INTU is essential for oncogenic Hh signaling through regulating primary cilia formation in basal cell carcinoma (BCC)

Abstract

Inturned (INTU), a cilia and planar polarity effector (CPLANE), performs prominent ciliogenic functions during morphogenesis, such as in the skin. INTU is expressed in adult tissues but its role in tissue maintenance is unknown. Here, we report that the expression of the INTU gene is aberrantly elevated in human basal cell carcinoma (BCC), coinciding with increased primary cilia formation and activated hedgehog (Hh) signaling. Disrupting Intu in an oncogenic mutant Smo (SmoM2)-driven BCC mouse model prevented the formation of BCC through suppressing primary cilia formation and Hh signaling, suggesting that Intu performs a permissive role during BCC formation. INTU is essential for IFT-A complex assembly during ciliogenesis. To further determine whether Intu is directly involved in the activation of Hh signaling downstream of ciliogenesis, we examined the Hh signaling pathway in mouse embryonic fibroblasts, which readily respond to Hh pathway activation. Depleting Intu blocked SAG-induced Hh pathway activation, whereas the expression of Gli2ΔN, a constitutively active Gli2, restored Hh pathway activation in Intu-deficient cells, suggesting that INTU functions upstream of Gli2 activation. In contrast, overexpressing Intu did not promote ciliogenesis or Hh signaling. Taken together, data obtained from this study suggest that INTU is indispensable during BCC tumorigenesis and that its aberrant upregulation is likely a prerequisite for primary cilia formation during Hh-dependent tumorigenesis.

INTRODUCTION

Inturned (INTU) is a tissue-specific planar cell polarity (PCP) effector. It mediates the polarity cues set up by the core PCP components in Drosophila and vertebrate animals ¹. During vertebrate morphogenesis, INTU also exerts prominent functions in primary cilia formation. Disrupting the Intu gene in mice can cause defects in neural tube formation, limb patterning, and hair follicle morphogenesis, all of which are associated with impaired cilia formation or function ^2–6. Mechanistically, INTU forms a molecular module with the other two PCP effectors, Fuzzy (FUZ) and WDPCP, to facilitate the assembly of intraflagellar transport A (IFT-A) peripheral proteins into the IFT-A complex to facilitate cilia formation during vertebrate morphogenesis ⁷. Because of their prominent functions in PCP and ciliogenesis, these PCP effectors are recognized as cilia and planar polarity effectors (CPLANEs). However, their roles in tissue homeostasis are unclear.

Basal cell carcinoma (BCC) of the skin is the most common malignancy in the United States, with more than 2.8 million new diagnoses annually. BCC originates from epidermal keratinocytes and is caused predominantly by loss-of-function mutations in the patched 1 (PTCH1) ^{8, 9} or gain-of-function mutations in the smoothened (SMO) gene ¹⁰, which encodes upstream transmembrane components of the hedgehog (Hh) signaling pathway. Oncogenic mutations in PTCH1 or SMO ultimately lead to uncontrolled activation of the glioma-associated oncogene (Gli) transcription factors, thus, the Hh signaling pathway ¹¹.

The activation of the Hh signaling pathway is dependent on the primary cilium, a cellular organelle that is essential for the translocation of Hh pathway components and the subsequent activation of the Hh target genes, such as Gli1 and Ptch1 ^{12, 13}. The functional importance of primary cilia for the Hh signaling pathway is well-established during both morphogenesis ^13–15 and the formation of Hh-dependent tumors, such as medulloblastoma ¹⁶ and BCC ¹⁷.

To gain insight into the function of the CPLANEs during mammalian tissue homeostasis, we explored the functions of INTU in the context of BCC formation. We found that the INTU gene is aberrantly upregulated in human BCC specimens and that genetically ablating the endogenous Intu in a BCC mouse model blocked BCC formation, likely upstream of the activation of Gli transcription factors. These observations suggest that INTU, which is essential for ciliogenesis, plays a permissive role during the development of BCC.

