Hoxb13 Up-regulates Transglutaminase Activity and Drives Terminal Differentiation in an Epidermal Organotypic Model*

Hox genes act to differentiate and pattern embryonic structures by promoting the proliferation of specific cell types. An exception is Hoxb13, which functions as a proapoptotic and antiproliferative protein during development of the caudal spinal cord and tail vertebrae and has also been implicated in adult cutaneous wound repair. The adult epidermis, which expresses several Hox genes including Hoxb13, is continually renewed in a program of growth arrest, differentiation, and a specialized form of apoptosis (cornification). Yet little is known about the function(s) of these genes in skin. Based on its role during embryogenesis, Hoxb13 is an attractive candidate to be involved in the regulation of epidermal differentiation. Here, we demonstrate that Hoxb13 overexpression in an adult organotypic epidermal model recapitulates actions of Hoxb13 reported in embryonic development. Epidermal cell proliferation is decreased, apoptosis increased, and excessive terminal differentiation observed, as characterized by enhanced transglutaminase activity and excessive cornified envelope formation. Overexpression of Hoxb13 also produces abnormal phenotypes in the epidermal tissue that resemble certain pathological features of dysplastic skin diseases. Our results suggest that Hoxb13 functions to promote epidermal differentiation, a critical process for skin regeneration and for the maintenance of normal barrier function.

During development, the skin begins as a single layer of multipotent ectodermal cells capable of producing either hair follicles or epidermis. Cells receiving an instructive signal from the underlying mesenchyme respond by producing hair follicles, but in the absence of the signal, they stratify and generate the morphologically distinct layers of the epidermis. During adult life, the developmental program is recapitulated in the form of a program of cellular differentiation in which the epidermal layers of the skin are continuously renewed. In this process, cells from the dividing basal layer undergo growth arrest and then begin an upward migration, differentiating as they proceed to build an impermeable layer of cornified cells (the stratum corneum) that arises through a coordinated proc-ess of programmed cell death (1,2) and protein cross-linking (3)(4)(5). This cornification process, the final step in epidermal differentiation, occurs when a series of defined structural proteins (including loricrin, involucrin, and various keratins) become tightly cross-linked via the activity of the transglutaminase (TGase) 1 family of enzymes (6,7). During skin wounding, once the epidermal cells have migrated laterally to repopulate the wound bed, they must vigorously activate the epidermal differentiation program to restore the cutaneous barrier. This process is critically important, because loss of barrier function and inadequate wound healing can result in life threatening fluid losses and infection (8).
A major objective in cutaneous biology is to understand the transcriptional machinery that regulates gene expression during epidermal differentiation and renewal. One class of transcriptional regulators that are attractive candidates for roles in epidermal renewal and repair is the highly conserved family of Hox transcription factors (9 -12). In mammals, the 39 Hox genes that have been identified to date are organized into four individual chromosomal clusters (A-D in humans; a-d in mice). They appear to have arisen from two duplication events of a single ancestral cluster. Based on sequence homology and location within the cluster, Hox genes have been assigned to 13 paralogous groups. During embryogenesis, Hox proteins function as important regulators in differentiation and cell fate determination, and they are critical for axial patterning. Because of their distinct spatio-temporal expression patterns in developing and adult skin, Hox genes have been implicated in the patterning and regenerative processes of skin and hair, as well as in certain skin pathologies including cancer (13)(14)(15)(16)(17). Interestingly, once the roles of certain Hox genes in axial patterning are complete, these genes may be utilized in a spatially ubiquitous way throughout tissues that are not segmentally restricted, such as the skin. For example, in mice, expression of Hoxb13 is restricted to the caudal region during early fetal development, and contributes to development of posterior organs such as the prostate (18) and the caudal spinal cord and tail vertebrae (19). However, the axial patterning rules are violated as the animal matures and Hoxb13 (20) as well as Hoxc13 (21,22) become expressed in the epidermis and/or hair follicles throughout the adult skin.
