Intrinsic genetic instability of normal human lymphocytes and its implication for loss of heterozygosity.

A combination of flow cytometry and microsatellite analysis was used to investigate loss of expression of HLA-A and/or HLA-B alleles and concurrent LOH at polymorphic chromosome 6 loci both in freshly isolated lymphocytes (in vivo mutations) and in lymphocytes cultured ex vivo. The fraction of in vivo mutants that showed LOH at 6p appeared to vary from 0%-49% for various donors. During culturing ex vivo, HLA-A(-) cells arose at a high rate and showed simultaneous loss of expression at the linked HLA-B locus. Up to 90% of the ex vivo arisen HLA-A2(-) cell population showed LOH of multiple 6p markers, and 50% had lost heterozygosity at 6q. This ex vivo spectrum resembles that found in HLA-A2 mutants obtained from lymphoblastoid cells. The HLA-A2 mutants present in vivo may reflect only a small fraction of the mutants that can be detected ex vivo. In normal lymphocytes, in vivo only mitotic recombination appears to be sustained, indicating the importance of this mechanism for tumor initiation in normal cells. Although mutations resulting in LOH at both chromosome 6 arms were shown to result in nonviable cells in normal lymphocytes, they have been shown to result in viable mutants in lymphoblastoid cells. We hypothesize that these types of mutations also occur in vivo but only survive in cells that already harbor a mutated genetic background. In light of the high rate at which these types of mutations occur, they may contribute to cancer progression.

Molecular methods for the detection of mutations.

We report the results of a collaborative study aimed at developing reliable, direct assays for mutation in human cells. The project used common lymphoblastoid cell lines, both with and without mutagen treatment, as a shared resource to validate the development of new molecular methods for the detection of low-level mutations in the presence of a large excess of normal alleles. As the "gold standard, " hprt mutation frequencies were also measured on the same samples. The methods under development included i) the restriction site mutation (RSM) assay, in which mutations lead to the destruction of a restriction site; ii) minisatellite length-change mutation, in which mutations lead to alleles containing new numbers of tandem repeat units; iii) loss of heterozygosity for HLA epitopes, in which antibodies can be used to direct selection for mutant cells; iv) multiple fluorescence-based long linker arm nucleotides assay (mf-LLA) technology, for the detection of substitutional mutations; v) detection of alterations in the TP53 locus using a (CA) array as the target for the screening; and vi) PCR analysis of lymphocytes for the presence of the BCL2 t(14:18) translocation. The relative merits of these molecular methods are discussed, and a comparison made with more "traditional" methods.

Heterozygous Aprt mouse model: detection and study of a broad range of autosomal somatic mutations in vivo.

During the development of cancer a series of specific genetic alterations have to occur in a stepwise fashion to transform a normal somatic cell into a malignant tumor cell. These genetic changes can be roughly divided in two groups: mutations in proto-oncogenes that result in a constantly activated gene product and mutations in tumor-suppressor genes that result in loss of function. While oncogenic mutations often have a dominant phenotype and mutation of one allele is sufficient for activation, in general both alleles of a tumor suppressor gene have to be disrupted to abolish its function. The requested specificity for activating mutations in proto-oncogenes is high, since only a limited number of mutations at specific sites result in an activated protein. In contrast, disruption of a tumor suppressor gene can be accomplished via various mechanisms. Familial cancers often contain a germline mutation in one allele of a tumor suppressor gene. In tumors, the second allele is then frequently lost by genetic alterations that also affect the heterozygous state of multiple loci adjacent to the tumor suppressor gene. Genetic events especially, such as mitotic recombination, chromosome loss and deletion, are frequently responsible for the loss of the functional allele of heterozygous mutant tumor suppressor genes. We generated an Aprt(+/-) mouse model that allows us to study in detail the nature of the alterations that lead to loss of the wild-type Aprt allele in somatic cells. These genetic changes are thought to be analogous to those occurring at autosomal tumour suppressor genes, where they may contribute to the development of cancer. Furthermore, this mouse model allows determination of the extent and mechanisms by which chemical carcinogens induce loss of heterozygosity and identification of the nature of the DNA adducts responsible.

Both the rate and spectrum of loss of heterozygosity differ between human lymphoblastoid cells derived from various donors.

