Clusters of CpG dinucleotides implicated by nuclease hypersensitivity as control elements of housekeeping genes

The synaptophysin-encoding gene in rat and man is specifically transcribed in neuroendocrine cells

Synaptophysin (SY) is an integral membrane protein of presynaptic small (30-80-nm) translucent vesicles also present in dispersed neuroendocrine cells. As the occurrence of this type of vesicle is specific for two major pathways of differentiation, the neuronal and neuroendocrine-epithelial information on the regulation of SY synthesis should contribute to an understanding of regulatory principles common to both pathways. Isolation and comparison of the complete rat and human single-copy genes showed that despite the difference in size (16 kb in rat vs. 13 kb in man) intron/exon boundaries are precisely conserved. Surprisingly, intron VI is located in the 3'-noncoding region in both species. The major transcriptional start point, as determined by primer extension and S1-nuclease protection analyses in rat pheochromocytoma-derived PC12 cells and rat brain, mapped to a site 27 nt 5' of the first methionine codon. Unexpectedly, the 5' upstream region is devoid of any TATA or CAAT boxes, but shows instead typical features of 'housekeeping' genes, i.e., G + C-rich islands and four Sp1-binding motifs. Using 'nuclear run-on' assays, we have identified examples in which SY synthesis is regulated at the transcriptional level. Reporter gene constructs showed that approx. 1.2 kb of the immediate upstream region contains promoter enhancer elements that were, however, insufficient to confer cell-type specific expression, whereas sequences farther upstream were able to suppress thymidine kinase promoter activity in a cell-type-dependent fashion.

Selection against CpG dinucleotides in lentiviral genes: a possible role of methylation in regulation of viral expression

Extremely low frequencies of CpG dinucleotides are found in the genomes of the lentivirus subfamily of retroviruses, including the human, simian and feline immunodeficiency viruses (HIV1, HIV2, SIV, and FIV, respectively), equine infectious anemia virus (EIAV), and the ovine lentivirus, Visna. The occurrence of CpG dinucleotides is greater in the 2-3 (NCG) than in the 1-2 (CGN) codon-defined frame, as well as in the gag and env genes, compared to the more conserved pol gene. These differences suggest that CpG depletion in lentiviruses occurs as a result of selection against CpG rather than due to mutational bias, the latter is responsible for low CpG frequencies in vertebrate genomes. CpG levels in the onco-retrovirus subfamily are reduced to a lesser extent, principally due to mutational bias. The difference between the retrovirus subfamilies appears to reflect their evolutionary origin, that is, lentiviruses have no known endogenous counterparts whereas most oncoviruses have endogenous cellular counterparts with which they can undergo recombination. Furthermore, we suggest that the number of CpG dinucleotides in a lentiviral genome determines the maximum potential DNA methylation level of the provirus, which in turn affects viral transcription in host cells.

Insights into X chromosome inactivation from studies of species variation, DNA methylation and replication, and vice versa

I am indebted to Mary Lyon as her X-inactivation hypothesis stimulated my mentor, Barton Childs, and in turn, myself, to think about the consequences of X-inactivation in heterozygous females. I often reread her original papers setting forth the single active X hypothesis, and still marvel at the concise and compelling exposition of the hypothesis and the logical predictions which seemed prophetic at my first reading, and have survived the test of time. My contribution to this Festschrift reviews evidence derived from studies of DNA methylation, species variation and DNA replication that reveals an important role for methylated CpG islands and suggests a role for late DNA replication in propagating X inactivation from one cell to its progeny. These studies also show that X inactivation is a powerful research tool for identifying the factors which program and maintain developmental processes.

Overlapping palindromic sequences associated with somatic deletion and meiotic recombination of MHC class I genes

H-2L-null variants were immunoselected from a transfected murine fibroblast cell line carrying a single copy H-2L gene, and were characterized to determine the basis for the loss of this MHC class I cell surface product. Molecular analysis indicated that inactivation of H-2L expression in nearly every null clone resulted from an apparent deletion or rearrangement of 5'-flanking and 5'-coding H-2L sequences, with breakpoints consistently mapping to within a 550 bp GC-rich region between exon 1 and the middle of intron 2. Notably, this region of the H-2L gene contains a large number of overlapping, inverted repeat sequences as well as potential topoisomerase I cleavage sites. Examination of several in vivo mutant class I genes, believed to have been generated by recombination, has revealed that each of these genes bears similar palindromic structures overlapping or adjacent to the regions of sequence exchange. These findings suggest that inverted repeat sequences may play a role in recombination and deletion within the MHC class I multigene family.

High levels of de novo methylation and altered chromatin structure at CpG islands in cell lines

CpG islands are normally methylation free in cells of the animal, even when the associated gene is transcriptionally silent. In mouse NIH 3T3 and L cells, however, over half of the islands are heavily methylated. Near identity of the methylated subset in the two cell lines suggested that methylation is confined to genes that are nonessential in culture. In agreement with this, islands at several tissue-specific genes, but not at housekeeping genes, have become methylated in many human and mouse cell lines. At the chromatin level, methylated islands are Mspl resistant compared with their nonmethylated counterparts. We suggest that mutation-like gene inactivation due to CpG island methylation is widespread in many cell lines and could explain the loss of cell type-specific functions in culture.

De novo methylation of the MyoD1 CpG island during the establishment of immortal cell lines

CpG dinucleotides are unevenly distributed in the vertebrate genome. Bulk DNA is depleted of CpGs and most of the cytosines in the dinucleotide in this fraction are methylated. On the other hand, CpG islands, which are often associated with genes, are unmethylated at testable sites in all normal tissues with the exception of genes on the inactive X chromosome. We used Hpa II/Msp I analysis and ligation-mediated polymerase chain reaction to examine the methylation of the MyoD1 CpG island in adult mouse tissues, early cultures of mouse embryo cells, and immortal fibroblastic cell lines. The island was almost devoid of methylation at CCGG sites in adult mouse tissues and in low-passage mouse embryo fibroblasts. In marked contrast, the island was methylated in 10T 1/2 cells and in six other immortal cell lines showing that methylation of this CpG island had occurred during escape from senescence. The island became even more methylated in chemically transformed derivatives of 10T 1/2 cells. Thus, CpG islands not methylated in normal tissues may become modified to an abnormally high degree during immortalization and transformation.

