Molecular and Genetic Characterization of an Ornithine Decarboxylase-deficient Chinese Hamster Cell Line*

The ornithine decarboxylase (ODC)-deficient Chinese hamster ovary (CHO) cell line C55.7 has normal amounts of ODC mRNA with very low amounts of immunologically detectable ODC protein, suggesting a structural mutation; however, 5-azacytidine treatment leads to phenotypical reversion (Steglich, C., and Scheffler, I. E. (1985) Somat. Cell Mol. Genet. 11, 11-23). We have demonstrated by chemical cleavage a single base mismatch in DNA heteroduplexes composed of wild-type and mutant cDNA strands. DNA sequencing showed that the mutant phenotype results from an aspartate-glycine substitution at amino acid 381 of the protein. When 5-azacytidine-revertant cell lines were selected for resistance to alpha-difluoromethylornithine, the resulting amplified ODC gene was structurally indistinguishable from the wild type gene. These results suggested the existence of a single active ODC locus in CHO cells. Using the methylation-sensitive restriction endonucleases AvaI and HpaII, we found evidence for two differentially methylated alleles in wild type, ODC-deficient and alpha-difluoromethylornithine-resistant cells. One of the alleles appeared completely inactivated by hypermethylation but could be reactivated by demethylation in spontaneous or 5-azacytidine-induced revertants.

When 5-azacytidine-revertant cell lines were selected for resistance to a-difluoromethylornithine, the resulting amplified ODC gene was structurally indistinguishable from the wild type gene. These results suggested the existence of a single active ODC locus in CHO cells. Using the methylation-sensitive restriction endonucleases AuaI and HpaII, we found evidence for two differentially methylated alleles in wild type, ODC-deficient and cY-difluoromethylornithine-resistant cells. One of the alleles appeared completely inactivated by hypermethylation but could be reactivated by demethylation in spontaneous or B-azacytidine-induced revertants.
Ornithine decarboxylase (ODC)' is the first enzyme in the pathway of polyamine synthesis in eukaryotic cells, and plays a key role in regulating intracellular polyamine levels (l-5). Our laboratory established an ODC-deficient line from Chinese hamster ovary (CHO) cells; it was isolated by ethyl methanesulfonate mutagenesis and ['Hlornithine suicide selection (6,7). These cells are putrescine auxotrophs with a low spontaneous reversion frequency (<IO-'/cell/generation portion of the gene; the mutation either renders the message untranslatable or generates an unstable protein product. Interestingly, we observed that treating the mutant cells with 5-azacytidine leads to recovery of putrescine prototrophs at a frequency of up to 16% of surviving cells, suggesting the existence of a silent gene or allele which was activated by decreased methylation (6). Analysis of mammalian genomic DNAs with ODC cDNA probes has suggested the existence of multiple ODC genes or pseudogenes (8-11).
Since a recessive mutation was isolated in a single step, the genotype of the wild type CHO cells could be either 1) haploid or heterozygous at the ODC locus with one allele deleted or permanently inactivated by mutation, respectively; or 2) functionally hemizygous at the ODC locus with one allele silenced by hypermethylation.
In either case, mutation of the remaining active allele would lead to total ODC deficiency, with 5azacytidine treatment either activating another, independent ODC gene (genotype 1) or activating the silent allele by hypomethylation (genotype 2; Table I). In this study, we defined the nucleotide change which leads to complete ODC deficiency in the mutant cell line C55.7. Moreover, we found that there is only a single transcribed ODC locus in CHO cells and that the expression of one of the alleles is strongly suppressed by methylation. Characterization of Ornithine Decarboxylase-deficient Mutant 5 ng of probe (5 x lo" cpm) was hybridized to 20 pg of total cytoplasmic RNA from C55.7 cells for 16 h at 45 "C. Samples were digested with RNase A and RNase Tl at 34 "C for 60 min and analyzed by denaturing acrylamide gel electrophoresis and autoradiography essentially as described (14). cDNA Synthesis and Amplification by the Polymerase Chain Reaction (PCRj-Poly(A)+-enriched RNA was prepared from cytoplasmic RNA on an oligo(dT)-cellulose column (14). First-strand cDNA synthesis was performed in a total volume of 20 ~1 with 1 pg of poly(A)' RNA, 50 pmol of random hexamers, 0.5 mM deoxynucleotide triphosphates, 50 mM KCI, 20 mM MgCl,, 20 mM Tris-HCl, pH 8.4 (room temperature), 1 mg/ml nuclease-free bovine serum albumin, 20 units of RNasin, and 50 units of Moloney murine leukemia virus reverse transcriptase for 10 min at 25 "C followed by 15 min at 42 "C.