A major difficulty in addressing Hox gene function, either in the fetus or in adult skin, is the fact that functional equivalency often exists within a paralogous group. Thus, with the singular exception of Hoxc13, whose loss in mice produces a fragile hair phenotype (21), no other skin mutant phenotype of any classical Hox gene has been reported. Adult mice that lack Hoxb13, a paralog of Hoxc13, do not display the altered hair phenotype. However, despite the lack of a skin phenotype in Hoxb13 knock-out mice, previous studies strongly suggest that Hoxb13 functions in both epidermal differentiation and wound repair. Hoxb13 is expressed in both fetal and adult skin (20,23). Wound healing studies conducted on human fetal and adult skin explants transplanted onto SCID mice revealed that Hoxb13 expression is significantly down-regulated in fetal skin that heals without a scar, but its expression does not change in adult wounds (23). These results suggested that suppression of Hoxb13 might be a necessary component to obtain optimal repair. To that end, we recently reported that adult cutaneous Hoxb13 knock-out wounds heal with less scarring compared with wild-type wounds (24). Examination of unwounded Hoxb13 knock-out skin revealed significantly higher levels of hyaluronan (HA, a known inhibitor of differentiation, Ref. 25 and thought to be an important factor in fetal scarless wound healing, Ref. 24) in both the epidermis and dermis, as well as reduced expression levels of several epidermal differentiation markers. Collectively, these data indicated that Hoxb13 might function to promote epidermal differentiation and renewal in adult skin.
To further investigate the role of Hoxb13 in epidermal differentiation, in this study, we describe the effects of Hoxb13 overexpression in an organotypic system that employs rat epidermal keratinocytes (REKs) (26,27). There are three significant advantages to using this rodent cell line as opposed to primary human keratinocytes or immortalized human cell lines. First, REKs can be stably transfected. Second, when lifted to the air-liquid interface, REKs fully stratify and differentiate, expressing all of the morphological and molecular markers normally observed in epidermis in vivo (25). Third, REKs grow and stratify without any need for fibroblast coculture. Here we show that overexpression of Hoxb13 dramatically alters the normal REK differentiation program as evidenced by abnormal accumulation of the cornified envelope, complete loss of the suprabasal granular layer, and several other morphological abnormalities that can also be observed in diseased skin. Together, these data support a role for Hoxb13 in suppressing proliferation, promoting apoptosis, and driving epidermal differentiation, and they strengthen the possibility that Hoxb13 is involved in wound repair and in the pathogenesis of certain skin diseases.

EXPERIMENTAL PROCEDURES
Generation of REK-Hoxb13 Stable Cell Lines-Full-length Hoxb13 in the pCMV-SPORTS6 expression vector (Open Biosystems) was subcloned into the retroviral vector pLPC. pLPC is a derivative of pLNCX (BD Biosciences). Either pLPCHoxb13 or pLPC alone was cotransfected with pCGP (encoding gag-pol) and pVSV-G (Stratagene) into 293T cells. At 48 h post-transfection, viral particles were harvested, filtered, and mixed with an equal volume of growth medium in the presence of 4 g/ml polybrene. 1 ml of the mix was added onto 5 ϫ 10 3 REK cells per well in a 24-well plate and the cells spin-inoculated at 2500 rpm for 1 h at room temperature. At 48 h post-transduction, the cells were trypsinized and seeded into 35-mm tissue culture dishes. Puromycinresistant cells were obtained after a week of selection in 2 g/ml puromycin. Cells were pooled and seeded onto three 96-well plates at a density of one cell/well and grown for 2 weeks in medium containing 1 g/ml puromycin. The presence of the retroviral vector or of Hoxb13 in the resultant clones was confirmed by RT-PCR.
Detection of Hoxb13 in Transiently Transfected REKs-Hoxb13 expression and cellular localization were examined by transiently transfecting REKs with a pLPCGFP-Hoxb13 fusion construct or with pLPCGFP alone using GenePORTER™ Transfection Reagent (Gene Therapy Systems, Inc.). Cells were examined using a Leica DMIL microscope and images captured with a Nikon E990 digital camera.