Methylation of mouse adenine phosphoribosyltransferase gene is altered upon cellular differentiation and loss of phenotypic expression

Morphologically differentiated cell lines were previously isolated from a mouse teratocarcinoma stem cell line exhibiting an unstable heterozygous deficiency for adenine phosphoribosyltransferase (APRT) expression. In this study, the methylation sensitive and insensitive isoschizomer restriction endonucleases HpaII and MspI, respectively, were used to demonstrate that the aprt gene in the heterozygous deficient stem cell line was hypomethylated. Loss of APRT activity in this stem cell line was not associated with DNA methylation change. However, differentiation of this stem cell line was associated with hypermethylation of three consecutive HpaII/MspI sites that were located in the second intron and the third exon of the aprt gene. A total of 15 independent APRT homozygous deficient cell lines were isolated from three differentiated heterozygous deficient cell lines, and in all 15 cell lines this differentiation-related methylation pattern was altered. Two classes of alterations were noted: (1) hypomethylation of a site located in the second intron or (2) the apparent spreading of methylation to downstream methylation sites. The CpG-rich promoter region remained hypomethylated in the APRT homozygous deficient differentiated cell lines and a methylation change affecting a specific CpG site upstream of the promoter region was noted in only two of the 15 homozygous deficient cell lines. It is proposed that methylation of the mouse aprt gene may be involved in controlling phenotypic expression in the differentiated cell lines, but not in the stem cell line they were derived from.

Cloning and analysis of the 5' region of the rat bone/liver/kidney/placenta alkaline phosphatase gene. A dual-function promoter

The rat bone/liver/kidney/placenta (BLKP) alkaline phosphatase (ALP) gene is expressed at high level in these particular tissues and at low levels in many other tissues. To study the mechanisms underlying the complex regulation of the rat BLKP ALP expression, we isolated a genomic clone, containing a 10.5-kb insert, which includes the promoter of the BLKP ALP gene with 2 kb of 5' flanking region, its first exon (84bp), and over 7 kb of the first intron. The promoter of the rat BLKP ALP displays features of a "housekeeping" gene promoter: an atypical TATA-box (TTCATAA); 3 potential Spl binding sites; high GC content (82% in positions-134 to -14); and a high CpG to GpC ratio (60:89 in the 0.85 kb promoter region), indicating an abundance of potential methylation sites. Likewise, transient transfection of CAT fusion genes into ROS 17/2.8 osteoblast-like cells reveals weak expression from the promoter and proximal 5' flanking sequences, which can be elevated by an SV40E enhancer. The homologous human bone/liver/kidney (BLK) ALP promoter, which demonstrates a similar combination of tissue-specific and housekeeping characteristics, shares close similarity (184 bp of 79% similarity excluding gaps) with the rat BLKP ALP promoter. The human placental ALP is encoded by a separate gene and its promoter, on the other hand lacks significant similarity to the rat BLKP ALP promoter despite their common expression in the placenta. This lack of similarity appears to reflect the close evolutionary relationship of the human placental ALP gene to the intestinal ALP gene. Significant sequence similarity was found between the rat and human BLK/BLKP ALP promoters and the human and mouse adenine deaminase promoters, and together they may represent a class of dual-function promoters, allowing both constitutive low-level, and tissue-specific higher levels of expression. A pentanucleotide with the consensus sequence 5'-GGCTC-3' is present in these promoters and in the promoters for the human fibronectin and the human alpha 1(II) procollagen genes in the region of maximal similarity with the rat BLKP ALP promoter, and in the vicinity of the Sp1-binding sites.

The human Na, K-ATPase alpha 1 gene: characterization of the 5'-flanking region and identification of a restriction fragment length polymorphism

We have determined the sequence of the 5'-flanking region and first three exons of the human Na,K-ATPase alpha 1 gene, ATP1A1. Primer extension and S1 nuclease protection analyses of RNA from human kidney, brain, and skeletal muscle indicate that transcription initiates 273 nucleotides upstream of the translation start site. The promoter region contains a potential TATA box at position -27 relative to the transcription initiation site; however, no CCAAT sequence is observed. The 5'-untranslated and 5'-flanking regions are G + C rich. Five sequence elements exhibiting similarity to binding sites for the transcription factor Sp1 are located within the 5'-flanking region. This region also contains potential binding sites for the transcription factors AP-1, AP-2, AP-3, and NF-1, as well as a site which exhibits perfect identity to an 8-bp sequence element important for calcium induction. A comparison of the 5'-flanking region of the alpha 1 and alpha 2 genes reveals differences in potential transcription factor and hormone receptor binding sites which may be important in mediating the tissue- and developmental stage-specific expression of these genes. We have also identified an intragenic DNA probe which detects a restriction fragment length polymorphism at the alpha 1 locus. This marker should facilitate genetic linkage studies designed to evaluate the role of the sodium pump in human disease.

Pyrimidine 5-methyl groups influence the magnitude of DNA curvature

DNA containing short sequences of the form (dA)n.(dT)n can exhibit pronounced degrees of stable curvature of the helix axis, provided that these homooligomeric stretches are approximately in phase with the helix repeat. However, the precise origin of this effect is unknown. We have observed that pyrimidine 5-methyl groups can have a significant effect on the degree of curvature, depending on their locations within the homooligomeric sequences. Such effects are observed in both (dA)n.(dT/dU)n and (dI)n.(dC/d5meC)n sequence motifs, arguing for a general structural perturbation due to the methyl group. The current observations suggest that pyrimidine methyl groups could influence protein-DNA interactions not only through direct protein-methyl group contacts but also by methyl group induced alterations in local DNA structure.

A nuclear protein that binds preferentially to methylated DNA in vitro may play a role in the inaccessibility of methylated CpGs in mammalian nuclei

The effects of DNA methylation on gene expression and chromatin structure suggest the existence of a mechanism in the nucleus capable of distinguishing methylated and non-methylated sequences. We report the finding of a nuclear protein in several vertebrate tissues and cell lines that binds preferentially to methylated DNA in vitro. Its lack of sequence-specific requirements makes it potentially capable of binding to any methylated sequence in mammalian nuclei. An in vivo counterpart of these results is that methylated CpGs are inaccessible to nucleases within nuclei. In contrast, non-methylated CpG sites, located mainly at CpG islands, and restriction sites not containing this dinucleotide, are relatively accessible. The possibility that DNA methylation acts through binding to specific proteins that could alter chromatin structure is discussed.

