|Home | About | Journals | Submit | Contact Us | Français|
The VNTR at c-Ha-ras resides in a hotspot for DNA methylation on chromosome 11 in human tumors, where it is flanked by two MspI restriction sites. We have investigated the nature of the MspI site polymorphism at the c-Ha-ras VNTR observed in variety of tumors including breast cancer. We find that the MspI site 5′ to the VNTR is present in a Non-B DNA structure with single-strand character that renders it accessible to bisulfite modification under native conditions, while the MspI site 3′ to the VNTR appears to reside in a normal B-form structure that is inaccessible to bisulfite. The non-B DNA structure accounts for the observed polymorphism since MspI cannot cleave single-stranded DNA and control experiments show that the MspI sites were neither mutated nor abnormally methylated. Southern blotting showed that structural polymorphism was present in tumor DNA and tumor adjacent normal tissue DNA but absent from lymphocyte DNA from the same patients. We conclude that the non-B DNA structural polymorphism detected in human tumors near the c-Ha-ras VNTR is a self-perpetuating epigenetic mark that manifests itself spontaneously during breast carcinogenesis in a methylation hot spot.
Non-B DNA structures are a plausible description of mutational hotspots (1–3), sites of triplet repeat expansion (2, 4, 5) and clustered sites of somatic recombination (6). However, many sites of unusual structure appear to be both stable and favored conformations that are present in living cells. These occur as structures associated with chromosomal centromeres (7), telomeres (8) and the control regions of genes (9–14). In general such sequences can be heavily methylated (15), highly polymorphic in the human population (16, 17), known breakpoint clusters in cancers (15), or unusual sequences associated with chromosome fragile sites (18).
One such carefully studied site is the VNTR at c-Ha-ras. This sequence of tandem 28 bp repeats resides in a known hotspot (11p15) for DNA methylation in tumors (19), and is highly polymorphic in the human population (20). Although unusual structures associated with the c-Ha-ras VNTR have not been reported, the presence of multiple length variants in the population suggests that the sequence is prone to the formation of unusual DNA structures associated with unequal recombination or slippage.
Since non-B DNA structures in isolated DNA appear to be exceptionally sensitive to modification by bisulfite, we have asked whether or not sites known to reside in the hotspot for DNA methylation on chromosome 11p15 that is associated with human solid tumors were sensitive to bisulfite modification. Our results show that a sequence adjacent to the highly polymorphic VNTR 3′ to the c-Ha-ras coding sequence is selectively sensitive to bisulfite modification in breast tumor DNA and in DNA from adjacent histologically normal margins, suggesting that the unusual structure is propagated clonally. The presence of an unusual structure in tumor and tumor adjacent tissue specimens was confirmed by restriction analysis using an enzyme that is unable to cleave single-stranded DNA. Inspection of patients’ blood lymphocyte DNA with the same methods showed that lymphocyte DNA from these same patients remained in the B-DNA conformation.
Adjacent specimens of malignant and normal breast tissues were obtained under an IRB approved protocol at resection and stored at −70°C prior to use. The tissue was pulverized under liquid nitrogen and then digested with proteinase K (Applied Biosystems, Foster City, CA, USA) overnight at 50°C. High-molecular-weight DNA was purified with an automated DNA extractor (Applied Biosystems). Purified DNA was treated with 0.05 μg of RNase (Sigma Chemical Co., St. Louis, MO, USA) per ml for one hour at 37°C and collected by ethanol precipitation.
