (0.52 ± 0.01 vs. 0.14 ± 0.02 BrdU labeled cells/mm of epidermis) (Fig. 2f) and skin tissue is infiltrated with various types of leukocytes. Barrier integrity is mildy compromised in P40 Gli2 ear skin, is again better preserved in Gli2 mice (Supplemental Fig. 4a). Thus, the marked reductions in inflammation and hyperplasia that define Gli2 ear skin occur as early as P40, ahead of progression to overt tumorigenesis in the Gli2 model.

Expression of inflammatory cytokines and chemokines was examined via qRT-PCR in ear tissue at P40 and P80. The findings are stratified according to specific classes of T-helper cytokines: Th1 (cellular immunity; generally “pro-inflammatory”), Th2 (humoral immunity; “anti-inflammatory”), and Th17 (anti-microbial immunity at epithelial barriers)17,18. Th1 and Th17 hyperactivity occur in psoriasis19. Absence of K17 in Gli2 skin correlates with a marked reduction in Th1- and Th17-related markers and induction of Th2-related markers (Table 1), many of which are prominently expressed by skin keratinocytes themselves. Expression of IL-1β, a keratinocyte mitogen20, is ~10 fold higher in Gli2 compared to Gli2 skin (Table 1). Immunostaining shows that IL-1β epitopes are strongly expressed in the skin epithelium (Fig. 2g). Spp1 (osteopontin), which acts to bias immune responses toward Th121, is reduced by ~15 fold and IL-6, associated with the acute phase response and upregulated in human BCC22, is lowered ~17 fold in Gli2 skin (Table1). The matrix metalloproteases MMP3, MMP9 and MMP13, whose expression is enhanced in BCC23, are downregulated in Gli2 ear tissue. Classical Th2 type cytokines primarily secreted by T-cells, e.g., IL-4 and IL-10, are modestly altered whereas Ccl24 and Ccr4, expressed by skin keratinocytes24, are respectively ~9 and ~3 fold higher in Gli2 ear tissue (Table 1). The expression of many of these cytokines and chemokines is already altered by P40. IL1β and Cxcl5 expression is enhanced in the presence of K17, while the Th2 markers IL20 and IL4 are enhanced in its absence (Table 1). Thus, the immunomodulatory influence of K17 is first manifested at an early stage in this model.

Table 1

Comparing the inflammatory and immune response in ear lesions, in male Gli2 relative to male Gli2 mice, at P80 and P40. The fold change reported represents alterations in mRNA levels due to loss of K17. Values reflect compiled data from three experiments involving distinct pools of cDNAs.

Postnatal day 80
Cytokine/Chemokine (Th1)	Fold Change	P-Value
Spp1	−14.83	0.013
Ccl3	−14.80	0.003
Cxcl5	−10.56	0.007
IL1β	−10.20	0.009
Ccl4	−8.78	0.016
Ccr1	−4.92	0.014
Cxcr2	−3.53	0.051
Ccr5	−2.53	0.054
Cxcl10	−1.73	0.036
Ccl5	−1.69	0.022
TNFα	−1.18	0.076
IFNγ	1.31	0.440
Cytokine/Chemokine (Th2)	Fold Change	P-Value
Ccl24	8.85	0.008
Ccl17	4.21	0.000
Ccr4	3.24	0.004
Ccl22	3.09	0.007
Ccl1	2.97	0.003
Ccl11	2.17	0.046
IL13	2.11	0,004
IL15	1.62	0.059
IL4	1.20	0.100
IL20	−1.06	0.900
Cytokine/Chemokine (Th17)	Fold Change	P-Value
Mmp13	−25.06	0.001
Csf3	−19.57	0.003
IL6	−17.12	0.001
Cxcl2	−9.99	0.017
Cxcl5	−9.37	0.001
Cxcl1	−6.52	0.002
Syk	−6.46	0.043
Mmp9	−5.62	0.004
Clec7a	−4.66	0.005
Mmp3	−2.67	0.023
IL10	−2.46	0.005
Cd3g	4.51	0.069
IL25	4.01	0.048
Cd3d	3.56	0.021
IL5	2.51	0.012
IL15	1.63	0.035
Postnatal day 40
Cytokine/Chemokine (Th1)	Fold Change	P-Value
IL1β	−4.21	0.028
Cxcl5	−3.52	0.005
Ccr1	−2.96	0.001
Cxcr2	−1.98	0.019
Ccl3	−1.51	0.233
Cytokine/Chemokine (Th2)	Fold Change	0.014
IL20	12.66	0.090
IL4	5.24	0.121
IL13	5.03	0.110
Ccl24	1.59	0.258
Ccl17	1.11	0.076

Topical application of the phorbol ester, 12-O-tetradecanoylphorbol-13-acetate (TPA)25, to ear skin induces acute inflammation and epidermal proliferation (Fig. 3a,b), providing a tumor-free, dermatitis-like setting in which to assess the impact of K17 loss. The latter curtails hyperplasia-driven epidermal thickening (wt ear tissue: 34.1±2.3 µm in TPA- vs. 10.4±0.3 µm in vehicle-treated; K17 ear tissue: 18.7± 0.8 µm in TPA- vs. 10.6±0.8 µm in vehicle-treated; Fig. 3a,b). Markers related to compromised skin barrier function (S100A826, thymic stromal lymphoprotein (TSLP)14, β-defensin 27) show elevated mRNA levels in TPA-treated wildtype skin (Fig. 3c). TSLP and β–defensin are markedly attenuated in K17 skin (Fig. 3c), suggesting better retention of barrier function. A partial shift toward a Th2-dominated cytokine profile is seen in TPA-treated K17 skin, though the magnitude of the changes is less than in Gli2 skin. The Th1 chemokines Cxcl5, Ccl3 and IL-1β are reduced 2.4-, 3.0- and 1.7-fold, respectively, and the Th2 cytokine IL-20 is 7.1 fold higher in TPA-treated K17 skin relative to control (Table 2). Thus, the K17 status exerts a similar immunomodulatory influence in acute dermatitis.

