Results

Generation of the ApcCKO and ApcΔ580 Mice
To investigate the role of Apc in development of skin and its appendages, we used the Cre/loxP technology to introduce a conditional mutation of the Apc gene in mice. We constructed embryonic stem (ES) cells and mice carrying an Apc allele harboring both a pair of loxP sites flanking Apc exon 14 and a pair of FLP recognition target (FRT) sites flanking PGK-neomycin selection cassette by recombineering [12,13] (Figure 1A, ApcCKON allele, N for neomycin cassette). A PGK-neomycin cassette was inserted in the same transcriptional orientation as Apc in intron 14 of the endogenous gene. The loxP and FRT sites were used to aid unidirectional recombination [12,13]. Two mouse lines containing the same modification were generated from two independent ES clones to ensure that these two lines behave in the same way. These ApcCKON/+ mice were crossed with FLPe-deleter to generate ApcCKO/+ mice that were heterozygous for the final Apc conditional (ApcCKO) allele that removed the PGK-neomycin cassette and contains only the loxP sites in the introns flanking exon 14. To assess the effect of deleting exon 14 in mice, both lines of ApcCKO/+ mice were crossed with the Cre-deleter to generate the germline knockout line of Apc, designated ApcΔ580/+. The mutant allele (ApcΔ580) lacks exon 14 (Figure 1A). The transcript from loss of exon 14 results in a shift in the normal reading frame, resulting in a premature chain termination codon which, if utilized, would result in a truncated polypeptide that is 605 aa in length, of which the first 580 aa correspond to the normal Apc protein.
Figure 1  Generation of the Conditional Apc Allele
(A) Schematic diagram of exons 14 and 15 of the mouse Apc gene, the targeting vector, and the resulting conditional allele with 2 LoxP sites sandwiching the exon 14. The PGK-neomycin cassette was inserted within intron 14 by recombineering technique. This cassette is sandwiched by 2 FRT sites that could be removed by crossing to FLPe-expressing mice. Positions of PCR primers used for genotyping PCR (F2, R2, R4) and RT-PCR (F546 and R721) are indicated. Positions of probe used for Southern blot analysis with NdeI sites are also shown. Upon Cre-mediated recombination, exon 14 is removed and leads to truncated Apc protein, of which the first 580 aa correspond to the normal.
(B) Southern blot analysis of NdeI-digested genomic tail DNA isolated from F1 mice of various Apc mouse lines (ApcCKON, ApcΔ580), hybridized to a 600-bp probe. Tail genomic DNA from ApcCKON F1 mice derived from a modified ES clone showed a 12-kb band for the ApcCKON allele and a 10-kb band for the wild-type allele, whereas genomic DNA from the ApcΔ580 mouse was heterozygous for the ApcΔ580 allele (9.2-kb band).
(C) Kaplan-Meier survival plot of ApcCKO/+ mice (thin solid line, n = 39), ApcCKO/CKO mice (thin dotted line, n = 57), ApcΔ580/+ mice (solid line, n = 51), and wild-type littermates (broken line, n = 21). Heterozygosity of the ApcΔ580 allele led to a significantly shortened survival (p < 0.0001), whereas those of heterozygous and homozygous ApcCKO mice had no significant difference to that of wild-type littermates. Southern blot analysis of tail DNA from F1 offspring of both ApcCKON and ApcΔ580 lines confirmed the germline transmission of modified Apc allele (Figure 1B). Mice that are heterozygous for ApcΔ580 mutation are viable but have a significantly reduced lifespan (Figure 1C). These results suggested that deletion of exon 14 indeed results in either loss or abnormal function of the Apc gene product. ApcΔ580/+ mice have median survival of 5 mo of age (Figure 1C), with progressive signs of rectal bleeding and anemia. Similar to the results reported with an independently generated ApcΔ580/+ conditional mouse strain [14], ApcΔ580/+ mice had more than 100 (120 ± 37, n = 11) intestinal tumors at the time of their death (Figure S1). Inactivation of wild-type Apc is an important prerequisite for tumor development. We analyzed 30 intestinal tumors from ApcΔ580/+ mice by in vitro transcription and translation assay, but none of them showed truncated Apc products (unpublished data), indicating that the most likely mechanism of wild-type Apc inactivation is by allelic loss. The mutant allele had to be maintained and transmitted through male mice, as ApcΔ580/+ females were frequently not healthy enough to successfully nurse their own pups because of their tumor burden.