Genomic DNAs isolated as described above from Human Breast Tumor, Tumor Adjacent, and Blood Lymphocyte specimens (1 μg each) were bisulfite treated using the EZ DNA Methylation Kit (Zymo, Orange, CA, USA) as described by the manufacturer except that, the sodium hydroxide denaturation step was omitted when non-denatured DNA was treated with bisulfite. 1 μg of DNA was added to a final volume of 45 μL, for both conditions. For denatured samples, 5 μL of M-Dil Buffer (NaOH) was added and was incubated at 37°C for 15 min. The non-denatured samples were likewise incubated for 15 min at 37°C. Then 100 μL of CT-reagent (bisulfite) was added to the denatured samples. 5 μL of M-dil buffer was added to 100 μL of CT-Reagent, mixed and then 105 μL added to the 45 μL non-denatured samples. A set of denatured and nondenatured DNAs was then incubated 55°C overnight protected from light, and likewise a similar set was incubated at 37°C overnight and protected from exposure to light. The DNA was subsequently desulfonated and purified per the manufacturer’s instructions. The DNA was eluted and adjusted with 10 mM Tris 1 mM EDTA, pH 7.5 to 40 ng/μg assuming 100% recovery. The resulting four samples were then amplified with 3′ MspI or 5′ MspI primers directed at the upper and lower strands of each region (Figure 1). The amplification reaction consisted of 0.25 μl Qiagen Hotstart Taq (Qiagen, Valencia, CA, USA), 2.5 μl 10x Qiagen buffer, 320 mM dNTPs, 2.0 mM added MgCl2, 1.0 μl Q-Solution, 900 nM each for forward and reverse primers, 5.0 μl DNA. The conditions for 5′ MspI were 95°C 10 min, then 35cycles of 95°C 30s, 50°C 30s, 72°C 60s, then extension for 7 min at 72°C. The conditions for 3′ MspI were 95°C 10 min, then 35 cycles of 95°C 30s, 60°C 30s, 72°C 60s, then extension for 7 min at 72°C. The PCR reactions were then run on a 2% agarose gel run at 100 volts for 1.5hrs, with the xylene cyanol dye marker at 6 cm. The Gels were then stained with ethidium bromide and visualized with a UV transilluminator. The sequences of the amplification primers were as follows: Lower Reverse 5′ MspI 5′-CTACTAAAAACCCACAACGC-3′; Lower Forward 5′ MspI 5′-TTTTGGTTAGTTAGCGGTAT-3′; Upper Reverse 5′ MspI 5′-CCAACCAACGACATACCCTA-3′; Upper Forward 5′ MspI 5′-TTTTGTTGGGGGTTTATAGCGT-3′; Upper Forward 3′ MspI 5′ATTAGTTTTTTTATCGATAGATTTTTC-3′; Upper Reverse 3′ MspI 5′-TCACCTACATCTAACGCCCCCAACA-3′; Lower Forward 3′ MspI 5′-ACCAACTTCCCCATCGATAAACTTCCC-3′; Lower Reverse 3′ MspI 5′-TTATTTGTATTTGACGTTTTTAGTA-3′. Untreated DNA PCR reactions were identical to those used for bisulfite treated DNA, except that for the 5′ MspI site, where the following PCR conditions were used: 95°C 10 min, then 35 cycles of 95°C 30 s, 60°C 30 s, 72°C 30 s, then extension for 7 min at 72°C, and for the 3′ MspI site 95°C 10 min, then 35 cycles of 95°C 30 s, 56°C 30 s, 72°C 30 s, then extension for 7 min at 72°C.
The bands corresponding to the 155 bp product from the upper strand and the 153 bp product from the lower strand of the 5′ MspI target and the 120 bp products from the upper and lower strands of the 3′ MspI target were gel-extracted using the QIAquick Gel Extraction Kit (Qiagen). The products of the untreated 3′ MspI (154 bp) and 5′ MspI (120 bp) were also gel-extracted in the same fashion. Once isolated, the PCR product was cloned from 4 μl of extracted DNA using the pCR 2.1-TOPO Cloning Kit (Invitrogen, Carlsbad, CA, USA), as described by the manufacturer. The resulting colonies were inoculated into liquid cultures and plasmid DNA was isolated from each. Plasmids with inserts of the appropriate size were sequenced at the City of Hope DNA Sequencing Lab as previously described (21).
DNA was suspended in a solution of 6 mM KC1, 10 mM Tris-HC1 (pH 7.4), 10 mM MgC12, 1 mM dithiothreitol, and 100 μg of bovine serum albumin per ml. Restriction endonucleases HpaII, MspI, and MboI (New England Biolabs, Beverly, MA) were used to digest the DNA. The DNA was pre-incubated for 15 minutes at 37°C. Separate digests containing HpaII (5 U/μg of DNA), MspI (10 U/μg of DNA), or MboI (5 U/μg of DNA) were incubated for two hours at 37°C. After one hour of digestion, additional MspI (5 U/μg) was added to the MspI digest. To ensure completeness of the digestion, co-digestion control experiments with ΦX174 DNA were performed for each DNA digest as previously described (22). Simian virus 40 DNA was used as a co-digestion control for MboI restriction digests since ΦX174 contains 6-methyladenine at the 5′ GATC recognition site and cannot be cleaved by MboI. MboI is not sensitive to the presence of 5-methylcytosine in DNA.
