Factors Affecting de Novo Methylation of Foreign DNA in Mouse Embryonic Stem Cells*

Integration of foreign DNA into an established host genome can lead to changes in methylation in both the inserted DNA and in host sequences and potentially alters transgene and cellular transcription patterns. This work addresses the questions of what factors influence de novo methylation, and whether the integration site or inserted DNA can affect de novomethylation. Homologous recombination was used to integrate foreign DNA into a specific gene, B lymphocyte kinase (BLK), in mouse embryonic stem (ES) cells. Two plasmids were chosen for integration; one contained the adenovirus type 2 E2AL promoter upstream of the luciferase reporter gene, and the second carried the early SV40 promoter. The methylation patterns were analyzed usingHpaII and MspI restriction endonucleases for both homologously recombined and randomly integrated foreign DNA in the ES cell clones. Upon homologous reinsertion of the BLK gene into the genome of mouse ES cells, methylation patterns in this gene were reestablished. In DNA segments adjoined to the BLK gene, the de novo patterns of DNA methylation depended on the viral sequences in these clones and on the locations of the inserts, i.e. on whether the insertions resulted from homologously recombined or randomly integrated foreign DNA. In homologously recombined DNA, sequences carrying the adenovirus type 2 promoter were heavily methylated, and those with an SV40 promoter and an SV40 enhancer element remained unmethylated or hypomethylated. Upon removal of the enhancer element, these inserted constructs also became heavily methylated. In addition, all randomly integrated constructs were heavily methylated independently of the promoter and enhancer element present in the construct. These results indicate that modes and sites of integration as well as the inserted nucleotide sequence, possibly promoter strength, are factors affectingde novo methylation.

Specific patterns of DNA methylation, an epigenetic modification in the mammalian genome (1), have been shown to affect gene expression (2)(3)(4)(5). DNA methylation is involved in a number of regulatory mechanisms, such as imprinting (6,7), X chromosome inactivation (8), carcinogenesis (9 -11), and em-bryonic development (12). Adenovirus (Ad) 1 DNA has been used as a model system to study the effects of foreign DNA integration and methylation on host and Ad gene expression (2,(13)(14)(15)(16). The Ad genome is 34 -36 kb in size and consists of a linear, double-stranded DNA molecule with a 55-kDa protein covalently attached at either 5Ј terminus. When this virus replicates in permissive cells, the replicated genome is not methylated (17,18). However, when the virus DNA integrates into the host genome, as in Ad-induced tumor cells or in Adtransformed cells, the viral genome becomes methylated by a de novo mechanism (19 -21). Frequently, multiple copies of Ad DNA are integrated at many sites in nontandem arrays in the host genome, but there seems to be a preference for sites with patchy homologies at the junctions between cellular and viral DNA (15,16,22,23). When adenovirus (foreign) DNA is inserted into the mammalian genome, there are changes in the methylation patterns of the flanking sequences due to the position of the integration and/or the inserted DNA sequences. After Ad integration, the methylation in the flanking host sequences can be altered (24) as well as other more remote genomic sequences such as major histocompatibility complex class I, Ig C and intracisternal A particle DNA sequences (25). In addition, when only bacteriophage DNA has been stably integrated into cellular DNA, again intracisternal A particle DNA sequences have shown changes in their methylation patterns (26).
Methylation of Ad or any other foreign DNA integration may be affected by the site of integration in the host genome. In addition, position effects on de novo methylation have been shown in several different systems including cell lines containing a variety of transgenes (27,28) and in transgenic mouse models (29,30).
Detailed analyses of DNA methylation and expression always reveal a more complex situation, because transgenic cell lines usually contain multiple copies of the gene of interest at different positions or in concatenated forms in the host genome, which is a common result of random integration. In order to understand better the regulation of de novo methylation, homologous recombination was used to insert different foreign DNAs into one site in the genome of mouse embryonic stem cells. The results show that random integration appears to promote de novo methylation. For homologous recombination with one allele of the same insert in the same cell type, the original methylation pattern for the genomic sequence is reestablished. However, for foreign DNA the extent of de novo methylation depends on the newly integrated sequence.