MATERIALS
For the polymerase chain reaction (18, 19), the total volume was 100 ~1, and 7 pmol of primer A and B was added. Primer A was an oligonucleotide with the sequence 5'-GAATTCTTTGTAGCA-CACCGAGAGCC-3'; it was derived from the hamster cDNA sequence 19 nucleotides upstream of the translation start site (plus EcoRI linker). Primer B was an oligonucleotide with the sequence GTCGACCTGCTCTTCCACTTCTGGTGG and it was complementary to the hamster cDNA sequence 97 nucleotides upstream from the stop codon (plus Sal1 linker) (15). After denaturating at 95 "C for 10 min, 2.5 units of Taq DNA-polymerase was added to initiate 30 cycles of denaturing at 92 "C for 1 min, annealing at 37 "C for 2 min, and extension at 72 "C for 3 min. The PCR products were purified from a 0.8% agarose gel and ethanol-precipitated.
Chemical Mismatch Cleauaye-Radioactive templates were generated by a second amplification of the wild type or mutant PCR product as described (20). A separate reaction was performed with either primer A or primer B radiolabeled with [r-"'P]ATP and polynucleotide kinase (14). This resulted in two strand-specific probes for wild type and mutant cDNA with either the sense or antisense strand labeled. About 10 ng of DNA probe (specific activity about 10' cpm/pg) was hybridized with 150 ng of wild type or mutant PCR product to form either a heteroduplex or control homoduplex (20). The chemical modification and cleavage reactions were performed using either 2.5 M hydroxylamine or 2.4% osmium tetroxide at 37 "C for 2 h (21). Samples were analyzed by denaturing acrylamide electrophoresis and autoradiography. Molecular weight markers were radioactively labeled with [T-"'P]ATP and polynucleotide kinase (14).
DNA Sequencing-The wild type and mutant PCR products were subcloned into pGem 32 and sequenced by the dideoxynucleotide chain-termination method using the Sequenase kit from United States Biochemical Corp. (USB). Two independently derived PCR products were directly sequenced: 300 ng of PCR product plus 1 pmol of primer B were heated to 98 "C for 10 min and allowed to cool to room temperature for 1 min; the extension reaction and termination reaction were each 3 min.
Restriction Endonuclease Digestion and Southern Analysis of DNA-High molecular weight DNA was prepared as described previously (22) and digested with restriction enzymes as recommended by the supplier. Since no difference was found between 6 or 16 h of incubation, digests were performed for 6 h at 37 "C. Digests were monitored by mixing 3 pg of @X174 RF DNA with the genomic DNA and checking for appropriately sized fragments of @X174 RF DNA. Samples were size-fractionated by electrophoresis in 0.6% agarose gels and blotted onto nylon membranes (14). Probes were prepared from hamster ODC cDNA coding region (15) or from cloned hamster ODC genomic sequences by the random hexamer priming method (16); filters were washed twice for 20 min at 42 "C in 0.3 M NaCI, 0.03 M Na:$citrate, 0.1% SDS, pH 7.0 (2 X SSC) and twice for 20 min at 68 "C in 0.1 x SSC.
Enzymes and Reagents-The following radionucleotides were from Amersham Corp.: [n-""P]dCTP, [tr-"'PICTP, and [-y-R'P]ATP and (San Pablo, CA). Osmium tetroxide, hydroxylamine, and piperidine were from Aldrich. Taq DNA-polymerase was from Perkin-Elmer-Cetus Instruments and T4 DNA polymerase from USB. Escherichia coli DNA-polymerase I (Klenow fragment), T4 polynucleotide kinase, SP6 and T7 RNA polymerases, Moloney murine leukemia virus reverse transcriptase, HpaII, MspI, AuaI, EcoRI, and molecular weight markers were from BRL. All other restriction endonucleases used were from New England BioLabs. RNase A and RNase T were from Sigma and RNasin and pGem 32 from Promega.

RESULTS
Messenger RNA in Wild Type, ODC-deficient, Revertant, and ODC-overproducing CHO Cells-Poly(A)' RNAs isolated from wild type CHO cells, the ODC-deficient line C55.7, a spontaneous revertant SRl, and several 5-azacytidine-induced revertants (ARl, AR2, etc.) were analyzed on Northern blots by hybridization with a hamster ODC cDNA probe. Each cell line contained approximately equal amounts of a 2.2-and 2.4-kb message derived from two alternative polyadenylation sites (23). In the cu-DFMO-resistant DF3 cells, the same mRNA species were about 50-fold overexpressed (Fig. 1).