Sequence signals in eukaryotic upstream regions

Two DNA sequence elements are known to recur frequently upstream of eukaryotic polymerase II-transcribed genes. The TATAAA, at position -40, specifies the transcription initiation site. The GGCCAATCT is less frequent around -80. Sequence analysis of upstream regions reveals that the underlined yeast UAS2 consensus sequence, TGATTGGT, is also very frequent at -80 in higher polymerase II-transcribed animal sequences. The underlined CCAAT box and yeast UAS sequences are complementary. Structural analysis suggests some symmetry in their DNA structures. Upstream of the TATAAT-rich region there is an abundance of GC sequences. Analysis of nucleotide tracts indicates that these are preferentially flanked by their complementary nucleotides with a pyrimidine-purine junction, i.e., TTAN, CCGn, CnGG, TnAA. Here, I discuss DNA structural consideration in upstream regions along with protein readout of the major and minor groove information content. These sequence-structure aspects are put in the general context of protein (factors)-DNA (elements) recognition and regulation.

In vitro methylation of CpG-rich islands

CpG islands are distinguishable from the bulk of vertebrate DNA for being unmethylated and CpG-rich. Since CpG doublets are the specific target of eukaryotic DNA methyltransferases, CpG-rich sequences might be expected to be good methyl-accepting substrates in vitro, despite their unmethylated in vivo condition. This was tested using a partially purified DNA-methyltransferase from human placenta and several cloned CpG-rich or CpG-depleted sequences. The efficiency of methylation was found to be proportional to the CpG content for CpG-depleted regions, which are representative of the bulk genome. However, methylation was much less efficient for CpG frequencies higher than 1 in 12 nucleotides, reaching only 60% of the expected level. That suggests that the close CpG spacing typical of CpG-islands somehow inhibits mammalian DNA methyltransferase. The implications of these findings on the in vivo pattern of DNA methylation are discussed.

Chromatin loop structure of the human X chromosome: relevance to X inactivation and CpG clusters

Part of the higher-order structure of chromatin is achieved by constraining DNA in loops ranging in size from 30 to 100 kilobase pairs; these loops have been implicated in defining functional domains and replicons and possibly in facilitating transcription. Because the human active and inactive X chromosomes differ in transcriptional activity and replication, we looked for differences in their chromatin loop structures. Since the islands of CpG-rich DNA at the 5' ends of X-linked housekeeping genes are the regions where functional differences in DNA methylation and nuclease sensitivity are found, we looked for scaffold association of these sequences after extraction of histones with lithium diiodosalicylate. Specifically, we examined the 5' CpG islands within the hypoxanthine phosphoribosyltransferase, glucose 6-phosphate dehydrogenase, P3, GdX, phosphoglycerate kinase type 1, and alpha-galactosidase loci in human lymphoblasts obtained from individuals with 1 to 4 X chromosomes. Although we detected no scaffold-associated regions near these genes, we found several such regions at the ornithine transcarbamylase and blood clotting factor IX loci. Our results suggest that the CpG islands are excluded from the nuclear scaffold and that even though transcriptionally active, housekeeping genes are less likely than X-linked tissue-specific genes to be scaffold associated. In all cases, the pattern of scaffold association was the same for loci on active and inactive X chromosomes.

Specific protection of methylated CpGs in mammalian nuclei

We have compared nuclear accessibility of methylated and nonmethylated sequences using restriction enzymes. MspI, which cuts CpG sites in naked DNA regardless of methylation, cut DNA in intact mouse liver or brain nuclei almost exclusively at CpG islands. Bulk chromatin was not significantly cleaved by MspI but was cleaved extensively by enzymes that do not recognize CpG. Quantitative analysis of limit digests showed that MspI and another methyl-CpG insensitive enzyme, Tth, have a strong bias against cutting methylated sites in these nuclei. Southern analysis confirmed this at three genomic loci. Our results suggest that resistance to nucleases is mediated by factors that are bound specifically to methylated CpGs. MeCP, a protein that binds to methylated DNA in vitro, may be one such factor, since nuclease resistance was significantly reduced in an MeCP-deficient cell line.

Structure of the alpha 1 subunit of horse Na,K-ATPase gene

Genomic DNA for Na,K-ATPase alpha 1 subunit was obtained from libraries of horse kidney genomic DNA in Charon 4A and in EMBL3 bacteriophages by screening with the full sized cDNA probe of the alpha 1 subunit of rat Na,K-ATPase as probe. The gene spans 30 kb and consists of 23 exons and 22 intervening sequences. Intron-exon boundaries were analyzed. The protein-coding nucleotide sequence encodes 1016 amino acids with an Mr of 112,264. The putative amino acid sequence of horse alpha 1 is 96-97% homologous to those of other mammalian species.

Chromatin loop structure of the human X chromosome: relevance to X inactivation and CpG clusters

Part of the higher-order structure of chromatin is achieved by constraining DNA in loops ranging in size from 30 to 100 kilobase pairs; these loops have been implicated in defining functional domains and replicons and possibly in facilitating transcription. Because the human active and inactive X chromosomes differ in transcriptional activity and replication, we looked for differences in their chromatin loop structures. Since the islands of CpG-rich DNA at the 5' ends of X-linked housekeeping genes are the regions where functional differences in DNA methylation and nuclease sensitivity are found, we looked for scaffold association of these sequences after extraction of histones with lithium diiodosalicylate. Specifically, we examined the 5' CpG islands within the hypoxanthine phosphoribosyltransferase, glucose 6-phosphate dehydrogenase, P3, GdX, phosphoglycerate kinase type 1, and alpha-galactosidase loci in human lymphoblasts obtained from individuals with 1 to 4 X chromosomes. Although we detected no scaffold-associated regions near these genes, we found several such regions at the ornithine transcarbamylase and blood clotting factor IX loci. Our results suggest that the CpG islands are excluded from the nuclear scaffold and that even though transcriptionally active, housekeeping genes are less likely than X-linked tissue-specific genes to be scaffold associated. In all cases, the pattern of scaffold association was the same for loci on active and inactive X chromosomes.