RESULTS

INTU is upregulated in human BCC

To determine whether INTU is involved in epidermal homeostasis, we constructed a tissue microarray (TMA) comprised of healthy human skin, BCC, and squamous cell carcinoma (SCC) specimens (Supplementary Fig. 1a), and examined the transcriptional levels of the endogenous INTU gene by in situ hybridization. In healthy adult epidermis, INTU is predominantly expressed in interfollicular and follicular epidermal keratinocytes at a relatively modest level in comparison to embryonic skin (Figure 1a and b and Supplementary Figure 1b). Strikingly, INTU expression was markedly elevated in BCC tumors (n = 19, P < 0.05), but not in SCCs (n = 3), when compared to healthy adult skins (n = 3) (Figure 1a and b). As expected, the transcription level of PTCH1, an Hh responsive genes, was also significantly elevated in BCC (P < 0.001), but not SCC, a type of keratinocyte carcinoma that is not dependent on Hh signaling (Figure 1a and c). GLI1, another Hh-responsive gene, exhibited a similar trend (Figure 1a and d). In contrast, the level of SMO, which is not an Hh responsive gene, was unchanged (Figure 1a and e). Moreover, the relative expression levels of INTU correlated with that of PTCH1 (Spearman r = 0.784, P < 0.001) or GLI1 (Spearman r = 0.858, P < 0.001), but not with SMO (Supplementary Figure 2a – c), suggesting that elevated level of INTU is associated with activation of the Hh signaling pathway in human skin.

Figure 1.

INTU, Hh signaling, and primary cilia in healthy skin, BCC, and SCC. (a) Representative H&E staining and in situ hybridization of INTU, PTCH1, GLI1, and SMO in healthy adult human skin (normal skin), BCC, and SCC. (b – e) Normalized scores of signal intensity of in situ hybridization in (a). (f) Representative images of primary cilia (ARL13B, green) and basal body (γ-tubulin (TUB), red) in normal skin, BCC, and SCC. Nuclei were stained with DAPI (blue). Lower panels are enlarged boxed areas above. (g) Percentage of ciliated keratinocytes in specimens in (f). n = 3 for normal adult skin or SCC, n = 19 for BCC. *, P < 0.05; ***, P < 0.001, two-way ANOVA and pair-wise posttest. Scale bar = 25 μm.

Human BCCs comprise abundant ciliated keratinocytes

It was reported that the primary cilia were frequently observed in BCC ¹⁷. To determine the prevalence of ciliated cells in human BCC, we evaluated the abundance of primary cilia in the same cohort of human skin specimens described above by immunofluorescence labeling. A significantly increased proportion (38.5 ± 13.3%) of BCC cells possesses a primary cilium in comparison to keratinocytes (2.5 ± 2.2%) in healthy skins or SCCs (2.4 ± 0.6%) (Figure 1f and g, P < 0.001), suggesting that the primary cilium may be a biomarker for BCC. Moreover, the abundance of ciliated cells also appeared correlated with the relative expression levels of INTU (Spearman r = 0.668, P < 0.001, Supplementary Figure 2d). This observation suggested that the ciliogenic program is likely upregulated in BCC cells, which may be related to the elevated levels of INTU and the activated Hh signaling pathway.

Intu is essential for BCC formation

INTU is essential for primary cilia formation and the activation of the Hh signaling pathway during skin morphogenesis ⁶. Elevated INTU expression, along with increased primary cilia formation and the aberrant activation of the Hh pathway, in BCC suggests that INTU may be functionally required for BCC formation. To test this idea, we engineered a BCC mouse model (Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox), which permits the simultaneous expression of an oncogenic mutant SmoM2 (SmoM2^cond) and the ablation of the endogenous Intu locus (Intu^flox/flox) through tamoxifen-induced activation of Cre recombinase in keratinocytes.

Initial tamoxifen treatment was carried out on 3-week old littermates. By five weeks, control mice that were heterozygous for the conditional allele of Intu (Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox) started to exhibit signs of various skin abnormalities, including hair loss, scaly skin, bumpy tails, and crumpled ears (data not shown). By 10 weeks, these skin phenotypes exacerbated and progressed to extensive skin erosion (Figure 2a and b), thus, the control mice had to be euthanized. Histologically, dorsal skins of control mice revealed remarkable downward growth of hyperplastic keratinocytes (Figure 2b). Consistent with previous reports ^17–19, these cells exhibited features of human BCC, such as basaloid appearance, high nuclear to cytoplasmic ratio, and palisading nuclei (Supplementary Figure 3a). Immunofluorescence labeling confirmed that these cells coexpressed KRT14 and KRT17 (Supplementary Figure 3b), which is characteristic of BCC. Moreover, these BCC-like cells were positive for yellow fluorescence protein (YFP, Figure 2c, upper and middle panels), which was coexpressed with SmoM2 as a fusion protein, thereby confirming that they were indeed derived from cells expressing SmoM2. Overall, these BCC-like lesions were reminiscent of human BCC. In situ hybridization confirmed that the endogenous Intu is abundantly expressed in these lesions (Figure 2c).

