Assessment of genomic instability in breast cancer and uveal melanoma by random amplified polymorphic DNA analysis


  • Sarantos Papadopoulos,

    1. Department of Obstetrics and Gynecology, Free University of Berlin, Berlin, Germany
    Search for more papers by this author
    • The first 2 authors contributed equally to this work.

  • Thomas Benter,

    Corresponding author
    1. Department of Hematology, Oncology and Tumor Immunology, Robert-Rössle Clinic, Charité, Campus Buch, Humboldt University Berlin, Berlin, Germany
    • Department of Hematology, Oncology and Tumor Immunology, Robert-Rössle Clinic, Charité, Campus Buch, Humboldt University Berlin, Berlin, Germany
    Search for more papers by this author
    • The first 2 authors contributed equally to this work.

    • Fax: +49-30-9417-1109

  • Gerasimos Anastassiou,

    1. Department of Ophthalmology, Essen University Medical Center, Essen, Germany
    Search for more papers by this author
  • Michael Pape,

    1. Division of Internal Medicine, Department of Nephrology, Hanover Medical School, Hanover, Germany
    Search for more papers by this author
  • Schaller Gerhard,

    1. Department of Obstetrics and Gynecology, Free University of Berlin, Berlin, Germany
    Search for more papers by this author
  • Norbert Bornfeld,

    1. Department of Ophthalmology, Essen University Medical Center, Essen, Germany
    Search for more papers by this author
  • Wolf-Dieter Ludwig,

    1. Department of Hematology, Oncology and Tumor Immunology, Robert-Rössle Clinic, Charité, Campus Buch, Humboldt University Berlin, Berlin, Germany
    Search for more papers by this author
  • Bernd Dörken

    1. Department of Hematology, Oncology and Tumor Immunology, Robert-Rössle Clinic, Charité, Campus Buch, Humboldt University Berlin, Berlin, Germany
    Search for more papers by this author


Some types of cancer have been associated with abnormal DNA fingerprinting. We used random amplified polymorphic DNA (RAPD) to generate fingerprints that detect genomic alterations in human breast cancer. Primers were designed by choosing sequences involved in the development of DNA mutations. Seventeen primers in 44 different combinations were used to screen a total of 6 breast cancer DNA/normal DNA pairs and 6 uveal melanoma DNA/normal DNA pairs. Forty-five percent of these combinations reliably detected quantitative differences in the breast cancer pairs, while only 18% of these combinations detected differences in the uveal melanoma pairs. Fourteen (32%) and 12 (27%) primers generated a smear or did not produce any band patterns in the first and second cases, respectively. Taking into account the ability of RAPD to screen the whole genome, our results suggest that the genomic damage in breast cancer is significantly higher than in uveal melanoma. Our study confirms other reports that the molecular karyotypes produced with random priming, called amplotypes, are very useful for assessing genomic damage in cancer. © 2002 Wiley-Liss, Inc.

Genomic instability is a hallmark of neoplastic transformation and a herald of genomic damage. The existence of an additional type of genomic instability has been described, the microsatellite mutator phenotype pathway. Progress in molecular cancer genetics has facilitated the detection of these mutations. In the last decade, representation differential analysis and comparative genomic hybridization (CGH) were added to methods like flow cytometry, restriction fragment length polymorphisms of polymorphic minisatellite loci and allelotyping, a PCR amplification of informative microsatellite loci.

Another promising approach has been introduced, arbitrarily primed PCR1 combined with random amplified polymorphic DNA (RAPD).2 Both PCR-based techniques, called amplotyping, use fingerprinting to quantitatively and qualitatively evaluate band alterations.3 This random priming method allows comparison of normal and tumor tissues at a minimum of 20–40 genome sites in a single PCR, though evidence suggests that the number of compared loci runs into the hundreds if we take into account that a gel band consists of many comigrating fragments.4, 5 It also includes internal controls and detects moderate gains of chromosomal sequences (trisomy/tetrasomy) that would otherwise escape detection.