Demethylation of specific sites in the 5' region of the inactive X-linked human phosphoglycerate kinase gene correlates with the appearance of nuclease sensitivity and gene expression

X8/6T2, a hamster-human hybrid cell line which contains an inactive human X chromosome, was treated with 5-azacytidine and selected for derepression of hypoxanthine-guanine phosphoribosyltransferase. Clones were examined for coreactivation of the phosphoglycerate kinase gene (Pgk). Of 68 of these hybrids, approximately 20% expressed measurable human phosphoglycerate kinase (PGK) activity. A 600-base-pair region of the Pgk 5' CpG cluster was examined for the methylation status of eight CCGG sites (site 1 being 5'-most) in a number of PGK-negative and PGK-positive cell lines. The inactive X chromosome is normally methylated at all eight sites, and this was also true for the majority of X8/6T2 cells. However, several PGK-negative hybrids were demethylated in the site 3 to site 6 region. PGK activity correlated with demethylation at both sites 6 and 7. The data for PGK-positive and -negative hybrids indicate that demethylation at or near site 7 was necessary for reactivation of Pgk. Chromatin sensitivity to MspI digestion in the nuclei of male lymphoblastoid cells and several PGK-positive and PGK-negative hybrids was examined. PGK-positive cell lines were hypersensitive to digestion, while PGK-negative hybrids were resistant. Cleavage at sites 6 and 7 was observed in all PGK-positive cell lines at each MspI concentration examined. Sites 7 and 8 were less accessible to digestion than site 6. Cleavage in the site 2 to site 5 region was observable at the lowest MspI concentration. In most PGK-positive hybrids, a nonspecific endogenous nuclease detected the presence of a hypersensitive region spanning at least 450 base pairs, bounded at the 3' end near HpaII site 6. Nuclease hypersensitivity appears to be related to promoter activity, because sites 7 and 8 are in transcribed regions of the gene. These data indicate that specific sites within the CpG cluster have a dominant controlling influence over the Pgk promoter conformation and the transcriptional activation of Pgk.

Demethylation of specific sites in the 5' region of the inactive X-linked human phosphoglycerate kinase gene correlates with the appearance of nuclease sensitivity and gene expression

X8/6T2, a hamster-human hybrid cell line which contains an inactive human X chromosome, was treated with 5-azacytidine and selected for derepression of hypoxanthine-guanine phosphoribosyltransferase. Clones were examined for coreactivation of the phosphoglycerate kinase gene (Pgk). Of 68 of these hybrids, approximately 20% expressed measurable human phosphoglycerate kinase (PGK) activity. A 600-base-pair region of the Pgk 5' CpG cluster was examined for the methylation status of eight CCGG sites (site 1 being 5'-most) in a number of PGK-negative and PGK-positive cell lines. The inactive X chromosome is normally methylated at all eight sites, and this was also true for the majority of X8/6T2 cells. However, several PGK-negative hybrids were demethylated in the site 3 to site 6 region. PGK activity correlated with demethylation at both sites 6 and 7. The data for PGK-positive and -negative hybrids indicate that demethylation at or near site 7 was necessary for reactivation of Pgk. Chromatin sensitivity to MspI digestion in the nuclei of male lymphoblastoid cells and several PGK-positive and PGK-negative hybrids was examined. PGK-positive cell lines were hypersensitive to digestion, while PGK-negative hybrids were resistant. Cleavage at sites 6 and 7 was observed in all PGK-positive cell lines at each MspI concentration examined. Sites 7 and 8 were less accessible to digestion than site 6. Cleavage in the site 2 to site 5 region was observable at the lowest MspI concentration. In most PGK-positive hybrids, a nonspecific endogenous nuclease detected the presence of a hypersensitive region spanning at least 450 base pairs, bounded at the 3' end near HpaII site 6. Nuclease hypersensitivity appears to be related to promoter activity, because sites 7 and 8 are in transcribed regions of the gene. These data indicate that specific sites within the CpG cluster have a dominant controlling influence over the Pgk promoter conformation and the transcriptional activation of Pgk.

CpG islands of the X chromosome are gene associated

Unmethylated CpG rich islands are a feature of vertebrate DNA: they are associated with housekeeping and many tissue specific genes. CpG islands on the active X chromosome of mammals are also unmethylated. However, islands on the inactive X chromosome are heavily methylated. We have identified a CpG island in the 5' region of the G6PD gene, and two islands forty Kb 3' from the G6PD gene, on the human X chromosome. Expression of the G6PD gene is associated with concordant demethylation of all three CpG islands. We have shown that one of the two islands is in the promoter region of a housekeeping gene, GdX. In this paper we show that the second CpG island is also associated with a gene, P3. The P3 gene has no homology to previously described genes. It is a single copy, 4 kb gene, conserved in evolution, and it has the features of a housekeeping two genes is within the CpG island and that sequences in the islands have promoter function.

Sequence and structure of the nucleolin promoter in rodents: characterization of a strikingly conserved CpG island

We report the isolation of the complete genes encoding nucleolin from rat and hamster. The DNA clones were obtained from partial genomic libraries by probing with a genomic DNA fragment containing the leader and promoter regions of the mouse nucleolin gene. We have determined the complete nucleotide sequence of the 5'-terminal region for the three rodent species. The sequenced regions extend over 1 kb downstream and upstream from the cap sites and include a conserved CpG island 1500 nucleotides (nt) long. The 5' end of the CpG island in each species has maintained a long alternating purine-pyrimidine sequence which could adopt a Z-DNA conformation. By sequence comparison, 42 blocks of homology are defined in the 5'-terminal region, of which 36 appear in the CpG island and contain numerous conserved CpG dinucleotides. Two blocks, 110 and 49 nt long, encompassing the cap sites and the region immediately upstream, respectively, present features characteristic of regulated genes: a possible TATA box (ATTA), two pyrimidine-rich nucleotide stretches and two inverted juxtaposed CCAAT-like boxes (GGTTGG). Furthermore, the adjacent upstream conserved region presents features characteristic of housekeeping genes: four G/C boxes, embedded in a high G + C-content sequence, among them one presenting a perfect consensus Sp 1-binding site (GCCCCGCCCC). Among unusual features, we report numerous large G + C-rich conserved sequences located in the first intron. One of these sequences contains two G/C boxes which border a sequence presenting a dyad symmetry (GCGCACGTGCTC). Our findings shed some light on the putative role of the CpG island. We show that CpG-rich sequence motifs are under strong selective pressure over the whole 5'-terminal region and are presumably involved in regulatory mechanisms.