Figure 2.

Disrupting Intu blocks BCC formation in mice. (a) Disease-free survival of SmoM2-driven BCC mouse models. Control, Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox (n=9); homozygous Intu mutants, Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox (n = 14). ***, P < 0.0001, Logrank test. (b) Representative gross appearance (upper panels) and H&E staining (lower panels) of the dorsal skin harvested 10 weeks after tamoxifen induction. Arrow points to a focal epidermal invagination. (c) Immunofluorescence images of YFP (green) in dorsal skin of control and homozygous mouse models in (B). KRT14 was labeled red. Nuclei stained with DAPI (blue). (d) Thickness of interfollicular epidermis (IFE) of specimens in (B). Mean ± SD. ***, P < 0.001, posttest, two-way ANOVA. (e) Percentage of BrdU and KRT14 double positive cells in specimens in (B). Mean ± SD. ***, P < 0.001, T-Test. Scale bar = 50 μm.

In contrast, littermates homozygous for the conditional allele of Intu (Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox) did not show gross signs of BCC formation even by 22 weeks (or 154 days) after tamoxifen induction (Figure 2a and b upper panels). Thus, disease-free survival of these mutant mice homozygous for Intu was significantly different than that of the control littermates (Logrank test P < 0.0001, Figure 2a). Histologically, the dorsal skins of homozygous mice exhibited rather mild hyperplasia and focal dysplasia (Figure 2b, lower panels, arrow). YFP and KRT14 staining confirmed that keratinocytes of these mice expressed oncogenic SmoM2 (Figure 2c), whereas in situ hybridization demonstrated markedly diminished expression of Intu (Figure 2c, lower panels), suggesting that Intu is required for the formation of oncogenic SmoM2-driven BCC.

Histologically, Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox mice exhibited consistently thinner interfollicular epidermis (IFE) in comparison to controls (Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox) (Figure 2d, P < 0.001). The thickness of IFE of BCC mouse models homozygous for Intu conditional allele remained constant up to 22 weeks after tamoxifen induction (Figure 2d), suggesting genetically ablating Intu blocked, rather than delayed, BCC formation.

Moreover, the epidermis of Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox mice maintained normal KRT17 expressing pattern (Supplementary Figure 3b). BrdU labeling demonstrated that Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox skins harbored substantially more proliferating keratinocytes (KRT14-positive) than those of Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox mice (7.06 ± 0.49% vs 1.61 ± 0.41%, Figure 2e and Supplementary Figure 3c). Taken together, these data demonstrated that genetically ablating Intu in the epidermis can protect mice from developing BCC-like lesions induced by the oncogenic SmoM2, suggesting that Intu plays a permissive role during BCC formation.

Intu is required for primary cilia formation in BCC

Primary cilia are abundant in human BCC (Figure 1f and g) and are essential for Hh signaling. Therefore, it was speculated that cells in the BCC-like lesions in control mice are likely to be ciliated. Indeed, BCC-like lesions in Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox mice contained abundant ciliated keratinocytes (34.9 ± 6.7%, Figure 3a), at a proportion similar to what was observed in human BCC (38.5 ± 13.3%, Figure 1g). These findings not only suggest that the SmoM2 mouse models are able to recapitulate cellular features of human BCC, but also suggest that upregulated cilia formation is an intrinsic feature of BCC.

Figure 3.

