Up-regulation of p27^Kip1, p21^WAF1/Cip1 and p^16Ink4a is associated with, but not sufficient for, induction of squamous differentiation

Squamous differentiation is a multi-step process in which irreversible growth-arrest is an early event followed by the sequential expression of many squamous cell-specific genes, including keratin 1, involucrin, cornifin α/SPRR-1 and transglutaminase type I (TGase I). These sequential events take place in epithelial tissues as cells in the basal layer lose their adhesiveness to the basement membrane (Jones and Watt, 1993) and move into suprabasal layers. Cessation of DNA synthesis and the appearance of the earliest markers of squamous differentiation occur in the innermost suprabasal layer (see Fuchs and Byrne, 1994, for review). NHEKs in culture have been used as a model system for in vivo squamous differentiation. Cell confluency, treatment with phorbol esters (Hawley-Nelson et al., 1982) or interferon γ (IFNγ; Saunders and Jetten, 1994) or growth in suspension (Adams and Watt, 1989) initiate effective signals which induce NHEK growth arrest and differentiation. Selective loss of cell:substratum adhesiveness and alterations in cell:cell interactions are early events during epidermal differentiation in vivo as well as in vitro (Adams and Watt, 1990; De Luca et al., 1990). Down-regulation of α6β4 integrins which bind to laminin in the basement membrane coincides with early differentiation (Tennenbaum et al., 1996) and expression of dominant-negative E-cadherin has been reported to inhibit proliferation and stimulate differentiation in NHEKs (Zhu and Watt, 1996). Such changes could serve as important signals that initiate the process of terminal differentiation since integrin-mediated signaling has been shown to contribute to the proliferative signal in NHEKs and prevent squamous differentiation. Links between adherence and the regulation of cell growth have also been indicated by the modulation of cadherin and integrin levels by growth factors (Sastry and Horwitz, 1996, and references therein) and the requirement for adhesion in cell cycle progression, in which adhesion promotes the expression of cyclin A and cyclin D1 (Böhmer et al., 1996; Schulze et al., 1996) and represses the levels of p21 and p27 protein which regulate cyclin E/cdk2 kinase activity (Fang et al., 1996; Zhu et al., 1996).

External signals which affect proliferation and growth arrest in mammalian cells have been shown to exert their influence primarily in regulating progression through the G₁ phase of the cell cycle (Graña and Reddy, 1995; Sherr, 1995). This progression is controlled by a series of cyclin-dependent kinases, cdk2, cdk4 and cdk6, which phosphorylate and inactivate the Rb family of tumor suppressor/growth inhibitors which include Rb, p107 and p130 (reviewed by Weinberg, 1995). The active, hypophosphorylated forms of Rb and p130 prevent the expression of a number of gene products, such as cyclin A, cdc2 and E2F-1, which are necessary for cell cycle progression (see Müller, 1995, for review). Cdk2 may also control progression into S phase via other critical target proteins (Heichman and Roberts, 1994; Strausfeld et al., 1994; Jackson et al., 1995; Hofmann and Livingston, 1996). Cdk activity is controlled at several levels (see Morgan, 1995, for review), including protein expression, phosphorylation and inhibition by small proteins that fall into two groups, the Ink and Kip families. Inhibition by the Ink family is limited to cdk4 and cdk6, while the Kip family can inhibit the activity of all three G₁ cdks (see Sherr and Roberts, 1995; Harper and Elledge, 1996, for reviews). Increases in the inhibitors p15^Ink4b, p21^Cip1/Waf1 or p27^Kip1 have been demonstrated during growth arrest in a large variety of systems (see Scherr and Roberts, 1995, for review; Koyama et al., 1996; Wu et al., 1996; Harvat et al., 1997).

It has been assumed that control of irreversible growth arrest and expression of a differentiated phenotype are closely linked in cells undergoing terminal differentiation (Scott, 1997). This is demonstrated by the requirement of Rb in normal myogenesis (Novitch et al., 1996) and activation of adipogenesis (Chen et al., 1996). In addition, combined expression of p16 and p21 activates muscle specific genes (Skapek et al., 1995) and expression of p27 can drive neuronal differentiation (Kranenburg et al., 1995). In a number of cell systems, increases in p21 or p27 protein levels have been associated with differentiation (Lois et al., 1995; Missero and Dotto, 1996; Wang et al., 1996; Durand et al., 1997, and references therein). However, in contrast to adipogenesis and myogenesis, little is known about the molecular events associated with irreversible growth arrest during squamous differentiation and how they relate to the expression of squamous cell-specific genes.

