Activation of Adenovirus Early Promoters and Lytic Phase in Differentiated Strata of Organotypic Cultures of Human Keratinocytes

Abstract

Human oncolytic adenoviruses have been used in clinical trials targeting cancers of epithelial origin. To gain a better understanding of the infectious cycle of adenovirus in normal human squamous tissues, we examined the viral infection process in organotypic cultures of primary human keratinocytes. We show that for the infection to occur, wounding of the epithelium is required. In addition, infection appears to initiate at the basal or parabasal cells that express the high-affinity coxsackievirus-adenovirus receptor, CAR, whereas the productive phase takes place in differentiated cells. This is due, at least in part, to the differentiation-dependent activation of the E1A and E2A early promoters and E4 promoters. We also show that adenovirus infection triggers a response mediated by the abnormal accumulation of cyclin E and p21cip1 proteins similar to the one previously observed in human papillomavirus-infected tissues. However, the virus seems to be able to overcome it, at least partially.

During the last decade, human adenoviruses have been exploited as vectors for gene transfer and gene therapy of various monogenic and polygenic diseases, with limited success. More recently, the potential of employing replication-competent adenoviruses as oncolytic agents has been put to the test in clinics, especially with neoplasias of epithelial origin, such as head-and-neck and prostate cancers (9, 23). The specificity of this new generation of anticancer agents relies on its preferential and efficient replication in tumors (1). The systems to study oncolytic adenoviruses ordinarily employ primary or transformed cell lines grown as monolayer cultures or as xenografts in nude mice. Comparatively few efforts have been made to expand our understanding of the viral life cycle in the normal stratified human epithelia, such as those encountered during clinical gene transfer. Thus, important aspects of the biology of the virus as related to vector safety could have been overlooked. Furthermore, differences in the physical characteristics of the epithelial tissue and submerged monolayer cells could also have important consequences for the infectivity of the virus in vivo. For instance, the protective cornified layer of a stratified epithelium may act as an effective barrier against adenovirus infection in addition to the possible heterogeneous distribution of the coxsackievirus-adenovirus receptor (CAR) across the epithelium (14, 15).

One alternative culture system that faithfully recapitulates the morphological and physiological characteristics of the normal stratified squamous epithelium is the organotypic raft culture of primary human keratinocytes (PHKs) (hereafter called raft cultures). This system is composed of a dermal equivalent consisting of fibroblasts embedded in a collagen matrix, on top of which neonatal foreskin keratinocytes are seeded. When the assembly is cultured at the air-medium interphase for 9 days, the keratinocytes proliferate, stratify, and differentiate, developing into a three-dimensional epithelium that closely resembles the native tissue. This culture can be harvested and studied by standard histological techniques (for a review, see reference 7). Hutchin et al. (15) have employed this system with nonreplicating adenoviruses for the study of gene transfer to oropharyngeal epithelial cells. A similar system was recently used to show that the binding of the fiber to CAR facilitates viral dissemination (37). On the other hand, we used the PHK raft culture system to screen and evaluate new conditionally replicating adenoviruses that would only proliferate in human papillomavirus (HPV)-positive cells on the basis of a conservation of adenovirus-HPV protein functions (described below). We have reported unexpected findings. For instance, an adenovirus type 5 E1A mutant, CB016, unable to bind pRb, p107, and p300, replicates in the fully stratified and differentiated squamous epithelium with delayed kinetics compared to replication in HPV E6/E7-transduced raft cultures (3). These observations indicate that adenoviruses have alternative means of mobilizing cellular genes needed to support viral DNA replication in differentiated keratinocytes, such as those required to supply deoxyribonucleoside triphosphates.