REK Lift Cultures-REK lift cultures were done according to the method described in Passi et al. (25) and Tammi et al. (27). Briefly, the collagen substrate was prepared using rat-tail collagen type I (BD Biosciences). 1 ml of the soluble collagen mixture was transferred into each tissue culture well on a nylon insert (Costar Transwell; 24-mm diameter, 3.0-m pore size), and polymerized for 3 h at 37°C. For preparation of the basement membrane, Maden Darby canine kidney (MDCK) cells were directly plated onto the collagen surface, grown for 18 -22 days, lysed, and removed. Control pLPC REKs or pLPCHoxb13 REKs were plated onto the basement membrane surface covering the collagen substrate, grown in submerged culture for 2 days, then raised to the air-liquid interface, and incubated for an additional 5 days. H&E Staining and Immunohistochemistry-5-day REK lift cultures were fixed in Histochoice Tissue Fixative MB (Amresco, code H120-1L) overnight at 4°C, dehydrated in isoparaffinic alcohols (Richard-Allan Scientific), and embedded in paraffin. 5-m sections were cut and then dried overnight at 40°C. H&E staining and immunohistochemistry were done using standard protocols. Primary antibodies utilized in this study were rabbit monoclonal antibody Ki-67 (1:200, NeoMarker, Inc.), rabbit anti-mouse K10 (1:500, BabCO), goat anti-rat filaggrin (1:2000, a generous gift from Dr. Beverly Dale-Crunk), and rabbit anti-mouse loricrin (1:500, BabCO). Immunofluorescent detection of HA was performed using aggrecan-derived biotinylated hyaluronan-binding protein (HABP; Seigagaku) as previously described (25,27). Sections were mounted in Vectashield (Vector Inc.), viewed with an Olympus BX50 microscope with epifluorescence attachments, and images captured digitally using a Polaroid DMC-2 camera system.
TUNEL Staining-Sections were labeled using the In Situ Cell Death Detection kit, POD (Roche Applied Science) per the manufacturer's protocol with minor modifications. Tissue sections were permeabilized with nuclease-free proteinase K (20 g/ml in 10 mM Tris-HCl, pH 7.4; Sigma-Aldrich) following incubation with the TUNEL reaction. Sections were mounted in Vectashield and viewed under epifluorescence as described above.
Preparation of mRNA and cDNA-Total mRNA was prepared using TRIzol reagent (Invitrogen) as per the manufacturer's protocol and quantified. First-strand cDNA synthesis was done using the Super-Script TM RNase H-Reverse Transcriptase kit (Invitrogen) following the manufacturer's protocol. 2 g of total mRNA were used with either oligo(dT) 20 primers (50 M; Invitrogen) or random oligo primers (Invitrogen). Identical controls with no reverse transcriptase were prepared at the same time.
Preparation of Protein Samples-REK tissues from submerged or lift cultures were harvested, lysed, centrifuged, and the supernatants transferred to fresh tubes. Protein concentrations were determined using the Bio-Rad protein assay based on the method of Bradford.
Corneocyte Analysis-Quantification of cornified envelope formation was done according to the method of Steven and Steinert (28). pLPC or pLPCHoxb13 REKs were plated at an equal density. After a 48 h period, cells shed into the culture medium were combined with those released from the dish by trypsinization, counted, collected by centrifugation, and lysed. Aliquots of the lysed samples were placed on a hemacytometer, and the total numbers of corneocytes determined for each clone.
Data from the pLPC and pLPCHoxb13 clones were pooled separately, and the results presented as percent cornified cells per total cell number.
Fluorescein-Cadaverine Uptake Study-pLPC control and pLP-CHoxb13 clones were grown in lift culture as described above. At 5-days post-lift, medium containing 100 M fluorescein-cadaverine (CadavF, Molecular Probes) or an equimolar amount of fluorescein (Molecular Probes) was added to the cultures, which were then incubated an additional 4 h and mounted in OTC (Electron Microscopy Sciences). 5-m frozen sections were air-dried and fixed in methanol. Sections were mounted in Vectashield, viewed with an Oympus BX50 microscope, and images captured using a Polaroid DMC-2 camera system.
Statistics-All statistical analyses were performed using the Stu-dent's t test. p Ͻ 0.05 was considered statistically significant. Values are presented as means Ϯ S.E.

Hoxb13 Is Differentially Expressed in REK Lift
Cultures-To begin our study of Hoxb13 in epidermal differentiation, we examined the expression pattern of Hoxb13 in REK lift cultures over time (Fig. 1A). At time 0 of lift, Hoxb13 is expressed at relatively low levels. By 3 h, the expression level increases, reaching a maximum at 12 h post-lift. After this point, Hoxb13 expression levels gradually subside, and by 120 h post-lift, a time when REKs are fully stratified (25), no significant message is detected. The dynamic expression pattern of Hoxb13 in REK lift culture suggests that it may participate in REK differentiation.