Figure 3

Absence of K17 blunts epidermal hyperplasia and alters inflammation in a chemical model of dermatitis. Wildtype and K17 mouse ears were treated with acetone (vehicle control) or TPA. (a) Hematoxylin-eosin stained tissue sections depicting the effect of TPA treatment on thickness of epidermis. Right panel: Expansion of basal layer visualized by immunostaining for K14. Scale bars: 20 µm. (b) Epidermal thickness (mean ± s.e.m.), as conveyed by K14 staining, in vehicle (−) and TPA-treated (+) male mouse ears. (c) Right: Semi-quantitative RT-PCR survey of targets associated with loss of barrier integrity. Left: Quantitation of RT-PCR results shown in c. (d) Cytokine/chemokine expression in primary cultures of Gli2 and Gli2 keratinocytes 12 hours after TPA. For d and e, fold change represents changes due to loss of K17 (compilation of 3 assays involving distinct pools of mRNAs). (f) Changes in cytokine and chemokine expression in Gli2 after reintroduction of K17 via transfection (triplicate).

Table 2

Comparing the inflammatory and immune response in TPA-treated ear tissue from K17 and wildtype mice. The fold change reported represents alterations in mRNA levels due to loss of K17. Values reflect compiled data from three experiments involving distinct pools of cDNAs.

Cytokine/Chemokine (Th1)	Fold Change	P-Value
Ccl3	−2.97	0.001
Cxcl5	−2.44	0.002
Cxcl1	−1.90	0.244
Ccl4	−1.86	0.009
IL1β	−1.70	0.010
Cxcl9	−1.20	0.070
IFNγ	1.48	0.355
Cytokine/Chemokine (Th2)	Fold Change	P-Value
IL20	7.10	0.003
Ccl22	1.88	0.004
IL15	1.40	0.001
IL4	1.20	0.100
IL10	1.11	0.356

Skin keratinocytes from Gli2 and Gli2 newborn mice were seeded for primary culture (48h), and treated with TPA (12h) to assess whether key changes in cytokine/chemokine expression are keratinocyte-autonomous. Under basal conditions, Gli2 and Gli2 cells show rates of proliferation similar to wt and K17 ones. TPA induces a two-fold enhancement in Gli2 keratinocyte proliferation by 12h, whereas Gli2 cells are unchanged (Supplemental Fig. 4b,c). Again, key chemokines are differentially expressed depending on K17 status. Levels of Cxcl11, Cxcl5, CxCl9 and Cxcl10 mRNAs, among others, are significantly lower in TPA-treated Gli2 keratinocytes (Fig. 3d and Supplemental Fig. 4d). These chemokines promote keratinocyte proliferation in skin tumors, and show a tight spatial correlation with K17 expression28,29. Re-expression of K17 into TPA-treated Gli2 keratinocytes markedly elevates the levels of Cxcl5, CxCl9, and CxCl11 mRNAs, relative to mock-transfected cells (Fig. 3d; Supplemental Fig. 4d). Thus, K17 impacts the TPA-induced expression of select chemokines relevant to BCC pathogenesis in both adult epidermis in situ and isolated newborn keratinocytes in culture, suggesting that the mechanism(s) involved are in part cell-autonomous.

Several NF-kB target genes show a modest but consistent reduction in their expression in Gli2 relative to Gli2 keratinocytes in TPA-treated cultures (Supplemental Figure 5a). This is consistent with the prominent role of NF-κB in skin inflammatory conditions 30 and, in particular, with its impact on Cxcl5, CxCl9, and CxCl11 expression31–33. Similar analyses of P80 whole ear skin tissue revealed no difference between the genotypes, likely reflecting the large complexity of these lesions in situ and the occurrence of secondary or compensatory changes (Supplemental Figure 5, b–c). Besides, K17 has been shown to promote anagen growth during hair follicle cycling34 and stimulates protein synthesis during tissue repair35. The phenotype reported here cannot be correlated to obvious alterations in these roles, again as inferred from analyses of skin tissue sections (data not shown) or extracts (Supplemental Fig. 5, b–d).

K17 is ectopically expressed in numerous settings associated with robust inflammation including cutaneous wounds, various carcinomas, psoriasis, and virus-induced warts10. High levels of K17 expression correlate with a poor prognosis in breast36 and pancreatic37 cancers – whether this phenomenon is related to altered inflammatory signatures represents an issue of interest. There exists a correlation between Th1 hyperactivity and K17 expression in psoriatic plaques19; plaque resolution coincides with a shift to a Th2 response and loss of K17 expression. We posit that the presence of K17 in epidermis (and related epithelia) promotes hyperplasia in BCC-like tumors (this study) and likely in additional tumors and inflammatory disease settings in part through its ability to promote a specific type of inflammatory response. Normal contexts in which prominent K17 expression is not correlated to local inflammation (e.g., hair follicles) may benefit from an immune-privileged status38 or reflect its regulation via post-translational modifications or interaction with other proteins34,35. A role for K17 as an immunomodulator, whether direct or indirect, provides a novel way of conceiving how SNPs affecting K5 influence the risk of developing BCC 2, and makes these keratins potentially attractive target for novel therapies aimed at curtailing conditions driven by or linked to chronic inflammation.

METHODS

Methods and any associated references are available in the online version of the paper at http://www.nature.com/naturegenetics/.