In this study, we examined the expression and activity of several cell cycle regulatory proteins, including Rb, p130, cdks and cdk inhibitors, during the commitment to squamous cell differentiation. We show that the irreversible growth arrest during squamous differentiation is associated with an inhibition of Rb and p130 hyperphosphorylation and cdk2 and -4 kinase activities. Furthermore, transient increases in p21^WAF1/Cip1 and increases in the levels of p130, p27^Kip1 and p16^Ink4a occur in a differentiation-specific manner, which raises the possibility that the cells may enter a distinct terminal growth-arrest state. A tight association between the expression of p27 and the crosslinked-envelope protein, cornifin α, was observed, suggestive of a close link between growth arrest and differentiation. However, over-expression of each of the cdk inhibitors, although sufficient to inhibit growth, did not promote the expression of cornifin α, suggesting that additional signals that can be activated by suspension culture are needed to induce differentiation.

Primary normal human epidermal keratinocytes (NHEK) were grown as described (Saunders and Jetten, 1994). For non-adherent growth, NHEKs grown to 60-70% confluency were trypsinized and placed in keratinocyte growth medium in culture dishes which had been coated with a layer of medium-conditioned, sterile 2% agarose. For colony forming efficiency assays, cells were placed in suspension culture for various times, then pelleted and replated in regular tissue culture dishes. After 48 hours, the number of colonies in a 4 cm² area with >4 cells were counted.

These procedures have been described previously (Harvat and Jetten, 1996; Harvat et al., 1997). Purchased antibodies include cdk2, cyclin D1 and cyclin E rabbit polyclonal antibodies from Upstate Biotechnology, Inc. (Lake Placid, NY); c-myc and p53 mouse monoclonal and Rb, p130 and cdk6 rabbit polyclonal antibodies from Santa Cruz Biotechnology, Inc. (Santa Cruz, CA); p27Kip1 mouse monoclonal antibody from Transduction Laboratories (Lexington, KY); Waf-1 mouse monoclonal antibody from Oncogene Science (Cambridge, MA); involucrin antibody (1:2,000) from Novocastra (Vector Laboratories, Burlingame, CA) and keratin I antibody (1:250) from Berkeley Antibodies (Richmond, CA). The antibody to cornifin α has been previously described (Marvin et al., 1992). The TGase I goat polyclonal antibody (1:50) was a gift from Peter Steinert (Kim et al., 1995), cdk4 rabbit polyclonal came from Yue Xiong (University of North Carolina, Chapel Hill, NC), p16/15 (JC6) mouse monoclonal from Jim Koh (Mass. General Hospital, Boston) and cyclin A rabbit polyclonal from Michele Pagano (New York University, NYC). Secondary antibodies used were peroxidase-conjugated mouse or rabbit anti-IgG (Fc fragment or species absorbed) from Chemicon (Temecula, CA).

Adherent and suspension cells were harvested, respectively, by trypsinization and centrifugation. Suspension cells were treated with trypsin for 2-4 minutes to dissociate cell clumps. Cells were then rinsed in cold phosphate buffered saline (PBS) once, suspended in 1 ml PBS, fixed for 15 minutes in 1% paraformaldehyde, washed once with PBS containing 1% bovine serum albumin (BSA) and resuspended in 1 ml PBS/1% BSA followed by the dropwise addition of 3 ml of ice-cold ethanol. Cells were then stored overnight at −20°C. Cell pellets were divided as needed and incubated with either p27/cornifin antibody or PBS/1% BSA for 1 hour, rinsed once, and incubated with either fluorescein (FITC) or phycoerythrin (PE)-conjugated secondary anti-mouse or rabbit IgG antibody for an additional hour. The cells were then double-labeled using another primary antibody and a secondary antibody with a different label or were treated with RNAse and propidium iodide for DNA staining using the flow cytometric analysis reagent set from Boehringer Mannheim (Indianapolis, IN). In experiments where BrdU incorporation was measured, BrdU was added to the culture medium at a final concentration of 10 μM for 30 minutes prior to harvest. Cells were harvested as above and fixed in ice-cold ethanol. BrdU detection used a FITC-conjugated anti-BrdU antibody according to the manufacturer’s instructions (Becton Dickinson, San Jose, CA). All analysis was done on a Becton Dickinson Facscan using ImageQuant software. For each sample, 10,000 or 30,000 registered events were gated on by PI (FL2-W) and forward light scatter signals for the single cell population. All data shown represent only the gated cells. All quantitation of protein expression was done with respect to baseline levels of controls treated with secondary antibody alone (data not shown).