ApcCKO/+ mice were intercrossed to generate ApcCKO/CKO offspring. Approximately one-quarter of the offspring (17 of 81) were homozygous for the ApcCKO allele. These mice as well as heterozygous mice for ApcCKO allele are normal, showing no differences in their survival to the wild-type littermates (Figure 1C). We tested whether our ApcCKO allele can compliment the wild-type allele by crossing the ApcCKO/CKO female with ApcΔ580/+ male mouse. The resultant ApcCKO/Δ580 offspring were viable and born in the Mendelian ratio, suggesting that the presence of loxP sites in introns flanking exon 14 have no adverse effect on the function of the Apc gene.

K14-Driven Loss of Apc Results in Severe Growth Retardation and Early Lethality
To introduce the mutation of Apc into cells expressing K14, we crossed WW6 ES cell–derived [15] ApcCKO/+ mice with K14-cre recombinase mice in FVB background [16]. The K14-cre; ApcCKO/+ mice were normal in appearance and were fertile. K14-cre; ApcCKO/+ males were crossed to ApcCKO/CKO females to avoid the potential deleter effect in oocytes of K14-cre–positive females [17]. The mice were intercrossed thereafter for maintenance; hence, the mice used for analysis were in a mixed background of FVB, 129/S, and C57BL/6 in similar proportions, with minimal contribution of SJ.
The K14 promoter is a commonly used epidermal cell promoter because of its expression by the mitotically active cells of the epidermis and its appendages in mature skin [18], but most notably it is active in embryonic ectoderm as early as the single layered ectodermal stage of embryonic day (E) 9.5 [19]. A restricted expression of K14 is also found in thymic epithelial cells (TECs) in the medulla of normal thymus [20].
We genotyped a total of 458 pups (8–10 d old) from 67 litters resulting from crosses between K14-cre; ApcCKO/+ and ApcCKO/CKO mice. The mutant mice of the genotype K14-cre; ApcCKO/CKO (hereafter, KA mutant) were born, but the observed frequency of KA mutants was much less than expected (78 of 458 [17.0%]; p < 0.0005 Chi-square analysis, Table 1). To assess the basis for the neonatal lethality of KA mutants, we monitored three litters from birth to weaning by measuring the body weight of each pup every day. A total of 25 pups were born from three litters, of which 7 (28%) were confirmed to be K14-cre; ApcCKO/CKO by genotyping, indicating that KA mutants were born in the expected Mendelian ratio. The KA mutant pups were nursed normally, and there was milk in their stomachs during the first 2 or 3 d after birth, but they failed to thrive (Figure 2). By postnatal day (P) 8–10, at the time of genotyping, many KA mutant pups were considerably smaller than their littermates (Figure 2B–2F) and some have died at or prior to this age. None of KA mutants survived to weaning age.
Table 1  Genotype Distribution of Progeny from the Matings
Figure 2  Postnatal Mortality and Stunted Growth in K14-cre; ApcCKO/CKO Mutant Mice
Animals whose genotype is either heterozygous or homozygous for the wild-type Apc allele are referred to as normal (N); those whose genotype are K14-cre; ApcCKO/CKO and show the presence of K14-cre–recombined mutant Apc allele are called mutant (M).
(A) Two P3 mutant mice, M1 and M2, and their normal littermates, showing size variation among mutants.
(B) P8 mutant mouse (right) and a normal littermate. Note sparseness of hair coat and abnormal ears.
(C–D) Vibrissae of whisker pads are short and oddly angled in a P12 mutant mouse (C), relative to control (D). Note the lack of incisors in the mutant.
(E) A P17 mutant mouse (right) with its littermate. Its bare forehead, dorsal median line, and abnormal ears are evident.