Loss of heterozygosity (LOH) contributes significantly to the inactivation of tumor suppressor genes and may involve a variety of mechanisms. Studying loss of HLA-A2 alleles in human lymphoblastoid cell lines, we previously showed that mitotic recombination and chromosome loss with concomitant duplication of the non-selected chromosome were the most frequent mechanisms of LOH. In the present study we used the HLA system to determine the rate and spectrum of LOH mutations in the EBV transformed lymphoblastoid cell line R83-4915. Spontaneous loss of HLA-A2 in R83-4915 occurred with a rate of 7.9x10-7 which was 5 to 10-times lower compared to the previously observed rate of loss of HLA-A2 in other lymphoblastoid cell lines. Among the HLA-A2 mutants, 27% did not show LOH of additional chromosome 6 markers. Molecular analysis showed that neither large deletion nor gene conversion was the cause for their mutant phenotype. The remaining mutants showed LOH, which was caused by mitotic recombination (40%) and chromosome loss (33%). However, the chromosome loss observed in mutants of R83-4915 was not accompanied by the duplication of the remaining chromosome. Instead 3 out of 5 mutants became polyploid suggesting that different mechanisms exist to compensate for chromosome loss. In conclusion, the rate and types of LOH that can be observed in cell lines obtained from various donors may depend on the genetic make-up or the transformation status of these cells

A search for novel tumour suppressor genes for adult acute leukaemia by allelotyping at sub-telomeric chromosomal regions.

A search was initiated towards the localization of novel mutated tumour suppressor genes that may be involved in adult leukaemia. For this purpose, we measured the occurrence of loss of heterozygosity (LOH) in nine patients with acute B-lineage leukaemia (ALL) and one with undifferentiated leukaemia (AUL). Eight leukaemias exhibited a diploid karyotype. For each patient, PCR products of 130 polymorphic microsatellite markers, located in subtelomeric areas of every autosomal chromosome arm were analysed to visualize LOH events resulting from reduplication of a single mutated chromosome or from mitotic recombination. These kinds of LOH events contribute most to LOH in model systems but cannot be detected by classical cytogenetic techniques. By comparing allelic PCR products in tumour cells with those in normal cells, LOH was found in tumour cells of one ALL patient at 9p which harbours the known p16INK4A tumour suppressor gene. In the AUL patient, however, LOH was detected at the telomeres of 4q and 21q, suggesting that these sites may contain novel tumour suppressor genes specifically involved in this form of leukaemia. In the DNA of tumour cells from eight out of 10 patients no LOH was detected. This is in contrast with the general assumption that LOH is a frequent phenomenon in ALL. However, some markers at telomeric regions of chromosomes were already homozygous in the control T-cells of several patients. For instance, we found in the DNA of control cells from one patient five consecutive microsatellites on 9p up to 9p43 which were homozygous and in three other patients homozygosity was observed in band 8q24, which includes the MYC gene. These observations indicate that LOH events already are present in non-cancerous putative stem cells and that mitotic recombination may be a very early event in leukaemogenesis.

Chromosome loss with concomitant duplication and recombination both contribute most to loss of heterozygosity in vitro.

Loss of heterozygosity (LOH) plays an important role in the expression of recessive mutations in mammalian cells. To gain insight into the rate and mechanisms of LOH the autosomal HLA-A gene was used as a model system. Spontaneous HLA-A2 mutants originated with a rate of respectively 4.1 x 10(-6) and 6.9 x 10(-6) per cell per generation in TK6 and WI-L2-NS, two isogenic lymphoblastoid cell lines which differ in TP53 status. The rate of loss of HLA-A2 is 10-50 times higher compared to the mutation rate of the X-linked HPRT gene. The homozygous TP53 mutation in WI-L2-NS had no effect on the rate of HLA-A2 loss or the spectrum of these mutations. Microsatellite analysis of most of the HLA-A2 mutants (84%) showed LOH for multiple markers on chromosome arm 6p telomeric of a recombination breakpoint, LOH for all 6p markers, or LOH for markers on both the 6p- and 6q-arms. Cytogenetic analysis showed that these mechanisms gave mutant cells which harbored two intact chromosomes 6 and which were indistinguishable from non-mutant cells. Therefore, loss of HLA-A2 is mainly caused by somatic recombination (33-50%) or chromosome loss with duplication of the remaining chromosome (34-40%). These findings correspond to the mechanisms behind loss of the wild-type RBI allele in retinoblastoma and suggest that both somatic recombination and chromosome loss followed by duplication contribute to tumorigenesis.

Isolation and molecular characterization of spontaneous mutants of lymphoblastoid cells with extended loss of heterozygosity.