Previously, we addressed the problems of reproducibility and contamination in random priming. In the present work, we used RAPD fingerprinting to assess the genomic damage present in breast cancer and in uveal melanoma cells. For this task, we designed primers based on sequences that are supposed to play an important role in the mechanisms leading to DNA alterations. Human gene mutations appear to be caused by multiple mechanisms whose relative importance is probably governed by local primary and secondary DNA structure.7 We have searched for consensus sequences located in the immediate vicinity of gene mutations and for “hot spots” that mediate illegitimate recombinations, a mechanism that is suggested to be crucial for the generation of genomic instability and cancer.


DNA samples

Breast tumor samples and corresponding EDTA blood were obtained from breast cancer patients who underwent surgery at the Friederickenstift Hospital (Hanover, Germany), while melanoma tumor samples and corresponding EDTA blood were obtained from uveal melanoma patients who underwent surgery at the Department of Ophthalmology, University Medical Centre of Essen (Essen, Germany). All tissue samples were shock-frozen immediately after removal and maintained at –70°C until use. The histologic profiles of the 6 breast carcinomas are shown in Table I. Genomic DNA from the tumor samples was extracted using the DNA Twin Prep DNA/RNA Kit (Invitek, Berlin, Germany). Genomic DNA from peripheral mononuclear white blood cells was extracted using the Invisorb Spin Blood Kit (Invitek). Both were eluted in 10 mM TRIS HCl/1 mM EDTA (pH 8.0).

Table I. Histopathologic Profiles of the 6 Breast Cancer Specimens (Pluses Describe Level of Protein Expression)
Stage/gradingpT2 GIIpT2 GIIpT2 GIIIpT2 GIIIpT1C GIIpT4 GIII
Estrogen receptor90%30%70%30%90%
Progesterone receptor90%30%30%30%90%
Cathepsin D+++++++++++


Amplification was achieved in an MJR thermocycler (Biozym Diagnostik, Hess Oldendorf, Germany). All PCRs were performed using 2.5 U/50 μl Taq DNA polymerase from GIBCO/Life Technologies (Karlsruhe, Germany) with a volume of 50 μl for the breast cancer DNA and a volume of 12.5 μl for the uveal melanoma DNA. Each primer had a final concentration of 0.5 μM and each deoxyribonucleotide (Pharmacia Biotech, Freiburg, Germany) was added to 0.2 mM. Reactions were conducted in 4 mM MgCl2, 50 mM KCl and 10 mM TRIS (pH 8.3) containing 80 ng DNA/50 Λl reaction volume. The PCR profile was 35 cycles consisting of denaturation at 94°C for 60 sec, annealing at 35°C for 60 sec, a ramping of 1°C/11 sec (420 sec) and extension at 72°C for 120 sec. After the cycling, a final extension for 10 min at 72°C was followed by slow cooling to 15°C, at which amplification products were maintained until being loaded onto gels. Reaction mixtures were overlaid with 2 drops of paraffin oil. Negative controls were used in every experiment to detect contamination. Controls contained all reaction components except DNA template, which was replaced by water. They were prepared in a laminar flow hood, according to suggested contamination precautions.8 For every trial, a master mix was made, to keep pipetting errors to a minimum. Each experiment was repeated twice. The primers used are listed in Tables II and III.