The northern procedure and most probes have been described previously (Harvat et al., 1997). The TGase I and cornifin α probes have been described (Saunders and Jetten, 1994). The p21 cDNA probe was obtained from Dr J. Fontana (University of Maryland, Baltimore, MD) and the p130 probe from Dr J. Nevins (Duke University, Durham, NC).

Cells were plated at least 48 hours prior to infection. At the time of infection, cells from one plate were trypsinized and counted. Based on this cell count, the rest of the plates were then rinsed once in medium lacking bovine pituitary extract and exposed to 0.5 ml (35 mm dish) or 2 ml (100 mm dish) of the same medium containing adenovirus at the desired m.o.i. with occasional rocking for 75 minutes in a 37°C, 5% CO₂ incubator. Following incubation, 3 or 10 ml, respectively, of complete medium was added to each plate and the cells were further incubated for 24 to 30 hours before harvest or trypsinization and transferred into suspension culture. Control plates lacking any adenovirus were subjected to the same procedure. Recombinant viruses Adnull, Adp27 and Adp21 have been described previously (Katayose et al., 1995; Craig et al., 1997; Harvat et al., 1997). Adp16 was constructed like Adp21 except p16^Ink4a cDNA (obtained from David Beach, Cold Spring Harbor Laboratory) was used.

Squamous differentiation can be induced in NHEKs by growing cells to confluence or in suspension culture or by treatment with IFNγ or phorbol esters. All of these methods rapidly elicit an irreversible growth arrest (see Jetten et al., 1995, for review). To better understand the relationship between growth arrest and expression of the differentiated phenotype, we characterized the growth arrest and examined the kinetic relationships between biochemical and molecular markers for the two events. Flow cytometric analysis of cell cycle distribution and BrdU incorporation revealed that within 24 hours of being placed in suspension, NHEKs exited S phase, accumulating primarily in the G₁ phase of the cell cycle, although nearly 30% of the cells remained in G₂/M. There was a concomitant 84% reduction in BrdU incorporation (Fig. 1).

Extracts from adherent and suspension cells examined by immunoblot analysis for markers of G₁ arrest showed that Rb increasingly accumulated in its hypophosphorylated form (Fig. 2). Consistent with reports that Rb autorepresses its own expression (Shan et al., 1994) and E2F-driven transcription (reviewed by Müller, 1995), Rb levels were later down regulated as the majority of Rb became active, i.e. hypophosphorylated. Similar effects were observed on the expression of cyclin A, an E2F-regulated gene, while down-regulation of c-myc occurred rapidly within 2 hours (Fig. 2). Dephosphorylation of the Rb family member, p130, occurred in parallel with the dephosphorylation of Rb, while increases in the total amount of p130, primarily in its hypophosphorylated form 2, began at 12 hours and continued for the duration of the time course (Fig. 2). This form of p130 has been described as being unique to cells in G₀/early G₁ (Mayol et al., 1996), but in this system the status of both Rb and p130 reflect the state in the 30% of cells with 4N DNA content as well as those in G₀/G₁.