The human major histocompatibility complex comprising the HLA class I and II genes provides a versatile source of natural heterozygous loci. This polymorphic genetic system allows analysis of the mechanistic aspects of loss of heterozygosity (LOH), a major phenomenon observed at tumor suppressor genes in human cancer cells. Four lymphoblastoid cell lines, ORI, TK6, WI-L2-NS and VH, were used to adjust current HLA immunoselection protocols to quantify loss of HLA-A2 in human lymphoblastoid cell lines. The modified selection protocol was used to isolate independent spontaneous HLA-A2 mutants from the lymphoblastoid cell line ORI. The frequency of spontaneous loss of HLA-A2 in ORI was 1.7 x 10(-5). By HLA typing 35 spontaneous HLA-A2 mutants, we showed that 74% of the HLA-A2 mutants also lost expression of the HLA-B allele, which is located on the same haplotype as HLA-A2. Microsatellites on both arms of chromosome 6 were used for molecular characterization of the spontaneous HLA-A2 mutants. Loss of heterozygosity at various loci on the p-arm or loss of an entire chromosome 6 was found in 80% of the mutants. Surprisingly, it appeared that a presumed mitotic recombination event in the cell line ORI itself had resulted in homozygosity of all markers distal from the HLA locus up to the telomere. This greatly limited the detection of mitotic recombination, resulting in LOH up to the telomere, on the short arm of chromosome 6 in this cell line. However, gene dosage analysis detected two copies of the remaining D6S265 allele in mutants which showed LOH at various loci along the p-arm. This suggested that recombination resulted in LOH in these mutants. The lymphoblastoid cell line TK6 did contain informative microsatellites along the complete chromosome 6. Mutants of TK6 either retained heterozygosity of all p-arm markers, showed LOH of all p-arm markers or showed loss from a breakpoint up to the telomere. These data indicate that recombination and chromosome loss both are important mechanisms involved in loss of the HLA-A2 allele in vitro. Such mechanisms may be involved in LOH in vivo and contribute to loss of tumor suppressor alleles.

Comparison of spontaneous hprt mutation spectra at the nucleotide sequence level in the endogenous hprt gene and five other genomic positions.

Mutation spectra at the nucleotide sequence level of five hprt cDNA genes integrated in different genomic positions of a HPRT(-) derivative of the human lymphoblastoid TK6 cell line were compared with each other and with the spectrum of mutations confined to the 657 bp coding region of the endogenous hprt gene in the parental TK6 cells. The mutation rates in these genomic positions vary significantly and also the mutation spectra are different. In each genomic position the majority of mutations are basepair substitutions and deletions. the ratios of which vary among the genomic positions. Although it is likely that the different rates of deletion are to a large extent the net result of different rates of misalignment and repair of these errors in the various genomic positions, for the basepair substitutions it is not possible to deduce which mechanisms have caused these mutations and what causes the differences among the genomic positions. Taken together, the differences in mutation rates and spectra cannot be explained by a single mutagenic process.

Spontaneous mutation spectrum in the hprt gene in human lymphoblastoid TK6 cells.

A spectrum of 100 mutations in the endogenous hprt gene of the human lymphoblastoid TK6 cell line is presented. The majority of the mutations originates in sequences outside the coding region of the gene. Large deletions are a major cause of inactivation of the hprt gene (57% of the mutants). Mutations in the splice sites that result in several forms of aberrantly spliced mRNA are relatively frequently recovered (16%) compared with mutants containing alterations in the coding region of the hprt gene (27%). The majority, but not all, of the splice mutants contain an alteration in the consensus sequences of the splice sites. A spectrum of mutations in the coding region of the hprt gene enlarged to a total of 42 mutants shows that basepair substitutions predominate (71%) and that small deletions and insertions are less frequently recovered. Basepair substitutions arise slightly more frequently at GC basepairs than at AT basepairs.

UV-induced mutagenesis in the endogenous hprt gene and in hprt cDNA genes integrated at different positions of the human genome.

