Discussion
Apc is implicated in the Wnt signaling pathway that is involved both in development and tumorigenesis. Human germline mutations in APC cause FAP [4,5], which is characterized by hundreds of adenomatous colorectal polyps, with an almost inevitable progression to colorectal cancer in the third and fourth decades of life. The phenotypical features of FAP and its variant, Gardner's syndrome, can be very variable. As well as colorectal polyps, these individuals can develop extracolonic symptoms, among which are upper gastrointestinal tract polyps, congenital hypertrophy of the retinal pigment epithelium, desmoid tumors, disorders of the maxillary and skeletal bones, and dental abnormalities [6]. While the heterozygous knockout mice for Apc develop adenomatous polyps predominantly in small intestine, the homozygous embryos die before gastrulation. To gain more insights into the effects of Apc loss in tissues other than gastrointestinal tract during life of animals and to circumvent the embryonic lethality associated with Apc nullizygosity, we created a mouse strain carrying a conditional allele of Apc (ApcCKO) in which exon 14 of the Apc is flanked by loxP sequences. The homozygous mice for the conditional allele are viable and indistinguishable to the normal mice, allowing us to study the roles of Apc in a tissue- and temporal-specific manner.
To assess the phenotypic consequences of inactivation of the Apc gene in cells that express K14, we created mice that are homozygous for the ApcCKO allele and contain a K14-cre transgene. These mice failed to thrive and died before weaning. They also exhibited hair, tooth, and thymus phenotypes.

Apc and Hair Follicle Morphogenesis
Current models of hair follicle development suggest that the establishment of a regular array of placodes in the surface epithelium in response to the first dermal message is achieved through the competing activities of molecules that promote or repress placode fate [24]. There is accumulating evidence that activation of the Wnt signaling pathway in the dermis may be involved in establishing the first dermal signal. Experimental activation of epithelial β-catenin signaling (by expression of N-terminal–truncated, constitutively stabilized forms of β-catenin or ectopic expression of Lef1) induces ectopic follicles in both mouse and chick skin [25–27]. Conversely, down-regulation of β-catenin signaling (through Lef1 knock-out, ectopic expression of Wnt inhibitor Dkk1 or conditional deletion of β-catenin in epidermis) results in loss of vibrissae and some pelage follicles in mice [28–30]. Our finding that K14-driven Apc loss in embryonic ectoderm resulted in irregularly spaced and often tightly clustered abnormal hair placode initiation and follicle morphogenesis (Figures 4F, 5B, 5D, and 5F) as well as in the development of multiple tooth buds (Figure 4O and 4P), is in line with the effects seen in activation of epithelial β-catenin signaling during placode formation, indicating the role of Apc in specification of embryonic ectodermal stem cells to produce a hair follicle. Given the role of Apc in down-regulation of β-catenin, loss of Apc would inevitably lead to altered expression of β-catenin in ectodermal cells during the hair placode formation, giving rise to aberrant follicular growth throughout the embryonic epidermis, including the footpads, where normally hairless (Figure 5B, 5D, 5F, and 5H). Of interest are the phenotypic similarities and differences between our KA mutant mice and the previously described K14-ΔN87βcat transgenic mice expressing stable β-catenin under the control of K14 promoter [25] and K14-Lef1 transgenic mice [27]. Surprisingly, our mutant mice shared more similarities with the K14-Lef1 mice than the K14-ΔN87βcat transgenic mice. For example, the KA mutant mice displayed the disoriented and short, curly whiskers seen in K14-Lef1but not in K14-ΔN87βcat transgenic mice. In addition, KA mutant mice displayed a disorganized hair coat beginning with the emergence of neonatal hairs, similar to that observed in K14-Lef1 mice, whereas in K14-ΔN87βcat mice, embryonic hair follicle morphogenesis was unaffected and other skin changes only emerged postnatally. The reason for this difference could be attributed to one or all of three possible contributing factors. (1) Although the K14 promoter itself is known to initiate the expression at E9.5 and elevated dramatically by E13.5–E14.5 [19,31], many factors could influence the expression of promoter/transgene construct in transgenic mice. One of the most important considerations has to do with the integration site of the transgene in the mouse genome. Depending on the location of integration, the transcriptional activity of even the same transgene could vary considerably. It is possible that the apparent lack of ΔN87βcat function in embryogenesis could be due to the delayed or weak expression of the K14-ΔN87βcat transgene compared to the intrinsic promoter. In contrast, Cre-mediated recombination is seen in E13.5 skin of K14-cre transgenic mice used in this study [16] and K14-cre–mediated truncation of the Apc gene is likely to have occurred by then, in time for the second wave of hair follicle morphgenesis. (2) Another important fact is that in the aforementioned transgenic model the overexpression of transgene is confined to the basal cells of the epidermis and the ORS of hair follicles, where the K14 promoter is active. The expression of transgene is hence transient and stops once K14-expressing cells terminally differentiate. In contrast, in our model, K14-cre–mediated deletion of Apc would result in Apc mutation not only in K14 promoter active cells, but remain so in all cell layers that derived from the K14 promoter active progenitors. These derivatives include the suprabasal epidermis and the entire epithelium of the hair follicle. Stabilization of β-catenin in a broader population of cells could account for some of the phenotypic differences. However, we did not detect any elevated β-catenin expression in either K1- or K6-positive differentiated cells in postnatal mutant skin. This indicates that despite the absence of Apc and potential stabilization of β-catenin in derivatives of K14-expressing progenitors, elevated β-catenin expression, and subsequent cell fate determination may only take place in basal cells. (3) We cannot exclude the possibility that given the multifunctional properties of Apc, disruption of its functions other than down-regulation of β-catenin may also have contributed to the observed overt phenotype.