Primary cilia, Hh and Wnt signaling in BCC mouse models. (a) Immunofluorescence labeling of primary cilia (ARL13B, green) and basal body (γ-tubulin (TUB), red) in dorsal skin of control (Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox, n = 4, and homozygous (Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox, n = 4) BCC mouse models and control mice (Krt14-Cre^ERT;Intu^+/flox, n = 3), and quantification of ciliated keratinocytes. A minimum of 150 keratinocytes were evaluated per mouse. ***, P < 0.001, two-way ANOVA and pair-wise posttest. Nuclei were stained with DAPI (blue). (b) In situ hybridization of Gli1 and Ptch1 in dorsal skin obtained from specimens in (a). (c) Immunofluorescence labeling of LEF1 (red) in dorsal skins obtained from specimens in (a). Keratinocytes were labeled with KRT14 (green). Nuclei were staining with DAPI (blue). Lower panels are enlarged boxed areas above. Scale bar = 25 μm.

Given the essential role of INTU in ciliogenesis, particularly in epidermal keratinocytes ⁶, we postulated that disrupting Intu might block BCC formation through impairing ciliogenesis. Indeed, homozygous Intu mutants (Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox) contained few ciliated epidermal keratinocytes (8.8 ± 1.4%), reminiscent of those not expressing SmoM2 (9.4 ± 1.1%, Krt14-Cre^ERT;Intu^+/flox, Figure 3a). These observations suggested that Intu might be required for BCC formation through facilitating ciliogenesis.

Intu is required for the activation of Hh signaling in BCC in vivo

Unrestricted activation of Hh signaling is the primary molecular driver for BCC. To determine whether Intu is required for the activation of the Hh signaling pathway during BCC formation, we examined Hh target genes (Gli1 and Ptch1) in the SmoM2-driven BCC mouse models. In situ hybridization demonstrated that Gli1 and Ptch1 transcripts were markedly elevated in BCC-like lesions of control mice (Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox), but attenuated in homozygous Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox mice (Figure 3b). In response to Hh pathway activation, canonical Wnt signaling is also activated in BCC. Indeed, nuclear LEF1, a marker for activated Wnt signaling, was expressed in cells of the BCC-like lesions of control (Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox) mice, but nearly undetectable in the skin of homozygous (Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox) mice, except in dermal papilla (DP) cells (Figure 3c). These data suggest that Intu is required for the activation of the Hh signaling pathway during the development of SmoM2-driven BCC in vivo.

Intu is required for SmoM2-mediated Hh pathway activation in keratinocytes in vitro

To further determine whether Intu is required for Hh pathway activation ensuing the expression of oncogenic Smo gene, we examined the expression of Hh target genes in primary keratinocytes harboring the inducible SmoM2 allele. Five days after 4-OH tamoxifen induction, Gli1 transcription was robustly upregulated in keratinocytes isolated from Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox mice in comparison to those isolated from control mice (SmoM2^cond;Intu^+/flox) (Figure 4a). Whereas such induction was essentially abolished in cells isolated from mice homozygous for the conditional Intu allele (Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox) (P < 0.05, Figure 4a). Quantification of the transcriptional levels of Intu confirmed that the attenuated induction of Gli1 was indeed associated with diminished levels of Intu (Figure 4a). These findings suggest that Intu is required for keratinocyte to active the Hh signaling pathway in response to the expression of oncogenic SmoM2.

Figure 4.

Hh signaling and primary cilia in primary epidermal keratinocytes expressing SmoM2. (a) Relative transcription levels of Gli1 and Intu in primary keratinocytes treated with 4-OH tamoxifen. n=3, *, P < 0.05, one-way ANOVA and posttest. (b) Representative immunofluorescence labeling of YFP (green) and primary cilia (acetylated α-tubulin (TUB), red, and ARL13B, blue) in 4-OH tamoxifen induced keratinocytes isolated from control (Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox) and homozygous (Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox) mouse models. n = 3 per genotype. Arrowhead points to a triple positive cilium in an induced cell; arrow points to a cilium in an uninduced cell. (c) Statistical analysis of ciliated cells based on the expression of YFP as shown in (b).