The influence of the genomic position of a gene on UV-induced mutations was studied in the endogenous hprt gene in human lymphoblastoid TK6 cells and in cell lines derived from TK6 each containing a single copy of a hamster hprt cDNA gene integrated on a retroviral vector in different positions of the human genome. Previous studies showed that the genomic sequences surrounding the integration site influence spontaneous mutagenesis, resulting in a 10-fold difference in mutation rates among the hprt cDNA genes. Here we demonstrate that the genomic positions of three integrated hprt cDNA genes do not influence UV-induced mutagenesis. The mutability by UV irradiation in these cell lines is approximately the same (16.0 x 10(-6) per J/m2). The nature of the UV-induced mutations determined in two of the cell lines containing the integrated hprt cDNA gene (approximately 30 mutants each) was also found not to be different. The endogenous hprt gene in the parental TK6 cells exhibits a significantly lower mutability (2.1 x 10(-6) per J/m2) than the cDNA genes, but the spectrum is very similar. The spectrum in TK6 shows no influence of strand-specific repair and resembles most closely the spectrum obtained by McGregor et al. after irradiation of human cells synchronized in S-phase. This suggests that mutations arising in cells that are in S-phase at the time of irradiation constitute the majority of the mutants in an asynchronous TK6 cell population. We hypothesize that repair in the endogenous hprt gene in TK6 cells is very efficient, removing virtually all lesions before replication takes place except in cells that were in S-phase at the time of irradiation when there is not enough time for repair. Furthermore we suggest that the higher mutability of the integrated hprt cDNA genes compared with the endogenous gene is caused by a less efficient repair in the cDNA genes.

Mutation induction by UV light in retroviral hprt cDNA integrated at various chromosomal positions in repair-deficient hamster cells.

Mutation induction by UV irradiation was studied in a retroviral vector integrated in one copy per cell at various chromosomal positions. As a mutational target, hamster hprt cDNA was present on the retroviral vector. To minimize the influence of repair we used repair-deficient hamster cells, V-H1 and UV5, as a recipient for the vector. There is no major influence of chromosomal position on UV-induced mutation frequency and spectrum because no statistically significant difference between mutation induction in retroviral cDNA copies integrated at different chromosomal sites was observed. However, a major difference was found in mutation induction between the endogenous hamster hprt gene and the retroviral cDNA copies. Most noticeable was the absence in the cDNA of the strong strand bias for mutation induction, which was reported for the endogenous hprt gene. Our results with the hprt cDNA exclude as a general phenomenon a difference in mutation induction for leading and lagging strand DNA replication, which was proposed as an explanation for this strand bias in the endogenous gene. The similarity of mutation induction in the different retroviral cDNA copies, all directly surrounded by the same DNA sequence elements, together with the marked difference between the mutation induction in the endogenous gene and the cDNA copies may point to an important role of chromatin structure in mutation induction.

Amplification of the 11q13 region in human carcinoma cell lines: a mechanistic view.

We previously proposed that a local duplication, not the loss of the subsequently amplified marker from its original site, might be the first step in gene amplification in human cells. It is important to investigate this issue in naturally occurring amplification and when copy numbers are relatively low. We have examined the location of single-copy and amplified 11q13 sequences in cell lines from human breast cancers and squamous cell carcinomas using fluorescence in situ hybridization both with a probe specific for the 11q13 amplifying region and with a chromosome 11-specific library. We show that in most cell lines the 11q13 amplicons are physically linked to chromosome 11 or to a chromosome derived from chromosome 11 by various rearrangements near the 11q13 region. In none of the cell lines were interstitial deletions of 11q13 detected. These results indicate that 11q13 amplification in human tumor cells generally does not involve deletion as the initial step. One cell line with chromosomally located amplified 11q13 sequences contained double minutes that harbored the MYC gene but no 11q13 sequences. This suggests that the genetic outcome and the mechanism of gene amplification are probably dependent on specific DNA sequences rather than on the origin of the cells.

Genome wide spontaneous mutation in human cells determined by the spectrum of mutations in hprt cDNA genes.

We have studied spontaneous mutagenesis in five hprt cDNA genes integrated at five different genomic positions in a human lymphoblastoid cell line (TK6). The spectra of 40 mutants from each position were combined to obtain a mutation spectrum of the overall genome. This collection of mutants was used to assess the contribution of several mutagenic processes to spontaneous mutagenesis. Deletions and single base pair changes account for the majority of the mutants and arise in approximately equal amounts (43 and 41%, respectively). The majority of the deletions and insertions are < 5 bp and are likely to be caused by template-directed misalignment (slippage) during replication. To account for frameshifts at non-iterated sites we propose a slightly different template-directed replication error model. A considerable amount of the observed base pair changes can also be explained by this last model, but several other processes leading to base pair changes such as depurination, deamination or spontaneously arising DNA damage are likely to contribute as well. We have compared this spectrum with mutation spectra in the endogenous hprt genes using published mutation data. It is shown that in the endogenous genes the contribution of base pair substitutions is much larger (71%) than in the hprt cDNA integrates and that deletions are less frequently observed (20%). The mutation rates of the integrated hprt cDNA genes show a mean increase of 30-fold as compared with the endogenous hprt gene. This results in a 60-fold increase of the absolute rate of deletion in the hprt cDNA genes and in a 15-fold increase of the base pair substitution rate. Replication errors such as slippage or the mechanism proposed in this study probably account to a large extent for this increase.

