Selection of mutant CHO cells with constitutive activation of the HIF system and inactivation of the von Hippel-Lindau tumor suppressor.

Hypoxia-inducible factor (HIF) mediates a widespread transcriptional response to hypoxia through binding to cis-acting DNA sequences termed hypoxia response elements (HREs). Activity of the transcriptional complex is suppressed in the presence of oxygen by processes that include the targeting of HIF-alpha subunits for ubiquitin-mediated proteolysis. To provide further insights into these processes we constructed Chinese hamster ovary (CHO) cells bearing stably integrated plasmids that expressed HRE-linked surface antigens and used these cells in genetic screens for mutants that demonstrated constitutive up-regulation of HRE activity. From mutagenized cultures, clones were isolated that demonstrated up-regulation of HRE activity and increased HIF-1alpha protein levels in normoxic culture. Transfection and cell fusion studies suggested that these cells possess recessive defects that affect one or more pathways involved in HIF-alpha proteolysis. Two lines were demonstrated to harbor truncating mutations in the von Hippel-Lindau (VHL) tumor suppressor gene. In these cells, defects in ubiquitylation of exogenous human HIF-1alpha in vitro could be complemented by wild type pVHL, and re-expression of a wild type VHL gene restored a normal pattern of HIF/HRE activity, demonstrating the critical dependence of HIF regulation on pVHL in CHO cells. In contrast, other mutant cells had no demonstrable mutation in the VHL gene, and ubiquitylated exogenous HIF-1alpha normally, suggesting that they contain defects at other points in the oxygen-regulated processing of HIF-alpha subunits.

In oxygen-replete cells HIF-␣ subunits are targeted for ubiquitin-mediated proteasomal destruction. Insights into the mechanisms regulating oxygen-dependent proteolysis have been provided by studies of the von Hippel-Lindau tumor suppressor protein (pVHL). In the presence of oxygen, pVHL binds and targets HIF-␣ subunits to the ubiquitin-proteasome pathway (11). This process involves interaction between conserved subsequences within the oxygen-dependent degradation domains of HIF-␣ subunits and the ␤-domain of pVHL, with pVHL acting as the recognition element of a multicomponent E3 ubiquitin ligase (12)(13)(14)(15). Recently, we and others (16,17) have demonstrated that the interaction of the HIF-␣ subsequence with pVHL is regulated through oxygen-dependent prolyl hydroxylation within the HIF-␣ degradation domain.
pVHL inactivation is commonly observed in both inherited and sporadic renal cell carcinoma (18), and studies of renal carcinoma cell lines show striking up-regulation of both HIF-␣ subunits and a broad array of HIF target genes (11, 19 -21). These studies demonstrate the critical importance of pVHL in regulation of the HIF transcriptional cascade in this cell background. Nevertheless, studies of co-activator recruitment (10) and nuclear localization (9) in other cell types have provided evidence for other regulatory steps in HIF activation that are not known to be dependent on pVHL function. In addition, inactivation of other tumor suppressors such as p53 and PTEN, and activation of a variety of oncogenic and growth signaling pathways can induce the HIF system in non-hypoxic cells or amplify the response to hypoxia (22)(23)(24)(25)(26)(27). The mechanisms by which these stimuli interact with HIF-␣ molecules are less clear than for pVHL, though p53 inactivation has been proposed to affect targeting of HIF-1␣ to an mdm-2-dependent ubiquitin ligase (23).
To provide further insights into the upstream pathways that regulate the HIF system, we have devised a genetic screen for mutant Chinese hamster ovary (CHO) cells with altered HRE activity. Previously, we described the selection of cells with reduced hypoxic HRE activity from a mutagenized culture of CHO cells, and the analysis of one clone that was functionally defective in HIF-1␣ (28). In this study we demonstrate the use of a similar strategy based on linking HREs to genes encoding cell surface antigens for selection of cells manifesting constitutive overactivity of the HRE in normoxia. We describe the selection and partial characterization of a series of mutant cell lines with significantly increased normoxic levels of HIF-1␣. Some but not all of these lines have been shown to possess inactivating mutations in the VHL gene.