The states of phosphorylation of Rb and p130 reflect the activity of G₁ cyclin dependent kinases (cdks) which target these proteins (Weinberg, 1995). G₁ kinase complexes were immunoprecipitated by specific antibodies and their activity toward a GST-Rb fusion substrate was analyzed. Cdk4 kinase activity was progressively down-regulated, starting by 4 hours after placing cells in suspension and decreasing to undetectable by 24 hours (Fig. 3A). Cdk2 activity was more abruptly inhibited between 4 and 8 hours in suspension. Levels of cdk2 and cyclin D₁ remained unchanged, cyclin E increased slightly and the amount of cdk2 that was in the phosphorylated, active form remained constant up to 8 hours, although it decreased by 16 hours (Fig. 3B), suggesting that initial inhibition of cdk activity at 8 hours was not the result of a block in CAK-mediated phosphorylation of cdk2. Levels of cdk4 in suspension decreased, but were still detectable by 16 hours and could therefore only partially explain the decrease in measurable cdk4 activity.

The results from Figs 2 and 3 suggested that inhibitors of the cdk’s are active in non-adherent conditions. When we analyzed the levels of the inhibitors p27^Kip1, p21^Cip1/Waf1 and p16^Ink4a, we found that all were significantly induced in non-adherent cells (Fig. 4A). p15^Ink4b levels were extremely low and did not change significantly (data not shown). An increase in the level of p21 was observed first, reaching a 7-fold induction within four hours after placing cells in suspension in the experiment shown in Fig. 4A. p21 expression is regulated by p53-dependent and -independent mechanisms. The suspension-induced increase in p21 levels appears to be p53-independent, because levels of p53 protein started to decrease between 4 and 8 hours and disappeared completely at the time when p21 levels were the highest. p21 levels began to decline between 16 to 24 hours in suspension, the time varying somewhat between experiments (compare Figs 4 and 9). Transient increases in p21 have also been reported for calcium-induced differentiation in murine keratinocytes (Missero et al., 1995) and human ectocervical keratinocytes (Kikuchi et al., 1997).

p27 protein levels demonstrated a biphasic response with reproducible twofold increases at 4 hours and further significant increases beginning at 12 hours (7-fold) and continuing beyond 24 hours (Fig. 4A). Increases in p16^Ink4a, on the other hand, were only first detected at 12 hours after shifting cells to non-adherent conditions, reaching 28-fold by 24 hours.

The kinetics of the up-regulation of these cki’s corresponded to growth arrest as measured by inhibition of thymidine incorporation (data not shown) and disappearance of hyperphosphorylated Rb and p130. To further determine the kinetic relationship between commitment to irreversible growth arrest and kinase inhibition, we replated cells which had been in suspension culture for various periods of time and counted colony plating efficiency 48 hours later. As seen in Fig. 4B, cells which had been in suspension for 4 hours were able to reattach and establish colonies with almost the same efficiency as cells which were merely trypsinized and replated immediately. However, following 8 hours in suspension, only 10-15% of the cells were able to establish colonies, indicating that commitment to irreversible growth arrest had occurred in most of the cells. Similar findings have been reported by others (Hotchin et al., 1993). This period of time when irreversible growth arrest occurs thus coincides with the time when cdk2 activity is essentially abolished and entry into S phase blocked. In summary, in addition to inhibition of cdk activity, the terminally differentiated cell cycle state is characterized by decreases in p21, p53, Rb and cyclin A expression and increases in p27, p16 and hypophosphosphorylated p130.

Increases in p27 levels during growth arrest have been reported to be primarily controlled at the protein level instead of the mRNA level (see Scherr and Roberts, 1995, for review). Northern analysis of mRNAs encoding p27, p21, p16 and p130 demonstrated that p21 and p27 mRNA levels, normalized by GADPH, increased minimally only after 24 hours of treatment, while there was essentially no change in levels of p16 or p130 mRNA levels (Fig. 5). The significant increases in levels of these four proteins, therefore, appear to be primarily regulated at the protein level and suggest that signals induced by suspension culture may rapidly affect mechanisms such as translational activity and/or protein stability which control protein levels directly.