To determine whether the Hh phenotype of Intu-deficient keratinocytes is corroborated with impaired ciliogenesis, we examined the primary cilia by immunofluorescence. In 4-OH tamoxifen activated control cells (Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox, left, indicated by *), SmoM2 was activated as demonstrated by YFP. Primary cilium was readily detectable in 49.9 ± 2.8% control cells (Figure 4b – c). Moreover, SmoM2 appeared enriched at the primary cilium such that triple fluorescence labeling makes the cilium appear white (Figure 4b, indicated by an arrow head). In contrast, in keratinocytes homozygous for the conditional allele of Intu (Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox), 4-OH tamoxifen activated the expression of SmoM2, as demonstrated by YFP expression (Figure 4b, right, indicated by *), but the formation of primary cilium was only observed in 25.8 ± 1.7% cells (P < 0.05, Figure 4c). Interestingly, primary cilia formation (Figure 4b, right, indicated by an arrow) appeared unaffected in Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox or Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox cells that did not undergo Cre-mediated recombination, as judged by the lack of YFP expression (42.3 ± 8.9% and 47.1 ± 8.8%, respectively) (Figure 4c). The ciliated Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox cells (Figure 4b, right, indicated by p) presumably retained the endogenous Intu due to insufficient Cre-mediated recombination. These results suggested that Intu is essential for primary cilia formation irrespective of the expression of oncogenic SmoM2.

In summary, these in vitro experiments suggested that Intu is a key component of both ciliogenesis and the SmoM2-activated oncogenic Hh signaling.

Intu acts upstream of the Gli transcription factors

Data obtained so far in BCC was in line with prior findings that Intu is required for Hh pathway activation at the level of primary cilia. To examine whether Intu is also involved in the activation of the Hh signaling downstream of primary cilia, we examined Hh pathway in mouse embryonic fibroblasts (MEFs) that readily form primary cilia after serum starvation and respond to Hh pathway activation. First, we found that Intu^−/− MEFs exhibited strong primary cilia phenotypes such that ciliogenesis was almost completely suppressed (0.52 ± 0.53%) when compared with wild type MEFs (61.2 ± 1.9%) (P < 0.001, Fig. 5a). Moreover, the Smo Agonist (SAG)-mediated Hh pathway activation was severely attenuated in Intu^−/− MEFs in comparison to control wild type MEFs, as determined by the processing of full-length GLI3 (GLI3-FL) to GLI3 repressor (GLI3-R) (Figure 5b), as well as the transcription levels of Gli1 and Ptch1 (Figure 5c – d, columns indicated by V). These results suggest that Intu is required for Smo-mediated Hh pathway activation.

Figure 5.

Effects of the expression levels of Intu on ciliogenesis and Hh signaling. (a) Primary cilia (ARL13B, green) and basal body (γ-tubulin, red) in wild type (WT) and Intu^−/− mouse embryonic fibroblasts (MEFs). (b) The relative expression levels of full-length GLI3 (GLI3-FL) and GLI3 repressor (GLI3-R) in DMSO and SAG-treated WT and Intu^−/− MEFs, and quantification (right). n = 3. *, P < 0.05; **, P < 0.01; ***, P < 0.001, two-way ANOVA and pair-wise post analysis. (c – e) Expression of Gli1 (c), Ptch1 (d), and Gli2 (e) in WT and Intu^−/− MEFs in response to the overexpression of full-length Gli2 (Gli2-FL) and a constitutively active Gli2 (Gli2ΔN), and SAG in a representative experiment. Vehicle (V) treated cells were used as controls. n = 4. (f) Primary cilia (ARL13B, green) in WT MEFs transfected with mouse Intu-GFP cDNA (mIntu), and quantification (right). (g – h) Relative expression levels of Intu and Ptch1 in mIntu-transfected MEFs. Mock transfected cells were used as controls. n = 4. Scale bar = 10 μm.

To determine whether Intu is also involved in the molecular signaling machinery downstream of Smo, we used plasmids expressing inactive full-length Gli2 (Gli2-FL) and a constitutively active Gli2 (Gli2_ Δ1-413 or Gli2ΔN) to rescue the Hh pathway in Intu^−/− MEFs. First, Gli2ΔN, but not Gli2-FL, consistently elicited Hh pathway activation in wild type and Intu^−/− MEFs, as determined by Gli1 and Ptch1 transcription (Supplementary Figure 4, and Figure 5c – d, open bars), suggesting that INTU functions upstream of GLI2 and is dispensable for Gli2ΔN-induced Hh pathway activation. When this experiment was conducted in conjunction with SAG, Gli2ΔN was unable to further activate Hh signaling in Intu^−/− MEFs, as determined by Gli and Ptch1 transcription (Figure 5c – d, solid bars). This result conforms to the prior findings that Intu is required for SMO activation. The overexpression of Gli2-FL and Gli2ΔN was robust as confirmed in Figure 5e. Collectively, these findings suggested that INTU acts upstream of Gli2 to facilitate oncogenic SMO-mediated Hh pathway activation.