Mutagenesis and Selection-Mutagenesis was performed by exposure of ϳ10 8 cells to ICR 191 (Sigma) at 1 g/ml for 5 h using a protocol modified from previous studies (28,29). This exposure resulted in ϳ50% cell lethality. Multiple exposures to ICR 191 (three for the selection experiment from C4.5 cells and two for the selection experiment from E48.4.51 cells) were performed by re-growing cells to the original number before repeating the procedure.
Immunoselection by "panning" was adapted from a previously described protocol, to permit selection of cells that overexpressed HRElinked markers in normoxia. Cells were labeled with mouse IgG1 antihuman E-Selectin (Serotec) diluted 1/1000. Panning was performed as described previously (28) using plates coated with goat anti-mouse IgG (Sigma). After each round adherent cells were retrieved, re-grown to similar numbers, and subjected to the next round of selection. At intervals, aliquots of cells were assayed for changes in HRE-linked reporter expression by FACS.
Magnetic immunoselection was performed using the MACS system (Miltenyi Biotec). Antibody labeling was performed for 15 min at 4°C with FITC-conjugated anti-human E-Selectin (R&D Systems) at 1:10 dilution. Cells were then washed, and resuspended in MACS anti-FITC microbeads and incubated for 15 min at 6°C. The cells were then washed, resuspended, and applied onto a prechilled 4°C VSϩ column with a 20G flow resistor. The effluent was discarded, the column was removed from the magnetic separator, and the retained fraction was eluted. This sequence was repeated immediately, and the final eluate collected as the positive fraction. Following selection, an aliquot of cells was analyzed by FACS analysis for E-Selectin and luciferase assay, while the remaining cells were grown to reach a population of ϳ1 ϫ 10 7 cells before undergoing further magnetic selection on a (smaller) MSϩ column using similar methods.
Experimental Conditions and Reporter Gene Assays-Comparison of normoxic and hypoxic gene expression was performed on parallel cultures maintained in an atmosphere of 5% CO 2 balance air, or 0.1% O 2 , 5% CO 2 balance nitrogen in a NAPCO 7001 incubator (Precision Scientific). Exposures were for 16 h (mRNA analysis, reporter gene analysis) or 4 h for HIF-1␣ protein analysis. Assays of luciferase reporter gene activity were normalized to that of a co-transfected control plasmid (pCMV-␤Gal) in transient transfections and to total extracted protein (Bio-Rad) in stably transfected cells. Luciferase activity was determined in cell lysates using a commercial assay system (Promega) and a TD-20e luminometer (Turner Designs). ␤-Galactosidase activity in cell lysates was measured using O-nitrophenyl-␤-D-galactopyranoside as substrate in a 0.1 M phosphate buffer, pH 7.0, containing 10 mM KCl, 1 mM MgSO 4 , and 30 mM ␤-mercaptoethanol. Expression of cell surface markers was analyzed by FACS after labeling cells with mouse IgG1 FITC-conjugated anti-human E-Selectin (R&D Systems) or mouse IgG2b phycoerythrin-conjugated anti-human CD2 (Serotec).
RNA Analysis-RNA was extracted by a modified acid/guanidinium thiocyanate/phenol/chloroform method (RNAzol B, Cinna/Biotec Laboratories), dissolved in hybridization buffer (80% formamide, 40 mM PIPES, 400 mM sodium chloride, and 1 mM EDTA, pH 8.0), and analyzed by RNase protection as described previously (31). Quantitation of the protected species was performed using a PhosphorImager (Molecular Dynamics) and was related to an internal control assay for a constitutively expressed U6 small nuclear RNA. The riboprobe template for Chinese hamster Glut-1 was described previously (32). The riboprobe template for Chinese hamster HIF-1␣ was generated by the polymerase chain reaction from an a23 Chinese hamster lung fibroblastoid cell cDNA library using oligonucleotides based on the mouse sequence (nucleotides 682-946; GenBank™ accession number U59496). The quantity of RNA analyzed was as follows: U6 small nuclear RNA, 5 ng; Glut-1, 20 g; and HIF-1␣, 50 g.