Unit-length line-1 transcripts in human teratocarcinoma cells

We have characterized the approximately 6.5-kilobase cytoplasmic poly(A)+ Line-1 (L1) RNA present in a human teratocarcinoma cell line, NTera2D1, by primer extension and by analysis of cloned cDNAs. The bulk of the RNA begins (5' end) at the residue previously identified as the 5' terminus of the longest known primate genomic L1 elements, presumed to represent "unit" length. Several of the cDNA clones are close to 6 kilobase pairs, that is, close to full length. The partial sequences of 18 cDNA clones and full sequence of one (5,975 base pairs) indicate that many different genomic L1 elements contribute transcripts to the 6.5-kilobase cytoplasmic poly(A)+ RNA in NTera2D1 cells because no 2 of the 19 cDNAs analyzed had identical sequences. The transcribed elements appear to represent a subset of the total genomic L1s, a subset that has a characteristic consensus sequence in the 3' noncoding region and a high degree of sequence conservation throughout. Two open reading frames (ORFs) of 1,122 (ORF1) and 3,852 (ORF2) bases, flanked by about 800 and 200 bases of sequence at the 5' and 3' ends, respectively, can be identified in the cDNAs. Both ORFs are in the same frame, and they are separated by 33 bases bracketed by two conserved in-frame stop codons. ORF 2 is interrupted by at least one randomly positioned stop codon in the majority of the cDNAs. The data support proposals suggesting that the human L1 family includes one or more functional genes as well as an extraordinarily large number of pseudogenes whose ORFs are broken by stop codons. The cDNA structures suggest that both genes and pseudogenes are transcribed. At least one of the cDNAs (cD11), which was sequenced in its entirety, could, in principle, represent an mRNA for production of the ORF1 polypeptide. The similarity of mammalian L1s to several recently described invertebrate movable elements defines a new widely distributed class of elements which we term class II retrotransposons.

Comparative study of DNase I sensitivity at the X-linked human HPRT locus

To examine the association between chromatin structure and gene expression at the human hypoxanthine phosphoribosyltransferase (HPRT) locus, DNase I sensitivity of active and inactive genes was analyzed. In a set of human-hamster hybrid lines containing either an active or an inactive human X chromosome, or a derivative of the latter in which the HPRT gene was reactivated by 5-azacytidine treatment, only the promoter region of the gene was found to contain a hypersensitive domain, and its presence was strictly correlated with gene activity. An S1 nuclease-sensitive site was mapped upstream from the DNase I hypersensitive domain using supercoiled plasmids. The overall level of DNase I sensitivity in the interior of the HPRT gene was also assessed by comparing the degradation of polymorphic restriction fragments on active and inactive alleles in both polyclonal and monoclonal lines of female human cells. In these internally controlled experiments, the active X chromosome was found to be approximately twofold more susceptible to DNase I digestion than the inactive X chromosome.

Unit-length line-1 transcripts in human teratocarcinoma cells

We have characterized the approximately 6.5-kilobase cytoplasmic poly(A)+ Line-1 (L1) RNA present in a human teratocarcinoma cell line, NTera2D1, by primer extension and by analysis of cloned cDNAs. The bulk of the RNA begins (5' end) at the residue previously identified as the 5' terminus of the longest known primate genomic L1 elements, presumed to represent "unit" length. Several of the cDNA clones are close to 6 kilobase pairs, that is, close to full length. The partial sequences of 18 cDNA clones and full sequence of one (5,975 base pairs) indicate that many different genomic L1 elements contribute transcripts to the 6.5-kilobase cytoplasmic poly(A)+ RNA in NTera2D1 cells because no 2 of the 19 cDNAs analyzed had identical sequences. The transcribed elements appear to represent a subset of the total genomic L1s, a subset that has a characteristic consensus sequence in the 3' noncoding region and a high degree of sequence conservation throughout. Two open reading frames (ORFs) of 1,122 (ORF1) and 3,852 (ORF2) bases, flanked by about 800 and 200 bases of sequence at the 5' and 3' ends, respectively, can be identified in the cDNAs. Both ORFs are in the same frame, and they are separated by 33 bases bracketed by two conserved in-frame stop codons. ORF 2 is interrupted by at least one randomly positioned stop codon in the majority of the cDNAs. The data support proposals suggesting that the human L1 family includes one or more functional genes as well as an extraordinarily large number of pseudogenes whose ORFs are broken by stop codons. The cDNA structures suggest that both genes and pseudogenes are transcribed. At least one of the cDNAs (cD11), which was sequenced in its entirety, could, in principle, represent an mRNA for production of the ORF1 polypeptide. The similarity of mammalian L1s to several recently described invertebrate movable elements defines a new widely distributed class of elements which we term class II retrotransposons.

Differential methylation of the ornithine carbamoyl transferase gene on active and inactive mouse X chromosomes

Ornithine carbamoyl transferase (Oct) is an X-linked gene which exhibits tissue-specific expression. To determine whether methylation of specific CpG sequences plays a role in dosage compensation or tissue-specific expression of the gene, 13 potentially methylatable sites were identified over a 30-kilobase (kb) region spanning from approximately 15 kb upstream to beyond exon II. Fragments of the Mus hortulanus Oct gene were used as probes to establish the degree of methylation at each site. By considering the methylation status in liver (expressing tissue) versus kidney (nonexpressing tissue) from male and female mice, the active and inactive genes could be investigated on active and inactive X-chromosome backgrounds. One MspI site, 12 kb 5' of the Oct-coding region, was cleaved by HpaII in liver DNA from males but not in kidney DNA from males and thus exhibited complete correlation with tissue-specific expression of the gene. Six other sites showed partial methylation, reflecting incomplete correlation with tissue-specific expression.

Molecular analysis of mutations induced by N-ethyl-N-nitrosourea at the HPRT locus in mouse lymphoma cells

The molecular basis of 29 N-ethyl-N-nitrosourea (ENU)-induced HPRT-deficient mutants of mouse lymphoma cells (GRSL 13-2) was investigated using nucleic acid blot hybridization techniques. DNA from all 29 mutants showed normal restriction patterns on Southern blots when probed with HPRT cDNA, but 10 mutants differed from wild-type cells in their cytoplasmic HPRT mRNA level. In 5 mutants we found 10-25% of the normal amount of HPRT mRNA, whereas in another 5 mutants no HPRT mRNA could be detected at all. These mutants do not seem to be induced by hypermethylation of regulatory sequences of the HPRT gene, since they could not be reverted to an HPRT-proficient phenotype by treatment of the cells with 5-azacytidine.

A promoter of the rat insulin-like growth factor II gene consists of minimal control elements

We have characterized the cis-control signals in one of the two promoters of the developmentally regulated rat insulin-like growth factor II gene (rIGF-II) by a combination of in-vivo transient expression, in-vitro transcription, footprinting, gel band-shifting and methylation-interference experiments, using a series of deletion mutant templates. Our results indicate that this simple (minimal) promoter (P2) consists of no more than 128 base-pairs, which include an ATA box and four proximal upstream GC boxes binding the general transcription factor Sp1. Three of the latter sites deviate from the known Sp1 consensus recognition sequence. The two types of cis-acting regulatory signals (GC/ATA motif) of the P2 promoter are inter-dependent and sufficient for transcription. A model for the operation of this type of minimal promoter is discussed. S1 nuclease-hypersensitive sites, localized by in-vitro mapping to the region of the P2 Sp1-binding sites, are also present in vivo and correlate with the transcriptional state of chromatin in the rIGF-II locus. We show that recognition sites for Sp1 binding are a subset of sequences that exhibit hypersensitivity to S1.