DNA was precipitated with ethanol and suspended in water, and its concentration was determined from its absorbance at 260 nm. DNA fragments were separated by electrophoresis on 1.2% agarose gels (Sigma Chemical Co.) in a solution of 50 mM Tris-acetate, 10 mM EDTA, and 30 mM sodium acetate (pH 8.7) overnight at 30 V.
Southern transfers (23) were prepared using the alkaline transfer procedure and nylon membranes (Zeta-probe, Bio- Rad Laboratories, Richmond, CA, USA). The 987 bp MspI fragment containing the VNTR region of plasmid pT24-C3 was isolated from MspI digests of plasmid DNA by the freeze-squeeze technique (24). The isolated DNA was subsequently 1abeled with 32P to a specific activity greater than 109 cpm/μg by using the random priming procedure (25) and was then used to probe Southern blots. Hybridization was performed (23) and the membranes were used for autoradiography.
We chose to inspect the regions flanking the VNTR since they have been shown to be refractory to restriction digestion by MspI in tumor DNA (26). In order to confirm that the reported polymorphism affects the DNA from the patients studied here, we performed a series of Southern blotting experiments on eleven different patient specimens.
Digestion controls like those depicted in Figure 2 were performed on each DNA specimen prepared for Southern blotting. These controls show that a limit digest of the human DNA had been achieved in each case. The number of bands in MspI digests confirms the unusual degree of polymorphism in sequences homologous to the c-Ha-ras VNTR region. Samples from individual patients showed either two or four prominent MspI bands (Figure 3). Additional (higher molecular weight) bands were often observed as minor bands with normal DNAs. Length polymorphism at these sites is well known, and allelic variants are well characterized (27). However, it seems unlikely that a normal individual would inherit four alleles for the locus. Thus, some of this polymorphism must be generated somatically.
Somatic insertion-deletion events at the VNTR could generate the observed polymorphism by creating length variants. Alterations at the MspI sites directly flanking the VNTR could generate the polymorphism by creating restriction site variants. To distinguish between these possibilities, DNA samples were digested with MboI and Southern blots were hybridized with the VNTR probe. MboI does not cleave within the smallest VNTR fragment produced by MspI (Figure 1). Thus, the number of MspI fragments produced from a given specimen must equal the number of fragments in MboI digests if the patterns observed with MspI reflect VNTR length. MboI cleavage experiments showed that this is not the case.
Patients exhibiting two prominent bands in MspI digests (Figure 3A) exhibited a single prominent band in DNA digested with MboI (Figure 3B). DNA from patients exhibiting four prominent MspI fragments generated only two fragments with MboI (Figure 3C and Figure 3D).
These findings suggest that the polymorphism is associated with unusual properties at MspI sites flanking the VNTR that render some of these sites refractory to MspI cleavage. Since cleavage kinetics for restriction enzymes in general and for MspI in particular are known to be altered by the sequence context near the recognition site (28) we tested the reproducibility of our Southern blotting results with selected patient samples. Those tests showed that patterns of cleavage and band intensity could be reproduced in experiments performed completely independently more than eighteen months apart. Moreover, sequential digestion with MspI followed by HpaII did not alter the patterns. These findings indicate that the patterns are produced from true limit digests of DNA.
Sequence information is available for the region in the vicinity of a representative VNTR. By using that data, the fragmentation pattern can be analyzed from the restriction map as shown in Figure 1. Fragments carrying a long VNTR are designated with the letter A. Fragments carrying a short VNTR are designated with the letter B. The fragments produced by cleavage of both of the sequenced MspI sites directly flanking the VNTR are designated A0 or B0. Fragments produced by cleavage of only one of these two sites, followed by cleavage of the next available MspI site, add 580 or 581 bp of DNA to the primary fragment. The two possible secondary fragments differ by only 1 bp, and were not separated from one another on these gels. These are designated A0′ or A0″ for the long variant and B0′or B0″ for the short variant in Figure 1. Fragments produced by MboI are designated A1 or B1.