EXPERIMENTAL PROCEDURES
Plasmids Used for Homologous Recombination and for DNA Probes-AdBLK and SVBLK (Fig. 1) were constructed to perform homologous * This research was supported by European Union Grant EN AX 101805 and Deutsche Forschungsgemeinschaft Grant SFB274-A1. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  recombination in mouse embryonic stem (ES) cells within the B lymphocyte tyrosine kinase (BLK) gene located on mouse chromosome 14 (31). The two constructs contain from the 5Ј-end, a 4.1-kb BLK left genomic fragment from the HindIII site in intron 1Ј to the BamHI site in intron 3, the phosphoglycerate kinase (Pgk-1; Ref. 32) promoter upstream of the neomycin gene, a viral promoter upstream of the luciferase gene, and a 3.3-kb BLK genomic fragment from the BamHI site in intron 3 to the EcoRV site in exon 6 ( Fig. 1). In addition, the thymidine kinase gene with the HSV-1 promoter was also present within the plasmids, but upon homologous recombination and selection with gancyclovir it would not be integrated. The BLK 5Ј-and 3Ј-genomic fragments were a gift of Gemma Texido and Alexander Tarakhovsky (University of Cologne, Germany). The numbering of the BLK exons is according to published data (33). The neomycin gene with the Pgk-1 promoter was derived from a published source (34). The luciferase gene was cloned from the PGL2 control vector (Promega). The only difference between these two plasmids, AdBLK and SVBLK, is in the luciferase promoter region (Fig. 1A). The AdBLK construct contains Adluc with the adenovirus type 2 (Ad2) E2A late promoter upstream of the luciferase gene (35). The SVBLK contains SVluc with both the SV40 early promoter 5Ј of the gene and the SV40 enhancer at the 3Ј-end derived from the PGL2 control vector (Promega). This enhancer contains both the 72-and 21-bp repeats. A third plasmid, SVpBLK, has been constructed that contains SVluc without the SV40 enhancer sequences at the 3Ј-end.
An additional BLK fragment, called BLK-out, was used to determine which clones contained homologously recombined DNA within the BLK gene (provided by Gemma Texido and Alexander Tarakhovsky), since this part of BLK was not present in any of the constructs used for transfection into ES cells. The BLK-5Јab probe is the fragment between the ApaI site in exon 2 and the BamHI site in intron 3, while the BLK-3Јxb probe is the DNA fragment from the same BamHI site to the XbaI in intron 4. The Adluc probe is the XhoI to SalI DNA fragment (3201 bp) of pAd2E2AL-LUX (35), and the same fragment was isolated from PGL2-control (Promega) to obtain the SVluc probe (3160 bp) (Fig.  1B). The neomycin (neo) probe is the 614-bp DNA fragment from BsmI to DdeI, which contains only the neomycin sequence.
Cell Culture-The embryonic stem cell clone BL/6-III from the C57BL/6 inbred mouse strain (36) and the clones derived from this parental clone were grown in Dulbecco's modified Eagle's medium with Glutimax (Life Technologies, Inc.), 15% fetal calf serum, 1 mM sodium pyruvate (Life Technologies), 1ϫ non essential amino acids (Life Technologies), 0.1 mM ␤-mercaptoethanol, and conditioned medium containing leukemia inhibitory factor (provided by the laboratory of Klaus Rajewsky, University of Cologne, Germany). In some experiments, penicillin and streptomycin were added. A modified solution was used to wash the cells, MT-PBS (4 mM NaH 2 PO 4 ⅐H 2 O, 16 mM Na 2 HPO 4 ⅐2H 2 O, 150 mM NaCl) (37). Cells were propagated on tissue culture plastic dishes or flasks at 37°C, 7.5% CO 2 , with 85-90% humidity. Medium was replaced every day, and the cells were passaged every 2-3 days to minimize differentiation of the cells (36).