Genomic position influences spontaneous mutagenesis of an integrated retroviral vector containing the hprt cDNA as target for mutagenesis.

We have used five isogenic human lymphoblastoid cell lines each containing a retroviral vector at a different position in the genome to assess the influence of these positions on spontaneous mutagenesis. The vector contains the hamster hprt cDNA and the neo gene, both genes are transcribed from the retroviral LTR promoter. The rates of mutation leading to a HPRT- phenotype during growth in non-selective medium differed up to 60-fold in the five retroviral integrates, ranging from 5.9 x 10(-6) to 3.5 x 10(-4) mutations per cell generation. From each of the cell lines approximately 20 independent mutants were analyzed by Southern blot analysis. In two cell lines all mutations were caused by inactivation of the LTR promoter (presumably by DNA methylation), whereas in another cell line the estimated rate of this mutation is 1000-fold lower. Another important class of mutation is homologous recombination between the LTRs. This accounts for at least half of the mutants in the other three cell lines. Mutants carrying deletions or point mutations form a minor fraction of the mutant distribution. Mutations confined to the hprt cDNA sequences only were studied by selecting HPRT- mutants in the presence of G418. Even for this subset of mutations the rates can vary at least 10-fold between the different genomic positions, ranging from 4.2 x 10(-7) to 5.1 x 10(-6). We conclude therefore that mutations leading to a HPRT- phenotype are quantitatively as well as qualitatively different in the studied cell lines. This suggests that spontaneous mutagenesis in a gene is dependent on its position in the genome.

Gene amplification in a human osteosarcoma cell line results in the persistence of the original chromosome and the formation of translocation chromosomes.

Although gene amplification, a process that is markedly enhanced in tumor cells, has been studied in many different cell systems, there is still controversy about the mechanism(s) involved in this process. It is still unclear what happens to the DNA sequences that become amplified, whether they remain present at their original location (conservative gene amplification) or whether gene amplification necessarily results in a deletion at the original location (non-conservative gene amplification). We have studied gene amplification in a human osteosarcoma cell line, starting from a cell clone which contains only one copy of a plasmid integrate. Independent amplificants, originating from this clone and containing elevated plasmid copy numbers, were isolated and analyzed. Based on previous observations, encompassing the persistence of single-copy DNA sequences besides amplified DNA sequences clustered at a different location in the independent amplificants, we proposed an amplification pathway including a local duplication step and transposition of the duplicated DNA to other chromosomal positions. Now we have extended our study to more independent amplificants. We prove that the single-copy plasmid-containing chromosomes in the different amplificants and the single-copy plasmid-containing chromosome in the original parental cell clone are indeed identical, namely a translocation chromosome composed of at least three parts of which two originate from chromosomes 14 and 17. We show that the unit of amplification and the unit of the proposed transposition event are at least 1.5 Mb. We also demonstrate that the amplified DNA sequences, present at genomic locations other than the original single-copy DNA sequences, are preferentially associated with chromosome 16. We find that the amplified DNA sequences are often located at or near a site of chromosome translocation involving chromosome 16. In one cell clone we detect the amplified DNA sequences in most of the cells to be located within a complete chromosome 16 while in a minority of cells the amplified sequences are located at or near a breakpoint on a translocation chromosome 16. This indicates that this amplification region is highly unstable and frequently gives rise to translocation events.

Gene amplification in human cells may involve interchromosomal transposition and persistence of the original DNA region.

In tumor cells in vivo and in vitro the amplification of large DNA sequences is a spontaneous and frequently occurring genetic event. We have used human cells to study independent events leading to a low level of amplification of a single copy of an integrated plasmid. Fluorescence in situ hybridization, chromosome banding, and chromosome painting revealed that the new amplified DNA sequences can become located on chromosomes that are totally unrelated to the chromosome that harbors the original DNA sequences, indicating that the transposition of amplified DNA sequences is interchromosomal. In cells containing amplified DNA sequences the integrated single-copy plasmid remained at its original location. The unit of amplification contained a DNA fragment of at least a 800 kb and the same fragment was also present in the parental single-copy cell clone. The data suggest that a doubling of the DNA region at the original location precedes or is coupled to gene amplification.