Table II. Primers and Results (Breast Cancer)
PrimerNameSequence (5′→3′)PCR productsDifferences/informative
10C 4/ PJaT GAA AAA G+
12R 12 A/ 267AGC GAG ACT CCG+
15Zn1 + CORE
16Zn1 + DCR4++
17Zn1 + R 12 A/ 267
18Zn1 + ACR3++
19Zn1 + TELO3+
20Zn1 + DHSmear
21TATA + R 12 A/ 267++
22ACR 3 + DCR 4+
23EP 7 + ACR 3+
24C 4/ PJa + ACR 3+
25C 4/ PJa + DCR 4++
26CHI + TELO 3++
27CORE + TELO 3++
28STARK + ACR 3++
29STARK + DCR 4++
30STARK + EP 2Smear
32STARK + R 12 A/267Smear
33STARK + TELO 3++
34STARK + C 4/PJaSmear
35DH + ACR 3++
36DH + DCR 4++
37DH + EP 2++
38DH + EP 6++
39DH + R 12 A/267++
40DH + TELO 3
41DH + C 4/PJa++
42DH + MHS++
43L1HS + MHS+
Table III. Primers and Results (Uveal Melanoma)
PrimerNamePCR productsDifferences/informative
1EP 2+
2EP 6++
3EP 7+
4Zn 1Smear
5Zn 2+
6ACR 3 + DCR 4+
7TELO 3 + CHI++
8TELO 3 + EP6++
10TELO 3 +DCR 4Smear
11C 4/PJa + DCR 4++
12C 4/PJa + R 12 A/ 267++
13C 4/PJa + ACR 3+
14L1HS + Zn 1Smear
15L1HS + Zn 2Smear
16L1HS + DH+
17Zn 1 + ACR 3+
18Zn 1 + EP 2Smear
19Zn 1 + EP 6+
20Zn 1 + TELO 3+
21Zn 2 + C 4/PJa++
22Zn 2 + ACR 3+
23Zn 2 + DCR 4+
24Zn 2 + MHS+
25Zn 2 + EP 2
26Zn 2 + EP6+
27Zn 2 + R 12 A/ 267+
28STARK + Zn 1Smear
29STARK + DCR 4+
30STARK + EP 2Smear
31STARK + EP 6Smear
32STARK + R 12 A/267Smear
33STARK + C 4/PJaSmear
34STARK + TELO 3++
35DH + ACR 3+
36DH + DCR 4+
37DH + MHS+
38DH + EP 6+
39DH + R 12 A/267++
40DH + Zn 1+
41DH + Zn 2+
42DH + TELO 3
44TATA + R12A/267+


Electrophoretic separation was carried out in a Horizon 10-14 horizontal electrophoresis apparatus (GIBCO/Life Technologies). Bands were separated on a 2% TBE agarose (GIBCO/Life Technologies) gel at 80 V for 4–5 hr.


Using RAPD analysis, we applied 17 primers and various combinations to detect genomic instability in 6 pairs of human breast cancer DNA/normal DNA. Fourteen of 44 primers or primer combinations led to smears or no band arrays at all because the targeted sequences were located at a distance that did not favor the generation of PCR products. Twenty (67% of pattern-producing reactions and 45% overall) of the remaining combinations were informative, revealing differences between the fingerprints in question (Table II). In the case of uveal melanoma, we applied the same 17 primers and various combinations to detect genomic instability in 6 pairs of uveal melanoma DNA/normal DNA. Twelve of 44 primers or primer combinations led to smears or no band arrays at all. Eight (25% of pattern-producing reactions and 18% overall) of the remaining combinations were informative (Table III).

After the initial screening, we used 5 primer combinations and 1 single primer (CHI+TELO3, DH+R12A/267, DH+C4, STARK+TELO3, EP6, STARK+EP6) to evaluate the effectiveness of primers at detecting differences in an additional 5 pairs of breast cancer DNA/normal tissue DNA. Five of the 6 reactions (CHI+TELO3, DH+R12A/267, DH+C4, STARK+TELO3, EP6) produced quantitative differences in the 5 breast cancer pairs, while STARK+EP6 led to a smear as it did in the screening. Interestingly, some differences were always present, especially with regard to the histopathologic profiles (Table I). By contrast, other differences characterized only a subgroup of the tissue pairs (Fig. 1b, compare bands marked with arrows 1 and 5 to those marked with arrows 2, 3 and 4).

Figure 1.

RAPD fingerprints of tumor DNA/white blood cell DNA pairs from 6 uveal melanoma patients (a) and from 6 breast cancer patients (b) generated by primer EP6. Odd numbers mark amplifications of tumor DNA; even numbers mark amplifications of white blood cell DNA. Arrows mark quantitative and qualitative differences. (a) Arrows point to high and low m.w. differences, which characterize either all pairs (arrows 3 and 4) or only a part of them (arrows 1, 2, 5). (b) Arrows 1 and 5 point to differences observed in all 6 pairs. Arrows 2–4 point to differences found only in a subgroup. The 1 kb ladder was used as a marker.