Nuclease sensitivity of the mouse HPRT gene promoter region: differential sensitivity on the active and inactive X chromosomes

We investigated the conformation of the X-linked mouse hypoxanthine-guanine phosphoribosyltransferase gene (HPRT) promoter region both in chromatin from the active and inactive X chromosomes with DNase I and in naked supercoiled DNA with S1 nuclease. A direct comparison of the chromatin structures of the active and inactive mouse HPRT promoter regions was performed by simultaneous DNase I treatment of the active and inactive X chromosomes in the nucleus of interspecies hybrid cells from Mus musculus and Mus caroli. Using a restriction fragment length polymorphism to distinguish between the active and inactive HPRT promoters, we found a small but very distinct difference in the DNase I sensitivity of active versus inactive chromatin. We also observed a single DNase I-hypersensitive site in the immediate area of the promoter which was present only on the active X chromosome. Analysis of the promoter region by S1 nuclease digestion of supercoiled plasmid DNA showed an S1-sensitive site which maps adjacent to or within the DNase I-hypersensitive site found in chromatin but upstream of the region minimally required for normal HPRT gene expression.

Directional fixation of mutations in vertebrate evolution

We have made pairwise comparisons between the coding sequences of 21 genes from cold-blooded vertebrates and 41 homologous sequences from warm-blooded vertebrates. In the case of 12 genes, GC levels were higher, especially in third codon positions, in warm-blooded vertebrates compared to cold-blooded vertebrates. Six genes showed no remarkable difference in GC level and three showed a lower level. In the first case, higher GC levels appear to be due to a directional fixation of mutations, presumably under the influence of body temperature (see Bernardi and Bernardi 1986b). These GC-richer genes of warm-blooded vertebrates were located, in all cases studied, in isochores higher in GC than those comprising the homologous genes of cold-blooded vertebrates. In the third case, increases appear to be due to a limited formation of GC-rich isochores which took place in some cold-blooded vertebrates after the divergence of warm-blooded vertebrates. The directional changes in the GC content of coding sequences and the evolutionary conservation of both increased and unchanged GC levels are in keeping with the existence of compositional constraints on the genome.

Differential methylation of the ornithine carbamoyl transferase gene on active and inactive mouse X chromosomes

Ornithine carbamoyl transferase (Oct) is an X-linked gene which exhibits tissue-specific expression. To determine whether methylation of specific CpG sequences plays a role in dosage compensation or tissue-specific expression of the gene, 13 potentially methylatable sites were identified over a 30-kilobase (kb) region spanning from approximately 15 kb upstream to beyond exon II. Fragments of the Mus hortulanus Oct gene were used as probes to establish the degree of methylation at each site. By considering the methylation status in liver (expressing tissue) versus kidney (nonexpressing tissue) from male and female mice, the active and inactive genes could be investigated on active and inactive X-chromosome backgrounds. One MspI site, 12 kb 5' of the Oct-coding region, was cleaved by HpaII in liver DNA from males but not in kidney DNA from males and thus exhibited complete correlation with tissue-specific expression of the gene. Six other sites showed partial methylation, reflecting incomplete correlation with tissue-specific expression.

Origin and evolution of a major feline satellite DNA

A major satellite DNA has been cloned from the domestic cat (Felis catus) and characterized. The satellite monomer, termed FA-SAT, is 483 base-pairs in size, 64% G + C, and represents about 1 to 2% of the cat genome. A consensus sequence based upon partial sequence data from 21 independently isolated clones demonstrates: (1) FA-SAT is not composed of a series of shorter repeats, although about 25 copies, primarily imperfect, of the hexanucleotide TAACCC appear in the sequence; (2) there are many more CpG dinucleotides present in FA-SAT than expected for a random sequence of its size; and (3) 61% of all base substitutions in FA-SAT involve the replacement of G and C residues by A and T residues, indicating that FA-SAT is rapidly becoming A + T-rich. FA-SAT-related sequences are found in many mammals, where they appear to be scattered throughout the genome and not tandemly arranged as in the cat. An FA-SAT-related sequence was cloned from the domestic dog genome and sequenced, and shown to contain multiple copies of the same TAACCC hexanucleotide found in the cat satellite.

A long polypyrimidine/polypurine tract induces an altered DNA conformation on the 3' coding region of the adjacent myosin heavy chain gene

A long (147 base pairs), natural A.T rich polypyrimidine/polypurine tract has been found 55 base pairs downstream of a chicken embryonic myosin heavy chain (MHC) gene. Analysis at the nucleotide level of nicks induced by S1 and Neurospora crassa nucleases indicate that this long interrupted polypyrimidine/polypurine tract exists in an alternate DNA structure in vitro at pH 4.5 and pH 7.5 in both supercoiled and linear plasmid DNA. The polypyrimidine/polypurine tract induces this alternate structure upon at least 200 base pairs of its 5' flanking DNA, and thus extends into the 3' coding and non-coding regions of the neighboring MHC gene. The different nicking patterns induced by the nucleases S1 and N. crassa on each strand of this alternate structure suggests that the polypyrimidine/polypurine tract may form heteronomous DNA. When this long polypyrimidine/polypurine tract is present in a supercoiled plasmid at low pH, a new and as yet undefined S1 hypersensitive DNA alteration was detected near the center of this tract.