Indeed, many characteristics of human adenoviruses are shared by other small DNA tumor viruses, such as HPVs. For example, both virus families encode proteins that inactivate p53, which normally prevents DNA synthesis in the presence of DNA damage or promotes apoptosis to maintain genome stability (29, 36). Both also inactivate the tumor suppressor retinoblastoma susceptibility protein (pRb), which controls G₁/S transition (4). The differentiation-dependent expression of the oncogenic HPV E7 protein reestablishes an S phase milieu in otherwise differentiated squamous cells, in which productive infection takes place (5, 10). Thus, it is possible that adenovirus replication is also differentiation dependent. As a matter of fact, CB016-induced cytopathic effects (CPE) in PHK raft cultures occurred in the differentiated strata, although the keratinocytes had been infected prior to being raised to the air-medium interphase (3). These observations strongly suggest that adenovirus infection is differentiation dependent in this setting. Several early reports suggest the importance of the differentiation status of the cell in promoting the activation of the early genes required to initiate the amplification of adenovirus (13, 18, 22, 34). The notion that the virus life cycle is affected by cellular differentiation is also suggested by the preferential expression of CAR, the high-affinity receptor of adenovirus, in the less-differentiated cells of the epithelium (15, 37).

Here we report the use of the raft culture system to study the infectious cycle of the wild-type adenovirus at high and low multiplicities of infection (MOI). Even though the adenovirus serotypes commonly used in virotherapy do not naturally infect cornified stratified epithelium, we show that the system can still provide important information on the virus life cycle and virus-host interactions that are likely to occur in a clinical setting. We also examined the transcriptional activities of several adenoviral early promoters by using a reporter assay. We show that in a productive infection, adenovirus has to have access to the basal layers of the epithelium where CAR is more abundant. However, at a low MOI, the virus remains at a low copy number until the infected cell or its daughter cell starts the differentiation program. Once the infected cell enters the mid-strata of the epithelium, the early promoters are activated, promoting the amplification of the viral genome. We also show that virus infection triggers an antiviral response that involves the costabilization of p21cip1 and cyclin E, analogous to the one exhibited against HPV-induced unscheduled DNA synthesis in raft cultures and in benign HPV lesions (16, 25, 30). However, unlike HPVs, adenovirus seems to be able to overcome this response, at least in part.

Adenovirus infects only lacerated raft cultures.

We first examined the infection process of human adenovirus type 5 in a fully stratified and differentiated squamous epithelium. The assembly of PHKs and dermal equivalent was lifted to the air-medium interphase, cultured for 9 days, and harvested. Twenty-four, 48, 72, or 96 h before harvest, adenovirus was added to the cultures by one of three routes. From below, 10⁶ PFU was mixed with the culture media. From above, 10⁶ PFU suspended in 40 μl of phosphate-buffered saline (PBS) was placed on top of stratified epithelial cultures. Alternatively, the culture was first lacerated with a scalpel before application of 10⁶ PFU in 40 μl of PBS from above. The raft cultures were formalin fixed, paraffin embedded, and cut into 4-μm sections. To reveal the CPE caused by adenovirus, the sections were stained with hematoxylin and eosin. Additional sections were probed for the expression of the early protein E1A and the late protein hexon by double immunofluorescence. Only cultures that were lacerated showed signs of infection in regions near the laceration (Fig. 1A). Apparently, the virus neither is capable of crossing the collagen matrix underneath the epithelium nor can penetrate the intact cornified layer that protects the surface of the tissue. It has been recently demonstrated that CAR lies below the tight junction seal in adherens junctions (37), in agreement with the notion that a wound in the epithelium is necessary to initiate the infection process.

FIG. 1.

