There is considerable debate over the role of genomic instability as a cause or consequence of tumorigenesis. By the time a tumor is clinically manifest, it has survived a checkered history of multiple rounds of cell division, cell death, clonal expansion and selective growth of some surviving clones. The altered nature of the tumor cells, and indirectly, the natural history of the tumor would be reflected in the magnitude of the overall genetic changes in a tumor. In this work RAPD-PCR of tumors and its comparison with the normal DNA has been used as a measure of genomic instability.
Genomic instability in the form of chromosome instability, aneuploidy, LOH, microsatellite repeat alterations are seen in almost all types of neoplastic and preneoplastic cells. All these mutations have different implications in the tumor development and progression. However the cause of genomic instability is still debated. Some authors find aneuploidy as the primary cause of the genomic instability in the tumors [23
]. Ninomiya et al [25
] have found chromosomal instability and LOH to play role in the development of lung cancer. Goel et al [26
] have characterized the role of chromosomal and microsatellite instability in the progression of colon cancer. In other systems, if one restricts to microsatellite repeat alterations, there has been some evidence linking genomic instability to tumor progression [27
]. Microsatellite instability was found to be associated with aggressive colorectal carcinoma [29
] and ovarian carcinoma [30
]. However in a broader sense, the association of generalized genomic instability with tumor progression would be somewhat in contradiction with the mutator hypothesis, in which the mutations are phenotypically neutral with no direct role in the progression of the tumors. A mutator phenotype might be a liability to an aggressive rapidly dividing clone arising during tumor progression as extensive mutations might adversely affect the growth properties of these cells [2
]. This apparent contradiction can be resolved if it is kept in mind that the assays for genomic instability measure different phenomena, with varying biological implications. For example, in colorectal tumors increased frequency of point mutations is associated with a stable karyotype [15
], which is less likely to be associated with progression.
There have been several attempts to estimate the extent of alterations in tumor genomes. Ionov et al [20
] used radioactive Arbitrarily Primed PCR (AP-PCR) and PAGE to study genomic mutations in colorectal carcinoma, documenting and characterizing bands altered in tumors. AP-PCR is similar to RAPD except the primers used are longer. They resolved PCR product on PAGE with a higher resolution for band sizes up to 1 kb and observed that altered bands in tumors were identical to the closest normal band except for deletions in repeats like poly A, dinucleotides like CT and CA or trinucleotide repeats. Their estimation of changes in the genome was about 1.3–1.5 per 105
bases in cancer tissue. This is very different from what we see. In our study, most of the altered bands do not have any homology with any other bands in the same RAPD profile as confirmed by Southern hybridization of the altered bands to the RAPD profile (Figure ).
Jackson and Loeb [31
] estimated ~100,000 genome wide alterations in cancers in the form of base substitutions, deletions, chromosomal translocations, and gene amplifications, and these mutations are found to accumulate as tumors progress. They proposed that the genomic instability caused by these alterations result from a mutator phenotype, which occurs early during tumor development and predisposes the tumor cell to the accumulation of further mutations. They [32
] emphasized the importance of considering different types of alterations in genome of malignant cells, like aneuploidy, translocations and amplifications along with those detected by molecular methods like analysis of loss of heterozygosity, microsatellite instability etc. Similarly Kerangueven et al [33
] have found extensive genetic diversity from a genome wide scan for loss of heterozygosity. Accumulation of simple random point mutations can lead to carcinogenesis [34
]. To predict the relative contribution of mutator mutants in cancer, Beckman and Loeb [35
] have developed a mathematical model that quantitatively determines the efficiency of carcinogenesis with and without mutator mutations. Recently Beilas et al [36
] have measured random mutations in normal and neoplastic human tissues and have found an elevation in random mutations of at least two orders of magnitude in neoplastic tissues.
There are studies using inter-simple sequence repeat PCR (ISSR-PCR), a DNA fingerprinting method to identify genomic instability in the tumors. Basik et al [37
] found that the extent of alterations detected in colorectal cancer using ISSR-PCR had no relationship with the tumor stage at diagnosis and microsatellite instability, though it was associated with loss of heterozygosity. The number of genomic alterations in colorectal cancer cells was more than expected and mean number of alterations per cell was ~11,000 [38
]. The lack of association of this phenomenon with tumor grade and observation of similar number of events in colonic polyps and colon carcinoma supports the theory that genomic instability occurring early during malignant change. Stoler et al [39
] using ISSR-PCR found genomic instability in invasive breast carcinoma. The mean instability index was 3.08%, which was similar to the mean value observed in colorectal and thyroid carcinoma studied by the same group. Using ISSR-PCR Rai et al [40
] showed genome wide instability with the mean instability index of 12% in chewing-tobacco associated oral cancer. However, ISSR-PCR measures only a specific type of regions in genome, between two defined repeats located end to end within a certain distance from each other. The RAPD analysis used in our study targets genomic loci of a different nature and we estimate that the number of overall alterations in tumor genome seems to be greater than earlier appreciated. While the estimated number of alterations in our study is different from that observed using ISSR-PCR, the observation of absence of increased genomic alteration by Basik et al [37
] in higher grade tumors is similar to our finding. Both ISSR-PCR and RAPD-PCR are techniques for measuring genomic changes that are likely to be phenotypically neutral.