Molecular
Analysis of the Mutation in C55.7 Cells-The presence of apparently normal amounts of ODC mRNA and low amounts of immunologically detectable protein and enzyme activity in the ODC-deficient C55.7 cells suggested either a point mutation or small deletion in the transcribed part of the gene. Initially, a ribonuclease (RNase) protection method was used to analyze ODC mRNA from C55.7 cells. In this method RNases A and Tl cleave single-stranded RNA at base pair mismatches in RNA:RNA hybrids (24). We observed full length protection of mutant mRNA by wild type antisense RNA probes of the entire coding region and the 5'-untrans-   (24)(25)(26). We therefore decided to use a chemical cleavage method which recognizes all variants of single-base mismatches by reacting with either mismatched thymine or cytidine (21). Wild type and mutant cDNA was amplified using the polymerase chain reaction with primers flanking almost the entire ODC coding region. Using a wild type probe radioactively labeled at the 3'-end in a heteroduplex with mutant cDNA, a distinctive mismatch site 142 nucleotides from the 3'-end of the probe was detected after hydroxylamine treatment and piperidine cleavage; this indicated a mismatch involving cytidine in the wild type antisense probe. The same probe hybridized with wild type cDNA showed the absence of significant nonspecific cleavage (Fig. 2). Using a mutant probe radioactively labeled at the 3'-end in a heteroduplex with wild type cDNA, a mismatch was found at the same site, but only Asp c55 after osmium tetroxide treatment and piperidine cleavage, indicating a mismatch involving thymine in the mutant antisense probe (Fig. 2). Although the cleavage with hydroxylamine was incomplete and the recovery of radioactivity after and hybridized to radiolabeled probes as described under "Materials and Methods." A, hybridization with the intronspecific probe III described in Fig. 5; B, hybridization with the ODC cDNA probe described under "Materials and Methods." osmium tetroxide treatment suboptimal, these results suggested a single-base mutation substituting adenine in the mutant for guanine in the wild type cDNA sense strand. The predicted sequence alteration was confirmed and the exact position of the mutation verified by DNA sequencing of three independently derived wild type and mutant PCR products. The wild type cDNA sequence confirmed the previously published hamster ODC cDNA sequence (15, 27) and the mutant showed adenine in place of guanine 1142 nucleotides from the translation start (Fig. 3). This resulted in a nonconservative substitution of aspartate for glycine at position 381 in the protein. This single-base mutation generated a new restriction site for the enzyme NsiI which recognizes the sequence ATGCA'T at a position 32 bp downstream from a NsiI site present in both wild type and mutant cDNA (15, 27). Digestion of radioactively labeled wild type and mutant PCR products with NsiI and analysis of the fragments by denaturing acrylamide gel electrophoresis and autoradiography revealed that the wild type fragment of 175 nucleotides was shortened to 143 nucleotides in the mutant PCR product. When ODC cDNA from a 5'-azacytidine-revertant cell line was amplified by PCR and the product digested with NsiI, both a 175-and a 143-nucleotide-long fragment were observed, indicating the presence of both wild type and mutant ODC mRNA in these cells (Fig. 4).

Structural
Analysis of the ODC Gene Activated by SAzacytidine Treatment of C55.7 Cells-Southern blot analysis of mammalian genomic DNAs hybridized with ODC cDNA probes show multiple bands suggesting the existence of multiple ODC genes or pseudogenes (8-11). In various cell culture systems only a single band, presumably representing the actively expressed gene, is amplified after selecting cells in a-DFMO (8-11). When a genomic probe from the first large intron of the hamster ODC gene, probe III in Fig. 5, was hybridized with EcoRI-digested genomic DNA from wild type and DF3 cells, only a single band was detected which was amplified in the DF3 cells (Fig. 6a). Thus, this intron-derived probe appeared to be specific for the active ODC gene in CHO cells.