Adenovirus infection of lacerated raft cultures at a high MOI. (A) Time course of infection. Fully differentiated raft cultures were lacerated with a sterile scalpel and infected with 10⁶ PFU of adenovirus grown in 293 cells and purified by CsCl gradient centrifugation as described previously (12). The rafts were raised to the air-medium interphase and cultured for 9 days. The cultures were harvested at 24, 48, or 72 h post-adenovirus infection (p.i.) as indicated, formalin fixed, paraffin embedded, and thin sectioned. The top row shows sections stained with hematoxylin and eosin (H&E) to reveal the morphology of the tissue and the extent of the CPE caused by adenovirus infection in the areas surrounding the lacerations. Ballooning degeneration can be observed at 48, 72, and 96 h postinfection. Lysed cells were displaced from the tissue, leaving empty spaces, particularly at the basal and parabasal layers 72 and 96 h postinfection. The bottom row shows double immunofluorescence for E1A (Texas red) (mouse monoclonal antibody M73 [Oncogene, Boston, Mass.]; 2-μg/ml final concentration) and hexon (fluorescein, green) (goat polyclonal antibody [Chemicon, Temecula, Calif.]; 1:300 dilution) performed essentially as described in reference 3. E1A was detected as early as 24 h postinfection next to the lacerated areas. The signal was mainly basal and parabasal at 24 and 48 h postinfection and became suprabasal by 72 and 96 h postinfection. Hexon is visible in the basal cells at 48 h and extended upwards by 72 and 96 h postinfection. (B) Electron micrographs of a nucleus from a 72-h postinfection raft culture. Paracrystalline arrays of adenovirus particles can be observed. Light particles presumably do not contain DNA (right panel). Bars, 1 μm. (C) Expression pattern of CAR in uninfected raft cultures of PHK or SiHa cells. Sections of uninfected rafts were probed with the 2A3 mouse monoclonal antibody (kindly provided by Igor Dmitriev). The signal was developed with AEC (Zymed, San Francisco, Calif.) and appears as a red cytoplasmic hue, mostly in basal cells and, weakly, in some suprabasal cells of PHK rafts. No signal was detected on SiHa raft cultures that were refractory to adenovirus infection (Banerjee et al., unpublished).

At a high MOI, infection begins in the CAR-positive basal and parabasal layers.

In the lacerated raft cultures infected with 10⁶ PFU, we observed the expression of E1A 24 h postinfection in the areas surrounding the cuts (Fig. 1A). By 48 h postinfection, the expression was stronger and localized to the basal cells of the raft. Hexon also became detectable at this time, as well as a fairly mild CPE. By 72 h, E1A expression was preferentially found in the mid-spinous cells, just above a region strongly positive for hexon. The CPE was more widespread, and empty spaces between the epithelium and the dermal equivalent were also observed. This space could have been generated when infected basal cells were lysed or detached from the collagen. By 96 h, the whole thickness of the epithelium in the area of laceration was affected. E1A was still detectable in cell layers above those positive for hexon. Most of the basal and parabasal cells were lysed by then (Fig. 1A).

It is interesting that cells weakly positive for hexon were also weakly positive for E1A. In contrast, most cells strongly positive for E1A were negative for hexon, and, conversely, cells strongly positive for hexon were negative for E1A. We suggest that these signal patterns reflect the progression of the infection process. At early times, E1A was strong, and there was no hexon (E1A+/hexon−). E1A expression started to decline when the late phase initiated (weakly positive for both). Finally, E1A expression was below detection, while hexon became highly abundant (E1A−/hexon+). It is likely that, at this late time, many cells were already dead and only the virus particles were stable, but other viral proteins were degraded. Paracrystalline arrays of adenovirus particles were observed in the nuclei of infected cells by electron microscopy of ultrathin sections of fixed tissues (Fig. 1B). From these experiments, it was quite clear that the primary infection of adenovirus takes place in the basal layer of the raft and extends upwards with time.

To investigate the reason for the preferential infection of the basal and parabasal cells early in the infection, we determined the expression pattern of CAR. By using immunohistochemistry of sections of normal raft cultures, we detected CAR mainly in the basal and parabasal cell layers (Fig. 1C, top panel), in agreement with a recent report on organotypic cultures of oropharyngeal cells and airway epithelia (15, 37). These results would explain, at least in part, the observed preference towards these cell layers during the initial infection. In the negative control, CAR was not detected in raft cultures of an HPV-associated cervical carcinoma cell line, SiHa (Fig. 1C). Interestingly, submerged cultures of SiHa cells are susceptible to adenovirus infection, but SiHa raft cultures were refractory to infection even after laceration (N. S. Banerjee, L. T. Chow and T. R. Broker, unpublished observations).