Glucose Transport Assay-Glucose transport was determined by measuring the uptake of 2-deoxy-D-glucose using modifications of a previously described method (33). Assays were performed on cells grown to near confluence in 60-mm dishes. The medium was aspirated, and cells washed three times with phosphate-buffered saline at 37°C. Cells were incubated for 2 min at 37°C in 900 l of Krebs-Henseleit buffer containing 2 mM pyruvate and 2-deoxy-D-[1-3 H]glucose 0.5 mM (1 Ci/ml), after which uptake was stopped by washing (three times) in ice-cold phosphate-buffered saline, and cells were lysed by the addition of 2% SDS. Uptake was calculated as nanomoles/10 6 cells/min following liquid scintillation counting. Nonspecific uptake was determined from parallel assays performed in the presence of 0.5 mM HgCl 2 , and subtracted. Cell numbers were determined by setting up parallel 60-mm dishes for Coulter counting (Coulter Electronics).
Cell Fusions-Cell fusions were performed using polyethylene glycol as described previously (28). To enable selection of heterokaryons following fusion, additional selectable properties were generated as follows. Hypoxanthine phosphoribosyl transferase-deficient (HPRT Ϫ ) derivatives of each line were obtained by selection in 6-thioguanine (30 M). Resistance genes encoding neomycin phosphotransferase (pSV2Neo) or puromycin acetyl transferase (pPur, CLONTECH) were stably expressed into E48.4.51 derivatives or C4.5 derivatives, respectively. This enabled fusions between parental and mutant cells or between different mutant cells to be selected in HAT medium (hypoxanthine (15 g/ml), methotrexate (1 M), and thymidine (10 g/ml)) containing either G418 or puromycin. Following fusions, pools of selected cells were assessed by FACS analysis of E-Selectin or CD2 expression and immunoblotting for HIF-1␣.
DNA Sequencing and Analysis of VHL Genotypes-Genomic DNA was prepared from CHO mutant cells by proteinase K digestion and phenol/chloroform extraction. Exons were PCR amplified individually using the following primers; exon 1, 5Ј-AAGGCAGCTGGTCAAGAG-GAGG-3Ј and 5Ј-GCCGAAAGCACAGGTAGATACTTTG-3Ј; exon 2, 5Ј-CCCGGCTCATGCTTCTCTTTTAGG-3Ј and 5Ј-GGTAACTCAGGGCT-GTAATCCCAG-3Ј; exon 3, 5Ј-CTGAAATGACACTATCACGTC-3Ј and 5Ј-GGCGCGCCGAAAGAAAGTGACCTGTCC-3Ј. PCR products were sequenced using BigDye terminators (Applied Biosystems) and an ABI 373 sequencer. Confirmation of heterozygosity of mutations was ob-tained by ligation of PCR products into pGEMTeasy, and sequencing the insert in cloned plasmids. Response to 5-azacytidine was tested by exposure of cells to 4 M for 24 h and assay of the phenotype by FACS analysis for E-Selectin expression.
Ubiquitylation Assays-In vitro assays of HIF-l␣ ubiquitylation in cell extracts were performed using modifications of a previously described method (12). In brief, cell extracts were prepared from wild type and mutant CHO cells in a hypotonic extraction buffer at 4°C, centrifuged at 10,000 ϫ g, and stored in aliquots at Ϫ70°C prior to analysis. Ubiquitylation reactions were performed at 30°C using cell extracts supplemented with an ATP-regenerating system and ubiquitin (Sigma). [ 35 S]Methionine-labeled human HIF-1␣ substrate was produced in coupled transcription and translation (IVTT) reactions in rabbit reticulocyte lysate (TnT, Promega) supplemented with ferrous chloride, 100 M, and programmed with pcDNA3-HIF-1␣. For complementation experiments, human pVHL was produced by IVTT in reticulocyte lysates and preincubated with the ubiquitylation reaction mix for 5 min prior to addition of the HIF-1␣ substrate.

Construction of CHO Cell Lines Bearing HRE-activated
Marker Genes-Our purpose in these experiments was to select cells with elevated HRE activity that might have trans-acting defects in regulation of the HIF system. Because we have not yet been able to develop a selection strategy based on the expression and/or activity of an endogenous HIF-1 target gene, we first created transfected CHO cell lines expressing HRElinked cell surface markers that permitted immunoselection. One such cell line (C4.5), which bears separately integrated HRE-linked genes, encoding the surface antigens E-Selectin and CD2, has been described previously (28). In an attempt to improve the selectable characteristics we created a second transfected CHO line (E48.4.51) during the course of this work. First, we constructed several configurations of HRE-linked promoters and tested then in transient transfection assays. From these assays the regulatory cassette PGK6TKp, in which 6 HREs were linked to a minimal TK promoter (Fig. 1), was chosen on the basis of very low activity in normoxia and a high level of induction by hypoxia. CHO-K1 cells were transfected with pPurPGK6TKpE-Sel, and transfectants were selected in puromycin. Clones were tested for hypoxia-inducible E-Selectin expression, and a clone showing high level induction was further transfected with pGL3PGK6TKpLuc and pSVHyg. Supertransfectants were then selected in hygromycin B. One clone (E48.4.51) was chosen for mutagenesis and selection that showed low baseline and highly inducible expression of both markers (Fig. 1).