To evaluate the kinetic relationship between the up-regulation of cdk inhibitors/growth arrest and differentiation, we compared the induction of squamous specific markers with that of growth regulatory proteins. Cornifin α (SPRR-1) and Tgase I proteins, early markers of squamous differentiation (see Jetten et al., 1992, for review; Marvin et al., 1992; Greco et al., 1995), were first detected 12 hours after shifting cells to non-adherent conditions at the same time in which p27 and p16 protein levels increased (Fig. 4). However, keratin 1, another early marker, was not induced until after 16 hours, while involucrin was expressed in untreated cells and was further induced only after 24 hours (data not shown). The up-regulation of cornifin and TGase I protein thus accompanied the predominant up-regulation of p27 levels and the first appearance of detectable levels of p16, but was preceded by cdk2 kinase inhibition and p21 up-regulation. We have previously reported a similar kinetic association of p27 expression with cornifin expression during squamous differentiation in mammary epithelial cells (Harvat et al., 1997).

The correlation of p27 and cornifin expression was corroborated on a single cell basis, using double immunostaining measured by flow cytometry. 76% of adherent cells contained little or no p27 nor cornifin α (p27⁻/cornifin⁻, Fig. 6A). Within 12 hours after being placed in suspension, 86% of the cells expressed both proteins at significant levels (p27⁺/cornifin⁺). Furthermore, as demonstrated in Fig. 6B, there was an extremely close correlation between the relative levels of the two proteins in individual cells, i.e. the relative intensity of the p27-FITC signal had a linear relationship with the relative intensity of the cornifin-PE signal. These findings strongly supported the hypothesis of coordinate regulation of activation of the differentiation program and growth arrest.

We had determined that 30% of cells were still in the G₂/M phase of the cell cycle after 12 hours in suspension (Fig. 1). The data in Fig. 6A suggests that these cells should express cornifin. Therefore, we analyzed adherent cells and cells in suspension for both DNA content and cornifin expression simultaneously. The population of non-adherent NHEKs arrested in G₂ expressed cornifin as efficiently as those cells that were arrested in G₁ (Fig. 6C). Similar cell-cycle independent expression has been shown for involucrin (Nakatani et al., 1992). These results suggest that the differentiation signal generated by non-adherence is not restricted to cell cycle position. Alternatively, the differentiation signal could be responsive to a block in cell cycle progression regardless of cell cycle position. The relationship of this finding to events in vivo is not known, since cells in suprabasal layers in vivo have been shown to be in a G₀/G₁ state by DNA content (van Erp et al., 1989).

We next evaluated the effect of non-adherent conditions on mRNA levels of cornifin α and TGase I. By 8 hours of treatment, both cornifin α and TGase I mRNA were induced (Fig. 5, data not shown). Induction of these squamous specific mRNAs were thus preceded by an increase in p21 protein levels, a 2-fold increase in p27 levels, and inhibition of cdk2 activity and were closely followed by the appearance of the major increases in p16 and p27 protein levels.

These kinetics, while indicating a close association between molecular changes tied to growth arrest and those initiating differentiation, did not give a clear answer as to whether there is a cause and effect relationship between the two events. Elevated levels of cki’s and the resulting kinase inhibition could serve as a signal leading to increased expression of cornifin α and TGase I, placing activation of the differentiation program directly downstream of the block in proliferation. Alternatively, non-adherent culture conditions could activate parallel pathways that: (1) affect the expression of differentiation markers such as cornifin α and TGase I; and (2) promote growth arrest, the coordinate action of which leads to terminal differentiation.

In order to distinguish between the two possibilities noted above, exogenous p27 was introduced into adherent NHEKs using an adenovirus carrying the p27^Kip1 cDNA (Adp27). Infection with Adp27 resulted in over-expression of p27 and growth arrest in a dose-dependent manner. This was illustrated by a decrease in thymidine incorporation (Fig. 7A), the disappearance of hyperphosphorylated forms of Rb (Fig. 7B) and p130 (Fig. 8, compare lanes 1,3 and 6) and down-regulation of cyclin A (Fig. 7B). The adenovirus vector alone (Adnull) had little effect, although the infection procedure itself had some effect, as seen by changes in the amount of hypophosphorylated Rb and expression of p27 and cornifin in control cells going through the procedure but receiving no virus (compare Fig. 7B, lane 1, with Figs 2 and 4, lane 1). Exposure to Adp27 did not increase cornifin α or involucrin levels in NHEKs above control levels (Figs 7B and 8, lanes 1, 3 and 6). Similar results were obtained when cells were infected with recombinant adenovirus carrying either p21 or p16 cDNA (Fig. 9A, lanes 1, 3 and 5, and data not shown). The higher doses of these latter viruses required for growth arrest of NHEKs had some growth inhibitory effects by themselves, as demonstrated by changes in protein levels in cells receiving Adnull (Fig. 9A, lane 1), but still did not effectively induce cornifin or involucrin. These results suggest that increases in the cyclin-dependent kinase inhibitors and the resulting growth arrest are not sufficient to induce the expression of differentiation genes.