In our time course experiments, we observed that after a few days, the infected cells were detected in the spinous and even the granular strata. It is not clear how the virus enters these suprabasal cells that are low on or devoid of CAR. One possibility is that progeny virus is transmitted to neighboring cells through gap junctions. Using electron microscopy, we examined numerous gap junctions, but we did not capture any viral particles in close proximity to them. This mode of transmission has not been described in monolayer cultures either (39). The second possibility is that the local high concentration of progeny virus released from the first round of infection makes possible the infection of surrounding cells in the presence of low CAR or even in the absence of CAR (2, 8, 11). Alternatively, in a fraction of the basal and parabasal cells, the virus may have remained inactive after gaining entry. Only when the infected cells or their daughter cells have ascended the epithelium and become differentiated does the virus enter into the productive phase. This possibility is consistent with the CPE found throughout the entire stratified epithelium when keratinocytes were infected with the wild-type virus prior to the development of raft cultures (3). Had the infection progressed into the late phase in all basal cells, there would not have been a stratified epithelium. This was not the case. This possibility of a delayed activation, as reflected by the progression of the CPE in Fig. 1A, is consistent with the observation that PHKs migrate up to three cell layers in 48 h under these culture conditions (6). Indeed, a differentiation-dependent virus production would be similar to the case with HPVs, the productive phase of which is strictly dependent on squamous differentiation (33).

At a low MOI, early signs of adenovirus infection are only found in the suprabasal cells after a lag.

To investigate the possibility that the productive phase of adenovirus infection is linked to differentiation, we repeated the infections of raft cultures with a 100-fold-lower MOI. It has been reported that at a high MOI, the gene expression program of the adenovirus is distorted and early promoters are derepressed (24, 32), thus obscuring a potentially inactive state.

We infected lacerated raft cultures with 10⁴ PFU as described above. Nine-day-old rafts were harvested at 24, 48, and 72 h postinfection. Paraffin-embedded sections were probed with E1A and hexon by double immunofluorescence. As shown in Fig. 2A, neither viral protein was detectable at 24 and 48 h postinfection. At 72 h postinfection, both proteins were detected in a small number of cells, mostly in the lower and midspinous strata. In situ hybridization of adenovirus DNA revealed very weak isolated signals in the basal and parabasal layers at 24 h postinfection (Fig. 2A, bottom panel). The isolated signals became stronger and clearly suprabasal at 72 h postinfection, consistent with viral DNA amplification and hexon expression. Thus, at the low MOI, fewer cells were infected and infection proceeded much more slowly; however, eventually late genes were expressed.

FIG. 2.

Differentiation-dependent adenovirus promoter activation and virus reproduction at a low MOI. (A) Time course of infection of lacerated PHK raft cultures. Fully differentiated raft cultures were lacerated and infected with 10⁴ PFU of adenovirus. The rafts were harvested 24, 48, and 72 h postinfection. (Top) The sections were probed for E1A and hexon proteins as described in the legend to Fig. 1. E1A (red) and hexon (green) were detected in a small number of suprabasal cells 72 h postinfection but not at earlier time points. (Bottom) In situ hybridization of adenovirus DNA was performed with biotinylated probes, synthesized by nick translation, and a signal amplification system based on fluorescein-labeled tyramide as described previously (26). White arrowheads point at cells in which viral DNA was detected. At 24 and 48 h postinfection, only very low levels of adenovirus DNA were detected, but by 72 h, some cells with very high viral DNA content were observed. White arrows point at the basement membrane. (B) Adenovirus E1 (nucleotides [nt] 1 to 532), E2 early (nt 26986 to 27336), and E4 (nt 35581 to 35935) promoters were each blunted and cloned at the NruI site in pLN-lacZ (27), a derivative of vector pLNSX (21). The adenoviral promoters were each placed upstream of the β-galactosidase gene but downstream of the LTR-driven neomycin resistance gene. (Top) Amphotropic retroviruses were generated as described previously (27) and used to infect primary human keratinocytes to make raft cultures as described previously (41). The raft cultures were fixed, stained with X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside), embedded, and sectioned. The sections were lightly counterstained with hematoxylin. (Bottom) The keratinocytes were first infected with pBabe^Pur-18URR-E7 (26) retrovirus and selected in puromycin-containing medium before infection with the pLN-derived retroviruses. Black arrows point at the basal membrane.