Tomlinson et al [41
] proposed that normal mutation rate is sufficient to explain a large number of mutations in tumors if the number of cell divisions during normal development is taken into account. We were not able to study normal brain tissue as control, and used normal leucocyte DNA. We checked the stability of the RAPD analysis by comparing the RAPD profile of buccal mucosal DNA with the leucocyte DNA from the same volunteer. Also the leucocyte DNA from individuals was obtained in intervals of 4–6 months over a period of two years, and RAPD profile of each sample were identical for the same person. The identical RAPD profile from buccal mucosal and leucocyte DNA, along with stability of leucocyte DNA profile over a period of 2 years, during which there would be a lot of cellular turnover, indicates that the alterations described in this study do not represent alterations accompanying cell division that might occur during normal development. These results, along with the universal forensic practice of taking the DNA fingerprinting profile of leucocyte DNA as representative of normal DNA for comparison with DNA of diverse tissue origins, body fluids etc. justifies the use of blood as constitutional DNA in our study.
Most of the published work till date on the extent of mutations in tumors is on colon carcinogenesis. Our study is on primary human glial tumors of astrocytic origin and we have taken DA and GBM as representative of low and high grade tumors of same cell origin. The mean estimate of alterations observed in tumor genomes varied between 4.42% in DA and 2.29% in GBM by method I and 11.96% (DA) and 6.03% (GBM) by method II. The difference between DA and GBM by both the methods was statistically significant. The increased extent of alterations occurring in tumors of a lower grade differs from the results obtained by some studies using AP-PCR [29
]. Both RAPD and AP-PCR are basically similar techniques, relying on a single primer to identify sites in both directions on double stranded DNA to prime a PCR reaction. However, there are basic differences in the methods employed by us and by the other researchers, which seems to have resulted in an altogether different form of clonal mutations identified in our study. In previous studies using AP-PCR followed by PAGE, the altered bands (as determined by sequencing) were found to differ from the closest normal counterpart only by small runs of simple repeat sequences. In our case the altered bands did not have any similarity with the normal bands on the RAPD profile. Five of the altered bands, characterized in detail, did not hybridize with any other band in the RAPD profile of either tumor or paired normal DNA (Figure ). Sequencing showed no microsatellite repeats in these altered fragments. These unique sequences were homologous to some recombinogenic sequences like HERV-K [22
], L1 repeat sequence and Chi
-core sequences (manuscript in preparation) and had the potential to mediate alterations in the genome by repeat mediated recombination and other mechanisms. Thus the changes identified by us are not due to microsatellite repeats being deleted from or added to the nearest normal band. The different nature of changes identified by us could be because of use of agarose gel to separate the PCR amplified bands. Agarose gel has a lower resolution range than PAGE, and while they have been extensively used in RAPD analysis, they do not resolve changes of few base pairs caused by microsatellite expansion and contraction. Hence, the changes detected by us were qualitatively different from the earlier studies. The size of bands studied by us is also different, up to 2.5–3 kb as opposed to much smaller bands studied by other groups. We have also used 10 mer primers, which would be expected to prime more promiscuously than longer primers used by others.
Our observation of a higher number of genetic changes in tumors of lower grade could be a consequence of an increased mutation rate in early tumorigenesis due to acquisition of a mutator phenotype. It is assumed that in absence of selection pressure the number of mutations would be more. However, the mutator phenotype would be at a disadvantage to a more aggressive rapidly growing clone [2
] as would be expected in a GBM, especially in a primary GBM. The higher number of mutations could also be a reflection of the longer time taken by a low grade tumor to clinically manifest, compared to a higher grade tumor. Mutations could accumulate in non-dividing cells [42
], thus making tumors with low proliferation indices e.g., grade II (DA) tumors accumulating more mutations than grade IV (GBM) tumors. Another factor could be the higher rate of apoptosis in low grade tumors, which would lead to cell populations in DA accumulating more clonal mutations, both as a function of time as well of cell divisions, before the tumor is manifested clinically.
We have earlier demonstrated that RAPD analysis reveals increased intra-tumor genetic heterogeneity in DA compared to GBM [7
]. When DNA was independently extracted from consecutive segments of 50 sections each, more differences between the segments of the same tumor were detected in DA compared to GBM. This result was also in concordance with the acquisition of a mutator phenotype early in tumorigenesis, leading to the co-existence of progeny cells with different DNA fingerprinting patterns, especially in low-grade tumors. It would be expected that heterogeneity might give the impression of increased change in a large tumor. In order to exclude such a possibility, we used DNA made from only one segment of 50 consecutive sections in each tumor used in our study. Similarly another study by our group [43
] has observed very significantly higher level of loss of heterozygosity (LOH) in the hMLH1 gene locus in DA compared to GBM using microsatellite markers, correlating with our earlier result which showed more intra tumor genetic heterogeneity in low grade gliomas.
The uniqueness of our study lies in the high estimated number of mutations. Also the estimated number of clonal mutations is greater than earlier reported, and not just restricted to microsatellite repeat sequences length. The extent of such changes is more in the low grade tumor type (DA) than in high grade GBM. Our evidence is in favor of dissociation of the dynamics of clonal mutations that are phenotypically neutral from those genetic changes leading to a phenotype with growth advantage. This is in concordance with the applicability of the mutator hypothesis. A further study comparing primary and secondary GBM would provide more evidence for its applicability in glial tumors. Similar approaches can also be used to study the role of mutations that do not have phenotypic consequences in other tumor types as well.