Gene Transfer, DNA Isolation, and Southern Blot Analyses-All plasmids used for transfection experiments were propagated in Escherichia coli strain XL1-Blue MRFЈ (McrA Ϫ , McrCB Ϫ , McrF Ϫ , Mrr Ϫ , HsdR Ϫ ), a derivative of XL1-Blue (Stratagene). Amounts of 30 g each of linearized plasmid DNA were electroporated into 1-2 ϫ 10 7 ES cells by using the Bio-Rad Gene Pulser apparatus as described previously (38). Medium was replaced daily. After 2 days, 300 g/ml Geneticin (G418sulfate; Life Technologies), and 2 days later, 2 ϫ 10 Ϫ6 M gancyclovir (Syntex/Roche) was added to the mediun to minimize the number of clones with random integration events (39). Individual clones were isolated from day 10 to 12, transferred into 48-well plates, and transferred several days later to 24-well plates. After passaging the clones into larger dishes, no further selection (G418 or gancyclovir) was used. ES clone names begin with an A when they were derived from transfection experiments using the construct containing Adluc. Those clone names that begin with an S were derived from the plasmid containing SVluc. ES clone names had an additional "o" when they were isolated from plates having media with only Geneticin. The first passage without selection is referred to as passage 1 (p1). Each clone analyzed was passaged either into T75 or T175 flasks to isolate DNA for Southern blot analyses by standard procedures. The DNA samples were resuspended in TE (10 mM Tris, pH 7.5, 1 mM EDTA, pH 8.0), and 10 g of each sample was used to identify the clones that contained homologously recombined (HR) or randomly integrated (RI) foreign DNA (Table I). For this analysis, the DNA was cleaved with HindIII (5 units/g DNA) for 6 h at 37°C. The fragments were separated by electrophoresis on a 0.8% agarose gel in 1 ϫ TAE (40 mM Tris acetate, 1 mM EDTA, pH 8.0) for about 15 h at 30 -40 V. The gel was stained with ethidium bromide (1 g/ml) for 10 min, destained in water, and then prepared for Southern transfer (40,41). Nylon Plus membranes (Qiagen) were used for the downward transfer. The membranes were then exposed to x-ray film for 1-5 days.
Within the BLK flanking sequences and the luciferase and neomycin regions, methylation levels were determined by cleaving 30 g of each DNA sample first with BamHI (10 units/g DNA) for 5 h followed by either HpaII or MspI (10 units/g) overnight. Both HpaII and MspI cleaved DNA at 5Ј-CCGG-3Ј sites. HpaII is methylation-sensitive and therefore cannot cleave this site when the internal (3Ј) C is methylated. When the samples were cleaved with BamHI only, 10 g of genomic DNA was used.
Some of the membranes were also exposed to the BAS-IIIS or BASMP phosphor image screen (Fuji), processed in the phosphor imager FUJIX BAS 1000 (Fuji), and analyzed with IPgel 1 software (Signal Analytics).
Luciferase Expression-Protein was extracted from the ES cell clones by standard protocols (Promega). Briefly, cells were washed twice with MT-PBS, scraped with a rubber policeman with MT-PBS, and transferred to an Eppendorf tube. The cells were washed and then resuspended in 1ϫ lysis buffer (25 mM Tris-phosphate, pH 7.8, 2 mM EDTA, 2 mM dithiothreitol, 10% glycerol, 1% Triton X-100). All samples were frozen once, centrifuged, and then assayed three times each with the luminometer Lumat LB 9501 (Berthold, Bad Wildbad, Germany) using the luciferase substrate and buffer system (Promega). In addition, protein concentrations were determined (42) for each sample to normalize the data to relative light units (RLU) per g of protein.

Homologous Recombinant and Randomly Integrated DNA in ES Clones
-Two different DNAs, AdBLK or SVBLK, were electroporated into mouse ES cells to obtain homologous recombinants and to investigate whether specific DNA sequences can affect de novo methylation. ES cells were chosen as the host, since they possess a de novo methylation mechanism (43)(44)(45). The BLK gene was chosen as the target, and both the genomic BLK DNA fragments and the ES cells (BL/6-III) are derived from the same mouse strain, C57BL/6. The plasmids, AdBLK and SVBLK, contain the neomycin and luciferase genes and are flanked by two BLK genomic DNA fragments. Therefore, after HR, the foreign DNA segment is inserted within intron 3 of the BLK gene (Fig. 1B). Two types of HR clones were obtained after electroporation into ES clones, and they differ only within the luciferase region. The A clones carry Adluc, the Ad2 E2A late promoter 5Ј of the luciferase reporter gene (Fig. 1A, Table I, left side). This promoter has been shown to be inhibited by sequence-specific methylation (46,47). The S clones have SVluc with the SV40 early promoter 5Ј of the luciferase coding sequence and at the 3Ј-end the SV40 enhancer element containing the 72-and 21-bp repeats (Fig. 1A, Table I, right side).