To further determine whether elevated INTU levels might be sufficient to drive ciliogenesis and Hh pathway activation, we overexpressed mouse Intu that was fused with GFP (mIntu) ⁵ in wild type MEFs. Intu was robustly overexpressed in MEFs (Figure 5f – g), but ciliogenesis was not increased (Figure 5f). Similarly, the level of endogenous Hh target genes, such as Ptch1, was also not increased (Figure 5h). These data do not support INTU as a driver for ciliogenesis or Hh pathway activation. Thus, the elevated expression of INTU in human BCCs might be required but not sufficient for the development of tumors.

DISCUSSION

The ciliogenic functions of the CPLANE genes (Fuz, Wdpcp, and Intu) are indispensable during vertebrate morphogenesis ^{4–6, 20–25}. The current study extended their biological functions to mammalian tumorigenesis by demonstrating that INTU plays a permissive role in BCC formation through regulating primary cilia formation and Hh signaling. Because that the CPLANEs form a molecular module by interacting physically to facilitate the assembly of the IFT-A complex during ciliogenesis ⁷, it is speculated that genetically disrupting other CPLANE genes may block BCC formation in a fashion similar to Intu.

The robust upregulation of INTU transcription in BCC cells was surprising. It is not only in concert with the idea that INTU is prerequisite for ciliogenesis, but also suggests that INTU is required for the propagation and maintenance of ciliated BCC cells. While we did not obtain evidence supporting a role of INTU in directly driving ciliogenesis or Hh signaling (Figure 5g–h), it is of interest to know that the expression level of INTU correlates with those of Hh responsive genes and ciliogenesis in BCC specimens (Supplementary Figure 3). These observations suggest that INTU acts in concert with other molecular events to promote ciliogenesis and Hh pathway activation in BCC.

It remains unknown how INTU is upregulated in BCC. It may be a mere reflection of the increased proportion of ciliated BCC cells that already express relatively high endogenous levels of INTU. It is also possible that the expression of oncogenic mutant Hh genes, such as PTCH1, SMO or Gli2, may drive de novo upregulation of INTU, and probably other ciliogenic genes, to deregulate the ciliogenic and retraction programs to favor cilia formation in keratinocytes. Another possible mechanism is UV, a primary risk factor for BCC ²⁶ and a robust inducer of ciliogenesis ²⁷. UV may exert previously unappreciated ciliogenic functions during the early stages of BCC formation. Understanding how INTU is upregulated may provide important insight into BCC tumorigenesis.

Previous observations suggested that very few interfollicular epidermal keratinocytes in adult skin possess a functional cilium, whereas BCC cells are frequently ciliated ^{6, 17, 28}. Indeed, quantifications obtained in this study confirmed these findings, thereby suggest that primary cilium is a candidate biomarker for BCC. But it is unclear how BCC cells acquired the cilia phenotype. In addition to the altered intrinsic ciliogenic programs discussed above, it is also possible that BCC originates from the few ciliated keratinocytes. It is conceivable that ciliated keratinocytes can activate the Hh signaling pathway more readily upon acquiring Hh pathway mutations, thereby, may gain growth advantage over unciliated keratinocytes. Thus, these findings provided insight into the debated origin of BCC ^{18, 19, 29–32} by suggesting that BCC may be more readily established from ciliated keratinocytes in the skin and hair follicles. Collectively, evidence supports the idea that ciliated keratinocytes are prone, whereas unciliated keratinocytes are resistant, to malignant transformation of keratinocytes.