Taken together, these observations support the interpretation that the virus preferentially infects the basal and parabasal cells but restricts its productive replication until the cell progresses in its differentiation program and enters the spinous layers. However, we cannot rule out the possibility that viral promoter activation and viral DNA amplification initiated in the basal or parabasal cells, but given the low MOI, the levels of viral protein and DNA were below the sensitivity of detection. By the time the amounts of protein and DNA were detectable, the cells had ascended through the stratified epithelium.

Activities of E1 and E2a early promoters are differentiation dependent.

To distinguish the possible interpretations discussed above, we conducted the following experiments to examine the viral promoter activities. Productive viral DNA amplification requires the orderly expression of the early proteins encoded by the E1A, E1B, E2, and E4 open reading frames (reviewed in reference 31). We cloned the β-galactosidase reporter gene under the control of the adenoviral promoters E1, E2 early (at coordinate 75), and E4. These reporter cassettes were introduced into the raft cultures by using a retrovirus vector that expresses the bacterial neomycin resistance gene. Submerged PHKs were infected with these retroviruses, selected with G418 for 2 days, and the collagen matrix was seeded with them. Usually, 50 to 70% of the cells survived the selection. We have previously demonstrated that in this setting, the presence of the retroviral LTR does not affect the activities of viral or cellular promoters in driving the β-galactosidase gene expression in the appropriate cell strata (27, 40, 41). In addition, a myc-tagged human PCNA gene introduced via a lentivirus also exhibits an expression pattern similar to that of the endogenous gene. Furthermore, both genes were activated in differentiated cells by the HPV-18 E7 gene driven by the differentiation-dependent HPV-18 enhancer promoter located in the upstream regulation region (HPV-18 URR-E7) (26).

The results of these experiments are shown in Fig. 2B (top row). For all three promoters, the reporter activity was observed only in the differentiated strata, with E4 being more sporadically distributed than the others. The E1 promoter became active in the lower spinous cells, while the E2 early promoter became active in the mid-spinous layers. Both promoters remained active throughout the upper layers of the raft. To verify that the retrovirus backbone did not affect the normal regulation of the promoters, we also transduced the PHKs with HPV-18 URR-E7 via a retrovirus. In monolayer cell lines, E7 transactivates the E2 early promoter through deregulation of the pRb/E2F pathway (4, 28). Upon E7 gene expression, the E2 early promoter expression became more widespread, beginning in the lower spinous layer (Fig. 2B, bottom row), coinciding with the expression pattern of the HPV-18 URR (27). In contrast, E7 had no effect on the reporter expression pattern of the E1 or E4 promoters. We were not able to test the activation of E2 early promoter by adenovirus E1A protein. We suspect that PHKs transduced with a retrovirus harboring an E1A expression cassette either expressed little or no E1A protein or underwent apoptosis, as reported previously in primary mouse fibroblasts (38).

From these observations, we conclude that the differentiation-dependent replication exhibited by adenovirus in the raft cultures is due, at least in part, to the differentiation-dependent activity of its early promoters. Particularly, in the case of E1, an AP3-like factor has been identified as being responsible for the inhibition of the promoter in undifferentiated cells (13). However, information on the function of this factor in uninfected cells is still lacking. The constitutive expression of the early promoters in differentiated cells is an important factor to take into account during the design of oncolytic vectors that rely on the abrogation of E1A functions to modulate their activities. The fact that the E2 early and E4 promoters are still active even in the absence of E1A may account in part for the ability of altered viruses, such as CB016, to retain the ability to replicate in normal epithelium (3).

Levels of p21cip1 and cyclin E are increased in adenovirus-infected cells.

