Coordinate Up-Regulation of Hypoxia Inducible Factor (HIF)-1α and HIF-1 Target Genes during Multi-Stage Epidermal Carcinogenesis and Wound Healing¹

Epithelial carcinogenesis and wound healing share biological and molecular features. Both processes develop in stages; epithelial carcinogenesis progresses from hyperplasia, to dysplasia, carcinoma in situ, and invasive malignancy, whereas wounds proceed through inception, reepithelialization, stromal remodeling, and scar formation (1, 2). Cancer formation and wound healing are regulated by a composite of alterations affecting both epithelium and stroma, with cross-talk between cell types in both compartments. Epithelial cells of both processes are proliferative and motile, and the stroma is the site of up-regulation of angiogenesis and a scaffold for epithelial cell migration or invasion (3). Common molecular features of carcinogenesis and wound healing include increased expression of growth factors, growth factor receptors, and angiogenic factors. The major obvious difference between carcinogenesis and wound healing is that the former is progressive with accumulation of biological, cellular, molecular, and genetic alterations, whereas the latter progresses and then recedes over time as the wound heals and is resolved by scar formation. Despite this difference, elucidation of genetic control of wounding and delineation of timing and cell-type specificity of changes in gene expression during distinct wound healing stages offer a potential window into similar functions required during neoplastic progression (3).

Regions of hypoxia are common in both healing wounds and growing masses of malignant tissue (4). In both cancers and healing wounds, hypoxia alters overall cellular behavior, as a consequence of,or in addition to, activating specific genetic pathways. In cancers,hypoxic regions are resistant to chemo- or radiation therapy, yet also harbor zones of enhanced apoptosis (4). In wounds, low oxygen appears to modulate the pace of healing and wound resolution and specifically to increase collagen synthesis and accelerate wound reepithelialization (5). This latter effect may be attributable to enhancement of keratinocyte motility and migration by hypoxia (6). At the level of gene expression and transcriptional activation, hypoxia alters the activity of selected transcription factors including HIF³-1, c-jun via ref-1 and/or c-jun NH₂-terminal kinase, and p53 (7, 8, 9, 10, 11). HIF-1 is a heterodimeric transcription factor composed of HIF-1α and HIF-1β(12). HIF-1α protein is ubiquitinated and degraded in normoxia (13, 14) but stabilized in hypoxia, whereas HIF-1β or aryl hydrocarbon nuclear translocator protein is constitutively expressed in both normoxia and hypoxia. HIF-1 transactivates a repertoire of genes facilitating metabolic and vascular adaptation to hypoxia (15). These genes include several glycolytic enzymes, GLUT1 and 3, VEGF, inducible nitric oxide synthase, and erythropoietin. Gene knockouts of either HIF-1 component demonstrate embryonic lethality at days 9–11 attributable to inhibition of vasculogenesis, mesenchymal cell death, and cardiac defects (16, 17, 18). Several studies document hypoxic induction of HIF-1α protein and up-regulation of HIF-1 target genes in established cancer cell lines. Immunohistochemical analysis demonstrates the presence of HIF-1 protein in variety of human cancers and in selected premalignant pathologies such as benign prostatic hyperplasia and colonic adenoma (19, 20). However, to date, no study has systematically investigated HIF-1α expression at each stage of carcinogenic progression. Similarly, despite evidence of hypoxia, increased angiogenesis, and VEGF up-regulation, there are no data regarding HIF-1α expression during wound healing.