Further, we used 4 primer combinations and 1 single primer (CHI+TELO3, DH+C4, ACR3+DCR4, EP6, STARK+EP6) to examine an additional 5 pairs of uveal melanoma DNA/normal tissue DNA. Two of the 5 reactions (CHI+TELO3, EP6) produced differences in the 5 uveal melanoma pairs, 2 did not reveal any differences (DH+C4, ACR3+DCR4) and STARK+EP6 led to a smear, confirming the results of the screening (Fig. 1a).

In our study, the most DNA genomic alterations were identified as band loss or gain. In some cases, we observed an increase or a decrease of the band intensity as well, which was also considered a criterion for classifying instability due to the low extent of tumor contamination with normal tissue. Our experiments suggest that the GC content is not important for the reproducibility of band arrays and that primers with 40% and 70% GC will yield reproducible results. We can confirm reports that primers with a higher GC content generate more PCR products than those with a lower content.9 A correlation between the ability of a primer to detect differences and its GC content could not be established. Some of our results were difficult to explain. For example, the primers CORE and R12A/267, although based on short interspersed nucleic element (SINE) sequences, gave different results, while the combinations Zn1 with CORE and Zn1 with R12A/267 did not produce any bands, despite the smear and the bands that were produced by Zn1 and R12A/267, respectively, when used alone. The kinetics of RAPD reactions is very complex and it appears that not only the primary but also the secondary DNA structure is of great importance. Data also suggest that the molecular nature of the flanking region of the target site determines the relative intensity of the RAPD bands.10 The proximity and/or the orientation of the primer sequences of Zn1, CORE and R12A/267 appear to be inadequate for the generation of amplimers. This is indeed hard to believe in light of the fact that the copy number of the major human SINE, the Alu elements, adds up to several hundred thousand to 1 million, which may account for approximately 6–10% of the whole genome and be distributed throughout the genome with an average spacing of 4 kb.11

Although RAPD analysis was first described in 1990 and has opened up new horizons in molecular diagnostics and genetics, this method has not been fully exploited in studies on humans due to concerns over reproducibility on a day-to-day basis and between laboratories and the problem of contamination. For many, it is also confusing that the several groups have acquired different experiences with random priming during their optimization procedures. Most researchers agree that the type of thermocycler12, 13 and the type of polymerase14 used in RAPD assays contribute to the variation in the PCR results. The underlying reasons are suboptimal thermocycler performance and the differences between the various strains/sources of the polymerases, the variable properties of the enzymes (e.g., proofreading activity, processitivity and Mg2+ optimum), the different unit definitions, the variations in the concentrations of buffer ingredients and pH and the presence or absence of additives (e.g., the various stabilizers). In our eyes, the only real obstacle in the standardization of RAPD terms is the polymerases. Singh and Roy15 demonstrated that detection of mutations in the genome of cancer tissue by random priming depends on the type of DNA polymerases used, a finding that was anticipated after the first articles that underlined the role of the enzymes in achieving reproducibility.14, 16 They argued that the ability of the polymerase to extend a particular locus of the gene is influenced by mutations in the primer binding sites in the competitive environment, where several genomic sequences flanked to a random primer are coamplified by RAPD. The conclusion that must be drawn is that absolute repeatability of results is achieved only when using the same polymerase from the same supplier, which will seriously hinder the standardization of random priming.

About the other factors that compromise the repeatability of results, such as the purity of the amplified template, different recommendations can be found in the literature. Singh and Roy,15 e.g., argued that the addition of primer in the master reaction mixture resulted in a smear and that DNA from the same tissue, which was isolated by different laboratories, needed different amounts of polymerase and MgCl2. Our experience suggests that the order of adding the reaction mixture components is not important and that the smear is characteristic for some primers (e.g., in the case of EP2+STARK). The quality of the template is of importance, but we have shown that DNA isolated by different persons on different days and with diverse methods provided reproducible patterns when the concentration ranged from 50–500 ng.6 Another element that we consider important for the uniformity of RAPD profiles is the slope, i.e., the slow transition between annealing and extension temperatures, which has been confirmed by other groups.17 We have demonstrated that a ramp slower than 1°C/11 sec provides no further benefits.6 It appears that with slow ramping/heating and the fact that the polymerase already starts extension at low temperatures the primer template complex is stabilized with the help of the added deoxyribonucleotides. However, fast transitions may result in premature detachment from the template if we consider that there might be no absolute homology between them. The positive impact of the ramp on reproducibility was also corroborated during experiments with 9 thermal devices from 6 different manufacturers (T. Benter et al., unpublished data). When a slope of 1°C/11 sec was used, 6 of them showed identical RAPD patterns.