Activation of the Hh pathway, which is a major driver of BCC, is dependent upon the primary cilium. The primary cilium is critical for BCC formation as we and others demonstrated that preventing cilia assembly by genetically ablating ciliogenic genes, such as Ift88, Kif3a, and Intu prevented BCC (¹⁷ and this study) could block the development of BCC in mice. These findings suggest that ciliogenic proteins might be therapeutic targets for BCC. However, systemically targeting INTU and other ciliogenic proteins is likely to result in significant adverse effects on the normal ciliary functions, such as those in the retina, kidney, and other organs. Therefore, this strategy is unlikely worthwhile. In contrast, targeting SMO with small molecules is remarkably effective for locally advanced or metastatic BCCs ³³. However, the development of treatment-resistant tumors in association with SMO inhibitor treatment ^34–37 presses for the identification of new therapeutic targets along the Hh signaling pathway ^{38, 39}. Thus, targeting Hh signaling components downstream of ciliogenesis and the smoothened receptor may become promising therapeutic strategies for BCC and other Hh-dependent cancers ^40–42.

In summary, this study provided evidence that ciliogenic genes, such as INTU, are dynamically regulated in the skin, and that the upregulation of INTU is required but not sufficient for ciliogenesis and oncogenic Hh signaling during BCC formation. Understanding the molecular mechanism underlying with the aberrant upregulation of ciliogenic genes, such as INTU, during early stages of BCC formation may shed light on the prevention of BCC and other Hh-dependent tumors.

MATERIALS AND METHODS

Skin and BCC specimens

Formalin-fixed and paraffin-embedded healthy adult human skin, BCC, and SCC specimens (n = 25) were obtained from the biobank of the Department of Pathology. All specimens were de-identified. The use of human specimens was approved by the Institutional Review Board of Stony Brook University. Tumor histology was confirmed and areas of interest were chosen by two pathologists (CL and KRS) before they were used for the construction of a tissue microarray (TMA). Two cores of each specimen were randomly placed in the TMA block. All the staining and analyses were performed using slides sectioned from the TMA. One BCC specimen was lost during this process and was excluded. The TMA was comprised of 3 healthy skins, 19 BCCs, and 3 SCCs.

Mouse models and tissue collection

The conditional mouse model of Intu (Intu^flox/flox) were reported previously ⁵. Tg(KRT14-cre/ERT)20Efu/J and Gt(ROSA)26Sor^{tm1(Smo/EYFP)Amc}/J (a conditional allele of Smo, also known as SmoM2^Cond) mice were obtained from the Jackson Laboratory (Bar Harbor, ME, USA). Mice were bred to generate the bigenic or trigenic mice used in the study. Littermates were used as controls wherever possible. Both male and female mice were included in this study, and housed in groups of the same age and sex regardless of genotypes. Cre recombination was induced by tamoxifen (T5648, Sigma-Aldrich, St. Louis, MO) as described previously ¹⁸. Briefly, 2.5 mg tamoxifen (dissolved in sesame oil) was injected intraperitoneally to 22 – 27 day old mice once daily for five days. BrdU labeling was performed by injecting 10 μg per gram of body weight of BrdU labeling reagent (Life Technologies, Waltham, MA) 2 hours prior to euthanasia. Injections were carried out blindly.

Based on our experience gained from a pilot study, we collected skin specimens from mice at two time points after tamoxifen induction. Additional time points were applied on Intu-deficient (Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox) mice up to 22 weeks after induction. We aimed to obtain a minimum of three mice for each time point. Freshly dissected dorsal skins were either frozen in optimal cutting temperature compound (Sakura Finetek USA, Torrance, CA) or fixed overnight in 10% buffered formalin and processed for routine histology. All procedures related to mice were approved by the IACUC of Stony Brook University.

Measuring the thickness of the epidermis

Three randomly chosen fields of each H&E section were photographed with a 20X objectives, and the thickness of the interfollicular epidermis (IFE) was measured using Nikon (Melville, NY) NIS-Elements Analysis D software. The average epidermal thickness from each mouse was used for analysis. Specifically, n = 9 or 18 for Krt14-Cre^ERT;SmoM2^cond;Intu^+/flox mice at 5 weeks or 10 weeks post induction, respectively; n = 9, 12, or 18 for Krt14-Cre^ERT;SmoM2^cond;Intu^flox/flox mice at 5, 10, or 22 weeks post induction.