K14-HPV16 transgenic mice develop spontaneous multistage epidermal carcinogenesis beginning with hyperplasia at 1–2 months of age,dysplasia at 4 months of age, followed by premalignant hyperproliferative follicular/interfollicular chest dysplasias or dysplastic sessile papillomas at 8 months of age, each of which lead to invasive squamous cancers (21). Previously, we demonstrated VEGF mRNA induction in keratinocytes, particularly in late-stage premalignant precursors in K14-HPV16 transgenic mice(22). Here, we show induction of HIF-1α mRNA at a threshold level in hyperplastic skin of transgenic mice, which is then markedly abundant in premalignant hyperproliferative follicular/interfollicular chest dysplasias (21) and further increased in cancers. Three targets of the HIF-1 transcription factor, GLUT-1, PGK-1, and VEGF are coordinately up-regulated at similar stages of premalignant neoplastic progression. However,although the expression pattern of GLUT-1 and PGK-1 parallel that of HIF-1α mRNA in neoplastic basal keratinocytes, VEGF mRNA is expressed in more superior upper spinous and granular epidermal cell layers. HIF-1α, GLUT-1, and PGK-1 mRNAs are also up-regulated in epidermal keratinocytes in a spatio-temporal expression pattern during wound healing. These genes are undetectable in nontransgenic skin, are expressed at high levels in basal keratinocytes that are reepithelializing the open wound, and then decrease to nearly undetectable levels when the wound is completely epithelialized and dermal scar formation is under way. Similar to carcinogenesis, the expression pattern of VEGF mRNA in reepithelializing keratinocytes appeared to be different compared with HIF-1α, GLUT-1, and PGK-1. Collectively, these data suggest coordinate gain of HIF-1α expression and HIF-1 function in proliferating basal squamous epithelial cells at specific stages of carcinogenesis and wound healing. Moreover, this study suggests that regulation of HIF-1α gene expression in cancer progression and wound healing may occur at the level of mRNA in addition to that of protein (8, 23).

HIF-1α and HIF-1 target gene expression during multi-stage carcinogenesis was determined using K14-HPV16 transgenic mice of the 1203#1 line backcrossed (n = 28) into the FVB/n inbred strain (21, 24). These mice develop spontaneous, invasive, epidermal squamous carcinomas predominantly from 8 to 12 months of age, most frequently in the ear or on the chest and lower neck (21, 25). Wound healing experiments were performed using 8–12-week-old nontransgenic male and female FVB/n mice. All procedures and protocols were reviewed and approved by the University of California Committee on Animal Research.

Mice were anesthetized with 2.5% Avertin, and their backs were shaved. Incisions ∼8 mm long were made through the skin, leaving the underlying fasciae intact, and spread apart to prevent healing by primary closure. Mice were housed individually and serially sacrificed under Avertin anesthesia on days 1, 3, 5, 7, 9, and 11 after wounding by bilateral thoracotomy, followed by intraventricular perfusion with ice-cold PBS and 3.75% paraformaldehyde. An ellipse of skin ∼1 cm² surrounding the wound was removed and postfixed overnight in 3.75% paraformaldehyde at 4°C. Wounds were bisected perpendicular to the axis of the incision and embedded in paraffin.

Routine histopathology was performed using H&E staining of 5-μm sections.

BrdUrd (100 mg/kg) was injected i.p., 2 h prior to sacrifice, and immunoperoxidase staining to detect BrdUrd incorporation into DNA was performed on paraffin-embedded tissue sections as described previously(21, 24).

Duplicate tissue sections from three different mice at each stage of carcinogenesis or wounding were hybridized to ³⁵S-labeled riboprobes as described previously(26). The 356-bp HIF1α riboprobe DNA template corresponded to nucleotides 513–869 of the mouse gene and was cloned into pGEM-T easy (Pharmacia, Madison, WI). An antisense riboprobe was generated by template linearization with SacII and in vitro transcription with SP6 polymerase. The GLUT-1 riboprobe template was 520 bp corresponding to nucleotides 275–794 of the mouse gene cloned into pBSSK (Stratagene, La Jolla, CA). An antisense riboprobe was generated by template linearization with PstI and T3 polymerase transcription. The PGK-1 riboprobe template probe was 485 bp, from 91 to 575 of the mouse gene, cloned into pGEM-T easy. An antisense riboprobe was generated by template linearization with PstI and transcription with T7 polymerase. An antisense VEGF riboprobe was generated as described previously (22). Sense riboprobes were generated from each template and used as negative controls. mRNA hybridization signals were determined from a qualitative estimate of silver grain density that was reproducible in duplicate sections from three different mice.