Another issue that must be addressed is contamination. Band arrays could be altered by adding foreign DNA of either human or bacterial origin, equivalent to 5–10% of the DNA to be amplified (T. Benter et al., unpublished data). Other groups made similar tests to exploit the potential of RAPD in mutation detection, finding that in mixing experiments the lowest proportion of mutant DNA at which polymorphic bands were detected was at a mixture ratio (wild-type DNA:total DNA) of 1:50.18 All of these results suggest that RAPD is a method of exceptional sensitivity. In the case of cancer genetics, tumor samples should be relatively free of normal tissue DNA to generate high-quality results. It is obvious that an admixture of normal tissue can give rise to differences in RAPD fingerprints from tumor to tumor depending on the extent of contamination. The purity is considered even more important when quantitative differences, i.e., differences in band intensities, are taken into account. Contamination of 50% of normal tissue, e.g., which is thought to be the theoretical detection limit,3 will make a tetrasomy of a DNA fragment appear as an intensity gain of 50%. This could indeed influence the results when tumors with different tendencies to infiltrate into normal tissue are compared.

For our comparison, we chose relatively large breast cancer tumor samples with a diameter >2 cm except one (T1) and uveal melanoma tumors with a diameter of 1 cm. DNA was isolated from the center of the tumors after histologic analysis, which demonstrated that the breast cancer tumors in these compartments had 5–10% contamination of normal tissue. In uveal melanoma tumors, contamination was even smaller.

Comparison of lanes 5 and 6 of Figure 1a raises another issue concerning DNA quality. Despite best efforts, there are sometimes differences in DNA quality when DNA is extracted from peripheral white cells and from tumor tissue. As tissue DNA tends to be more sheared, high m.w. bands tend to be better amplified in leukocyte DNA and low m.w. bands tend to be better amplified in tumor DNA. This appears to be the case with the third uveal melanoma tumor (lane 5), where the large band (arrow 1) and the fragments above 1.6 kb were not produced. In contrast, generation of the lower bands (arrows 3 and 4) was not compromised but rather facilitated with the amplification of a new band (arrow 5). In all other tumor tissues, the high m.w. band (arrow 1) was produced, confirming the good quality of the extracted DNA. In Figure 1b, more differences were detected between 700 and 400 bp but not in all tumors. In our opinion, all differences are convincing, both the larger (arrow 1) and the smaller (arrows 2–5), since the preceding and succeeding bands are of similar intensity in tumor and normal profiles (e.g., bands under band 1 and bands between 3 and 4), a criterion that was previously applied by other groups.19

In a previous article,6 we tested different DNA isolation methods; e.g., we prepared DNA by embedding lymphocytes in low-melting agarose for pulsed field gel electrophoresis and by the phenol/chloroform/proteinase K procedure, which enable the isolation of high m.w. DNA as with commercial kits and, in the forensic science broadly used, by the chelating resin Chelex (Bio-Rad/Greiner, Flacht, Germany), which is thought to extract low m.w. DNA. All methods gave reliable and identical results except Chelex, even after a trial to optimize this crude process using proteinase K. In the third uveal melanoma, the shearing may have exceeded a certain degree, which was reflected in the absence of band profiles over 1.6 kb.

Interesting aspects were gained with the digestion of the template prior to RAPD analysis.20 Treatment with restriction enzymes had an impact on RAPD profiles for some, but not all, primers. In some cases, larger products were formed from restricted than from unrestricted template. Frequently, but not always, the treatment simplified the profiles. Simple loss of those sequences that would normally be amplified, as a result of encompassing a restriction site, was not observed. In conclusion, it appears that primer annealing is more efficient along shorter DNA fragments, where a simplified secondary DNA structure is less likely to interfere with the process. Weining and Langridge21 suggested that template digestion facilitates more effective denaturation and thereby allows better access for the polymerases.