The analysis is illustrated by data in Figures 1 and and3.3. In the first patient, the MspI digest shows two bands at about 2720 and 3320 bp (Figure 3A). This might be interpreted as heterozygosity for length alleles at this locus; however, the intensity of the 3320 bp band as determined from gel scans is inappropriately lower than that of the 2720 bp band. Heterozygosity is ruled out for this patient by the MboI digestion pattern (Figure 3B), which shows a single length variant of 3650 bp. Since the distance between the MspI site and the MboI sites that flank the VNTR is 589 bp on the left and 317 bp on the right (Figure 1), the length of the MspI fragment containing the VNTR (i.e., fragment A0) should be 906 bp shorter than the 3650 bp MboI fragment containing the VNTR. Since the expected value of 2734 bp agrees well with the observed value of 2720 bp for the shorter of the two MspI fragments, we conclude that the short fragment is fragment A0 with reference to the restriction map. Since the long MspI band is about 580 bp longer than the shorter band, we conclude that this band represents the A0′or A0″ fragments produced by lack of cleavage of one of the MspI fragments immediately flanking the VNTR.
The second patient (Figure 3C) was clearly heterozygous for allelic length variants, as judged from the MboI digests. This enzyme produced fragments of about 3175 bp (A1) and 2050 bp (B1). Moreover, these appear to be true length alleles, since scans of the hybridization signal in the autoradiograph showed that the two fragments were present at about equivalent ratios.
Four bands were detected in MspI digests of DNA from the same patient. The fragment sizes were again completely consistent with the map if one assumes that a small fraction of the MspI sites immediately adjacent to the VNTR are refractory to MspI digestion. The largest fragments (designated A0′ or A0″) were about 2900 bp long and differed from the next largest fragment A0 (2325 bp) by about 580 bp. B0′ or B0″ fragments were about 1650 bp long and differed from the shortest fragment, B0 (1050 bp) by about 580 bp. This is again consistent with the restriction map, and is confirmed by comparing the two primary MspI bands (designated A0 and B0 in Figure 1) with the corresponding MboI fragments (A1 and B1). In each case the observed difference was very close to the expected 906 bp.
MspI cleavage might be blocked by point mutation at either of these two sites. Thus, we asked whether or not the sites were mutated by sequencing representative clones of the PCR amplified regions surrounding the affected sites (Figure 1). Mutations were not observed in any of the 20 clones sequenced from each site, all of which exhibited the normal 5′CCGG 3′ sequence that is susceptible to MspI cleavage (Figure 4).
Since single-stranded DNA is known to block MspI cleavage, it seemed likely that one or both of the restriction sites near the VNTR might possess single-stranded character that would render it accessible to bisulfite modification under native conditions. Bisulfite attacks cytosines and 5-methylcytosines that are accessible to the reagent and water (i.e. present in single-strands or exposed in unusual DNA structures). Cytosines are rapidly deaminated while 5-methylcytosines are deaminated much more slowly (29). In general cytosines can be converted to uracils under conditions where 5-methylcytosines remain intact. This results in strands that are no longer complementary so that PCR amplification can focus on either strand (Figure 5) when primers complementary to one or the other of the converted strands are used (30).
PCR systems (Figure 1) specific for each of the two converted strands (Upper and Lower) were used to amplify regions containing the MspI site 5′ to the VNTR (5′ MspI) and the MspI site 3′ to the VNTR (3′ MspI). Only 5′ MspI was amplified under native conditions at 37°C indicating that it was in a structure accessible to bisulfite under these conditions (Figure 6). In order to assess the degree of accessibility at 5′ MspI under native conditions, representative amplicons corresponding to both the upper and lower strands of 5′ MspI were cloned and sequenced from the amplification product obtained after bisulfite treatment under native conditions (non-denatured DNA treated with bisulfite at 37°C). The results are depicted in Figure 7A. Nearly all cytosines except those at the two CG sites in the amplicon were converted to uracils in the region. When amplification was carried out after bisulfite treatment under the denaturing conditions used in determining CG methylation state (30) sequencing of the cloned representatives of the amplification products from the upper and lower strands showed that the cytosines in the two CG sites in the region were symmetrically methylated (Figure 7B). Thus the methylated state of the two CG sites in the region is associated with bisulfite accessibility in the region.
We next asked whether or not tumor tissue differed from lymphocyte DNA from the same patient. When PCR amplification was carried out after bisulfite treatment under native conditions using DNA isolated from lymphocyte or from tumor DNA isolated from the same patient, tumor DNA was found to be about 2.5 fold more accessible to bisulfite than lymphocyte DNA. This finding was confirmed by Southern blotting (Figure 3D and 3E). In these experiments, limit digests of both blood and tumor DNA were obtained (see representative digestion control in Figure 2). Southern blotting demonstrated partial digestion of the tumor and tumor adjacent DNA contrasted with complete digestion of DNA from the blood lymphocyte DNA.