To determine alterations in expression of HIF-1α and the HIF-1 target gene GLUT-1 during multistage carcinogenesis, we used ³⁵S mRNA in situ hybridization to tissue sections from control nontransgenic and transgenic mice with intermediate (16 weeks of age) and late-stage (32 weeks of age)precursor lesions and invasive malignancies (Fig. 1). Precursor lesions from the 16-week-old transgenic mice were predominately hyperplasias and intermediate-grade dysplasias, whereas those of the 32-week-old transgenic mice were hyperproliferative follicular/interfollicular dysplasias (Ref. 21; Fig. 1, top row). Chest malignancies were moderately to poorly differentiated squamous carcinomas (Fig. 1, top row),whereas ear cancers were well-differentiated sessile papillary malignancies (data not shown). Neither HIF-1α nor GLUT-1 mRNA was detectable in control, nontransgenic skin (Fig. 1, left panels). Hyperplastic chest skin displays a 2-fold increase in all epidermal cell layers and a moderate infiltration of the dermis with mononuclear cells (Fig. 1). Both HIF-1α and GLUT-1 mRNAs were just detectable in the basal and immediate suprabasal epidermal cell layers in these lesions (Fig. 1). Hyperproliferative follicular/interfollicular chest dysplasias consist of marked enlargement of the hair follicle ORS, suprabasal extension of basal-type keratinocytes, and replacement of the uppermost stratum corneum with nucleated parakeratotic cells (Fig. 1; Ref.21). HIF-1α and GLUT-1 mRNAs were markedly increased in these lesions (Fig. 1). Moreover, nearly all of the hybridization signals for both mRNAs were localized to the basal layers of the dysplastic interfollicular epidermis and the hair follicle ORS (Fig. 1). Hyperproliferative follicular/interfollicular chest dysplasias immediately adjacent to chronic superficial wounds displayed the highest level of signal for both HIF-1α and GLUT-1 mRNA (data not shown). No hybridization of either HIF-1α or GLUT-1 mRNA was detected in the dermis (Fig. 1). In moderate-poorly differentiated chest cancers(Fig. 1, right panels), hybridization signals for HIF-1αand GLUT-1 mRNAs further increased compared with hyperproliferative follicular/interfollicular chest dysplasias, and was similarly localized to epithelial squamous carcinoma cells compared with stromal cells (Fig. 1). Both HIF-1α and GLUT-1 mRNA levels were increased in dysplastic ear papillomas and papillary ear cancers; however,hybridization signals were much fewer compared with the chest lesions,suggesting that the carcinogenesis at this site was associated with a lower induction of these mRNAs (data not shown).

Analysis of VEGF expression during multistage carcinogenesis revealed both similarities and interesting differences compared with the pattern of HIF-1α mRNA (Fig. 2). Both VEGF and HIF-1α mRNAs were initially, and barely, detectable in the hyperplastic stage of carcinogenesis (Fig. 1 and Ref.22). Similarly, the hybridization signals for both VEGF and HIF-1α mRNAs were further increased in premalignant hyperproliferative follicular/interfollicular chest dysplasias (Ref.21; Fig. 2). However, the distribution of VEGF and HIF-1α mRNAs was markedly different in these precursor lesions. HIF-1α expression was localized to the basal and immediate suprabasal epidermal layers, whereas VEGF mRNA was predominantly detectable in the upper spinous and lower granular epidermal layers (Fig. 2, A and B). In cancers, both HIF-1α and VEGF hybridization signals further increased compared with hyperproliferative follicular/interfollicular chest dysplasias; however, the expression patterns of the mRNAs appeared to have subtle differences. Although both HIF-1α and VEGF mRNAs were diffusely present throughout the squamous epithelial component of the cancers, VEGF expression was punctate and markedly increased (Fig. 2, C and D)in what, on higher magnification, were individual malignant squamous cells (data not shown).