Selection of Clones with Enhanced HRE Activity in Normoxic Culture-Following mutagenesis different selection procedures were used for the C4.5 and E48.4.51 cells.
C4.5 cells were subject to positive immunoselection using E-Selectin and panning. Cells were maintained in 10 separate pools throughout the selection procedure, and aliquots of each pool were assayed at intervals for high expression of the second (integrated but unselected) HRE-linked marker, CD2. After 15 rounds of panning, aliquots of the five pools that showed the greatest increase in normoxic expression of CD2 were cloned by limiting dilution. Clones were re-assayed with respect to expression of cell surface markers, and those showing constitutive up-regulation of both stably integrated markers were tested further for enhanced normoxic HRE activity by transient transfection of an HRE-linked luciferase reporter. Of 39 clones tested in this way, two clones, M2.18 and M6.19 (derived from separate pools), were selected for further analysis on the basis of manifesting large increases in normoxic activity of both the stably integrated and transiently transfected HREs. Results of FACS analyses of both normoxic and hypoxic cells are shown in Fig. 2A. In comparison with C4.5, both lines show a clear up-regulation of both markers in normoxia.
E48.4.51 cells were subject to positive selection for increased expression of E-Selectin by magnetic cell sorting. After each round of selection, an aliquot of cells was removed from culture and tested for expression of the HRE-linked luciferase marker. Increased normoxic expression of luciferase was observed after the third round of magnetic selection suggesting that the selection procedure was somewhat more efficient than the panning protocol. After five rounds the increase was much more marked. Aliquots of cells were cloned by limiting dilution after the 5th, 6th, and 7th magnetic selection rounds. Individual clones were then tested for normoxic up-regulation of the stably integrated HRE-linked luciferase and E-Selectin genes. The stably integrated luciferase gene prevented the further assay of HRE activity by transient transfection of an HRE-linked luciferase reporter, as was used on the C4.5 derivatives. We therefore sought a different screening method that could be used for rapid assessment of clones likely to possess the desired property of a trans-acting abnormality in HIF/HRE activation. Dur- ing the course of this second selection experiment we determined that the clones selected from C4.5 in the first experiment (M2. 18 and M6.19) showed significantly enhanced rates of uptake of labeled glucose analogues. We therefore screened the E48.4.51 derivatives using an assay of 2-deoxy-Dglucose transport. From ϳ200 clones tested, we selected for further analysis three (10.8, 10.10, 14.2) that showed the greatest and most stable normoxic up-regulation of both the HRElinked reporters and normoxic up-regulation of glucose transport.
Results of FACS analysis of E-Selectin expression in normoxia and hypoxia are shown in Fig. 2B. In comparison with the parental cell line, E48.4.51, the up-regulation of E-Selectin expression in normoxic cells is clearly apparent. In hypoxia, different responses were observed. Although some lines showed a further increase in E-Selectin expression, others (e.g. clone 10.8) showed an inversion of the normal response to hypoxia with a decrease in E-Selectin expression in hypoxia. Similar effects were seen with assays of expression from the stably integrated HRE-linked luciferase reporter gene. Some lines again showed significant, although more modest, reductions in reporter gene activity in hypoxia.
HIF-1␣ Expression and Activity in Cells with Enhanced HRE Activity-Because regulation of the HIF system occurs through the alpha subunit, we tested clones showing up-regulation of HRE activity for HIF-␣ expression. Because our previous stud-ies indicated that CHO cells express undetectable levels of HIF-2␣ (28), we focused on analysis of HIF-1␣. HIF-1␣ mRNA levels were first measured using RNase protection. No significant differences from the parental line were seen for either series of cells. Results for the C4.5 derivatives are shown in Fig.  3A. In contrast, immunoblotting for HIF-1␣ showed substantially increased HIF-1␣ protein levels in normoxia in all lines of each series (Fig. 3B).