Hoxb13 Translocates to the Nucleus and Is Overexpressed in Stably Transfected REKs-A large body of evidence supports a role for Hox proteins as transcription factors. Nuclear localization of Hoxb13 in REKs was evaluated in cells transiently transfected with a retroviral vector (pLPC) expressing either a GFP-Hoxb13 fusion protein, or GFP alone. In cells expressing only GFP, fluorescence was restricted to the cytoplasm (Fig.  1B, left panel). In cells expressing GFP-Hoxb13, fluorescence was strictly nuclear (Fig. 1B, right panel), supporting a role for Hoxb13 as a transcription factor in keratinocytes. We then generated stable REK clones carrying the pLPC vector alone (control clones) or full-length Hoxb13 (experimental clones). Six clones were chosen for analysis; control clones 10, 25, and 30 and Hoxb13 clones 13, 36, and 41 (Fig. 1C). The clones were grown for 5 days in lift culture and then evaluated for Hoxb13 expression. Whereas there was no detectable message at day 5 post-lift in the vector-only controls (similar to untransfected cells, e.g. Fig. 1A), Hoxb13 was highly expressed in the experimental clones at day 5 post-lift (Fig. 1C).
Overexpression of Hoxb13 in REK Lift Culture Dramatically Alters the Normal Differentiation Program-When grown at the air-liquid interface, REKs stratify and differentiate, producing the same morphologically distinct layers of the epidermis that are observed in vivo (25,27). We first determined that the differentiation program was not altered in REKs stably transfected with the vector alone. Fig. 2A shows an H&E-  stained section of a vector-only control (clone 10) obtained after 5 days in lift culture. The normal morphological phenotype, identical to that observed in untransfected REKs (25), features a well formed basal, spinous and cornified layer ( Fig. 2A), and the presence of discrete keratohyalin granules in the granular layer (white arrows). Identical results were obtained with control clones 25 and 30 (data not shown). In contrast to the vector-only controls, overexpression of Hoxb13 in lift culture, as shown with clones 13 and 36, profoundly affected the overall epidermal phenotype (see Fig. 2, B, BЈ and C, CЈ). The first notable change was an increase in overall thickness in the Hoxb13-overexpressing tissue. Other abnormal phenotypic features were a severely disorganized basal layer (yellow asterisks), an occasional suprabasal mitotic figure (m), complete absence of a granular layer (upward white arrows) and an abnormal accumulation of what appeared to be cornified tissue (white asterisks) in which numerous nuclei were retained (parakeratosis; p).
Overexpression of Hoxb13 Results in Decreased Proliferation and Increased Apoptosis-Given the apparent overall thickening of the Hoxb13-overexpressing lift cultures, we asked whether increased cellular proliferation versus decreased cellular loss (via retention of cornified keratinocytes) might be responsible for the increased epidermal thickness observed. To evaluate the effect of Hoxb13 on cell division, we stained 5-day lift tissue with the proliferation marker, Ki-67 (Fig. 3, A and B). We observed a 2-fold reduction in the number of proliferating cells in the Hoxb13-overexpressing clones, as quantified by counts of Ki-67-positive cells (Fig. 3F). Clearly, epidermal thickening in the Hoxb13 lift cultures cannot be attributed to enhanced cell proliferation in the Hoxb13 clones. Rather, suppression of cell proliferation by Hoxb13 would be consistent with a role in promoting growth arrest and subsequent cellular differentiation, as has been reported for Hoxb13 in epithelial cells of the prostate (29).
Another striking abnormality observed in Hoxb13-overexpressing lift cultures was the presence of numerous apoptotic cells, characterized on H&E stains by their shrunken size, condensed nucleus, and pink cytoplasmic staining (e.g. see Fig.  2, B and CЈ). To confirm this morphologic finding, we used TUNEL staining, which demonstrated a large increase in cells with degraded nuclear DNA (compare Fig. 3, C and D). In the vector-only tissue, as shown in clone 10 (Fig. 3C), a few TUNEL-positive cells were detected in the region of the granular/cornified layers, as expected. In contrast, Hoxb13 clones, as shown in clone 36 (Fig. 3D), showed numerous cells that were TUNEL-positive in both the basal and suprabasal layers. Cell counts of all the clones demonstrated a significant, Ͼ5-fold increase in the prevalence of apoptotic cells in the Hoxb13driven cultures (Fig. 3E).