While some features of our mutant mice were similar to these transgenic mice, other phenotypic aspects were largely distinct. In addition to aberrant hair follicle morphogenesis, K14-driven loss of Apc caused formation of multiple tooth buds that, like hair follicles, were known to develop through inductive interactions between the epithelium and mesenchyme. This observation was similar to the ectopic tooth buds found in animals misexpressing Lef1 [27], but more severe and was also present at birth, indicating the effect of Apc loss during the initiation of embryonic tooth development, which was evident by aberrant Shh expression in E15.5 embryonic oral epithelium (unpublished data). It should be noted that although multiple tooth buds were histologically visible (Figure 4O and 4P), these teeth never broke out and the KA mutant mice appeared toothless. This unusual severe tooth defect is unique to these mutant mice. In addition, neither of the two transgenic mice was postnatally lethal as in the KA mutant mice. We did not find any obvious histopathological abnormalities in the internal organs of KA mice that could contribute to the lethality. However, the fact that all the mutants had lower weight (Figure 2F) with hardly any evidence of solid food in their stomach indicates that the mutants might have died of starvation. Dermal fat was reduced in the mutant skin, possibly as a consequence of poor nutrition caused by the absence of teeth. Since the weight loss in KA mutants started from P8–P10 while pups were still nursed by their mothers, starvation due to lack of teeth cannot be the sole cause of death, but is likely to be a contributing factor. Absence of teeth and mammary glands have been observed in mice deficient in Lef1 and ectopically expressing Dkk1 [29,30] but their absence was due to the block in development before the bud stage. Hence, neither loss nor excess of tooth bud formation allows proper development of teeth for mice to have a healthy diet and normal life. Mechanistic studies to understand how increased levels of β-catenin leads to altered skin and tooth phenotypes are under way.

Apc and Thymus Organogenesis
In this study, we observed that K14-driven loss of Apc resulted in a small thymus with severe squamous metaplasia leading to the formation of numerous pyogranuloma and loss of proper meshwork structure for thymocyte maturation, rendering the mice athymic. Previous studies have shown that in normal adult thymus K14 expression is found together with K5 in the stellate medullary TECs, but not in association with K5+ TECs in the cortex or at the cortico-medullary junction. In addition, it has been demonstrated that K14+ TECs do not coexpress K8; hence, there are two distinct medullary subsets, namely a K8−K14+ stellate subset and a K8+K14− globular subset [32]. In agreement with the previous results, in P3 normal thymus K14 expression was restricted to stellate medullary TECs (Figure 6D), whereas K8 expression was found throughout the TECs (unpublished data). We could not clarify whether these two keratins were coexpressed in the same TECs without double-staining. There were individual differences among mutants and these were prominently reflected in the histological abnormalities of the thymus at P3, but as the older surviving mutants all showed the same histopathologies of the thymus, the mutant thymi eventually seem to result in the same fate. It is unclear when K14-cre induction actually takes place in the mutant thymus, but as the population of cells showing a strong nuclear β-catenin staining as well as the cells expressing K14 were small and thymic epithelial compartments still existed in a mild P3 mutant thymus (Figure 6E and 6F), it seems that initial differentiation to medullary and cortical TECs and thymocyte colonization have already taken place prior to the major effects of K14-driven Apc loss. However, cells with nuclear β-catenin and K8+K14+ double-positive epithelial cells increased subsequently, associated with active proliferation in the latter group of cells and loss of TEC compartments. With age, β-catenin expression pattern became more diffuse and fewer epithelial cells showed nuclear localization but K8+K14+ cells remained, forming concentric structures of epithelial cells filled with hard keratin deposits and infiltrated with vast number of neutrophils and macrophages. Only a few epithelial cells were dividing and hardly any thymocytes were present by this time. These results suggest that with the deletion of Apc in TECs, stabililization and nuclear localization of β-catenin took place and subsequently these cells differentiated into keratinocytes that expressed both K14 and K8, similar to the basal cells of the skin, instead of TEC subsets. The expansion and differentiation of these keratinocytes lead to loss of proper thymic epithelial compartments and in turn produced and deposited the hard keratins that consequently caused vast amount of neutrophils and macrophages to infiltrate the thymus. These aberrant epithelial structures eventually have overtaken the whole of the thymus and driven out the colonized thymocytes. K14-driven loss of Apc and subsequent constitutive expression of β-catenin in TECs have therefore misdirected them to wrong epithelial cell fate, not allowing proper differentiation to either cortical or medullary TECs, which is essential for normal thymocyte development. This is not only evident from the lack of dividing thymocytes in the mutant thymus by P13, but also by the differential expression pattern of keratins, which were more skin-like than TEC-like. The importance of Apc function in thymic development has been demonstrated by thymocyte-specific loss of Apc by crossing a different strain of Apc conditional mice and LckCre transgenic mice [33]. Here, by K14-driven TEC-specific loss of Apc, we have demonstrated its importance in thymus development not only in thymocytes but also in TECs.
It is of interest that dental abnormalities, such as supernumerary and impacted teeth similar to those observed in our mutant mice, are frequently seen in patients with Gardner's syndrome, carriers of APC germline mutation [6]. The importance of odonto-stomatological examinations should hence be pointed out as a means of reaching a presumptive diagnosis, whose confirmation is vital to the patient. Further characterization of the mechanism of such developmental defects using our mouse model should provide important insights into Apc function in multiple organ systems and to give better insights into potential adverse events in human subjects.
In conclusion, we have shown that loss of Apc in K14-expressing embryonic cells causes aberrant morphogenesis in various skin appendages, including hair follicles and teeth, and abnormal thymus organogensis. Our results provide genetic evidence that expression of Apc is essential for regulation of proper cell fates in these organs that require epithelial–mesenchymal interactions.