To determine if up-regulation of HIF-1␣ and HRE activity had effects on endogenous genes known to be regulated by the HIF system, we analyzed effects on glucose transporter-1 (Glut-1) expression. Glut-1 mRNA was assayed by RNase protection and was found to be consistently up-regulated in normoxic mutant cells showing up-regulation of HRE activity ( Fig.  3C and data not shown).
To better characterize the nature of the abnormality in HIF/ HRE activity in certain cell lines, we tested the activity of heterologous HIF-1␣ genes in transfection studies. These studies were restricted to lines M2.18 and M6.19, which did not contain a stably integrated luciferase reporter. Cells were cotransfected with pGL3PGK6TKp and either the HIF-1␣ expression plasmid pcDNA3-HIF-1␣ or empty vector. Both lines showed increased HRE activity, which was strongly augmented by co-transfection of pcDNA3-HIF-1␣, suggesting that the cells supported greater activity of both endogenous and heterologous HIF-␣ genes (Fig. 4A). This implied a defect in the regulatory pathway affecting HIF-␣, rather than in the HIF-␣ subunits themselves. To confirm this, the activity of a transfected Gal4/ HIF-1␣ fusion gene was tested. Cells were co-transfected with the Gal4 reporter UAS-tk-luc and plasmids expressing either Gal4, or a Gal4/HIF-1␣ fusion protein. Whereas Gal4 transfections manifest equivalent activity between cell lines, the activity of the Gal4/HIF-1␣ fusion was much greater in normoxic M2.18 and M6.19 cells than the parental line C4.5 (Fig. 4B).
Cell Fusion Studies-After introduction of suitable resistance markers, each mutant line was fused to its wild type parental line (C4.5 or E 48.4.51) and tested for regulation of the HIF system by FACS analysis of an appropriate HRE-linked surface marker or immunoblotting of HIF-1␣ levels. All fusions showed wild type regulation indicating that the defects were recessive in nature (Fig. 5 and data not shown). This suggested that all the mutant cells contain defects in the signal pathway affecting HIF rather than a stabilizing mutation affecting the HIF-1␣ molecule itself. Further fusion studies between the mutants provisionally assigned the cells to four complementation groups, M2.18, M6.19, 10.8/10.10, and 14.2.
Complementation with the von Hippel-Lindau Tumor Suppressor (pVHL)-During the course of this work, studies of renal carcinoma cell lines that are defective for pVHL demon-strated a key role for pVHL in the proteolytic regulation of the HIF system. We therefore sought to determine whether defective pVHL function might account for up-regulation of the HIF system in any of the mutant CHO cells. Because we were unable to detect Chinese hamster pVHL with sufficient sensitivity with any of the available antibodies, we first addressed this question by transfecting the cells with a wild type human VHL gene. Probably because of the number of resistance markers, reliable production of stable transfectants proved difficult in some cases. Together with some variation in the level of up-regulation of HIF/HRE activity this made analysis by transfection difficult to interpret for some clones. Nevertheless, stable transfectants expressing an HA-tagged wild type or mutant (N90I) pVHL at equivalent levels were produced from lines 10.8, 10.10, and M6. 19. Results are shown in Fig. 6. Restoration of regulation of HIF-1␣ by oxygen was observed with wild type but not mutant pVHL in lines 10.8 and 10.10, whereas neither transfection altered the phenotype of line M6. 19. This result indicated that some but not all of the mutant CHO lines were defective in pVHL function.
Definition of Mutations in the VHL Gene-To further define the reason for defective pVHL function in the CHO lines, the DNA sequence of the Chinese hamster VHL gene was determined, and compared in each of the cell lines. The Chinese hamster VHL cDNA sequence was first defined from PCR products generated from a Chinese hamster lung fibroblastoid cell (a23) cDNA library. This sequence was then used to design PCR primers that permitted amplification of both introns from genomic DNA. Further primers designed from the 5Ј-untranslated region, intronic sequence, and 3Ј-untranslated region were then used to amplify and sequence the entire coding region within each of the three exons of the Chinese hamster VHL gene from genomic DNA. Over these regions complete agreement was obtained between the cDNA sequence from a23 cells and that predicted from the genomic sequence obtained from the parental CHO line E48.5.41. The Chinese hamster VHL gene encodes a 167-amino acid product that omits all the GXEEX repeated sequences found in the N-terminal region of the human gene (35) and is initiated at a site corresponding to the internal initiation site at Met 54 in the human gene (Fig.  7A).