Histopathological analysis of spontaneous healing of unsutured back incisions revealed that on day 1 after incision, the epidermis at the wound edge is slightly thickened and the open wound is covered by inflammatory cells (Fig. 3). Epidermal thickness at the wound edge further increased by day 3(data not shown). Depending on the age of the mouse, by days 5–7, the wound was nearly covered by a sheet of reepithelializing epidermal cells migrating inward from both edges (Fig. 3). This epidermal sheet was six to eight cell layers thick, whereas epidermis at a distance from the wound edge was four cell layers in thickness (data not shown). By days 11–13 after wounding, the epidermis remained thickened;however, contracture markedly decreased the overall wound area with epidermis of normal thickness drawn closer to the wound edge (Fig. 3). Fibroblasts, destined to create the wound scar, were also prominent in the underlying dermis (Fig. 3).

On day 1 after wounding, BrdUrd incorporation was focally increased in keratinocytes at the wound edge and in the adjacent hair follicle ORS(Fig. 3). On days 5–7, BrdUrd incorporation was seen in the basal keratinocytes of the reepithelializing epidermal sheet and in scattered cells in the dermis (Fig. 3). Elevation in BrdUrd incorporation was restricted to the wound area and was similar to normal skin only 200–300 μm from the wound edge at all stages of wound healing (data not shown). At resolution of wound reepithelialization by days 11–13,BrdUrd incorporation was similar to unwounded skin, with only an occasional positive basal keratinocyte (Fig. 3).

On day 1 after wounding, HIF-1α mRNA was barely detectable in epidermal keratinocytes at the wound edge (Fig. 4). By day 5, HIF-1α message was abundant in the sheet of keratinocytes reepithelializing the wound, in the same basal cell population, in adjacent sections, as was incorporating BrdUrd (compare Figs. 3 and 4). By day 11 after wounding, HIF-1α message was undetectable in epidermis (Fig. 4). At no time during wound healing was HIF-1α mRNA detectable in dermis. Similar to DNA synthesis (see previous section),HIF-1α expression also fell off precipitously such that no message was detectable 200–300 μm from the wound edge (data not shown).

Spatio-temporal expression patterns of the HIF-1 target genes GLUT-1 and PGK-1 were also similar to that of the HIF-1α subunit during each stage of wound healing (Fig. 4). GLUT-1 and PGK-1 mRNAs were detectable in basal and immediate suprabasal keratinocytes on day 1 after skin incision at the wound edge. During wound reepithelialization, there was a marked increase in GLUT-1 and PGK-1 message, again predominantly localized to the basal and immediate suprabasal layers. By day 11, GLUT-1 and PGK-1 mRNAs were barely detectable in occasional basal keratinocytes. Similar to HIF-1α,GLUT-1 and PGK-1 mRNAs were not detectable in the dermis at any stage of wound healing.

Comparison of HIF-1α and VEGF expression patterns during wound healing also revealed similarities and differences, although the differences were less striking than in carcinomas (Fig. 2). During wound reepithelialization, VEGF mRNA displayed both a basal and a suprabasal/granular cell layer component in the reepithelializing sheet of keratinocytes at day 5 after wounding (Fig. 2,F). Similar to carcinomas, the VEGF expression pattern was punctate with marked increases in VEGF mRNA in individual keratinocytes (Fig. 2,F). In contrast, HIF-1α expression was diffusely distributed throughout the basal and immediate suprabasal layers of the reepithelializing keratinocyte sheet (Fig. 2 E).