Genomic DNA was obtained from about 40 different ES clones from each transfection to determine which clones had incorporated the neomycin and luciferase sequences within the expected BLK site by homologous recombination and which clones contained part or all of these sequences in other locations due to random integration events. The genomic DNA was cleaved with HindIII, and the fragments were separated by electrophoresis on a 0.8% agarose gel for Southern blot analysis. Two different probes were used to analyze the location of the inserts and the presence of the luciferase sequence (Fig. 2). In Fig. 2A, the DNA on the membrane was probed with the EcoRV-HindIII fragment of BLK, designated BLK-out, since this BLK sequence was outside the region used for homologous recombination (Fig. 1B). For the Adluc-containing clones (e.g. A1) and for the SVluc-containing clones (e.g. S4), the expected 8.5-kb fragment for the endogenous allele is detected as well as the 7.9-kb fragment for the homologous recombinant allele. Using the luciferase sequence probes (Fig. 1B, Adluc or SVluc), the expected size fragment of 7.9 kb is also detected and is the same size as that present with the BLK-out probe (Fig. 1B), since these fragments contain both the BLK and the luciferase sequences. When the DNA on the same membrane is probed with left and right BLK fragments, all BLK fragments are present as expected, the endogenous allele (8.5 kb) and the two fragments from the HR allele (5.9 and 7.9 kb).
In addition to the clones that contain foreign DNA homologously recombined within intron 3, a number of ES clones with randomly integrated foreign DNA have been isolated. All RI clones are designated with an asterisk. No 7.9-kb band is detected in the DNA of these clones when BLK-out is used as the probe. In addition, an off size DNA fragment is visible when the DNA on the membrane is hybridized with the luciferase probe. Table I lists the Adluc-containing clones, referred to as A clones, and the SVluc-containing clones, referred to as S clones, which have been chosen for further studies. At the time points indicated, genomic DNA was extracted for Southern blot analyses, and in addition, protein extracts were prepared at different times to determine the luciferase activity in the different cell populations at various passages. In this study, we have analyzed a number of different ES clones to ascertain representative sampling.
Luciferase Expression of ES Clones-For the protein extracts of all HR A and S clones, the RLU/g of protein decrease with initial passaging as seen in Fig. 3. Throughout passaging, the S clones exhibit 10 -50-fold higher RLU/g of protein than the A clones (Fig. 3, compare A and B). In addition, the constitutive RLU/g of protein in later passages is higher for the S clones than for the A clones. Both sets of A* and S* clones with RI foreign DNA exhibit a 100-fold reduction or no luciferase activity at all from early on (passage 3) to the latest passage analyzed (passage 31; data not shown). After an initial drop in luciferase expression, both the A and S clones maintain a constitutive level even to late passages. We have not related these expression levels to the patterns of DNA methylation, because correlations between promoter activity and DNA  a ES clones with homologously recombined DNA chosen for further analyses. These clones were derived from cells transfected with a DNA plasmid containing Adluc (see Fig. 1).