Sequencing of genomic DNA prepared from each of the mutant cell lines demonstrated mutations in lines 10.8, 10.10, and 14.2 but not M2.18 or M6.19 (Fig. 7B). Two mutations were frameshifting nucleotide insertions affecting codon 53 in line 10.8 and codon 101 in line 10.10, whereas the third was a non-conservative base substitution resulting in a missense mutation Ser 58 3 Asn in line 14.2. All mutations appeared to be heterozygous in the original direct sequencing reactions of PCR product from genomic DNA and were confirmed to be so by ligation of the PCR products into pGEMTeasy and further sequence determinations of the cloned products. This suggested that the recessive phenotypes most likely arose through silencing of the second VHL allele by DNA methylation. This was tested by exposure of cell lines 10.8 and 10.10 to 5-azacytidine (4 M), which restored wild type behavior.
Ubiquitylation of HIF-1␣-Finally, cell extracts were prepared from each clone and tested for the ability to support ubiquitylation in vitro of an exogenous human HIF-1␣ substrate prepared by IVTT in reticulocyte lysates. Because recent experiments have shown that HIF-1␣ requires modification by an iron-dependent HIF prolyl hydroxylase to promote interaction with the pVHL ubiquitylation complex (16,17), IVTTs were supplemented with iron (ferrous chloride, 100 M) to maximize modification of the exogenous HIF-1␣ substrate. Under these conditions, extracts from lines M2.18 and M6.19  (Fig. 8A). In keeping with functional data and DNA sequence analysis, addition of exogenous human pVHL prepared in a parallel IVTT complemented the defect in in vitro ubiquitylation manifest by cell extracts from 10.8, 10.10, and 14.2 cell lines ( Fig. 8B and data not shown). DISCUSSION In this work we have demonstrated the successful use of transfected HRE-linked immunoselectable reporter genes to obtain CHO cell mutants showing constitutive up-regulation of HIF activity. Two different selection methods, magnetic capture using antibody-conjugated beads and "panning" using antibody-coated plates, were used. Each method was successful in obtaining mutant cells with the desired property, although some difficulties were encountered. To distinguish variants displaying enhanced expression of the transfected reporter gene alone from cells bearing mutations that affected the HIF pathway, we used two separately integrated HRE-linked reporters. In this way, cells that were selected on the basis of up-regulation of one reporter could be assayed rapidly for upregulation of the second marker. We postulated that this should provide a rapid means of defining cells bearing mutations that affect HIF activation, because such mutations should affect the regulation of both, rather than just one, of the integrated reporters. Cells with these properties were then assessed either by transient transfection of a third HRE-linked reporter or by assay of an endogenous HRE-linked property. Somewhat surprisingly, many cells that demonstrated up-regulation of both stably integrated HRE-linked reporters did not manifest any other evidence of up-regulation of the HIF system and expressed a similar HRE-linked reporter gene normally after transient transfection. The mechanisms by which the two independently integrated reporters genes were up-regulated are unclear but are apparently unrelated to HIF activation.
Despite this unexpected difficulty, by screening substantial numbers of selected clones, we were able to define other cells that manifest clear up-regulation of the HIF transcriptional response when assayed either by the activity of transfected reporter genes or endogenous HIF target genes. We focused the selection procedure on obtaining cells that manifest the most marked up-regulation of HRE activity and obtained several such clones from each mutagenized culture. The exact frequency of such mutants is difficult to calculate, because we could not be sure to have retrieved all the mutants from each pool of mutagenized cells. However, the approximate frequency, which we estimate is in the range of 1 in 10 7 -10 8 , was similar to that observed in other selection experiments from mammalian cells (36). Much larger numbers of cells, manifesting minor enhancement of HRE activity, were observed but have not been analyzed in the current study.