The Detection of Epidermal Differentiation Markers Is Significantly Reduced in Hoxb13-overexpressing REKs-We previously reported that the expression levels of several epidermal differentiation markers were significantly reduced in murine Hoxb13 knock-out skin as assayed by microarray analysis (24). Here, we utilized immunohistochemistry, RT-PCR, and Western blot analyses to examine the effects of Hoxb13 overexpression upon epidermal marker expression in REK 5-day lift cultures. Fig. 4A shows immunostaining for the intermediate differentiation marker keratin 10 (K10) and the late differentiation markers filaggrin and loricrin. In the vector-only control (clone 10), K10 staining was robust and confined to the suprabasal and granular layers, excluding the cornified layer. Filaggrin and loricrin staining were confined to the region of the granular layer. The expression patterns for all three proteins are identical to those seen in untransfected REKs and in mouse epidermis in vivo (data not shown). Similar results were obtained with the other vector-only controls (data not shown). In contrast to the control clones, expression levels of K10, filaggrin, and loricrin in the Hoxb13 overexpression clones were significantly reduced after 5 days in lift culture (Fig. 4A, clones  13 and 36). Occasional areas of detectable staining were focal and sparse when examining the entire length of the tissue.
Western blot analyses of protein from the 5-day lift tissue gave results in agreement with the immunohistochemical data. Abundant K10, filaggrin, and loricrin signals, detected in the three control clones, were lost in the Hoxb13-overexpressing clones (Fig. 4B). An analysis of K10 and filaggrin transcriptional activity revealed that the mRNA levels were the same or only slightly reduced in Hoxb13-overexpressing clones, as compared with controls (Fig. 4C). Thus, mRNA levels in individual Hoxb13 clones did not appear at face value to correlate with the amounts of protein detected in the Western blot analyses. For example, the intensity of RT-PCR bands for filaggrin was similar in the Hoxb13 clones to that in the control clones, yet little to no protein was detected on the Western blot (compare Fig.  4B with Fig. 4C). A reasonable hypothesis to explain this finding (discussed below) is that Hoxb13 induces TGase activity, causing proteins to become cross-linked into the insoluble cornified envelope, with subsequent loss or masking of the protein epitopes.
Hyaluronan Levels Are Significantly Decreased in 5-Day Hoxb13 Overexpression REK Lift Cultures-We have previously shown that Hoxb13 knock-out mice display higher levels of HA in the epidermis and dermis (24). To evaluate the effects of overexpression of Hoxb13 in REK lift tissue, we treated sections from REK lift cultures with a biotinylated hyaluronan binding protein. In control clones, high levels of HA were present between cells of the basal layer (Fig. 5A) as previously reported (25,27). In contrast, HA levels were substantially reduced in the basal layer of the Hoxb13-overexpressing clones (Fig. 5B). Some accumulation of HA was observed in the suprabasal layers of the Hoxb13 clones (Fig. 5B); this is most likely related to an ER stress response (30) in the tissues experiencing forced Hoxb13 expression (see "Discussion").

Overexpression of Hoxb13 in Submerged REK Cultures Induces Corneocyte Formation and Cross-linking-
The presence of what appeared to be excessive cornified tissue in Hoxb13 clone lift culture and lack of detection of cornified envelope precursors suggested that Hoxb13 may be influencing the production of cornified envelopes. TGases are a family of enzymes that assemble differentiation-and cell death-associated structures, including the cornified envelope, through the formation of isopeptide bonds (7). The resulting cross-linked products are highly resistant to mechanical challenge, proteolytic degradation, and immunodetection. Formation of corneocytes, the final stage in epidermal differentiation, is normally induced in REK cultures by exposure at the air-liquid interface, but also occurs spontaneously in scattered stratifying islands in monolayer cultures. To test the possible influence of forced Hoxb13 expression on TGase activity, we evaluated submerged cultures for the presence of corneocytes, using a method of hypotonic lysis (Fig. 6). Corneocytes, the only cells that withstand the lysis intact, appear as round gray ghosts. In control clones, as shown in clone 10, only occasional individual corneocytes were observed (Fig. 6A, arrowheads). However, in the Hoxb13 overexpressors, as shown in clones 13 and 36, we not only observed individual corneocytes (Fig. 6, B and C, arrowheads) but more frequently large clusters of corneoctyes that were impervious to lysis, suggesting extensive cross-linking (Fig. 6, B and C, asterisks). Evaluation of all clones confirmed that corneocyte formation was significantly higher in the Hoxb13-overexpressing cells as compared with the vector-only controls (Fig. 6D).