Random priming has been used successfully to assess the degree of genomic heterogeneity of tumors. Jotwani et al.22 isolated DNA from several sectors of a pituitary adenoma and identified alterations at specific 9p loci and in RAPD profiles in only one of them, demonstrating the localized nature of genomic instability in a morphologically relatively homogenous tumor. Taking into account that adenoma is a benign tumor and that we examined the DNA from only 1 sector of the breast cancer and uveal melanoma tumors, it is highly unlikely that other sectors would produce RAPD fingerprints identical to those of the leukocyte DNA. Sirivatanauksorn et al.23 offer some hints in this direction, using RAPD to investigate clonal diversity, which is seen with progression in neoplasms. Their studies suggest that as the hepatocellular carcinoma nodules expand beyond 6 mm in diameter, the single initiating clone evolves into multiple distinct derivative clones, which can be recognized by their DNA fingerprints. We assume that experiments with other cancer types will have similar results.

In our experiments, we compared breast cancer, one of the most studied cancer types, with uveal melanoma, our understanding of which has made great progress in the last decade. We have chosen these tumor entities to confirm by RAPD analysis the higher number of genetic aberrations seen in breast cancer compared to uveal melanoma during CGH experiments.24, 25, 26, 27 Uveal melanoma is the most common primary intraocular tumor, with an annual incidence of 6 per 1 million.28 Although the genes responsible for this cancer type are unknown, cytogenetic analyses and CGH have revealed abnormalities of chromosomes 3, 6 and 8. Loss of chromosome 3, overrepresentation of 6p, loss of 6q, amplification of 8q and overrepresentation of chromosome 8 are the most characteristic.24, 29, 30 Emphasis has been given lately to the monosomy of chromosome 3 since it has been correlated with metastasis and poor prognosis.31, 32 The finding that this monosomy is the most frequent aberration suggests the presence of tumor-suppressor genes on this chromosome. Cytogenetic studies and microsatellite analysis were used to restrict the candidate regions for these genes, suggesting that loci 3q24-q26 and 3p25 might harbor them.33

If we examine the existing methods of mutation detection, we find that every method has both advantages and disadvantages. Loss of heterozygosity, e.g., cannot identify moderate gains of genetic material; CGH can detect only alterations of relatively large chromosomal regions, with the limit currently at 5–10 Mb;34 and representation differential analysis can identify only homozygous aberrations and requires separate experiments for detection of DNA gains and losses.35 Although the random priming approach is not perfect, e.g., it cannot detect loss of an allele followed by duplication of the remaining allele,3 it provides several advantages over the aforementioned methods. First, it is the only method that permits the cloning, in a single step, of DNA sequences that have undergone the 2 most common alterations in the cancer cell genome: loss of heterozygosity and gain of extra gene sequences.36 The differences in RAPD fingerprints arise from nucleotide substitutions that create or abolish primer sites and from either deletion, insertion or inversion of a priming site or of a fragment between priming sites.37

If we consider that the number of differences in all tumor–blood pairs was maximal at 6 (with an average of 3) and that the average number of the generated bands was 25, the genomic damage fraction (GDF) index,38 would be ≤0.2 (with an average of 0.1). Interestingly, the number of changes in uveal melanoma tumor–blood pairs was about the same or slightly smaller than that in breast cancer pairs (e.g., we did not observe any uveal melanoma pair displaying 8 or 10 differences).

Primers were chosen based on consensus sequences involved in DNA alterations. Taking into account the similar GDF indices and the 18–45% discrepancy in the detection of differences between breast cancer and uveal melanoma pairs, our experiments suggest that, despite the small number of samples, the grade of genomic instability in breast cancer is much higher than that of uveal melanoma.

Here, it would be useful to examine the results of Singh and Roy,39 who used RAPD analysis on breast cancer samples. They used the Stoffel fragment, a recombinant polymerase that lacks the N-terminal 289 amino acid portion of the full-length Taq polymerase, which is thought to facilitate amplification of the low m.w. bands and indeed the RAPD profiles ranged from 1 and 1.5 kb to 300 bp and consisted of an average of 6 bands (30 primers led to amplification of 190 bands). The authors observed that 34.2% of the fragments showed differences and we conclude that in the most cases (82%) 1 or 2 bands were changed per pair and primer, resulting in a GDF index of 0.1–0.3. Although the range of the band arrays is about half and the number of bands per primer is only one-fourth of what we achieved, GDF values are about the same. Despite its use, we believe that until standardization of RAPD is achieved the GDF index will not be widely applied. The GDF index was first used in 1997 and has not gained ground yet.