Studying DNA mutations in human cells with the use of an integrated HSV thymidine kinase target gene.

A shuttle vector carrying the origin of SV40 replication, the thymidine kinase (tk) gene of herpes simplex virus and the E. coli xanthine guanine phosphoribosyl transferase (gpt) gene has been introduced into human TK- cells. A transformed cell line containing only one stably integrated copy of the shuttle vector was used to study mutations in the introduced tk gene at the molecular level. Without selection for gpt expression, spontaneous TK- mutants arose at a frequency of approximately 10(-4)/generation, and were caused by deletion of plasmid sequences. However, when selection for expression of the gpt gene was applied, the background level of mutations at the tk gene was below 4.10(-6). From this cell line, TK- mutants were obtained after treatment with N-ethyl-N-nitrosourea (ENU). COS fusion appeared to be an efficient method for rescue and amplification of the integrated shuttle vector from the human chromosome. After further amplification and analysis in E. coli, rescued tk genes were easily identified and were shown to be physically unaltered by the rescue procedure. In contrast to rescued tk genes from TK+ cells, those obtained from the ENU-induced TK- mutants were unable to complement thymidine kinase-negative E. coli cells. Two such tk mutations were mapped in E. coli by marker rescue analysis. A GC----AT transition was the cause of both mutations. We show here that plasmid rescue by COS fusion is a reliable system for studying gene mutations in human cells, since no sequence changes occurred in rescued DNA except for the 2 ENU-induced sequence changes.

Cytosine methylation in the EcoRI site of active and inactive herpesvirus thymidine kinase promoters.

The herpesvirus thymidine kinase (tk) gene integrated in the human cell line, 2.1-a, can be inactivated by limited de novo methylation. All these TK- clones show partial EcoRI digestion of the recognition site (cGAATTCg) in the tk promoter in contrast to complete digestion of this site in the original cell line. Studies on well-defined substrates prepared in vitro showed that methylation of one cytosine in the EcoRI recognition sequence resulted in partial and methylation of both cytosines in severe inhibition of digestion by EcoRI. This characteristic was used to determine whether no, one or both cytosines in the EcoRI site of the tk promoter were methylated in various TK- clones derived from 2.1-a and in TK+ clones re-expressing the gene after 5-azacytidine treatment. A high correlation was found between inactivity of the tk gene and methylation of only one of the two cytosines in the EcoRI recognition site. The results also show that the tk promoter can be active despite the presence of a methylated cytosine.

De novo methylation as major event in the inactivation of transfected herpesvirus thymidine kinase genes in human cells.

Spontaneous inactivation of integrated thymidine kinase genes was studied in three human cell lines, one with multiple copies and two with a single copy of a transfected shuttle plasmid containing two selectable genes: the HSV tk gene and the Eco gpt gene. Selection for gpt expression prevented the isolation of TK- mutants which are the result of plasmid loss. Under these conditions TK- clones were isolated with a frequency of 5.10(-6) both with the cell line containing 5 or 6 copies of the tk gene and with one of the two cell lines containing one copy of this gene. This inactivity of the tk gene was associated with de novo methylation as the number of HAT-resistant (TK+) clones strongly increased after growth of the TK- derivatives in the presence of the demethylating agent, 5-azacytidine. Digestion with methylation-sensitive restriction enzymes revealed two different patterns of DNA methylation in the genomic DNA of TK- variants. In the TK- derivatives of the cell line containing multiple copies of the tk gene many HpaII restriction sites in the gene copies were insensitive to digestion. These HpaII sites were, however, not methylated in TK- variants of the cell line containing one copy of the plasmid, and methylated CpGs could be detected only with EcoRI which recognizes the cGAATTCg sequence in the tk promoter region. With the other of the two single-copy TK+ cell lines no TK- mutants were obtained, suggesting that the position of a gene in the genome is an important factor in determining the frequency and the extent of de novo methylation. Additionally, we observed that remethylation is an even more efficient process of gene inactivation as TK+ clones reactivated with 5-azacytidine can become TK- again at a 100-fold higher rate than the original TK+ cell line.