To distinguish between the activation of an independent ODC gene and the activation of a silent allele of a single locus (Table I), a number of 5-azacytidine-revertant clones were selected for a-DFMO resistance. Genomic DNAs from these revertants and their a-DFMO-resistant variants were digested with EcoRI and analyzed on Southern blots using the intron-derived probe or a cDNA probe for hamster ODC (Fig.  6, a and b). The same bands were observed in all of the variant cell lines with the molecular weight of the amplified fragment corresponding to that of the active locus in wild type and ODC-overproducing DF3 cells. A total of 10 independently derived 5-azacytidine-revertant clones and their cr-DFMOresistant variants were examined with the same result. The additional, nonamplified bands identified by the cDNA probe were similar in their intensity and demonstrated equal loading of DNA in all lanes (Fig. 66) plete hamster ODC gene was previously isolated in our laboratory from the ODC-overproducing cell line DF3." We mapped exons by hybridization with hamster cDNA sequences and found the organization of the gene to be very similar to that in other mammalian species (28)(29)(30)(31)(32)." Subsequently, we mapped restriction sites for the methylationsensitive endonucleases AvaI and HpaII and the methylationinsensitive isoschizomer of HpaII, MspI. About 2 kb of the 5'-region of the gene were sequenced including the promoter region, exon I and part of intron I to accurately determine the high density of restriction sites in this region.' The ODC gene contained 10 AvaI sites and 33 HpaII/MspI sites, most of which were clustered within a region of 1 kb surrounding exon I (Fig. 5).

Methylation
Profile of the ODC Gene in Wild Type, Mutant, Revertant, and ODC-overproducing CHO Cells-To examine whether we could distinguish two ODC alleles present in CHO cells on the basis of their methylation profile, we digested genomic DNA with the methylation-sensitive restriction endonucleases AvaI and HpaII. Southern blots were probed with genomic ODC sequences derived from the promoter region, the first exon, the first intron, the coding region, and the 3'flanking region of the gene. In some cases DNA was digested with AvaI or HpaII and a second, methylation-insensitive restriction endonuclease to allow more accurate fragment size determination. Control experiments demonstrated that all digests were complete. Since instability of methylation patterns in cultured cells has been described (33)(34)(35), the cell lines used in these studies were subcloned and DNA was isolated at various times thereafter. The results obtained with DNA samples prepared in 3-6-month intervals were fully reproducible. Fig. 7b shows a Southern blot of genomic DNA from various cell lines digested with AvaI and hybridized with the intronspecific probe derived from a sequence between the AvaI sites A6 and A7 (probe HI in Fig. 5). As previously demonstrated, this probe did not detect ODC pseudogenes (Fig. 6~). Wildtype, mutant, and ODC-overproducing DF3 cells showed two distinct bands: one of 1-kb size indicating that neither A6 nor ' A. Grens, unpublished results. " R. B. Pilz and A. Grens, unpublished results.
A7 were methylated in this allele (this band is amplified in the ODC-overproducing cells) and a second band of ~12 kb indicating a high degree of methylation in the second allele.
In wild type and mutant cells, the intensity of both bands was similar and comparable to the intensity of the nonamplified band in DF3 cells. Digestion of wild type and DF3 DNA with both AvaI and EcoRI reduced the size of the high molecular weight band to 3 kb, indicating that Al was unmethylated while A2-A6 were completely methylated in this highly methylated allele. The EcoRI/AvaI double digest of C55.7 DNA produced a 6-kb band indicating that all AvaI sites were methylated between the EcoRI sites El and E2. The second, 0.5-kb band represented the AuaI/EcoRI (A6-E2) fragment of the hypomethylated allele (Figs. 5 and 7b). The spontaneous revertant SRl as well as the 5-azacytidine-induced revertants and-their cu-DFMO-resistant derivatives showed loss of methylation in the second allele: A6 and A7 appeared to be unmethylated in both alleles (Fig. 7~). Interestingly, AvaI-digested genomic DNA from the hamster lung fibroblasts line V79 demonstrated only a single l-kb band, suggesting that both ODC alleles were hypomethylated in these cells (Fig.   7a).
Similar results as described for AuaI were obtained with the methylation-sensitive restriction endonuclease HpaII; control digests with the methylation-insensitive isoschizomer MspI showed identical restriction patterns for all cell lines examined (data not shown). Tables II and III summarize the methylation profile of AvaI and HpaII sites in the ODC gene. In the wild type, mutant and ODC-overproducing DF3 cells there was clear evidence for an almost completely methylated allele which is presumed to be the silent allele; this allele was not amplified in the DF3 cells. The other allele was partially methylated in wild type and mutant C55.7 cells. Since the amount of ODC mRNA was comparable in wild type and mutant cells, the minor differences found in the methylation profile of wild type and mutant cells did not appear to correlate with transcriptional activity. In the ODC-overproducing variant DF3, the amplified allele was globally hypomethylated. Hypomethylation of the amplified ODC gene in a-DFMO-resistant human myeloma and mouse Ehrlich ascites cells has been reported (36). Theoretically, the putrescine overproduction in these cells could lead to a depletion of Sadenosylmethionine, thus causing generalized hypomethylation of DNA. However, we have found no evidence for hypomethylation of the nonamplified ODC allele nor of two inde-  DISCUSSION We have examined the nature of the genetic and epigenetic changes in the ODC-deficient Chinese hamster cell line C55.7 and its &azacytidine-induced revertants.