Overexpression of Hoxb13 in REK Lift Cultures Does Not Influence TGase Transcriptional Activation-Results thus far suggested that Hoxb13 leads to increased TGase activity in the overexpressing clones. Two possible routes to such enhanced enzymatic activity could be: 1) activation of transcription/ translation of one or more of the TGase genes, or 2) posttranslational modulation of one or more of the TGase enzymes. Of the TGases identified to date, at least four are expressed in stratified squamous epithelia such as the epidermis; TGases 1, 2, 3, and 5. Semi-quantitative RT-PCR was used to evaluate mRNA levels for each of these TGases in lift cultures of control versus Hoxb13-overexpressing cells. No differences in transcript levels for TGase 1 and 3 were detected (Fig. 7). The mRNA levels for TGase 2 and TGase 5 actually decreased in the overexpressing clones, relative to controls. Overall, these results indicate that increased cross-linking activity observed in Hoxb13-overexpressing clones is not occurring through induced accumulation of specific TGase mRNAs.
Overexpression of Hoxb13 Results in Increased Fluorescein-Cadaverine Uptake in 5-Day REK Lift Cultures-To determine whether Hoxb13 influences TGase enzymatic activity, we supplemented the culture medium in REK lift cultures with Ca-davF or with an equimolar amount of fluorescein alone, 4 h prior to harvest of the lift tissue. Cadaverine serves as a TGase substrate. One representative vector-only control (clone 10) and two Hoxb13-overexpressing clones (clones 13 and 36) were evaluated. As shown in Fig. 8A, CadavF incorporation in clone 10 was specific and limited to a region corresponding to the granular/cornified layer as demonstrated by comparison with a phase-contrast image of the same region (Fig. 8AЈ). This is the epidermal region where TGase activity is expected. In some regions of the tissue, CadavF uptake was discontinuous but still limited to the granular/subcorneal region (Fig. 8, B and BЈ). In contrast, CadavF incorporation was dramatically increased in Hoxb13 clones (Fig. 8, C and D) showing areas of oversaturation (upward arrows) when images were taken at the same exposure as for the control clone. CadavF uptake in the Hoxb13-overexpressing lift tissue was discontinuous along the entire length of the culture. Lift cultures treated with fluorescein alone were negative for incorporation (data not shown).

Fluroescein-Cadavarine Incorporation in Hoxb13-expressing Lift Cultures Coincides with Regions of Thickened Cornified
Tissue-In unstained tissues examined under phase contrast, it appeared that areas of intense CadavF incorporation coincided with dense, possibly cornified regions, in the uppermost epidermal layers (Fig. 8, CЈ and DЈ). To ascertain whether areas of CadavF incorporation actually corresponded to areas of hypercornification, we H&E-stained sections from lift cultures of Hoxb13-overexpressing clones that had previously been treated with CadavF. If the reagent were indeed tightly cross-linked into cornified tissue, then it should be able to withstand the rigors of this additional staining procedure. Results are shown in Fig. 8, E to FЈ. Comparison of fluorescent images from CadavF-treated Hoxb13 lift cultures (Fig. 8, E and  F) with light microscopic images of the same sections (Fig. 8, EЈ and FЈ) shows that the areas of fluorescence correspond exactly to the regions of thickened cornified tissue (asterisks). Taken together, the data indicate that overexpression of Hoxb13 in REK lift cultures results in heightened TGase activity and excessive cross-linking, with abnormal accumulation of cornified envelope.

DISCUSSION
Hox genes function as critical regulators of cell specification and fate determination during embryonic development across the animal kingdom. Once this task is accomplished, Hox genes do not return to a quiescent state but continue to be expressed in almost every organ of the adult body. Yet little is known about how these master regulators of differentiation might function in the maintenance and renewal of tissues in the FIG. 6. Hoxb13 overexpression in submerged REKs results in excessive cornified envelope formation. Medium plus trypsinized cells were centrifuged and the resulting pellet lysed. A, lysed cell preparation from vector-only control clone10. Corneocytes appear as large gray ghost cells (arrowheads), among broken cell debris. B and C, lysed cell preparations from Hoxb13-overexpressing clones 13 and 36. Individual corneocytes are observed (B and C, arrowheads), along with groups of cells resistant to lysis suggesting extensive corneocyte cross-linking (B and C, asterisks). D, total cell numbers prior to lysis, and corneocyte cell numbers after lysis, were evaluated by counting with a hemocytometer; triplicate independent counts were performed on each of the 3 control clones and the 3 Hoxb13 clones. Pooled results (mean Ϯ S.E., fraction of corneocytes relative to total cells prior to lysis) show that Hoxb13-overexpressing cultures generate a significantly higher number of corneocytes; p Ͻ 0.0001. adult. The involvement of Hox genes in the renewal of adult tissues is of particular interest, because aspects of the molecular program that generated the tissue during development are likely recapitulated during post-neonatal life. The epithelia of the skin (23), gut (31), and prostate (18,31), all of which express Hoxb13, are prime examples of tissues that are continually renewed throughout the life of the adult.