This study demonstrates stage-specific up-regulation of HIF-1αmRNA during epidermal carcinogenesis in K14-HPV16 transgenic mice and wound healing in nontransgenic mice. Similar to a previous report(20), HIF-1α mRNA is induced in the proliferative compartment of the epidermis in both healing wounds and during carcinogenesis. The proximity of these epidermal basal cells to the underlying dermal capillaries (22) suggests that this compartment may not necessarily be hypoxic, and that increases in HIF-1α mRNA may be induced by signals activating proliferation, such as growth factors. Our data suggest that spatio-temporal gain of HIF-1 function may contribute to malignant conversion of late-stage premalignant epithelial dysplasias and to wound reepithelialization.

Previous reports suggest that the predominant regulation of HIF-1αexpression is via inhibition of protein ubiquitination and breakdown by hypoxia (13). Ubiquitination of HIF-1α protein is coordinated by a 200-amino acid “oxygen dependent degradation domain” (13). Both von Hippel-Lindau protein and p53 have been shown to bind HIF-1α and may contribute to normoxic protein instability by recruiting ubiquitin ligases such as CUL-1 or mdm-2,respectively (27, 28). p53 may also inhibit HIF-1-mediated target gene transactivation (29). Modulation of HIF-1α function by p53 is a link to HPV, because the E6 oncoprotein of high-risk viruses, such as type 16, bind to p53 with high affinity and catalyze its ubiquitination and destruction(30). As such, HPV16 E6 expression has been shown to increase HIF-1α protein expression, DNA binding, transcriptional activity, and inhibit HIF-1α ubiquitination and destruction(28). In contrast, other work demonstrates that hypoxia inhibits E6/E6-associated protein-mediated p53 destruction, suggesting that augmentation of HIF-1α protein levels in hypoxic HPV16-expressing cells may occur by mechanisms independent of loss of p53 function (11). A major difference between these experiments is that the level of hypoxia was greater in the Alarcon et al. (0.02% O₂; Ref.11) compared with the Ravi et al. (1.0%O₂; Ref. 28) study, with the possibility that an additional “stress” response coordinated by c-jun, c-jun NH₂-terminal kinase, or ref-1, may have been induced by more profound hypoxia (10, 31, 32). Regardless of these discrepancies, increases in HPV E6oncogene expression may explain the up-regulation of HIF-1 target genes in late-stage precursor lesions in K14-HPV16 transgenic mice. Previously, we demonstrated increased basal keratinocyte E6and E7 transgene expression during multistage epidermal carcinogenesis in this model (25). Thus, incremental E6 expression would be consistent with the marked up-regulation of GLUT-1 and PGK-1 in late-stage premalignant precursor lesions and in invasive epidermal cancers by stabilization of the HIF-1α component of HIF-1.

HIF-1α and HPV16 may also cooperate at the level of cellular metabolism. Alterations in the activity and oligomerization of the m2 isoform of pyruvate kinase (m2PK) by the HPV16 E7 oncoprotein is a potential link between enhanced aerobic glycolysis, HIF-1α function,and HPV carcinogenesis. Binding of HPV16 E7 to m2PK favors formation of m2PK dimers rather than tetramers (33). Dimerization of m2PK decreases its affinity for pyruvate, shifting enzyme activity toward lactate production via aerobic glycolysis (33). Although pyruvate kinase is not a direct HIF-1 target gene,transactivation of lactate dehydrogenase would further facilitate lactate production and aerobic glycolysis in cells expressing HPV E7. As such, coordinate functions of both HPV E7 and HIF-1/HIF-1α could enhance the metabolic adaptation of HPV expressing neoplastic or malignant cells to focal regions of hypoxia within growing cancers.