The mutant phenotype in C55.7 cells is the consequence of a single-base change resulting in the substitution of aspartate for glycine at position 381 in the ODC protein. Our results suggest that ODC mRNA is transcribed from a single locus in wild type and 5-azacytidine-revertant cells. Evidence was found for two differentially methylated alleles at the ODC locus in CHO cells; one of the alleles appears to be inactivated by hypermethylation but can be activated by demethylation in the event of spontaneous or 5-azacytidine-induced reversion. The mutation leading to complete ODC deficiency in C55.7 cells was identified by three independent methods: chemical cleavage of mismatched DNA strands, DNA sequencing, and demonstration of a new NsiI restriction site generated by the mutation.
All three methods were applied to at least three independently derived PCR products, thus excluding the possibility of an artifact generated by infidelity of the Taq polymerase (37).
The nonconservative substitution of glycine with aspartate at position 381 of the hamster ODC could either prevent formation of a functional active site or prevent formation of active homodimers.
To our knowledge, the active site of the enzyme has not yet been identified.
The mutation is situated within a region of extremely high sequence homology among the rat, mouse, and hamster proteins; computer analysis of the mouse sequence strongly predicts a-helix secondary structure for this region (27,29,38).
A change in the surface charge of this a-helix could interfere with the folding of the protein into a tertiary structure or the assembly of the quaternary complex required for enzymatic activity.
Using a sensitive radioimmunoassay, very little ODC protein could be detected in C55.7 cells (6); these results have been confirmed by Western blot analysis (data not shown). Apparently, the mutation results in a protein with a shorter half-life than that of wild type ODC. The mutation described here falls outside of the region near the carboxyl terminus of the protein which appears to regulate the short half-life of wild type ODC (39). However, misfolded proteins can be unstable.
The ODC gene appears to be part of a dispersed family of related loci; some pseudogenes have been identified (40,41). In all cases studied only a single locus is amplified under selection for a-DFMO resistance (8-11). We were able to distinguish the transcribed ODC gene in hamster cells from the other genes or pseudogenes in two ways: 1) by amplification of the active ODC gene in cells selected for cu-DFMOresistance and 2) by specific hybridization to an intronderived ODC probe. Ten independent 5-azacytidine-induced revertants were subjected to a-DFMO selection, and our results indicate the existence of a single active ODC locus in CHO cells. The existence of other, potentially functional ODC loci in the hamster genome is not disproved but unlikely.
The potent hypomethylating agent 5-azacytidine can activate silent genes both in vivo and in tissue culture (42,43). Experiments with cell lines that lack specific enzymes but can regain a normal phenotype after treatment with 5-azacytidine suggest that genes can be stably but reversibly inactivated by methylation (42-44); in thymidine kinase and arginase-deficient mutants, a correlation between the degree of cytosine methylation and gene expression has been documented (45-47). Direct evidence that DNA methylation can repress gene transcription comes from transient expression assays with cloned genes; in vitro methylation renders some genes transcriptionally inactive after transfection (44,48). For example, Halmekyto et al. (49) recently showed that in vitro methylation of I-ZpaII and HhaI sites of the cloned human ODC gene virtually abolished the transient expression of human ODC in CHO cells. These findings support our hypothesis that the highly methylated ODC allele found in wild type and C55.7 cells represents a transcriptionally inactive allele. Spontaneous and 5-azacytidine-induced reversion to wild type phenotype correlated with demethylation of the previously hypermethylated allele. Furthermore, [3H]ornithine suicide selection of CHO cells resulted in ODC-deficient mutants at a relatively high frequency, suggesting functional hemizygosity of the CHO cells at the ODC locus (6, 7). In contrast, ODC-deficient mutants could not be isolated from the hamster lung fibroblast cell line V79 using the same protocol (7). This result can be explained by our demonstration of two active, hypomethylated alleles in these cells (Fig.   71.
We have noted partial methylation of individual AuaI and MspI/HpaII sites on one ODC allele (Tables III and III). Although this phenomenon is not understood, other investigators have made the same observation in other systems, and recently it has been documented that a partial methylation profile appears to be stably maintained in cloned, cultured cell lines (45, [50][51][52][53][54].
DNA methylation has been implicated in the control of