We previously reported that the expression levels of several epidermal differentiation-regulated genes are significantly decreased in Hoxb13 knock-out adult skin (24). This led us to hypothesize that one role for Hoxb13 in skin is to promote epidermal differentiation. In this study, we evaluated this possibility by examining the effects of constitutive overexpression of Hoxb13 in a REK organotypic lift culture system. The data show that excessive Hoxb13 activity produces several abnormal phenotypic changes in the differentiated tissue. The most remarkable change, and the strongest indicator that Hoxb13 is driving differentiation, is the presence of excessive and ectopic cornified tissue in Hoxb13-overexpressing lift cultures (Fig. 2). In the epidermis, cornified envelope formation normally occurs late during terminal differentiation, and is characterized by cross-linking of cornified envelope precursors (such as loricrin, involucrin, etc.) through the action of calcium-dependent TGases (discussed below).
Other phenotypic changes driven by Hoxb13, which include an apparent reduction in epidermal differentiation proteins (Fig. 4) and loss of the granular layer (Fig. 2), are also consistent with accelerated differentiation. When monolayer cultures of murine keratinoctyes are stimulated to differentiate with relatively high levels of calcium, later differentiation events (such as cornified envelope formation) occurs at the expense of earlier differentiation events (such as the expression of K1 and K10) (32,33). Hoxb13 overexpression may be inciting a similar phenomenon, wherein the expression of genes for early keratinocyte differentiation becomes curtailed by a massive prodifferentiating stimulus.
Other interpretations are also possible. Whereas expression of K10 and filaggrin does decrease at both the mRNA and protein level in Hoxb13-overexpressing REKs, the decline in mRNA expression is relatively small compared with a much larger reduction in measurable K10 and filaggrin protein (Fig.  4). At least two possible explanations might explain these data. First, K10, filaggrin, and loricrin protein may be produced in relative abundance but then become immediately cross-linked by TGase. In this case, the cross-linked tissue would be impermeable to antibodies, or might remain in the pellet during protein preparation. The second possibility is that K10, filaggrin, and loricrin protein levels are indeed greatly reduced as a result of Hoxb13 overexpression, suggesting that other proteins are being aggressively cross-linked. Indirect support for this observation comes from the loricrin knock-out mouse. Loricrin normally makes up 60 -80% of the total protein content of the cornified envelope (28,34). However, loricrin-null mice actually make a quasi-normal cornified envelope, implying that alternative protein substrates can be employed in the absence of loricrin (35,36). We do not yet know the protein content of cornified cell envelopes in the Hoxb13-overexpressing epidermal tissue.
The increases in cross-linked corneocyte formation observed in Hoxb13-submerged cultures (Fig. 5) and lift REK cultures at regions corresponding to increased TGase substrate incorporation (Fig. 7), indicate that Hoxb13 is driving the final crosslinking stage of terminal differentiation. Establishment of the cornified envelope is normally accomplished through the catalytic activity of TGase by means of the formation of covalent epsilon-(␥-glutamyl) lysine protein-protein bonds (6,37).
TGases are involved in the formation of differentiated structures and require the binding of Ca 2ϩ for their activity, usually at concentrations well above the typical range for most other intracellular processes. Keratinocytes express several types of transglutaminases, including TGase 1, TGase 2, TGase 3, and TGase 5, but TGase 1 appears to be the major player in corneocyte assembly in the differentiating keratinocyte (7). An examination of TGase 1, 2, 3, and 5 mRNA levels in lift cultures (and in submerged cultures; data not shown) revealed no increase in expression between vector-only controls and Hoxb13overexpressing clones (Fig. 6), indicating that Hoxb13 is upregulating TGase activity post-transcriptionally. Whereas these data do not rule out the possibility that Hoxb13 may be transcriptionally activating some other TGase gene(s), it seems unlikely, because only TGase 1, 3, and 5 are known to induce cornification in epidermal keratinocytes.