b ES clones with homologously recombined DNA chosen for further analyses. These clones were derived from cells transfected with the SVluc-containing plasmid (see Fig. 1 (Fig. 1B). In this way, each DNA sequence was analyzed individually. In addition, one of the two restriction endonuclease isoschizomers MspI or HpaII was used to analyze the DNA methylation levels. Unlike MspI, HpaII is unable to cleave the sequence 5Ј-CCGG-3Ј when the internal (3Ј) C is methylated. Therefore, the difference in DNA fragment patterns generated with these two endonucleases represents differences in DNA methylation at 5Ј-CCGG-3Ј sites. When control genomic DNA (C in Fig. 4) from untransfected cells is cleaved with BamHI (lanes B), a single fragment is detected as expected when BLK-3Јxb was used as a probe (Fig. 4, C, lanes  B). MspI cuts the DNA into much smaller fragments (lanes M) which are not resolved by electrophoresis on a 0.8% agarose gel. There are larger molecular mass fragments present after BamHI/HpaII cleavage (Fig. 4, lanes H) as compared with BamHI/MspI cleavage (lanes M). This finding demonstrates that many HpaII sites within the BLK 3Ј-flanking sequence (see Fig. 1B) are methylated. A comparison of the control (C) DNAs to the DNAs from A and S clones reveals that the HpaII patterns are very similar or identical (Fig. 4, H lanes). Moreover, for a given DNA fragment within an H lane of any of the ES clones, the relative intensities are similar as for the corresponding bands in the control DNA H lane (Fig. 4, e.g. C and A17 or S11; H lanes). This similarity of HpaII patterns within the BLK 3Ј-flanking sequence has been seen for all A and S clones analyzed. With a few exceptions, similar results have also been obtained for the 5Ј-flanking BLK sequence (data not shown). First, as in the BLK 3Ј-flanking sequence for the control (C) DNA from untransfected ES cells, there are many larger molecular mass fragments present in the H lanes as compared with the M lanes. This finding again implies that the BLK 5Ј-flanking sequences are methylated. The HpaII patterns of DNA from one clone, S4, show some differences in banding patterns of the 5Ј-flanking sequence, both between passages (p6 to p20) and compared with DNA from the parental untransfected ES cells (data not shown). Slight differences in the HpaII patterns may not be detectable, since the 5Ј-and 3Ј-BLK probes hybridize to both the endogenous and HR allelic fragments. We conclude that the homologously reintegrated BLK sequences are remethylated in 5Ј-CCGG-3Ј patterns very similarly to those of the endogenous BLK gene.
A and S Clones Develop Different Methylation Patterns in the Luciferase Region-The luciferase regions of the A and S clones were analyzed to investigate whether foreign DNA sequences influence de novo methylation after their integration. Fig. 5 shows the autoradiograms of Southern blots of genomic DNA isolated at different passages for a number of the A clones. The entire luciferase region including the Ad promoter was used as the probe (Adluc; Fig. 1A). As expected, there is no detectable hybridization to the DNA in the control lane (C, lanes B, M, and  H). However, for the DNAs from the A clones, the expected 5-kb fragment is detected after BamHI cleavage (B lanes). There are 15 5Ј-CCGG-3Ј sites (vertical lines in Fig. 1A) that can be differentially methylated. Different cleavage patterns are generated by MspI versus HpaII (M and H lanes, respectively). Upon HpaII cleavage of the DNA from all passages analyzed, very large HpaII fragments, ranging in size up to the 5-kb BamHI fragment in the DNA of many clones (e.g. A4, A5, A8, A16, and A18) are generated (Fig. 5).
For the S clones, a very different trend is seen. SVluc (Fig.  1A) was used as a probe to determine the methylation patterns of these clones. The DNA samples of only a few clones, such as S5 and S11, exhibit larger HpaII fragments at molecular masses higher than those for the MspI fragments. For the majority of the S clones, the MspI and HpaII cleavage patterns are quite similar (Fig. 6, M versus H lanes). These data demonstrate that the HpaII patterns of DNA from A and S clones are strikingly different. The foreign luciferase gene thus becomes extensively de novo methylated when it is located downstream of the Ad2 promoter but remains unmethylated or

FIG. 2. Southern blot analyses (40) of the DNAs from ES clones transgenic for AdBLK and SVBLK constructs.
10 g of genomic DNA from each ES clone was cleaved with HindIII. The fragments were electrophoresed through a 0.8% agarose gel and transferred to a Qiagen Nylon Plus membrane. The DNA probes were radiolabeled with ␣-32 P to detect DNA fragments after hybridization. In all figures, all ES clones that contain Adluc DNA have designations beginning with an A, while those clones that contain the SVluc DNA begin with an S (see Fig. 1). A, BLK-out (map in Fig. 1B) was used as a probe to detect two expected DNA fragments at 8.5 kb, representing the endogenous BLK allele, and 7.9 kb for the homologously recombined allele within intron 3 (see Fig. 1B). C represents the control lane, ES cells that were not transfected. B, the same membranes as in A were reprobed with either Adluc (left panel) or SVluc (right panel) to detect the 7.9-kb fragment representing the luciferase region within BLK intron 3. C, both membranes were reprobed again with BLK-in left and right together to detect all BLK fragments integrated in the A and S clones. Randomly integrated ES clones are marked with an asterisk. A minus symbol represents clones that lack detectable integrated foreign DNA. hypomethylated when the same gene is within the context of the SV40 promoter and enhancer.
