Hypoxia-inducible Factor-1α Is a Positive Factor in Solid Tumor Growth

To understand the biology of tumor growth, it is critical to also understand the cellular response to changes in oxygen tension. The process of tumor expansion is characterized by rapid growth as a tumor establishes itself in the host. Accompanying this rapid growth are alterations in the microenvironment of the tumor cells, typically caused by an inability of the local vasculature to supply enough oxygen and nutrients to rapidly dividing tumor cells (1). Whereas the hypoxia that results may inhibit new cell division or even lead to cell death (2, 3, 4, 5), it can also lead to adaptive responses that will help the cells survive (6). These responses include an induction of angiogenesis and a switch to anaerobic metabolism; in addition, hypoxia can act as a selective force within a tumor for cells that harbor mutations, further improving their chance for survival (7). Hypoxia thus represents a paradox for those studying tumor growth: although oxygen deprivation can have negative effects on cell growth, the hypoxic response can mitigate those effects and even drive critical tumorigenic adaptations.

Control of the hypoxic response in mammalian cells occurs through a number of mechanisms, primarily transcriptional and posttranscriptional mechanisms (8). The transcription factor HIF-1²is one of the major regulators of hypoxic response (reviewed in Ref.9) and was first identified by Semenza and colleagues(10, 11, 12) as a regulator of hypoxia-induced erythropoietin expression. The HIF-1 binding site was then found on a wide range of promoter elements of genes up-regulated by hypoxia. This provided the first indication that there was a common mechanism regulating hypoxic response via transcription. The activation of transcription by HIF-1 occurs through the oxygen-regulated stabilization of HIF-1α, followed by its dimerization with ARNT, a constitutively expressed protein. Two other hypoxia-responsive homologues of the HIF-1α gene have been cloned, yet there appears to be little redundancy in hypoxic response(13, 14, 15, 16, 17). In cells examined thus far, the loss of HIF-1αresults in a total loss of binding to HIF-1 response elements(18, 19).

The hypoxic response would appear to promote tumor growth by promoting cell survival; this likely occurs through its induction of angiogenesis and its activation of anaerobic metabolism. Initial data have indicated that this is likely the case because loss of either HIF-1α or ARNT has been shown to retard tumor growth (19, 20). The mechanism for this retardation appears to be decreased vascularization accompanied by an increase in apoptosis. The decrease in vascularization presumably occurs in part through a loss of HIF-1-mediated expression of a critical effector of tumor angiogenesis,VEGF. These tumor data support a model in which the primary role of tumor hypoxia and the hypoxic response is to promote tumor angiogenesis. The role of HIF-1α as a tumor-promoting factor has become a controversial point, however, because recent work by one group has indicated that HIF-1α acts as a tumor suppressor, or negative factor, in ES cell-derived tumors (21).

We have further explored this important issue through the generation of differentiated, genetically manipulated cell lines nullizygous for HIF-1α. We report here the generation of wild-type and HIF-1α-null H-ras- transformed mEFs. Our findings confirm that HIF-1αacts as a positive regulator of tumor growth in this cell type as well. Surprisingly, we found no difference in vascular density between wild-type and null tumors, despite the fact that VEGF induction under hypoxia was significantly reduced both in vitro and in vivo. Our data demonstrate that HIF-1α is a positive regulator of tumor growth after H-ras transformation of fibroblasts and that the loss of HIF-1α alters VEGF expression in vivoduring solid tumor formation without a concomitant effect on vascular density in null tumors. Furthermore, we demonstrate that the cre/loxP system will provide a useful way to understand the role of HIF-1α and the hypoxic response in other processes.

Genomic DNA was obtained as described previously (19). A loxP site was engineered in the first intron through PCR, and a loxP-flanked neomycin resistance cassette was cloned into a SphI site in the second intron. The targeting vector was linearized, and 20 μg of the vector were electroporated into R1 ES cells (22). The neomycin resistance cassette was removed by electroporation of 30 μg of a cre-expressing plasmid into targeted cells [pML 78 (23, 24)]. PCR was used to identify cell lines that had maintained loxP sites on either side of exon 2. Chimeric mice were generated by injection of ES cells into C57Bl/6 blastocysts (25).