Cell culture and in vitro assays

The isolation of primary mouse epidermal keratinocyte was conducted as described elsewhere ⁴³. Briefly, skins of E18.5 embryos were digested by dispase II (Roche, Indianapolis, IN) to separate epidermis and dermis. Epidermis was then digested briefly with trypsin (Invitrogen) to dissociate keratinocytes. Keratinocytes were suspended in defined keratinocyte serum-free medium (Gibco, Life Technologies Waltham, MA) and plated on collagen I-coated tissue culture plates (Thermo Fisher, Waltham, MA). MEFs were cultured with Dulbecco’s modified Eagle’s medium (4.5 g/l glucose) supplemented with 10% fetal bovine serum and 100 U/ml penicillin and 100 U/ml streptomycin (Gibco, Life Technologies). To induce Cre recombination, subconfluent keratinocytes were treated with 0.5 μM 4-OH tamoxifen (Sigma-Aldrich) for 5 hours before grown to near-confluent, and starved in medium without growth supplement for 72 hours before harvested for RNA extraction or immunofluorescence. To activate Hh signaling, MEFs were grown to confluency and serum-starved in medium containing 0.25% serum for 24 hours, then treated with 400 nM smoothened agonist (SAG, Calbiochem, San Diego, CA) for 4.5 hours before harvested for mRNA extraction.

DNA constructs and transfection

Expression plasmid encoding Intu, full length mouse Gli2 (Gli2-FL), and the constitutively active Gli2 lacking the coding sequence for the first 413 amino acids (Gli2ΔN) were described previously ^{5, 44}. GLI2 was expressed as a fusion protein with YFP. The transfections were carried out by Lipofectamine 2000 (Thermo Fisher Scientific, Waltham, MA) per manufacturer’s instructions.

Quantitative RT-PCR

RNA isolation and quantitative RT-PCR analyses were performed as previously described ⁶. The following TaqMan probes were used: Ptch1, Mm00436026_m1; Gli1, Mm00494645_m1; Gli2, Mm01293111_m1; Intu, Mm01284303_m1; and β-actin, Mm00607939_m1 (Life Technologies). Results were analyzed using the ΔΔCt method. Relative expression levels of target genes were normalized with β-actin and compared with controls.

In situ hybridization and scoring

In situ hybridization was carried out on formalin-fixed paraffin embedded tissue sections using the RNAScope system (Advanced Cell Diagnostic, Inc, Hayward, CA) per manufacturer’s instructions. The probes used are INTU (413571), GLI1 (310991), PTCH1 (405781), SMO (405831), Intu (413601), Gli1 (311001), and Ptch1 (402811).

Semi-quantitative scoring was performed for signaling generated by each target probe on all TMA specimens. Scoring was based on the average number of signal puncta within a cell’s boundary. POLR2A (Cat 310451) was used as an internal positive control for the quality of each specimen for this assay. Scores of each target probe were then normalized by POLR2A. Normalized scores of the randomly placed cores of the same specimen were averaged and presented. Scoring was performed by the investigators blind to the protocol.

Immunofluorescence labeling and microscopy

Immunofluorescence labeling was performed as described previously ⁶. The following primary antibodies were used: antibody to KRT1 was produced in the Roop lab⁴⁵; KRT14 (Covance; Princeton, NJ); GFP/YFP (Aves Labs Inc; Tigard, Oregon); KRT17 (Abcam, Cambridge, MA, USA); LEF1 (Cell Signaling, Danvers, MA); BrdU (Life Technologies); acetylated α-tubulin (Sigma); γ-tubulin (Abcam, Cambridge, MA), ARL13B (NeuroMab, 73-287; Davis, CA); AlexaFluor-conjugated secondary antibodies were from Life Technologies. Mounting medium with DAPI (Vector Laboratories, Burlingame, CA) was used. Images were acquired by Nikon 80i fitted with Nikon DS-Qi1Mc camera and processed with Photoshop 5.5 CS.

Statistical analyses

All quantifications are presented as mean ± SD. Student t-test (two-tail, unpaired, assuming unequal variance) was used unless otherwise stated. Kaplan-Meier curve, one-way ANOVA, two-way ANOVA analyses were generated by the GraphPad software (GraphPad Software, Inc., San Diego, CA). P < 0.05 is considered statistically significant.