While not being sufficient, growth arrest induced by over-expression of a cki could have some affect on expression of squamous specific genes, perhaps by mediating early steps in a multi-step process. However, when placed in suspension, p27-expressing, growth-arrested cells demonstrated no difference from control cells or cells exposed to the Adnull virus in their ability to induce the expression of cornifin α by 16 hours (Fig. 8). In particular, neither the timing nor the level of cornifin or involucrin up-regulation in suspended cells was altered by p27-mediated growth arrest. Up-regulation was not a function of time alone, because p27-arrested cells failed to induce cornifin, even when these adherent cells were followed for an additional 16 hours (Fig. 9B). The ability of suspension to activate cornifin and TGase I also demonstrates that exposure to adenovirus did not have deleterious effects on gene expression in general. Importantly, up-regulation of p16 and hypophosphorylated p130 protein levels in p27-arrested cells, as well as further increases in p27 itself despite its high level of expression, still correlated with the induction of cornifin α in suspension culture (Fig. 8), suggesting that these events are regulated by a differentiation signal and not a growth arrest signal.

Up-regulation of squamous specific genes was similarly induced by suspension of NHEK in which p21 and p16 were over-expressed (Fig. 9A). It should be noted that p21 levels in this experiment were already decreasing by 16 hours and that the regulation of this decrease affected protein from recombinant adenovirus as well as the endogenous gene (compare lanes 1, 3 and 5 with lanes 2, 4 and 6 of the p21 band). Taken together, these results suggest that a second signal independent of cki-mediated growth arrest is generated during suspension, which is essential for the induction of squamous specific genes. This signal(s) presumably also mediates changes associated with the irreversible cell cycle state; i.e. increases in p27, p16 and p130 protein levels and a down-regulation of p21.

The relationship of growth arrest to terminal differentiation has generally been assumed to be critical, with growth arrest closely associated with or perhaps even triggering early differentiation events. Squamous differentiation in the epidermis occurs as a multistep process that begins with detachment of the cell from the basement membrane and its subsequent migration into suprabasal layers (Fuchs and Byrne, 1994 for review; Jones et al., 1995). Irreversible growth arrest occurs as cells exit the basal layer, but the precise relationships between adhesion, growth arrest and expression of the differentiated phenotype are not yet fully understood. Placement of cultured NHEKs in suspension in many respects mimics the early stages of epidermal differentiation in vivo. Cells cease to proliferate and rapidly terminally differentiate, even though kept in the presence of growth factors which support proliferation of adherent cells (Hotchin et al., 1993). Suspension culture is thus a physiologically relevant in vitro model to study the molecular mechanisms that control growth arrest, expression of squamous-specific genes and the relationship between these two processes and adhesion. Our studies show that lack of adherence promotes an irreversible growth arrest within 8 hours that is characterized by a reduction in cdk activity and transient up-regulation of the cdk inhibitor p21, followed by up-regulation of p27 and p16 as well as an increase in levels of p130 protein. Control of the level of these growth inhibitory proteins appears to be primarily at the level of rate of translation or protein stability, as their mRNA levels do not change significantly in suspension culture.

These cell cycle regulatory events were closely associated with the induction of several squamous-specific genes, supporting the hypothesis that the loss of adherence to substratum is sufficient for terminal differentiation in NHEKs and produces coordinate regulation of entry into the irreversible growth arrest state (IGA) and expression of squamous specific genes. The kinetics of the changes that are involved in suspension-induced differentiation are summarized graphically in Fig. 10. Commitment to irreversible growth arrest and initiation of transcription of squamous-specific genes occur within the same narrow window of time between 4 and 8 hours in suspension. Inhibition of cdk2 activity also occurs during this period as well as an increase in p21 and a 2-3 fold increase in 27 levels, although these do not peak until 12 and 24 hours, respectively. Rb is partially activated at this time, its phosphorylation state appearing to correlate closely with progressively diminishing cdk4 activity.