Polymorphism of the MspI sites near the VNTR region at c-Ha-ras has been observed a variety of tumors (19, 26, 31). In general polymorphism at the site could result from point mutation, non-CG methylation, or an unusual DNA structure affecting the expected site of MspI attack. The Southern blotting results reported here (Figure 3) extend the results reported by others (19, 26, 31) by demonstrating that the polymorphism affects one or both of the two MspI sites immediately adjacent to the VNTR since partial digestion products were about 580 bp longer than the primary bands detected in the blot (Figure 1). With reference to Figure 1 homozygous individuals were found to possess two fragment lengths in limit digests that differed by about 580 bp in length (Figure 3A and Figure 3B), while heterozygous individuals (Figure 3C, Figure 3D and Figure 3E) were found to possess four fragment lengths among which two sets were found to differ by about 580 bp each. Partial digestion could be ruled out by the digestion controls performed on each digest (Figure 2). Sequence polymorphism could also be ruled out by inspecting cloned representatives of the two sites (Figure 4). Non-CG methylation was ruled out by sequencing the methylation-sensitive PCR products of each strand (Figure 7B). Thus we conclude that the data demonstrating bisulfite modification of native DNA at 5′ MspI (Figure 7A) coupled with the Southern blotting results outlined above indicate the presence of an unusual DNA structure that spans this 5′ MspI site.
A gap in the duplex in the region is ruled out because both strands were amplifiable and both were accessible to bisulfite (Figure 7A). This suggests a possible cruciform structure at the site or any one of the canonical unusual structures that can be adopted by cruciform DNA (e.g. quadruplex or triplex). However the data do not allow us to determine the exact nature of the unusual structure present at this site. Although the non-B DNA structure survives DNA isolation as has been shown for the structure observed by Raghavan et al. (6) at the Bcl2 locus, it remains possible that the structure present in vivo is altered somewhat during DNA isolation. Moreover, given the close proximity to the VNTR it is possible that secondary structure extending into VNTR itself contributes to the unusual structure at 5′ MspI.
Since the structure is present at high levels in both tumor and tumor-adjacent (histologically normal) DNA but present in much reduced levels in lymphocytes it constitutes either a tissue specific or tumorigenesis specific epigenetic mark. Preliminary evidence (Clark unpublished) in prostate cancer specimens suggests that it is specific to tumorigenesis. The presence of the structure in non-tumor tissue in the breast suggests a field cancerization of the cells involved in clonal propagation of the structure. A clonal loss of function associated with the inability of the cells to suppress the formation of the unusual structure after replication would provide a simple explanation for its propagation. Preliminary experiments with nuclease S1 sensitivity suggest that the breast tumor DNA is rapidly digested indicating that additional sites of non-B DNA structure are present in the breast cancer DNA (Smith unpublished).
Finally, we note that this unusual structure is methylated at both CG sites inspected by the methylation sensitive PCR (Figure 7B). Given recent reports demonstrating that Dnmt1 is responsible for both de novo methylation (42) and the maintenance of methylation patterns (43) in human cancer cells, it seems likely that the unusual structure promotes de novo methylation by Dnmt1 (39, 44, 45). Thus the hypermethylation of these sites is consistent with a simple loss-of-function hypothesis in which a component of the cellular system that suppresses unusual structure formation is lost permitting the spontaneous formation of unusual structures that are known to be activated substrates for human Dnmt1 (2, 34, 46, 47). This is consistent with reports that the 3′ MspI fragment resides in a known hot spot for DNA methylation (19). Moreover, it suggests that the known tendency of DNA binding proteins important in epigenetic signaling in both mice and humans [e.g. DNA (Cytosine-5) Methyltransferase I [Dnmt1] (32–39) and Methylated Cytosine Binding Protein 2 [MeCP2] (40, 41)], to interact strongly with unusual structures present in DNA is biologically meaningful.
Supported primarily by grant l571AR2 from the Council for Tobacco Research, grant GM32863 from the National Institutes of Health and grant 1-1130 from the March of Dimes Birth Defects Foundation, with additional support from Training Grant CA-09477 from the National Institutes of Health and the City of Hope Cancer Center CA-335172.
The authors would like to thank David Baker and Lori Jardines for their help with DNA isolation and southern blotting.