All five mutants analyzed behaved recessively in cell fusion studies. Such signal pathway mutants may be of utility either in analyzing the importance of a candidate molecule in the pathway or in the definition of new components of the pathway by genetic complementation. The HIF pathway mutants selected in this study should be useful in both respects.
Surprisingly, in several of the CHO mutant cell lines, assays of HRE-linked reporter gene expression commonly demonstrated a clear reduction after exposure to hypoxia. The reasons for these findings are unclear. One possible explanation is that such effects represent compensatory mechanisms that are normally induced by hypoxia and act to limit HIF activation. If so, such a response must be independent of HIF. Feedback mechanisms that are activated by HIF transcription have been demonstrated (10,37), but such processes should limit or attenuate the response over time rather than invert the response to hypoxia. Despite variations in this behavior that we cannot currently explain, all cells showed substantial up-regulation of normoxic HRE activity.
The functional and DNA sequence analyses indicated that three lines contained defects in the pVHL tumor suppressor. As expected from the known action of ICR 191, lines 10.8 and 10.10 had frameshifting mutations due to base inserts into GC-rich sequences. These mutations produce early truncations that may be predicted to inactivate pVHL. In keeping with this, they displayed a striking increase in HIF-1␣ protein in normoxic cells that was fully reversed by transfection of a wild type human VHL gene and a defect in the ubiquitylation of exogenous human HIF-1␣ that was complemented by pVHL. Additionally, DNA sequencing of line 14.2 defined a mutation (G 3 A) in the pVHL gene that encodes a non-conservative amino acid substitution (Ser 58 3 Asn). This substitution is equivalent to the human mutation Ser 111 3 Asn, which results in the VHL syndrome. In addition to G:C base pair insertions, previous studies of ICR 191 indicate that a significant minority of associated mutations are base substitutions (38). Thus it is likely that this VHL mutation was induced by ICR 191 exposure. Its relation to the cellular phenotype is, however, less secure than for 10.8 and 10. 10. In vitro studies indicated that the defect in HIF ubiquitylation may be incomplete, transfection studies were inconclusive, and cell fusion studies apparently assigned the cell to a different complementation group. Although intragenic complementation has been described in mutational analysis of other pathways (39), and might arise if different mechanisms are involved in suppression of the second allele, these results make it difficult to be confident about assigning the defect in HIF regulation solely to the VHL mutation. Thus our studies defined two lines in which defective HIF regulation appeared securely related to pVHL inactivation.
Although pVHL and HIF-␣ are expressed and interact in a wide range of cell types, studies of the HIF transcriptional response in pVHL-defective cells have previously been confined to the renal carcinoma cell background. These pVHL-defective renal carcinoma cells show striking up-regulation of HIF and HIF target genes in normoxia so that regulation by hypoxia is minimal or absent (11,19,20). In view of the multiple mechanisms known to be involved in HIF activation, this result is unexpected. Given the tissue specificity of the VHL syndrome, it has been suggested that in the renal carcinoma cell background regulation of the HIF transcriptional response is unusually dependent on proteolytic regulation of HIF-␣ subunits by pVHL. Such a hypothesis would predict that the consequences of pVHL inactivation on the HIF system should be less striking in other types of cell. Our results show that this is not the case, at least in CHO cells. Both 10.8 and 10.10 cells showed striking up-regulation of both HIF-1␣ protein levels and HRE-dependent transcription under normoxic conditions, and both parameters were fully normalized by expression of a functional VHL gene. This indicates that pVHL is also a critical regulator of the HIF response in CHO cells and that the position in renal carcinoma cells is certainly not unique. In addition to providing new insights into the response to hypoxia, the derivation of pVHL-defective CHO cells should be of use in the analysis of other pVHL functions. Studies of pVHL-defective renal carcinoma cells transfected with either wild type or mutant pVHL have provided important insights into tumor suppressor function. These studies have established that restoration of wild type pVHL function abrogates tumorigenicity, whereas growth under standard tissue culture conditions is largely unchanged (40). In addition to effects on HIF regulation, alterations in extracellular matrix involving fibronectin (41), alterations in cell cycle, and apoptotic responses (42-44) have been described, and it remains unclear precisely how these functions relate to tumor suppressor activity. The availability of a new type of pVHL-defective cell may provide  2 and 3 with lanes 4 and 5). Durations of incubation of the ubiquitylation reactions are indicated.