Wild-type and HIF-1α-null embryos were harvested at embryonic day 9.5, dissociated by incubation in 0.25% trypsin (Life Technologies,Inc.), and cultured. Embryos carrying two loxP-flanked alleles of HIF-1α were harvested at embryonic day 13.5, dissociated by passage through an 18-gauge needle, and cultured. Cells were immortalized by stable transfection of SV40 large T antigen, using Superfectamine(Qiagen) according to the manufacturer’s instructions and transformed by infection with a retrovirus expressing H-ras(26). The+^f/+^fras/TAg cells were infected with adenovirus expressing either β-galactosidase or cre recombinase.

The wild-type or null status of cells was confirmed by standard Southern blotting. Nuclear extracts were isolated from normoxic and hypoxic (4 h) cells by incubation in cell lysis buffer [10 mm Tris-HCl (pH 8.0), 1 mm EDTA (pH 8.0), 150 mm NaCl, 0.5% NP40, 1 μg/ml aprotinin, 1 μg/ml pepstatin A, 1 μg/ml leupeptin, and 1 mmphenylmethylsulfonyl fluoride] and separation of the nuclei by centrifugation. Nuclei were lysed by incubation in a buffer containing 20 mm HEPES (pH 7.9), 400 mm NaCl, 1 mm EDTA (pH 8.0), and 1 mm DTT. Extracts were analyzed by SDS-PAGE, electroblotting, and immunodetection with an anti-HIF-1α IgY antibody (27). Detection of HIF-1α was performed using a horseradish peroxidase-conjugated goat anti-IgY(Promega) secondary antibody and SuperSignal West Femto reagent from Pierce.

Cells were cultured for 0 or 8 h under hypoxia, and RNA was extracted with Trizol reagent (Life Technologies, Inc.) according to the manufacturer’s protocol. Approximately 15 μg of total RNA were loaded per lane, run on a 1% denaturing agarose gel, and hybridized with cDNA probes. Probes were generated as described in Ryan et al. (19).

A total of 1 × 10⁷ cells were injected s.c. intrascapularly into immunocompromised mice, either RAG1−/− mice (28) or nu/nu mice from Charles River. Tumors were harvested 16–18 days after injection,weighed, and processed for histology.

Sections were cut from frozen tissue and stained for CD31 as described previously. Vessel density in the most vascular regions of the tumor was determined with a Chalkley eyepiece graticule as described by Fox et al. (29).

In situ hybridization was performed on paraformaldehyde-fixed, paraffin-embedded sections using ³⁵S-UTP-labeled riboprobes as described previously (30, 31).

To carry out initial characterization of differentiated cells that lack a functional HIF-1α allele, we generated lines directly from HIF-1α wild-type and null embryos. Embryos were harvested at embryonic day 9.5 and cells were immediately immortalized by stable transfection with SV40 large T antigen (32). The cells were transformed by infection with a retrovirus expressing the activated H-ras allele (26). Analysis of these cell lines by DNA and protein blotting confirmed their wild-type or null status, and they are referred to hereafter as +/+ras/TAg or −/− ras/TAg cell lines (Fig. 1, C and D).

Homozygous deletion of HIF-1α results in embryonic death between embryonic days 9 and 10 (18, 19). Cell lines isolated from this early stage of development might differ from standard mouse fibroblasts in a number of respects. To control for these variables in this study, we also created HIF-1α-null cell lines via conditional targeting of the HIF-1α locus using the cre/loxP system(33).