There have been few reports focusing on HIF-1α expression at the level of mRNA. In most instances, HIF-1α mRNA levels are found to be similar in transformed cells compared with nontransformed cells. However, retroviral oncogenes, v-src in particular, have been shown to increase HIF-1α mRNA in cells after transfection(34). Our study clearly demonstrates incremental increases in epidermal basal cell mRNA during both epidermal carcinogenesis and wound healing. Although this increase in mRNA may occur by both transcriptional as well as posttranscriptional mechanisms, examination of the HIF-1α enhancer-promoter DNA sequence provides a provocative link to epidermal carcinogenesis and HPV disease in particular. There are at least two Sp-1 sites 5′ to the start site and an AP-1 site in the 5′ untranslated region (35, 36). The presence of the AP-1 site may be significant. Enhanced AP-1 expression and transcriptional activity has been demonstrated in multi-stage skin carcinogenesis in mice transgenic with bovine papillomaviral oncogenes(37). Moreover, loss of c-fos function (38)or dominant-negative inhibition of c-jun (39) each prevent epidermal carcinogenesis induced by chemical carcinogens, further highlighting the role of AP-1-mediated signaling in skin carcinogenesis. As such, it will be of interest to determine whether loss or inhibition of AP-1 function in these models is also associated with less HIF-1α expression.

In addition to hypoxia, another level of HIF-1α regulation is growth factor signaling. Insulin-like growth factor-1, insulin-like growth factor-2, epidermal growth factor, and basic fibroblast growth factor have all been shown to increase HIF-1α expression(40, 41, 42). Moreover, growth factor and hypoxic control of HIF-1α protein stability appear to be independent and parallel pathways (41). We have demonstrated constitutive basic fibroblast growth factor expression and up-regulation of acidic fibroblast growth factor in the basal cell layer during epidermal carcinogenesis in K14-HPV16 transgenic mice (26). Moreover, we have unpublished data showing marked elevation of basal keratinocyte transforming growth factor-α expression in the dysplastic stage and in invasive epidermal cancers in this model. In our current study, up-regulation of HIF-1α mRNA in the basal cell epidermal cell layer in both multi-stage carcinogenesis and wound healing was correlated with increased proliferation during both neoplastic progression in K14-HPV16 transgenic mice (24, 25) and wound healing in nontransgenic mice. Additionally, the highest level of HIF-1α mRNA hybridization signals were in premalignant precursor lesions and cancers adjacent or in the midst of chronic superficial ulcers, which would be expected to contain both stromal and epidermal cells with increased wound-associated growth factor expression (1). Thus, our work, similar to previous reports, (20) appears to link HIF-1α expression, and possibly HIF-1 transcriptional activity, to epithelial cell proliferation and elevated growth factor expression.

Several studies, including our own, highlight angiogenesis, and in particular VEGF up-regulation, as frequent and important components of both multistage epidermal carcinogenesis and wound healing (22, 43, 44, 45). In the current study, the stage specificity of up-regulation of HIF-1α and VEGF mRNAs in K14-HPV16 transgenic mice during carcinogenesis and wound healing were similar, although the distribution and patterns of expression of the respective mRNAs were different. Because HIF-1 is a nuclear transcription factor, detection of VEGF and HIF-1α mRNAs in both similar and distinct epidermal cell layers is consistent with both HIF-1α-dependent and -independent pathways of regulation of VEGF expression during carcinogenesis and wound healing. In particular, macrophage VEGF expression(46) during wound healing is likely to be HIF-1αindependent, because we did not detect macrophage HIF-1α mRNA at any stage of wound healing.

In summary, we have demonstrated increased basal epidermal keratinocyte HIF-1α mRNA expression in the skin of K14-HPV16 transgenic mice during multi-stage carcinogenesis and during wound healing in nontransgenic mice. Coordinate increase of HIF-1 target gene expression in the same cell population, and at similar stages of either carcinogenesis or wound healing, is consistent with increased HIF-1αfunction with the caveat that independent and parallel mechanisms for up-regulation of these target genes may also exist. Ultimately,engineering targeted gain or loss of keratinocyte HIF-1α function will establish the necessity and role of this molecule in epithelial carcinogenesis and wound healing.

We thank James Cleaver, Allan Balmain, and Joe Gray for manuscript review and comments.