(F) Growth curve of mutants and normal littermates. Mutants exhibit stunted growth, which became more prominent as they aged, and weigh significantly less than littermates from P8 (p < 0.05).
(G) Comparison of mutant and normal thymus from P3 mice. The mutant thymus (left) is dramatically smaller for its age compared to the normal littermate (right). The scale bar equals 1 mm.
(H) Skeletal preparations of normal (left) and mutant (right), showing differences in development of both incisor (I) and molar (M) teeth. The ability of whole embryos to exclude blue dye was used to examine the epidermal barrier, normally acquired beginning at E16 and complete by E18.5 [21]. Analyses of E17.5–E18.5 KA mutants showed that they were able to exclude blue dye, indicating that the epidermal barrier was intact (Figure S2). At these embryonic ages, there were no differences in size between the mutants and their littermates, but the mutants showed a patch of “birthmark” or dark pigmentation on their foreheads and a dark median line that ran caudally from head to tail. Their external ears or pinnae were shriveled in appearance and pigmented compared to those of littermates.
External characteristics of KA mutants that were evident at E18.5 persisted after birth and became more prominent as they grew (Figure 2A–2F). Growth of pelage hair was generally delayed in the mutants. At around P8, the KA mutants were hairless and had wrinkled skin while their phenotypically normal littermates had a smooth thin coat of hair (Figure 2B). At this age, two lower incisors start to erupt in normal littermates and these were absent in the KA mutants (Figure 2C and 2D). Animals also tended to be smaller and around P10–P12 displayed abnormally short and misshapen vibrissae and short, shaggy pelage hairs (Figure 2C and 2D). Development of thick ridges in their skin, particularly around the ears, eyelids, forehead, nose, and paws, became noticeable (Figure 2E). These regions looked scaly, and these animals hardly kept their eyes open. In contrast to the normal littermates that consistently increased their body weight with age, surviving KA mutants started to lose weight from P10 onwards; by P16–P17 they were all lethargic, and none of them survived to weaning (Figure 2E and 2F). At the time of autopsy all the mutants were toothless, without incisors or molars, and their stomachs were consistently small and had no solid food, unlike their age-matched littermates, suggesting that the observed weight loss could be the result of failure to ingest solid food (Figure 2F). Interestingly, changes in body weights and timing of hair growth varied considerably among mutant pups even if they were from the same litter, whereas those of phenotypically normal littermates tended to be similar. This difference was also reflected in the variation in timing of death in mutants: some mutant pups were born alive but died within a day or two, some survived close to the weaning age. This variability of the mutant phenotypes suggests possible variation in the timing and efficiency of cre-mediated Apc deletion. It is possible that the genetic background has a role to play in this variability.
Gross examination of internal organs also showed that the mutants' thymi were consistently inconspicuous and were very small for their age, whereas those of their littermates were very prominent in size (Figure 2G). This difference was evident as early as P3. Quite frequently mutant thymi in P12–P17 mutant mice also contained black deposits within the tissue (unpublished data). Mutant mice were also examined for any skeletal abnormalities by preparing skeletal specimens of P16–P17 mice stained with Alizarin red. No differences between the normal and KA mutant mice in the mandibular bone can be detected, but the mutant mice lacked or had underdeveloped set of maxillary incisors and molars (Figure 2H). We detected no other major skeletal abnormalities.

Genotype- and Tissue-Specific Expression of the Truncated Apc Transcripts
To assess the molecular effects of the K14-cre–mediated recombination, we screened for the presence of deleted Apc (ApcΔ580) alleles. Genomic DNA was extracted from liver, thymus, and skin from all 4 possible genotypes: K14-cre; ApcCKO/CKO, K14-cre; ApcCKO/+, ApcCKO/CKO, and ApcCKO/+. Genotyping on genomic DNA from these tissues showed that the ApcΔ580 allele (500-bp product) was detected only from the skin and thymus of the K14-cre–positive mice. The presence of mutant Apc allele in the thymus of K14-cre; ApcCKO/+ mice was consistently much less than the DNA from the skin of the same animal or other tissues from the KA mutants. In addition, this product was not detected at all in either the liver of K14-cre–positive or in any of the K14-cre–negative mouse tissues samples, establishing that Cre-mediated recombination has taken place in the tissue-specific manner in the mice that inherited K14-cre (Figure 3A).