We designed a conditional allele of HIF-1α in which the second exon is flanked by loxP sites. The second exon encodes the helix-loop-helix motif, which has been shown to be essential for HIF-1α dimerization with ARNT and subsequent transcriptional activation (34). The targeting vector contains a loxP site in the first intron and a loxP-flanked neomycin resistance gene in the second intron (Fig. 1,A). Homologous recombination at the HIF-1α locus results in an allele with a loxP site 5′ of exon 2 and the loxP-flanked neomycin resistance gene 3′ of exon 2 (Fig. 1 B). It is possible that the presence of the neomycin resistance gene in the second intron could affect proper expression of HIF-1α; to safeguard against this possibility, we transiently expressed cre recombinase in the targeted ES cells. A fraction of the ES cells excised the neomycin resistance gene but retained 5′ and 3′loxP sites, whose presence was confirmed by PCR (data not shown). These ES cell clones were used for the generation of chimeras and mouse strains via blastocyst injection (25).

Mice containing loxP-flanked alleles of HIF-1α were crossed, and mEFs were harvested from embryonic day 13.5 embryos that were homozygous(+^f/+^f) for the conditionally targeted allele. The cells were transformed with SV40 large T antigen and H-ras as described above.

We next generated HIF-1α-null cells by transiently expressing cre recombinase to delete the loxP-flanked second exon. We used adenovirus to transiently express cre recombinase in+^f/+^f ras/TAg cell lines. An adenovirus expressing β-galactosidase was also used to infect +^f/+^fras/TAg cells. These β-galactosidase-infected cells were then used as wild-type controls in all of the following experiments and are referred to as +^f/+^fras/TAg cells. Cre-infected cells are referred to as− ^f/−^f ras/Tag cells. Four to five days after infection, cells were assayed for excision of the second exon. DNA analysis via Southern blotting (Fig. 1,C) and PCR (data not shown) indicated complete excision of the second exon in the entire cell population. This was further confirmed by analysis of nuclear extracts from the conditionally targeted cells, which showed an absence of HIF-1α protein under hypoxic conditions (Fig. 1,D). β-Galactosidase-infected cells maintained a wild-type Southern profile and expressed HIF-1αunder hypoxia (Fig. 1 and D). Both cell lines grew at similar rates in culture and formed similar numbers of colonies in soft agar assays (data not shown).

HIF-1α regulates a wide array of genes in response to hypoxia. The loss of HIF-1α in ES cells leads to a reduction in hypoxic expression of VEGF and a number of other genes at the mRNA level (18, 19). We assayed both of the HIF-1α-null mEF cell lines for loss of target gene expression by Northern blotting (Fig. 1 E). In the− ^f/−^f ras/TAg cell line, the absence of HIF-1α lessens the hypoxic induction of VEGF and completely inhibits that of the hypoxia-responsive genes phosphoglycerate kinase, lactate dehydrogenase, and glucose transporter-1. The situation is similar in the −/− ras/TAg cell line.

The wild-type and null transformed cell lines described above were used to create fibrosarcomas in immunocompromised animals. Cells were injected s.c., and at 16–18 days after injection, the tumors were harvested and weighed. In both sets of cell lines, the absence of HIF-1α resulted in a significant decrease in tumor mass (Fig. 2).

Hypoxia is considered to be a major stimulus for tumor angiogenesis,and VEGF has been shown to be crucial for this process(35, 36, 37). Considering both the role of HIF-1α in hypoxia-induced VEGF expression and previously published tumor data, we looked at the degree of vascularization within wild-type and HIF-1α-null tumors. Tumor sections were stained for the endothelial cell marker CD31. This revealed no obvious difference in tumor vasculature between wild-type and null tumors (Fig. 3,A). Quantification of vascular density through Chalkley analysis confirmed that there was no significant difference in vessel density between wild-type and HIF-1α-null tumors (Fig. 3 B).