We have previously shown that HPV induces S-phase reentry in a fraction of the differentiated cells in the squamous epithelium in patient specimens. HPV-18 URR-E7 is sufficient for this induction in the raft culture setting (5). However, HPV-18 URR-E7 also induces a host defense response in a separate population of differentiated keratinocytes. This involves the inhibition of cyclin E/cdk2 in a complex with the cdk2 inhibitor, p27kip1 or p21cip1 (25). In normal squamous epithelium and in epithelial raft cultures, p21cip1 mRNA is up-regulated in differentiated cells; however, the p21cip1 protein is quickly degraded and is below the level of detection in situ (16, 30). Its level rises dramatically along with that of cyclin E in a subset of the differentiated cells of warty patient lesions or of HPV-18 URR-E7-transduced raft cultures. No accumulation is observed in the basal strata. Differentiated cells that exhibit this abnormal accumulation are unable to reenter S phase, as revealed by the lack of bromodeoxyuridine (BrdU) incorporation. This stabilization of cyclin E and p21cip proteins has been attributed to the differentiation-induced stable expression of the p27kip1 protein, which sequesters E7-induced cyclin E into a stable inactive complex with cdk2. In turn, cyclin E then stabilizes p21cip1 protein in an analogous kinase-inactive complex (25). Since HPV DNA replication also depends on an active cyclin E/cdk2 (19, 20), viral DNA is unable to amplify in cells that stably accumulate p21cip1 and cyclin E (17).

We asked whether this cellular response could also be observed during adenovirus DNA infection and whether adenovirus could overcome it. To answer these questions, we examined raft cultures infected with adenovirus for 3 days at a high MOI. In the last 12 h prior to harvesting, the cultures were exposed to BrdU. In a series of double immunofluorescence studies, we probed for p21cip1, cyclin E, or BrdU. As reported previously, neither protein was detected in the mock-infected raft cultures, and BrdU incorporation was restricted to the basal layer of the epithelium. Similar observations were made in the areas away from infection sites in the lacerated raft cultures (Fig. 3A, top row). In contrast, high levels of p21cip1 were found in and around the regions that show evidence of CPE (Fig. 3A, bottom row). Cells that showed high levels of p21cip1 also had high levels of cyclin E (Fig. 3B). BrdU was also detected in suprabasal cells in areas of CPE. However, unlike the HPV-host interactions, a significant fraction of cells were positive for both p21cip1 and BrdU in the lower spinous cells (Fig. 3A, bottom row). Similarly, some of the cyclin E-positive cells were also positive for BrdU (Fig. 3C). Thus, DNA replication took place in some of the cells with high levels of cyclin E and p21cip1.

FIG. 3.

Double-immunofluorescence detection of cellular proteins, viral proteins, or DNA replication. Assays were performed on lacerated rafts infected with 10⁶ adenovirus PFU and harvested 72 h postinfection as described previously (25). In all parts, the center panels are merged images. (A) BrdU incorporation (fluorescein, green, left panels) and p21cip1 (Texas red, right panels). The top row depicts a region of the raft away from the laceration area, free of adenovirus infection. BrdU incorporation is only observed in the basal cells, and p21cip1 is undetectable, as observed in normal uninfected raft cultures (16). The bottom row shows a region close to the laceration area. BrdU incorporation and p21cip1 protein accumulation are both detected in suprabasal cells. (B) p21cip1 (green, left panel) and cyclin E (red, right panel). A representative areais presented. Most p21cip1-positive cells are also positive for cyclin E. (C) BrdU incorporation (green, left panel) and cyclin E (Texas red, right panel). A region close to the laceration area is shown. High levels of cyclin E negatively correlate with high levels of BrdU incorporation, as observed previously for HPV (17). (D) Hexon (green, left panel) and BrdU (red, right panel). Hexon was detected as described in the legend to Fig. 1. BrdU was probed with a mouse monoclonal antibody (1:50 dilution [Zymed Laboratories, Inc., South San Francisco, Calif.]) followed by a 1:75 dilution of biotinylated anti-mouse IgG (Vector Laboratories, Burlingame, Calif.) and Texas red-streptavidin (Vector Laboratories). (E) p21cip1 (green, left panel) and E1A (red, right panel). E1A was detected first as described in the legend to Fig. 1A. The slide was then probed with anti-p21cip antibody directly conjugated with fluorescein (OP64F; dilution, 1:50 [Oncogene]). (F) Hexon (green, left panel) and p21cip1 (red, panel). The experiment was similar to that described for panel D, but p21cip1 was detected with a monoclonal antibody (OP64, dilution 1:25 [Oncogene]).