Figure 3  Tissue-Specific Detection and Expression of Deleted Apc Alleles
(A) Tissue-specific genotyping PCR. Only genomic DNA samples from the skin (S) and thymus (T), but not liver (L) of mice positive for K14-cre show the presence of deleted ApcΔ580 allele.
(B) Genotype- and tissue-specific expression of the truncated Apc transcripts. A representative gel of RT-PCR using primers F546 and R721, showing that only RNA from the skin and thymus but not liver of mice positive for K14-cre have transcripts from both wild-type (528 bp) and deleted (313 bp) Apc alleles.  Apc transcripts were also analyzed by RT-PCR with primers spanning exon 14 (Figure 1A) using total RNA isolated from the corresponding tissue samples. We detected the expected RT-PCR product (313 bp) from the truncated Apc (ApcΔ580) allele only in the tissues where Cre recombinase is known to be expressed in the K14-cre–positive mice. However, this product was not detected in either the K14-cre–negative mouse tissues samples or the liver of K14-cre–positive mice, and only the product from the wild-type allele (528 bp) was detected from these RNA samples, further confirming that Cre-mediated recombination has taken place in the tissue- and genotype-specific manner (Figure 3B).

K14-cre–Driven Apc Loss Induced Aberrant Hair Follicles throughout the Epidermis
To understand the basis for delayed and abnormal hair development in the KA mutants, we conducted a histological and immunohistochemical examination (Figure 4). The hair follicle is an epidermal appendage that consists of an upper permanent portion, and a lower cycling portion that produces the hair [22,23]. The outer root sheath (ORS) is contiguous with and biochemically similar to the basal layer of the epidermis. The inner layers of the hair follicle include three concentric layers of inner root sheath and three concentric layers of hair-producing cells. At the base of the hair follicle is the germinative hair follicle bulb, which contains rapidly proliferating “matrix” cells that differentiate to populate all of the layers of the inner root sheath and the hair shaft itself [22]. During the anagen phase of the hair cycle (until P15), hair follicles of phenotypically normal mice grew deeply into the subcutaneous fat and were uniformly spaced and aligned in parallel arrays at a specific angle relative to the skin surface (Figure 4A). In contrast, KA mutant follicles were irregularly spaced and often seen as disoriented and clamped invaginations at P3 that became even more remarkable at P12 when the mutant mice were covered by fur coat (Figure 4F). Bulbs were often bent in addition to being irregularly angled to one another and their sizes and locations were often variable. Clusters of multiple invaginations or dysplastic follicular structures were frequently observed throughout the epidermis, whereas other regions showed gaps with no follicles. Serial sectioning indicated that some of the hair follicles in the P12 mutant skin were not properly formed or shorter than normal. Taken together, these features could account for the apparently delayed, followed by outgrowth, of the short and shaggy-looking fur coat of these mutant mice.
Figure 4  Histological and Immunochemical Examination of P12 Skin and Teeth
(A–E) P12 normal skin. (F–J) P12 mutant skin. (K–N) P12 normal oral cavity. (O–R) P12 mutant oral cavity. Stained with H&E for histology (A, F, K–L, O–P), Ki67 (B, G), β-catenin (C, H, M, Q), K14 (D, I, N, R), and K6 (E, J). Aberrant follicular morphogenesis, characterized by formation of irregularly spaced, nonpolarized hair follicles, in mutant skin is evident. Despite the abnormal histology, proliferation seems to be confined to hair bulb-like structures (arrows in [G], inset [G′] at higher magnification), but in mutant skin (arrows in [H], inset [H′] at higher magnification) and oral cavity (arrows in insets [Q′] at higher magnification) elevated cytosolic localization of β-catenin is detected in some cells. Scale bars: 50 μm for (A–F), (H–J); 250 μm for (K) and (O); 100 μm for (G), (L–N), (P–R); 20 μm for (Q′). Apc is a regulator of β-catenin that is important for Wnt signaling. We examined the patterns of expression of β-catenin in the affected tissues. In the normal skin, β-catenin, a member of the adherens junction complex, was found in the ORS of hair follicles and basal layer of epidermis, where K14 expression is also observed (Figure 4C and 4D), whereas the expression of K1, involucrin, and loricrin (markers for spinous and granular layers of epidermis) was only observed in the nonbasal epidermis (unpublished data). The patterns of expression of K14, K1, involucrin, and loricrin, in skin from mutant and normal littermate mice at P3–P17, showed no significant differences in the terminal differentiation (Figure 4A–4D, 4F–4I). Similarly, the pattern of expression of K6, which is normally only expressed in the suprabasal or inner layer of the ORS of the hair follicle but not in the epidermis (Figure 4E), did not change. Due to the abnormal and disorganized structure of hair follicles themselves, K6 localization highlighted the histological abnormality (Figure 4J). Yet as in the normal skin, K6 was principally seen only in the suprabasal layer of the ORS that did not colocalize with the basal markers, K14 or K5 (Figure 4I and 4J).