At this time, we do not know how overexpression of Hoxb13 leads to increased TGase activity. Several possibilities exist. TGases require the binding of Ca 2ϩ for their cross-linking function, and in keratinocyte culture systems the extent of cornification and TGase activity correlate with Ca 2ϩ concentration of the medium (38,39). Overexpression of Hoxb13 in REK lift cultures may increase the amount of available Ca 2ϩ , leading to heightened TGase activity and excessive cross-linking. Overexpression of Hoxb13 could also lead to disproportionate TGase activity by directly or indirectly up-regulating other molecules, which then act upon TGase. One theoretical possibility is tazarotene-induced protein (TIG3), described in humans to be expressed in the suprabasal epidermis (40). Human foreskin keratinoctyes overexpressing TIG3 exhibit reduced proliferation, enhanced differentiation, increased apoptosis, and heightened TGase activity (41), similar to the situation observed with Hoxb13 overexpression in REKs. At present, no rodent homolog to human TIG3 has been described. A third possibility is that Hoxb13 could be influencing TGase activity by promoting its proteolysis. During terminal differentiation, full-length TGase 1 proteins undergo proteolytic cleavage, leading to ϳ100 to 1000-fold higher specific activity (42,43). Calpain, a Ca 2ϩ -dependent protease, may activate TGase 1 in vivo (44,45), and microarray analyses of Hoxb13 knock-out versus wild-type skin (24) revealed that calpain expression is significantly reduced in the Hoxb13 knock-out (data not shown). Finally, it is interesting to note that another Hox protein, HoxA7, down-regulates TGase 1 expression at the transcriptional level, and inhibits differentiation (46). It is tempting to speculate about a counteractive network of Hox genes, acting together to fine tune epidermal differentiation.
In addition to our results, several other studies and observations support a role for Hoxb13 in promoting terminal epidermal differentiation. (i) In the murine caudal spinal cord and tail vertebrae, Hoxb13 functions as an inhibitor of cell proliferation, activator of apoptotic pathways, and promoter of terminal differentiation (19). (ii) In mouse cutaneous wounds, Hoxb13 epidermal expression is up-regulated at a time when epidermal proliferation is down-regulated and terminal differentiation is up-regulated (24). (iii) HA, which has been shown to inhibit differentiation in REK cultures (25) and is also thought to inhibit differentiation in vivo (47), is significantly reduced in the basal layer of Hoxb13-overexpressing REK lift tissue. Parenthetically, cells undergoing stress produce high levels of HA (30). We believe that the increased HA levels in the suprabasal layers of the Hoxb13-overexpressing lift tissue compared with the controls is the result of ER-stress and is independent of a role for Hoxb13 in differentiation. (iv) Forced expression of Hoxb13 in PC3 (29) and LNCaP (48) human prostate epithelial cancer cell lines induces terminal differen-tiation. (v) Hoxb13 expression is up-regulated during calciumand confluency-induced differentiation of human foreskin keratinocytes (20). As a critical regulator of epidermal formation and renewal, Hoxb13 cannot be the only gene involved because its loss does not affect the health of the skin in Hoxb13 knockout mice. Yet it remains attractive to think that Hoxb13 may be functioning in concert with other Hox genes to help orchestrate epidermal differentiation.
Exactly which genes are regulated by Hoxb13 to affect terminal differentiation, and the mechanisms by which Hoxb13 regulates these genes, remains an open question. The bulk of the evidence suggests that Hoxb13 acts as a transcription factor. Various studies in which epitope-tagged versions of the Hoxb13 protein were overexpressed as a GFP fusion (Fig. 1 in this study), as a ␤-gal fusion (18), or as a FLAG fusion (48) all showed strong nuclear localization. The one study so far that examined Hoxb13 expression in adult native skin by immunochemistry (20) reported Hoxb13 expression both in the cytoplasm (strong) and in the nucleus (weaker), a pattern very similar to two other classes of transcription factors involved in epidermal differentiation, namely C/EBP (33,49) and NF-kB (50). For all of these factors, the signals and interactive partners that determine the protein intracellular locations in native tissue have yet to be fully explored.
As a final point, several skin diseases are characterized by aberrant differentiation. Forced expression of Hoxb13 resulted in the absence of a granular layer, parakeratosis (retained nuclei in the cornified envelope), and increased apoptosis; these histopathologic features are typical in psoriasis vulgaris, a common skin disorder (51,52). Overexpression of Hoxb13 also produced several changes common to Bowen's disease, a precancerous form of squamous cell skin cancer. These include hyperkeratosis (excessive cornified tissue), parakeratosis, and disordered epidermal architecture (53). Another feature of Bowen's disease histopathology is the presence of multinucleated epidermal cells containing clusters of nuclei (53), a feature that we observed in submerged cultures of Hoxb13-overexpressing REKs (data not shown). Together these data raise the possibility of the involvement of Hoxb13 not only in promoting epidermal differentiation, but in skin pathogenesis as well.