Both the A and S clones were passaged for an extended period of time to determine the regulation of de novo methylation for a given site in the genome, i.e. within the BLK gene. For the A and S clones analyzed, once the HpaII pattern was established, it remained unchanged with passaging, e.g. A15 and S14 (Figs. 5 and 6). In addition, longer exposures were made, and phosphor images were analyzed that confirmed these results (data not shown).
All Randomly Integrated Sequences Have a High Degree of Methylation in the Luciferase Region-As a control, genomic DNAs from a number of A* and S* clones with randomly integrated DNA were also analyzed (Fig. 7). In contrast to the A and S clones with the HR foreign DNA, all HpaII patterns in RI clones contained only very high molecular mass fragments starting at 1 kb up to the size of the 5-kb BamHI fragment. These RI clones carried from about five to 15 copies of the foreign DNA. However, detailed mapping was not done, and the sites of integration remained unknown. We conclude that randomly integrated foreign DNA becomes extensively de novo methylated.

The DNAs in the A and S Clones Differ in Their Methylation
Patterns in the Neomycin Gene-Since the HpaII patterns for the DNAs from the homologously recombinant clones with Ad2 or SV40 promoters in front of the luciferase gene are strikingly dissimilar, the neomycin regions just 5Ј of the luciferase sequence have also been analyzed in DNA from the same ES clones. Since the nucleotide sequence of the inserted foreign DNA is identical in both sets of these HR clones, the same neomycin probe was used to determine the HpaII cleavage patterns. In the control lanes (Fig. 8, C, lanes B, H, and M), there are no detectable bands at the sizes expected for the neomycin fragments. For the DNAs from the A clones, the HpaII patterns (H lanes) show various intensities of molecular fragments larger than the MspI patterns (M lanes) (e.g. clones A5, A8, and A18) (Fig. 8) neomycin region. Hence, there appears to be a congruence in the levels of de novo methylation between the luciferase and neomycin regions; the DNAs from clones that are strongly methylated in the luciferase region are also more highly methylated in the neomycin region.
If this correlation holds for the S clones as well, we would expect very low levels of methylation in the neomycin regions, since there is little to no methylation present in the luciferase sequence. As expected for the S clones, the MspI and HpaII patterns are quite similar (Fig. 9, compare the M and H lanes) (i.e. no or very low methylation is detected).
The DNA in All RI Clones Has a High Degree of Methylation in the Neomycin Region-Next, the neomycin region was investigated for differences between the HpaII patterns of the RI A* and S* clones. As observed in the luciferase region (Fig. 7), only very large molecular mass fragments are detected in the HpaII patterns of the DNA on Southern blots hybridized with the neomycin probe (data not shown). These data confirm that all Probe BLK-3Јxb (see Fig. 1B) was used to determine the methylation patterns of the DNA from the transgenic ES clones as well as from the parental ES cells (C) that had never been transfected. Both HpaII and MspI cleave DNA at 5Ј-CCGG-3Ј, but HpaII is methylation-sensitive. Sites that are not cleaved by HpaII are detected on the autoradiogram when DNA cleavage fragments are present in an H lane that have higher molecular mass as compared with the M lane in the same passage. These larger fragments are due to methylation at these sites.  Fig. 1, A and B) was used to determine the methylation patterns within this sequence of integrated DNA in the transgenic ES clones by comparing the M and H lanes after each passage. segments of the randomly integrated foreign DNA become extensively de novo methylated both in the luciferase and neomycin genes.
DNA Methylation Patterns and Luciferase Expression of the SVpBLK Clones Lacking the Enhancer Element-Sp1 sites have been implicated in inhibiting de novo methylation of 5Ј-CG-3Ј-rich regions (48,49) and in increasing gene expression (50). The Sp1 sites could negatively affect the de novo methylation of the entire luciferase and neomycin genes within the S clones. To test this hypothesis, new ES clones were established using another plasmid, SVpBLK, which contains SVluc as in the S clones, but without the SV40 enhancer at the 3Ј-end (see Fig. 1A). The 21-and 72-bp repeats that are part of the enhancer were also removed.