In normal skin, proliferating cells were detected in either the basal layer of epidermis or in germinative hair follicle bulbs at the base (Figure 4B). In the mutant skin, either BrdU incorporation or Ki67 expression was observed not only in cells in bulbs at the base of the hair follicle but also in bulb-like structures that were budding out from the ORS of the existing hair follicles (Figure 4G and 4G′). Each budding tip was becoming like a hair follicle bulb containing proliferating cells. Hence, despite the abnormal histology in the mutant skin, proliferation seems to be confined to bulb-like structures as in the normal skin (Figure 4G and 4G′). The exact locations of hair follicle bulbs were not as easy to define for some mutant follicles due to their disorganized structures. Interestingly, in the mutant skin, in addition to diffuse membrane-bound localization as in the normal skin, cells with strong cytosolic β-catenin localization were also observed frequently (Figure 4H and 4H′). These elevated β-catenin–expressing cells were usually surrounded by proliferating cells, forming bulb-like structures. Comparison of immunochemically stained serial sections showed that these intense cytosolic β-catenin stainings were usually found in either K14-positive K1-negative basal epidermis or K14-positive K6-negative basal ORS cells, and are surrounded by proliferating cells.
To determine the initiation of hair follicle morphogenesis in these mutants, we examined the expression pattern of Sonic hedgehog (Shh), a factor expressed in hair bulbs in embryonic skin (Figure 5). The aberrant hair follicle morphogenesis is evident as early as E14.5 in mutant embryonic skin, by multiple apolarized expression of Shh throughout the epidermis (Figure 5B), whereas that of control embryos was well polarized and regularly spaced (Figure 5A). With development, control mouse hair follicles invaginate downward in a polarized manner (Figure 5C), whereas those of mutant embryos were completely irregular and apolarized (Figure 5D). It was also noted that the size of each “budding” follicle, as detected by Shh expression, was variable (Figure 5D). The intensity of Shh staining was generally stronger in mutant skin than in the normal skin. The aberrant initiation of multiple hair placodes during early hair follicle morphogenesis was also evident by the whole-mount in situ hybridization (ISH) of E15.5 mutant embryos for β-catenin (Figure 5F and 5F′). The expression pattern of β-catenin in embryos clearly demonstrated the formation of regular arrays of hair placodes in the normal embryonic skin (Figure 5E and 5E′), but such regular patterning was lost, and often tightly clustered abnormal hair placodes were initiated in mutant embryonic skin (Figure 5F′). Aberrant hair placodes were also evident throughout the skin surface of limbs in E15.5 mutants (Figure 5F), whereas those of the control embryos had not yet formed (Figure 5E). Most interestingly, in the mutant footpads, where hair placodes do not normally form (Figure 5G), we also found ectopic irregularly sized and spaced hair placodes, indicating that the footpads still have the potential to form hair placodes in the absence of the Apc gene (Figure 5H).