DNA methylation stabilizes X chromosome inactivation in eutherians but not in marsupials: evidence for multistep maintenance of mammalian X dosage compensation

In marsupials and eutherian mammals, X chromosome dosage compensation is achieved by inactivating one X chromosome in female cells; however, in marsupials, the inactive X chromosomes is always paternal, and some genes on the chromosome are partially expressed. To define the role of DNA methylation in maintenance of X chromosome inactivity, we examined loci for glucose-6-phosphate dehydrogenase and hypoxanthine phosphoribosyltransferase in a North American marsupial, the opossum Didelphis virginiana, by using genomic hybridization probes cloned from this species. We find that these marsupial genes are like their eutherian counterparts, with respect to sex differences in methylation of nuclease-insensitive (nonregulatory) chromatin. However, with respect to methylation of the nuclease-hypersensitive (regulatory) chromatin of the glucose-6-phosphate dehydrogenase locus, the opossum gene differs from those of eutherians, as the 5' cluster of CpG dinucleotides is hypomethylated in the paternal as well as the maternal gene. Despite hypomethylation of the 5' CpG cluster, the paternal allele, identified by an enzyme variant, is at best partially expressed; therefore, factors other than methylation are responsible for repression. In light of these results, it seems that the role of DNA methylation in eutherian X dosage compensation is to "lock in" the process initiated by such factors. Because of similarities between dosage compensation in marsupials and trophectoderm derivatives of eutherians, we propose that differences in timing of developmental events--rather than differences in the basic mechanisms of X inactivation--account for features of dosage compensation that differ among mammals.

Nuclease sensitivity of the mouse HPRT gene promoter region: differential sensitivity on the active and inactive X chromosomes

We investigated the conformation of the X-linked mouse hypoxanthine-guanine phosphoribosyltransferase gene (HPRT) promoter region both in chromatin from the active and inactive X chromosomes with DNase I and in naked supercoiled DNA with S1 nuclease. A direct comparison of the chromatin structures of the active and inactive mouse HPRT promoter regions was performed by simultaneous DNase I treatment of the active and inactive X chromosomes in the nucleus of interspecies hybrid cells from Mus musculus and Mus caroli. Using a restriction fragment length polymorphism to distinguish between the active and inactive HPRT promoters, we found a small but very distinct difference in the DNase I sensitivity of active versus inactive chromatin. We also observed a single DNase I-hypersensitive site in the immediate area of the promoter which was present only on the active X chromosome. Analysis of the promoter region by S1 nuclease digestion of supercoiled plasmid DNA showed an S1-sensitive site which maps adjacent to or within the DNase I-hypersensitive site found in chromatin but upstream of the region minimally required for normal HPRT gene expression.

The distribution of the dinucleotide CpG and cytosine methylation in the vitellogenin gene family

Sequence data from regions of five vertebrate vitellogenin genes were used to examine the frequency, distribution, and mutability of the dinucleotide CpG, the preferred modification site for eukaryotic DNA methyltransferases. The observed level of the CpG dinucleotide in all five genes was markedly lower than that expected from the known mononucleotide frequencies. CpG suppression was greater in introns than in exons. CpG-containing codons were found to be avoided in the vitellogenin genes, but not completely despite the redundancy of the genetic code. Frequency and distribution patterns of this dinucleotide varied dramatically among these otherwise closely related genes. Dense clusters of CpG dinucleotides tended to appear in regions of either functional or structural interest (e.g., in the transposon-like Vi-element of Xenopus) and these clusters contained 5-methylcytosine (5 mC). 5 mC is known to undergo deamination to form thymidine, but the extent to which this transition occurs in the heavily methylated genomes of vertebrates and its contribution to CpG suppression are still unclear. Sequence comparison of the methylated vitellogenin gene regions identified C----T and G----A substitutions that were found to occur at relatively high frequencies. The predicted products of CpG deamination, TpG and CpA, were elevated. These findings are consistent with the view that CpG distribution and methylation are interdependent and that deamination of 5 mC plays an important role in promoting evolutionary change at the nucleotide sequence level.

CpG deficiency, dinucleotide distributions and nucleosome positioning

The dinucleotide CpG is deficient in (A + T)-rich regions of vertebrate DNA in both coding and non-coding sequences and there is a corresponding increase above expectation in the occurrence of TpG and CpA. By contrast in (G + C)-rich regions no deficiency of CpG is found. Such (G + C)-rich sequences, containing the expected number of CpG dinucleotides, alternate along the genome with (A + T)-rich sequences which have a lower than expected CpG content. The G + C content of vertebrate DNA can oscillate with a period of 150-200 bp and this may be a factor in positioning nucleosomes. The role of mutagenesis in loss of CpG and increase of A + T, particularly in non-coding regions, is discussed.

Two distinct compositional classes of vertebrate gene-bearing DNA stretches, their structures and possible evolutionary origin

Genomes of vertebrates are built of long, compositionally uniform DNA regions differing in guanine and cytidine (G + C) content. Examination of G + C distribution and CpG dinucleotide frequency in the longest stretches of vertebrate DNA base sequences available show that the long-range structural features are correlated with the structure of genes. Two classes of DNA stretches are conspicuous: (i) the stretches having low G + C content and low CpG doublet frequency and (ii) stretches rich in G + C containing CpG-rich islands. Both classes show other compositional islands containing exons. These structural features result from evolutionary pressures acting on the DNA or RNA level, as well as from mutations and repair differently biased in different genomic compartments. The analysis presented provides a rationale for a discussion of evolution of the long-range structural characteristics of DNA.

Possible origin of a calmodulin gene that lacks intervening sequences

The divergent, muscle-specific allele of the chicken calmodulin gene contains no intervening sequences and apparently was produced by a reverse transcriptase-mediated event. The nucleotide and deduced amino acid sequences of this gene were compared with nucleotide and amino acid sequence data of other known calmodulin genes in order to investigate its evolutionary history. These comparisons, as well as the CpG dinucleotide content, support the conclusion that this highly divergent chicken calmodulin gene did not exist for any significant period of times as a pseudogene and suggest plausible alternative genetic histories. The most parsimonious history involves the viral import of a very old foreign gene of high CpG content.

Nucleotide quartets in the vicinity of eukaryotic transcriptional initiation sites: some DNA and chromatin structural implications

Examination of all mammalian, nonmammalian vertebrate, and invertebrate genomic sequences present in the GenBank data shows a striking distribution of the G + C (A + T) content. It has been known for a few decades that the G + C content in higher organisms averages 42%. The average of the sequenced genomic regions is significantly higher, with a distinct peak near the transcription initiation site. This distribution, as well as the overall A + T genomic incidence may be directly related to DNA geometry. Specific A-T sequences may facilitate DNA curving, allowing more tight packaging of chromatin in eukaryotic genomes. To examine further which particular oligomers may potentially contribute to recognition of transcription initiation sites, an extensive search was instituted for detecting recurrences of particular oligonucleotides in the regions surrounding these sites. Accordingly, the analyzed mammalian genomic sequences have been scanned for the occurrences of each of the 256 quartets. The CAAT box components yield moderate peaks. As expected, the TATA box is much more pronounced, though the ATAA sequence yields a more striking peak than the TATA. Tetranucleotides containing solely A + T are relatively rare downstream from coding regions. Upstream, their level is low except for the distinct peaks; the level of A4 is very high upstream. All significant signals are enumerated. Of the G + C-rich quartets, CCCC and AGGG exhibit large peaks near transcription initiation sites. These may produce straight DNA segments.