The ability of random priming to detect moderate gains of chromosomal fragments, which cannot be identified by restriction fragment length polymorphism and microsatellite allelotyping, underlines the potential of the method in tumor studies.3 Further, specific bands could be used as markers once correlation with phenotypes and other clinical parameters is attained. Other groups have demonstrated that the degree of genomic damage assessed by arbitrarily primed PCR DNA fingerprinting correlates with genotypic, phenotypic and clinical variables in colorectal cancer and may be useful in assessing prognosis in that cancer type.38

In contrast, use of traditional PCR amplification of multiple microsatellite markers to detect microsatellite instability is labor-intensive and may be marker-dependent. Although microsatellite instability has been detected in small fractions in some cancer types, e.g., lung cancer40, 41 and several forms of leukemia,42, 43 reports of the microsatellite instability phenotype in breast cancer are inconsistent. Microsatellite instability in breast cancer appears to be an early event in carcinogenesis. Its incidence varies between 0% and 28%.44, 45, 46, 47, 48, 49 One group even observed microsatellite alterations in all 13 breast tumors studied.50 There is no association between microsatellite instability and patient age, histotype, histologic grade, tumor size or lymph nodes. Only one group has found an association between microsatellite instability and the absence of both estrogen and progesterone receptors.51

The high genomic instability detected in breast cancer by RAPD analysis is in line with results on other tumors.52, 53 In ovarian cancer, the incidence of abnormal arbitrarily primed PCR patterns was 53% compared to a microsatellite instability incidence of 37%. In addition, high-grade lesions and family history of cancer were associated with abnormal randomly primed DNA fingerprints. In lung cancer, the ability of arbitrarily chosen primers to detect genomic instability had a sensitivity ranging from 15–75%. In azoxymethane-induced F344 rat colon tumors, K-ras and microsatellite instability were often not involved in the carcinogenesis, though genomic instability was always present and detectable by RAPD analysis.54 These findings prove that random priming achieves broader coverage of the genome than microsatellite analysis with fewer reactions, facilitates screening and increases sensitivity in identifying genomic instability.

Most groups working with RAPD are choosing their primers randomly, e.g., by simply purchasing a set of predesigned primers by Operon (Alameda, CA).39 Since we were by definition free to choose our primers, we designed oligonucleotides that included sequences engineered to amplify a predefined motif. We analyzed the possible role of the chosen motifs/sequences in mechanisms of mutations, such as chromosomal translocations and deletions, which have an impact in tumorigenesis. We examined how informative the chosen primers were, i.e., if the generated patterns could reveal differences between the fingerprints of breast tumor DNA and its normal counterparts.

Genomic instability in cancer cells could arise by several mechanisms, involving either trans- or cis-acting factors. Cis-acting DNA sequences that promote rearrangements in the surrounding DNA could also play an important role in genomic instability and cancer.55 Deletions of tumor-suppressor genes and amplifications and altered expression of proto-oncogenes are caused by mitotic nonhomologous recombinations, i.e., the exchange of nonhomologous DNA regions.56, 57 This gene destabilization might be explained by preexisting recombination hot spots or by de novo hot spots arising from chromosomal exchanges.58, 59 Specific DNA sequences enhance the rate of recombination in the genomes of many different organisms.