Figure 5  Expression of Shh and β-catenin Transcripts in Normal (ApcCKO/CKO) and Mutant (K14-cre; ApcCKO/CKO) Embryonic Skin
(A–D) Section ISH with Shh probe in E14.5 normal (A), E14.5 mutant (B), E16.5 normal (C), and E16.5 mutant (D) skin. Broken lines indicate the interface between epithelium and mesenchyme. Scale bars: 50 μm. Whole mount in situ detection of β-catenin in E15.5 normal (E, G), mutant (F, H) embryos. Aberrant initiation of multiple hair placodes is evident at E14.5. Loss of K14-driven Apc loss caused aberrant pattern formation (F′) and formed ectopic hair placodes in normally hairless foot pads (H, arrows) which are absent in normal (G). These results collectively suggest that the terminal differentiation does take place normally in the mutant skin, but initiation of embryonic hair follicle morphogenesis is severely disrupted, accompanied by a continuous ectopic hair follicle morphogenesis in postnatal mutant skin.

Effects of K14-cre–Driven Loss of Apc in Other Epidermal Appendages
Similar to the biology of hair follicles, K14-cre–driven loss of Apc also affected the development of other epidermal appendages that depend on epithelial–mesenchymal interactions for their formation. The most striking of these was dental dysplasia (Figure 4K–4R). Tooth development is normally initiated between E11 and E12 by invagination of ectodermally derived oral epithelium into the underlying cranial neural crest–derived mesenchyme, generating a tooth germ. Despite the grossly toothless phenotype of KA mutants, histological analysis of their oral cavities revealed the formation of multiple tooth buds at each location. These aberrant teeth obviously failed to grow out during the dietary transition from milk to solid food. Analogous to the expression patterns of K14 and β-catenin in the normal skin, diffuse membrane-bound expression of β-catenin was detected in K14-expressing oral epithelium and ameloblasts of normal mice. In mutants, some of the K14-expressing cells also showed strong cytosolic/nuclear β-catenin staining, as observed in the mutant skin (Figure 4Q, 4Q′, and 4R). Initiation of ectopic tooth buds in the mutant mice was evident at E15.5 by extra dots of Shh expression adjacent to the primary teeth (data not shown).
Loss of Apc also leads to hyperplasia in squamous epithelia of cornea, oral, salivary, and Hardarian glands (unpublished data). Squamous metaplasia to hair follicle–like structures, ectopic hair follicle morphogenesis was also observed in these epithelia.

K14-cre–Driven Apc Loss Results in Hypoplastic/Athymic Mice
Thymus is an organ that is also known to have K14 expression [16]. It represents the primary lymphoid organ for thymocyte development and selection. Distinct population of TECs of cortex and medulla mediates both of these critical functions. Cortical and medullary TEC subsets are characterized by differential expression of four keratin species: K8, K18, K5, and K14.
The normal thymus is a lobulated lymphoid organ, each lobule clearly showing the two distinct TEC compartments, an outer cortex and an inner medulla (Figure 6). There were no major differences in the histology of thymus between the ages P3 to P17 in phenotypically normal littermates. As shown in the H&E staining of thymus, the cortex was formed of dense lymphoid tissue that lacks nodules (Figure 6A). Since the stroma of the medulla is less heavily infiltrated with lymphocytes than the cortex, the medulla stained more lightly than the cortex. In normal mice, the thymus retains its size until the young adult age and regresses thereafter by atrophy. In the normal young mice we examined (P3–P17), it is evident that thymocytes were mitotically active in the cortex as determined by BrdU immunostaining (Figure 6B). Immunohistochemistry of normal thymus from P3 to P17 mice showed a similar staining pattern for K14 in that its expression was restricted to a small population of TECs in the inner medullary region and in the keratinocytes in Hassall's corpuscles (Figure 6D). Diffuse cytoplasmic staining for β-catenin was also detected in the medullary epithelial cells (Figure 6C). In contrast to K14 expression, diffuse staining for K8 was observed in epithelial cells both in the medulla and cortex (Figure S3). K1 staining was not detected in young mice at P3 but in older mice it was detected in differentiated keratinocytes in some of Hassall's corpuscles (Figure S3).