In contrast to the HR clones from the S group, these new SVp clones (HR and RI clones) have genomic DNA that is not completely cleaved with HpaII as seen with the A clones (data not shown). These results indicate that the DNA has become highly de novo methylated in all ES clones lacking the enhancer and the Sp1 sequences. While promoter strength is  (M lanes). Probes SVluc or Adluc (see Fig. 1, A and B) were used to determine the methylation patterns of the S* or A* clones, respectively, within this sequence of integrated DNA by comparing the M and H lanes within each passage. Randomly integrated ES clones are marked with asterisks. The results for the DNAs from clones S18 and So5, which show low levels of DNA methylation and are derived from HR experiments, are shown for comparison.  probably a factor influencing de novo methylation, these data imply that cis-acting sequences can inhibit de novo methylation over several kb, as suggested elsewhere (51).
The luciferase expression from the SVp HR clones with the highly de novo methylated sequences is at least 100-fold lower (10 -25 RLU/g of protein) as compared with the original S clones (data not shown). Similarly, the SVp RI clones show very low to no luciferase activity (data not shown).
De Novo Methylation after Release and Reestablishment of Selection-Can this experimental mode lead to the high levels of methylation in the Adluc-containing clones? An early passage of cell clone A8 was used for the release and reselection experiment and passaged either in the presence or the absence of G418. Genomic DNA from both clonal sublines was then isolated after 8 days. A comparison of the two sets of A8 DNA shows that there is no change in the methylation patterns in either the luciferase or neomycin region. This finding suggests that the expression of neomycin does not play a role in the inhibition or activation of de novo methylation of its sequence or of the luciferase gene.

De Novo Methylation of Foreign DNA Is Dependent on Sev-
eral Factors-For many problems in molecular biology, it is of considerable interest to understand the mechanisms and the regulation of de novo methylation of foreign DNA that has been inserted into an established mammalian genome, haphazardly or by experimental design. One of the major unresolved questions about the mechanism of de novo DNA methylation addresses the factors determining the sites and extent of foreign DNA methylation upon insertion into the recipient genome. The following factors could be of relevance: (i) the site of insertion, (ii) the nucleotide sequence of the foreign DNA, (iii) specific motifs in the foreign DNA, (iv) the timing of the insertion event relative to the cell cycle, and (v) promoter strengths and/or elements (e.g. enhancers) in the inserted foreign DNA.
We have devised a series of experiments in which part of the murine BLK gene on chromosome 14 was reinserted into one of its authentic allelic genomic sites by homologous recombination. Mouse ES cell clones with the reinserted DNA in the correct position were selected. ES clones with the BLK DNA inserted into randomly targeted positions were also analyzed as controls. Foreign DNA sequences have been attached to the BLK gene in the constructs used in the electroporation-driven transfection experiments and have been placed inside the endogenous BLK gene (Fig. 1B). Three different constructs were transfected into ES cells to create three types of ES clones. In all of them, the neomycin gene has the same promoter, phosphoglycerate kinase, but the luciferase gene has a different promoter and/or enhancer sequence in each construct. The A clones carry the luciferase gene under the control of the Ad2 E2A late promoter. For the S clones, the SV40 early promoter is located 5Ј of the luciferase gene, and the SV40 enhancer sequences are located 3Ј.
The DNAs in all of the A clones become de novo methylated in both the luciferase and neomycin genes, while the S clones stay hypomethylated even after 30 passages. The third type of ES clone, SVp, is identical to the S clone, except that it contains the promoter without the SV40 enhancer sequences at the 3Ј-end. Unlike the original S clones, these HR clones become de novo methylated within the luciferase and neomycin sequences at levels as seen for the A clones. These results indicate that the enhancer element, containing the 21-and 72-bp repeats, is required to inhibit de novo methylation within the entire luciferase and neomycin regions. The protection of sequences from de novo methylation may be due to factors binding these sequences, e.g. multiple Sp1 sites that are present in the en-hancer sequence. Inhibition may occur at sites where there is an interaction between the promoter and the enhancer element within the transgene. In addition, the strength of the SV40 promoter, as compared with the Ad2 E2A late promoter, could also be a factor.