To determine whether the BrdU incorporation represented, at least in part, viral DNA replication, we carried out hexon and BrdU double immunofluorescence, because hexon is expressed late in the infection after the initiation of viral DNA replication. As shown in Fig. 3D, BrdU incorporation was observed in unaffected basal cells (the right side of the panel) as well as in basal cells positive for hexon (the left side of the left panel). However, suprabasal BrdU incorporation often coincided with hexon expression (the left side of the panel). Thus, viral DNA had replicated in these cells.

To investigate the kinetics of p21cip1 induction, we performed double immunofluorescence for E1A and p21cip1 and for hexon and p21cip1. A fraction of cells strongly positive for E1A were also positive for p21cip1, whereas those weakly positive for E1A were negative for p21cip1 (Fig. 3E). All cells positive for hexon were also positive for p21cip1, but some p21cip1-positive cells were negative for hexon (Fig. 3F). Taken together, these data indicate that p21cip1 appears at or shortly after E1A is expressed and is maintained throughout the remainder of the adenovirus life cycle.

Collectively, these results suggest that as in the case of HPV, the response against unscheduled DNA synthesis in postmitotic differentiated cells mediated by p27kip1 and p21cip1 is also triggered by the adenovirus infection. However, this mechanism seems less effective than in the case of HPV. Although a number of p21cip1-positive cells were negative for BrdU incorporation and hexon, indicative of a lack of viral amplification, many p21cip1- and cyclin E-positive cells were positive for BrdU and hexon. Evidently, adenovirus is better equipped than HPV to overcome this cellular response. For example, adenovirus carries its own DNA polymerase, terminal protein, and single-stranded DNA binding protein to support viral DNA replication. Currently, there is no information on whether adenovirus needs an active cyclin E/cdk2 for efficient viral replication. Nevertheless, these results stress the fact that the stable expression of p27kip1 protein and the abundance of p21cip1 mRNA in the differentiated keratinocytes (16, 25) serve the purpose of attempting to prevent nonscheduled DNA replication in differentiated cells. As such, they act as an antiviral mechanism to temper infections. Clearly, this scheme is quite successful during HPV infection, but is less effective during adenovirus infection.

In conclusion, we have shown that epithelial raft cultures are suitable tools for the study of the different aspects of the adenovirus infectious cycle. First, we have shown that the cornified layer effectively prevents adenovirus infection, and wounding of the basal and parabasal strata is required to effect an infection. Second, we have shown that the primary targets of Ad5 are the basal and parabasal cells, where CAR is expressed, in agreement with the results obtained with oropharyngeal cells (15). Third, at low MOI, the virus may remain in a relatively inactive state until the infected cell or its daughter cell differentiates and enters the spinous layers. It is in these differentiated layers where the viral early promoters are activated and the productive viral DNA amplification takes place. This would explain why this virus encodes gene products that can reestablish the S-phase milieu in cells that are postmitotic or in G₀ (35). In contrast, at high MOI, the viral gene expression program is altered. Expression of early genes occurs right after infection, bypassing the lag period. Thus, viral amplification takes place in the basal and parabasal cells. Fourth, the host responds to the infection by accumulating p21cip1 and cyclin E in a way resembling the response to another small DNA tumor virus, HPV, although this strategy is much less effective for adenovirus. In summary, this work introduces new tools and concepts to the study of the basic biology of adenovirus in the belief that a better understanding of how the virus infects and interacts with its host cells will greatly improve gene delivery methods and help in the design of new, clinically safer vectors.