The first motif was the “cross-over hot spot instigator” (chi) sequence (primers CHI and STARK).60, 61 DNA sequence analysis of the major breakpoint cluster region of Philadelphia chromosome in chronic myelogenous leukemia has revealed chi-like octamers.62, 63 Chi-like octamers were also identified in the bcl-2 major (mbr) and minor (mcr) breakpoint regions as well as their IgH reciprocal counterparts, which constitute the chromosomal translocation t(14;18)(q32;q31), the pathogenetic hallmark of follicular lymphoma.64

Other sequences that favor chromosomal recombinations are repetitive ones,11 like the SINEs65, 66, 67, 68, 69 and the long interspersed nucleic elements (LINEs).70, 71, 72, 73 The major human SINE is the Alu DNA sequence family (primer R12A/267 and CORE)74, 75 and one of the major LINEs is L1 (primer L1HS, 5′-ATGTAACAAACC-3′).76

The genomes of higher primates are able to duplicate or transpose gene-rich genomic segments into the pericentromeric and telomeric regions of various chromosomes.77 Interspersed repeated sequences, such as β satellites, found at the junctions of the duplications indicate that they play a functional role in mediating the interchromosomal transfer of genetic material. A 9 bp alphoid-derived direct repeat, found at the junction of α satellite and classical satellite 3 DNA in various chromosomes, was used in our RAPD experiments as the basis for the primer C4/Pja.78

The variable number of tandem repeats or minisatellites and microsatellites might also be sequences that function as recombination hot spots by serving, e.g., as a resolution point for Holliday junctions.59 Nuernberg et al.79 demonstrated highly amplified fragments in the fingerprints of several intracranial tumors in hybridization experiments using (GTG)5 and (GT)8 oligonucleotide probes and localized them in the EGFR gene (primers EP2, EP5, EP6 and EP7).

Telomer-like sequences (telomer-like repeats) have also been observed in interstitial chromosomal sites and involve head-to-head arrangements of telomere sequences, i.e., 5′-(TTAGGG)n-(CCCTAA)n-3′ (primer TELO3).80, 81 Hastie and Allshire82 suggest that telomer-like repeats are hot spots for recombination, breakage and fragility.

By studying deletion breakpoint junction regions, Krawczak and Cooper7 found the consensus sequence 5′-TGA/GA/GG/TA/C-3′ (primer DH 5′-GGTCTGAAGA-3′) in deletion hot spots of 5 human genes (AT3, F8, HBA, HBB and HPRT) and in almost half of 60 sporadic deletions of diverse genes.

A 5 kb deletion in mitochondrial genomes was found in patients with Kearns-Sayre syndrome and progressive external ophthalmoplegia. This common deletion flanked by a perfect 13 bp direct repeat (primer MHS 5′-ACCTCCCTCACCA-3′) suggests that homologous recombination also occurs in mammalian mitochondrial genomes and is at least one cause of the deletions found in these 2 related mitochondrial myopathies.83

Other primers were devised based on the splice junction consensus sequences,84 both the 3′ splice acceptor and 5′ splice donor sites (primers ACR3 and DCR4) and 1 based on the TATA box (primer TATA).

Stone and Wharton85 used a modification of an RNA fingerprinting method that readily identifies cDNA fragments expressed differentially and enriched for members of a specific gene family. They successfully isolated cDNAs encoding a protein kinase and a gene that contained zinc finger domains (primer Zn1 and Zn2 5′-GCGAAGACATAC-3′) using both arbitrary primers and primers specific for a preserved region of a gene family.

Our approach relied on the sequences having a sufficient proximity, the appropriate orientation and adequate homology to our chosen primers to support amplification of the encompassed DNA. Although the mechanism underlying RAPD is not fully understood, sequencing of RAPD fragments revealed that multiple sites in the genome are flanked by inverted repeats, which may be perfect or imperfect and which permit multiple mismatches to occur, paving the way for exponential amplification of the encompassing fragment.10 We assume that when primer pairs like DH and R12A/267 or Zn and R12A/267 were used, the targeted primer was able to identify a specific target sequence and the other primer acted as a nonspecific “walking” primer. By attaching to ubiquitous genomic sequences, this nonspecific walking primer could support the first primer in generating fragments within a 5 kb size range, a strategy that was first applied in targeted gene walking PCR.86 Although we did not demonstrate that the generated PCR products are within genes or that the PCR products generated, e.g., with the TATA primer indeed encompass TATA boxes, other groups have shown that PCR with degenerate primers that encode conserved amino acid sequences can be used to screen animal genomes reliably for likely gene family members.87 In this way, we think that we increased the possibility of amplifying these regions.

Nevertheless, our results demonstrate that RAPD can measure generalized genomic damage in cancer cells, thus providing a molecular approach to cancer cytogenetics.