Figure 6  Histological and Immunochemical Examination of Thymus
(A–D) P3 normal thymus. (E–G) Mild P3 mutant thymus. (H–K) Severe P3 mutant thymus. (L–O) P13 mutant thymus. Stained with H&E for histology (A, E, H, L), BrdU (B, I, M), β-catenin (C, F, J, N), and K14 (D, G, K, O). (B) Actively dividing thymocytes are visible at the superficial edge of cortex of normal P3 thymus. Note the progression of histological abnormalities in the mutant thymus from mild P3, severe P3 to P13 (A, E, H, L). Scale bars, 20 μm. The histological abnormalities of thymus were evident as early as P3 in KA mutants (Figure 6E and 6H). The thymus was made of two lobules as in the normal mice but the mutant thymus was significantly smaller in size than that of the age-matched controls (Figure 2G). Interestingly, variations in the phenotypic severity of the mutant pups at P3 were prominently reflected in the extent of histological abnormalities of thymus. A P3 KA mutant pup that showed milder phenotype with a comparable body weight to its normal littermates (Figure 2A, M2) showed milder thymus abnormalities (Figure 6E) compared to its more severe mutant littermate (Figure 2A, M1; Figure 6H). The milder P3 mutant thymus was already much smaller in size compared to those of normal littermates (data not shown) but two epithelial compartments of thymus were histologically still distinguishable, with colonization of thymocytes evident in the cortex. However, there were small populations of lightly stained cells by H&E extending from the edge of the outer cortex towards inner medulla (Figure 6E), and these cells showed intense nuclear β-catenin staining whereas the rest of the medullary cells showed diffuse β-catenin staining pattern similar to that of the control (Figure 6F). Localization of K14 was limited to a few cells in the medulla and some overlapped with K8 localization (Figure 6G).
In the other P3 mutant thymus the distinct thymic epithelial compartments have been lost completely, and only a few lymphocytes were remaining at the edges and some in the middle (Figure 6H). Proliferative activities were no longer observed in thymocytes as prominently as in the normal thymus, but the epithelial cells seemed to be forming concentric structures (Figure 6I). Unlike in the normal or mild mutant thymi, the severe P3 mutant thymus showed extensive K14 expression that overlapped with K8 expression (Figure 6K). These cells were more like basal cells of the skin than TECs and were adjacent to the most immature looking cells that were showing strong nuclear and cytoplasmic β-catenin staining (Figure 6J). The nuclear staining of β-catenin was not observed in the normal age-matched thymus (Figure 6C). Most notably, the K14 and β-catenin staining patterns were mutually exclusive (Figure 6J and 6K).
At P10–P13, the mutant thymus consisted of numerous enlarged Hassall's corpuscle–like structures, made of arrays of K14- and K8-expressing keratinizing epithelial cells surrounding large keratin deposits (Figure 6L and 6O). There were numerous neutrophils and macrophages infiltrating the thymus in response to these keratins; hence, these structures could be called pyogranuloma. Varying degrees of differentiation-specific markers depending on the age of mice, in this case K1 and involucrin that are normally present only in Hassall's corpuscles, were also detected in mutant thymi (Figure S3). In these mice no thymocytes were detectable. BrdU incorporation was only observed in very few keratinizing epithelial cells, looking somewhat similar to the pattern of mature skin (Figure 6M). The diffuse expression of β-catenin was also present in these epithelial cells, and at this age fewer cells were positive for nuclear β-catenin staining (Figure 6N). As in the younger mutant mice, however, nuclear localization of β-catenin was only observed in K14-negative cells that looked like undifferentiated basal cells. In older P17 mice, the histopathology and keratin expression pattern of the mutant thymus was similar to that of P13 except for the fact that β-catenin expression became increasingly diffuse and appeared to colocalize with K8/K14 expression (data not shown). This coincided with fewer immature cells in the older mutant thymus.
Collectively, these results suggest that loss of Apc and consequent stabilization of β-catenin in K14-expressing TECs lead to their aberrant proliferation and differentiation to keratinocytes, causing massive squamous metaplasia, rather than to form either medullary or cortical TECs. Loss of proper TEC compartments consequently resulted in loss of thymocytes for maturation and the mice to be “athymic.”