These closely associated kinetics led us to investigate further whether there was a cause/effect relationship between the up-regulation of growth control proteins and the activation of squamous-specific genes. We used recombinant adenoviruses to over-express the cdk inhibitors p27^Kip1, p21^Cip1/WAf1 or p16^Ink4a in adherent, proliferating NHEKs. Exogenous expression of each inhibitor induced growth arrest which included accumulation of hypophosphorylated Rb and down regulation of cyclin A, but was ineffective in inducing increased expression of squamous specific genes above controls. When these arrested cells were placed in suspension culture, expression of squamous specific genes was activated with the same kinetics and to the same extent as control cells. The up-regulation of p27, p16 and p130 proteins also occurred in a manner identical to control cells, strongly suggesting that these events are the result of a differentiation-specific signal.

These studies clearly demonstrated that the mid-G₁ arrest provoked by inhibition of the G₁ cdks is not sufficient to cause terminal differentiation. Furthermore, adhesion must regulate an independent signal which presumably activates a critical, and perhaps rate-limiting, step in differentiation, because the kinetics of squamous gene activation by suspension were not altered in the presence of growth arrest induced by over-expression of p27.

These findings imply that an integral part of terminal differentiation involves a shift in the growth status of the cell, a withdrawal from the cell cycle into an irreversibly-arrested state. The fact that roles for p130, p27 and p21 have been defined for withdrawal from the cell cycle into G₀ (Hengst et al., 1994; Mayol et al., 1996; Harvat et al., 1997) and the irreversible growth arrest in senescence (Afshari et al., 1996; Alcorta et al., 1996) suggest that these proteins exert similar functions during other forms of growth arrest, but it may be their specific combinatorial action which produces IGA in NHEKs. In support of this interpretation, 30% of NHEKs in suspension for 24 hours appeared to be arrested in G₂/M, yet express cornifin α. This implies that a G₁ arrest is not required for the induction of the differentiation program in keratinocytes. It is important to note, however, that there is no detectable hyperphosphorylated Rb or p130 at this point, even though Rb and p130 are normally hyperphosphorylated in G₂/M, suggesting that cells arrested with a G₂/M DNA content may nevertheless be in a unique cell cycle state.

The establishment of IGA appears to correlate with the induction of squamous genes, suggesting that the associated cell cycle regulatory changes we observe may serve to maintain irreversible growth arrest rather than promote differentiation. Further studies are necessary to verify this hypothesis. However, additional evidence supports such a role for p130, p27 and p16 in keratinocytes: (1) TGFβ, which induces a reversible growth arrest in NHEK and antagonizes terminal differentiation (Saunders and Jetten, 1994), blocks the hyperphosphorylation of p130 but does not increase the levels of p130 protein (Herzinger et al., 1995). On the other hand, p130 protein levels are increased during myogenesis (Kiess et al., 1995) and unique E2F-binding complexes containing p130 appear during muscle (Shin et al., 1995) and neuronal (Corbeil et al., 1995) differentiation. (2) p27 may likewise have a role in maintaining growth arrest as it has been shown to be essential for cell cycle exit (Rivard et al., 1996). (3) p16 deletions and mutations have been associated with squamous cell carcinoma (Loughran et al., 1996; Olshan et al., 1997; Reed et al., 1996). The ability of p16 to suppress these tumors in vitro (Liggett et al., 1996) suggests that the presence of p16 may be essential to maintain irreversible growth arrest as part of complete terminal differentiation.