The DNAs from all of the RI clones in our study, regardless of the foreign sequence integrated, become highly methylated. These sequences can be recognized as foreign and may activate a host defense de novo methylation reaction (52,53). Thus, the enhancer element that is present in some of the RI clones (data not shown) or promoter strength does not suffice to inhibit de novo methylation after random integration. Topologically correct insertion, as after reintegration by HR, apparently does not activate this defense mechanism. Therefore, our data indicate that genome location plays a major role in the regulation of de novo methylation.
The results presented in this report lead to the following conclusions. (i) The homologously recombined BLK sequence becomes remethylated in patterns very similar to the authentic preexisting cellular patterns of the endogenous alleles. (ii) Randomly integrated foreign DNA becomes very heavily methylated. Thus, position and vicinity effects including sequence and structure of the insert must be of considerable importance in determining de novo methylation patterns. (iii) The extents of de novo methylation of the foreign luciferase and neomycin genes reintegrated homologously with the BLK sequence depend on the sequences present in these constructs and/or on promoter strength. The early SV40 promoter with the enhancer sequences seems to prohibit de novo methylation, and the luciferase gene is expressed in the transgenic ES cells. Constructs under the control of the E2A late promoter of Ad2 DNA or under the SV40 promoter without the enhancer elements become extensively de novo methylated even in the authentic BLK position, and luciferase expression is reduced. However, in randomly integrated constructs the nature of the promoter or presence of the SV40 enhancer element is not decisive for de novo methylation. Thus, their role in determining de novo methylation may not be fundamental, at least not in randomly integrated foreign DNA. (iv) Removal and addition of the selective drug G418 does not alter the methylation levels of the examined DNA sequences in the ES clones that contain the Ad2 E2A late promoter. (v) After homologous recombination into BLK genes, nucleotide sequences in the transgenic foreign DNA can affect de novo methylation. One can ponder the possibility that promoters with their transcriptional potential in transgenic sequences within a specific chromatin arrangement interact with the cellular machinery that is responsible for de novo methylation or its inhibition.
A Working Model-The mechanism of de novo methylation of foreign DNA inserted into an established mammalian genome cannot yet be explained in detail. De novo methylation may be related to an ancient cellular defense system targeted against foreign DNA (52,53). In the present study, we have investigated some of the factors influencing de novo methylation when an authentic cellular DNA sequence is reinserted into the mouse genome at different locations. The experimental approach chosen exposes the authentic unmethylated BLK sequence to the de novo methylation system in two different ways. In one set of experiments, the BLK DNA has been reinserted into its original chromosomal location by homologous recombination. Alternatively, upon random integration by heterologous recombination, the BLK and adjacent sequences are located at several randomly selected sites where they are apparently recognized as foreign by the de novo methylation machinery. In the latter context, promoter strength and presence or absence of the SV40 enhancer sequence seem to be of lesser importance in determining the extent of de novo methylation.
As a working model, we are pursuing the possibility that the enzymatic process of de novo methylation recognizes and is dependent upon the specific chromatin structure at the site of an individual gene or DNA segment. Thus, the endogenous BLK DNA at its authentic site or the same DNA inserted at randomly chosen locations would look very different to the DNA methyltransferase system. When the BLK gene and adjoining sequences in the construct are homologously reinserted into one of the authentic allelic BLK positions on mouse chromosome 14, the preexisting chromatin configuration can be reconstituted. Under these conditions, the de novo methylation system then imprints the methylation pattern specific for the BLK locus that had preexisted prior to the transfection of ES cells. In contrast, after random integration, when the BLK construct arrives in an alien position with a locus-specific, but not BLK-typical, chromatin arrangement, a different pattern of de novo methylation will be imposed upon the integrated DNA that is recognized as foreign in this location with a heterologous sequence context. Of course, this model still leaves the question unanswered of how de novo methylation and chromatin structure are interdependent. Differences in the accessibility of individual 5Ј-CG-3Ј dinucleotides to the DNA methyltransferase system may be one of the important parameters.