Multiple transcription start sites, DNase I-hypersensitive sites, and an opposite-strand exon in the 5' region of the CHO dhfr gene

Transcription of the 26-kilobase (kb) dihydrofolate reductase (dhfr) gene in CHO cells is initiated at two sites: a major site (approximately 85% of the dhfr mRNA) at -63 relative to the translation start and a minor site (approximately 15%) at -107. Transcription also occurs from the opposite DNA strand in the dhfr 5' region, with a probable initiation site at approximately -195 relative to the dhfr translation start. A 4-kb polyadenylated RNA that is derived from the opposite-strand transcription increases threefold in abundance after serum starvation of CHO cells for 24 h. dhfr mRNA levels do not change during this time. The first dhfr exon lies within a 1-kb genomic region marked by exceptionally high G + C content and lack of DNA methylation. This region also includes a 214-base-pair (bp) exon for the opposite-strand transcript and five of the six DNase I-hypersensitive sites identified at the dhfr locus. Analysis of the DNA sequences of hamster, human (M. Chen, T. Shimada, A. D. Moulton, A. Cline, R. K. Humphries, J. Maizel, and A. W. Nienhuis, J. Biol. Chem. 259:3933-3943, 1984), and mouse (M. McGrogan, C. C. Simonsen, D. T. Smouse, P. J. Farnham, and R. T. Schimke, J. Biol. Chem. 260:2307-2314, 1985) dhfr genes reveals the presence of a 29-bp unit that is conserved 45 to 49 bp upstream of major and minor dhfr transcription start sites. This unit follows the consensus: GRGGCGGTGGCCTNNNNTGTCRCAARTRGGTR. The 5' part of the 29-bp unit contains a GC box that agrees with the GGGCGG consensus-binding site for the RNA polymerase II transcription factor Sp1 (D. Gidoni, W. A. Dynan, and R. Tjian, Nature (London) 312:409-413, 1984). Each of the three mammalian dhfr genes has several G-rich GC boxes proximal to the major dhfr transcription start site and several GC boxes of the opposite orientation (C rich) in a distal region about 500 bp upstream.

A calcium ionophore-inducible cellular promoter is highly active and has enhancerlike properties

We examined the regulatory/promoter sequence of a calcium ionophore-inducible gene isolated from the rat genome. Whereas the promoter of this ubiquitously expressed gene is active under noninduced conditions, after induction by calcium ionophore A23187 this promoter is 10- to 25-fold more active than the simian virus 40 early promoter, as measured by chloramphenicol acetyltransferase activities. Within this regulatory/promoter region, we have identified a DNA fragment with enhancer-like properties immediately 5' to the TATA sequence. This 291-nucleotide fragment acts in cis to enhance expression of the neomycin phosphotransferase (neo) gene driven by the herpes simplex virus thymidine kinase promoter in an orientation-independent manner. In addition, this fragment can confer A23187 inducibility to the neo gene and effectively compete for positive regulatory factors involved in A23187 induction. Sequence analysis of this promoter reveals homology with viral core enhancer sequences, and the apparent organization of direct repeat domains is similar to those observed in viral enhancers.

Structural analysis of the 5' region of the chromosomal gene for hamster histidyl-tRNA synthetase

The chromosomal gene (HRS) coding for hamster histidyl-tRNA synthetase, like many other housekeeping genes, lacks many of the features associated with promoters of RNA-polymerase-II-transcribed genes. HRS transcripts have multiple start points. Using RNase protection analysis, we also identified a 300-bp exon located only 36 bp away from the 5'-most start point of the HRS transcript. This exon hybridizes to a 3.5-kb transcript which transcribes from a different strand of DNA in the 5' region of the HRS gene. This divergent 3.5-kb transcript also has multiple transcription start points. The identity and function of this 3.5-kb transcript is not known.

Upstream regions of the human cardiac actin gene that modulate its transcription in muscle cells: presence of an evolutionarily conserved repeated motif

Transfection into cultured cell lines was used to investigate the transcriptional regulation of the human cardiac actin gene. We first demonstrated that in both human heart and human skeletal muscle, cardiac actin mRNAs initiate at the identical site and contain the same first exon, which is separated from the first coding exon by an intron of 700 base pairs. A region of 485 base pairs upstream from the transcription initiation site of the human cardiac actin gene directs high-level transient expression of the bacterial chloramphenicol acetyltransferase gene in differentiated myotubes of the mouse C2C12 muscle cell line, but not in mouse L fibroblast or rat PC-G2 pheochromocytoma cells. Deletion analysis of this region showed that at least two physically separated sequence elements are involved, a distal one starting between -443 and -395 and a proximal one starting between -177 and -118, and suggested that these sequences interact with positively acting transcriptional factors in muscle cells. When these two sequence elements are inserted separately upstream of a heterologous (simian virus 40) promoter, they do not affect transcription but do give a small (four- to fivefold) stimulation when tested together. Overall, these regulatory regions upstream of the cap site of the human cardiac actin gene show remarkably high sequence conservation with the equivalent regions of the mouse and chick genes. Furthermore, there is an evolutionarily conserved repeated motif that may be important in the transcriptional regulation of actin and other contractile protein genes.

Methylated DNA-binding protein from human placenta recognizes specific methylated sites on several prokaryotic DNAs

Methylated DNA-binding protein (MDBP) from human placenta recognizes specific DNA sequences containing 5-methylcytosine (m5C) residues. Comparisons of binding of various prokaryotic DNAs to MDBP indicate that m5CpG is present in the recognition sites for this protein but is only part of the recognition sequence. Specific binding to MDBP was observed for bacteriophage XP12 DNA, which naturally contains approximately 1/3 of its residues as m5C, and for Micrococcus luteus DNA, M13mp8 replicative form (RF) DNA, and pBR322 when these three DNAs were methylated at CpG sites by human DNA methyltransferase. Five DNA regions binding to MDBP have been localized by DNase I footprinting or restriction mapping in methylated pBR322 and M13mp8 RF DNAs. A comparison of their sequences reveals a common 5'-m5CGRm5CG-3' element or closely related sequence in which one of the m5C residues may be replaced by a T. In addition to this motif, one upstream and one downstream m5CpG as well as other common residues over an approximately 20-bp long region may be recognized by MDBP.