Because we did not observe any decrease in vascular density in the null tumors, we analyzed VEGF expression in+^f/+^f ras/TAg and −^f/−^fras/TAg tumors by in situ hybridization (Fig. 4). In tumors derived from− ^f/−^f ras/TAg cells, which show a significant difference in hypoxic regulation of VEGF under hypoxia in culture, there was a clear difference in the expression pattern of VEGF. In wild-type tumors, high levels of VEGF are expressed in wide swaths throughout the tumor, similar to the expression patterns in solid tumors reported elsewhere for this gene(20). The null tumors also have regions of very high VEGF expression, but it is more circumscribed, in some cases to single cells as opposed to large numbers of cells in a single region (Fig. 4 B, magnification, ×400); this results in a more punctate pattern of VEGF expression in the HIF-1α-null tumors.

Varying reports in the literature have made it unclear as to what role the HIF-1-mediated hypoxic response might have in tumor growth (19, 20, 21). Conflicting evidence from ES cell-derived tumors has indicated, on one hand, that HIF-1α acts as a positive regulator of tumor growth (19), most likely through its activation of VEGF, and, on the other hand, that HIF-1α acts as a negative regulator of tumor growth, possibly through its stabilization of p53 in hypoxic cells (21, 38). A primary point of contention is whether HIF-1α promotes or inhibits tumor growth. To address this issue, we have generated a tumor model using H-ras-transformed fibroblasts from which HIF-1α can be genetically removed. With this model, we have demonstrated that HIF-1α clearly acts as a positive regulator of tumor growth.

Data published previously on the mechanism by which HIF-1 positively regulates tumor growth have focused on angiogenesis (19, 20, 21, 39). The consensus has been that the absence of either HIF-1αor its dimerization partner, ARNT, leads to reduced vascularization within a tumor due to a reduced capacity to hypoxically induce VEGF expression. We were therefore surprised to find that loss of HIF-1αdid not alter tumor vascularization in H-ras-transformed fibrosarcomas. This is in contrast to experiments in which VEGF itself is deleted from tumor cells, causing a large decrease in vascular density (35), and in contrast to the experimental evidence from ES cell-derived tumors, where significant, albeit subtle,differences in vascular density are seen (19, 21).

We considered the possibility that our observation might be due to differences between in vitro and in vivo VEGF expression, where the tumor environment works in such a way as to make expression differences seen in culture inconsequential in vivo. This may explain our observations; however, within the tumors, in situ hybridization demonstrates that there are clear differences in VEGF expression between+^f/+^f and− ^f/−^f tumors. This difference is best seen in the higher magnification in Fig. 4 B, which shows that VEGF expression within these tumors is more punctate and restricted. A possible explanation for the different expression pattern seen in the− ^f/−^f tumors is that a higher degree of hypoxia is required to stimulate VEGF expression in these cells and that this more restricted expression pattern is an indication of those regions. This high level of VEGF expression from a smaller number of cells may be able to compensate for a general reduction in VEGF expression, ultimately resulting in an adequate degree of vascularization.

The work presented here has clearly demonstrated that HIF-1 acts to promote tumor growth. What has become less clear is the exact mechanism by which HIF-1 functions in this capacity. Our fibrosarcoma model has provided the first indication that hypoxic induction of angiogenesis may play a smaller role in tumor growth than previously thought and that HIF-1-mediated regulation of VEGF is not crucial for tumor vascularization. With the multitude of HIF-1 target genes, there are a number of possible mechanisms yet to be investigated, and in all likelihood, the effect of HIF-1 on tumor growth is complex and involves the activation of several adaptive pathways. We also report for the first time the development of conditionally targeted mice and transformed cell lines from which HIF-1α can be easily excised via cre recombinase expression. These should prove invaluable reagents to investigators wishing to study the role of HIF-1α during normal development and the role of hypoxic response during tumorigenesis.

We thank members of the Johnson laboratory for their support and advice throughout the course of this work and Keith Laderoute, Amato Giaccia, Robert Warren, Michael Karin, Gregg Semenza, Ronald Wisdom,and Frank Giordano for reagents and helpful advice.