The results presented here raise the question of which events are necessary for terminal differentiation. With regard to Rb, we have found that NHEKs infected with human papilloma virus, in which Rb and p53 are undetectable, differentiate very well in suspension culture (unpublished observations), suggesting that Rb activation is not essential. On the other hand, although the inhibition of cdk2 activity is clearly not sufficient to induce terminal differentiation, the question of whether it is required for differentiation remains intriguing, given the kinetics of its down-regulation. In addition to our data showing close association of p27 and cornifin α expression on a single cell basis, several other findings appear to support the hypothesis that cdk inhibition or an unknown alternative function of the cki’s may be required. Antisense p27 has been shown to repress keratin 1 expression (Hauser et al., 1997), as well as basal levels of cornifin α and involucrin (B. L. Harvat, unpublished observations) and keratinocytes derived from p21 null mice express lower levels of differentiation markers than wild-type mice (Missero et al., 1996). Cell cycle arrest and increases in p27, p21 and p15 also correlate with expression of squamous differentiation markers in mammary epithelial cells (unpublished observations). In addition, increases in p21 levels occur during differentiation in mouse keratinocytes (Missero et al., 1995) as well as in mouse skin (Parker et al., 1995).

It is possible that cki’s may have a role early in the commitment to differentiation, prior to the proposed critical step activated by lack of adhesion. It has been proposed that keratinocyte stem cells, which remain in the basal layer of the epidermis and have a high proliferative capacity, undergo division to give rise to an intermediate cell, the transit amplifying cell (see Fuchs and Byrne, 1994, for review; Jones et al., 1995). These cells in turn undergo a few cell divisions in the basal layer, but are already committed to give rise to differentiated cells that move into the suprabasal layer. p21 and/or p27 levels may have a role in regulating the change in proliferative capacity that is part of this first step toward differentiation. The early and transient increase in p21 expression may lower the threshold of other cdk inhibitors needed for kinase inhibition or affect transcription of particular genes via participation in cyclin/cdk2/E2F/p107 or p130 complexes as has been recently reported (Afshari et al., 1996; Shiyanov et al., 1996). Similarly, Durand et al. (1997) recently proposed that increases in p27 levels serve as a timing mechanism for oligodendrocyte differentiation during development as well as the mediator of growth arrest. The interconnection between regulation of growth control and activation of differentiation markers becomes most important in vivo, where a quiescent precursor keratinocyte population must be preserved and distinguished from irreversibly arrested differentiating cells. Adhesion, as such, most likely has a role in mediating repression of the differentiation process as well as maintaining a signal for proliferation while changes in relative levels of the cki’s could affect the balance between quiescence with proliferative capacity or commitment to terminal differentiation.

Viable p27, p21 and p16 null mice with apparent normal epidermal development argue that these gene products are not essential for squamous differentiation (Deng et al., 1995; Fero et al., 1996; Serrano et al., 1996). However, our findings suggest that cki’s may share redundant functions. In particular, there is a reciprocal relationship between p21 and p27 levels in NHEKs. Over-expression of p27 lowered p21 levels in NHEK cells (Fig. 7, compare lanes 1 and 6) and exogenous p21 expression had the same effect on p27 levels (Fig. 9a, compare lanes 1 and 2 with lanes 3 and 4). We have also observed that anti-sense p27 which decreases p27 protein levels produces increased p21 levels in NHEKs (unpublished data).

Our results support the hypothesis that a rate-limiting signal(s) for differentiation is generated by loss of adherence to substratum and is required for driving the cell into an irreversible growth arrest state as well as activation of squamous specific genes. Cornifin α, TGase I, keratin 1 and involucrin have been reported to be under the control of a variety of transcriptional factors and in particular, the AP-1-like sites in their promoters (An et al., 1993; Saunders et al., 1993; Lu et al., 1994; Lopez-Bayghen et al., 1996; Rutberg et al., 1997), while several members of the AP-1 family have been reported to be induced during terminal differentiation of NHEKs by suspension culture (Gandarillas and Watt, 1995). The activities of such transcription factors appear not to be affected by cki-mediated growth arrest in adherent NHEKs and may be targets for the signal(s) controlled by adherence and other agents which induce differentiation. The exact nature of these additional signals is under current investigation.

We thank Drs P. Steinert, Y. Xiong, J. Koh and M. Pagano for antibodies and Drs J. Fontana, D. Beach and J. Nevins for cDNAs. Thank you also to David Alcorta and Ron Cannon for critical reading of this manuscript.