Dr Steven C. Pruitt, Department of Molecular and Cellular Biology, Roswell Park Cancer Institute, Buffalo, NY 14263, USA. Tel.: +1 716 845 3589; fax: +1 716 845 8169; e-mail: email@example.com
Genomic instability within somatic stem cells may lead to the accumulation of mutations and contribute to cancer or other age-related phenotypes. However, determining the frequency of mutations that differ among individual stem cells is difficult from whole tissue samples because each event is diluted in the total population of both stem cells and differentiated tissue. Here the ability to expand neural stem/progenitor cells clonally permitted measurement of genomic alterations derived from a single initial cell. C57Bl/6 × DBA/2 hybrid mice were used and PCR analysis with strain-specific primers was performed to detect loss of heterozygosity on nine different chromosomes for each neurosphere. The frequency with which changes occurred in neurospheres derived from 2-month- and 2-year-old mice was compared. In 15 neurospheres derived from young animals both parental chromosomes were present for all nine chromosome pairs. In contrast, 16/17 neurospheres from old animals demonstrated loss of heterozygosity (LOH) on one or more chromosomes and seven exhibited a complete deletion of at least one chromosomal region. For chromosomes 9 and 19 there is a significant bias in the allele that is lost where in each case the C57Bl/6 allele is retained in 6/6 neurospheres exhibiting LOH. These data suggest that aging leads to a substantial mutational load within the neural stem cell compartment which can be expected to affect the normal function of these cells. Furthermore, the retention of specific alleles for chromosomes 9 and 19 suggests that a subset of mutational events lead to an allele-specific survival advantage within the neural stem cell compartment.
The possibility that age-related changes in the genome contribute to aging phenotypes, including malignancy and other aspects of tissue dysfunction, has been recognized previously (e.g. Boveri, 1914; Failla, 1958). Several indirect lines of evidence support a link between tumorigeneis, aging and a common mechanism involving genome instability. For example, progeroid syndromes in humans result in elevated levels of cancer and several (Blm, RTS and Wrn) have been shown to result from defective recQ class helicases, which are known to function in genome maintenance (Mohaghegh & Hickson, 2001; Nakayama, 2002). A variety of direct experimental approaches have been employed to measure the extent to which various types of genomic alterations occur in aging tissues. These include cytogenetic analysis (Curtis & Crowley, 1963), the use of selectable marker genes including HPRT (Martin et al., 1996) and APRT (Shao et al., 1999) or transgenic models with integrated plasmid vectors containing reporter genes (Boerrigter et al., 1995; Vijg et al., 1997). Each of these types of studies has shown that mutations occur in vivo and accumulate as a function of age. However, although it has been difficult to extrapolate from these studies to establish the genome-wide mutational load carried in aging tissues, it has been argued that, in the absence of a mutator phenotype, the level of point mutations in particular is not sufficient to account for the observed tumour frequencies (e.g. Schneider & Kulesz-Martin, 2004) and by extension other age-related phenotypes.
Vertebrates, and particularly mammals, have evolved cellular strategies to allow tissue replacement while minimizing the accumulation of genetic damage. Specifically, all epithelial tissues and, with only a few exceptions, other tissue types, utilize a tissue replacement strategy in which an asymmetrically, but slowly, dividing stem cell gives rise to a replacement stem cell and a rapidly dividing proliferative progenitor (or transient amplifying cell). Divisions of the proliferative progenitor are symmetric in that all of the progeny are programmed to differentiate and exit the cell cycle within a few divisions. Thus, for many tissues the somatic stem cell is the only population of cells that maintain the ability to proliferate over the lifetime of the organism. Consequently, the assumption has been made that the site of accumulation of at least some of the mutations that ultimately lead to a tumour is the stem cell. A similar rationale is applicable to the accumulation of mutations leading to tissue dysfunction during aging. However, a direct test of the level to which mutations accumulate in stem cells during aging has been difficult to obtain as each stem cell would be expected to carry a different set of mutations that would not be detected in measurements using total tissue samples.
Here we have taken advantage of the ability to expand neural stem cells or progenitor cells derived from the mouse brain in the form of neurospheres (Reynolds & Weiss, 1992; Doetsch et al., 2002), following seeding at clonal density, to survey the frequency with which deletions or loss of heterozygosity (LOH) occur in young and old mice. Neural stem cells were derived from 2-month- or 2-year-old hybrid mice (C57Bl/6 × DBA/2), and strain-specific primers, based on known single nucleotide polymorphisms, were designed and used to assess the presence or absence of C57Bl/6 or DBA/2 alleles in individual neurospheres using a PCR-based method.
Stem cell populations exist within both the subventricular zone (SVZ) of the lateral ventricles (Doetsch et al., 1997) and the dentate gyrus of the mammalian brain (Kuhn et al., 1996) and contribute to neurogenesis in the adult. Neurogenesis at each of these sites declines as a function of age such that 2-year-old animals exhibit approximately half the level observed in 2-month-old animals (Kuhn et al., 1996; Tropepe et al., 1997). Within the SVZ, this decline has been shown to result from a proportional loss of neural stem cells using both in situ detection and clonal growth of neural stem cells (NSCs) in the form of neurospheres from tissue explants (Maslov et al., 2004). One hypothesis to account for the loss of neural stem cells within the SVZ is that these cells have suffered genomic damage leading to cell cycle arrest and/or apoptosis.
Not all genomic changes are anticipated to result in cell death. Hence, if the hypothesis that genomic damage is responsible for the loss of NSCs in old mice is correct, evidence for genomic changes should be present in surviving NSCs. To survey progeny from individual neural stem cells, neurospheres were cultured from the brains of C57Bl/6 × DBA/2 hybrid mice in microtitre plates. A seeding condition that results in neurosphere growth in fewer than one-third of the wells was utilized. Cultures were established from the brains of three 2-month-old and three 2-year-old hybrid mice. Although the absolute number of neurospheres was reduced, similar to previous observations using C57Bl/6 mice (Maslov et al., 2004), neurosphere recovery from the older animals was approximately half that from the younger animals (young = 27 ± 5.1, old = 12 ± 4.4; Fig. 1). Despite the reduction in number, neurospheres derived from older animals were qualitatively similar to those from younger animals in both size and morphology.
To detect deletions or LOH, primers were designed to amplify a region of DNA containing a known single nucleotide polymorphism (SNP) between C57Bl/6 and DBA/2. Two sets of primers were designed for each SNP. In each primer set the 3′ nucleotide of one primer reflects the polymorphism present in the relevant strain, thus allowing strain-specific amplification from genomic DNA. This is possible since Taq polymerase will not extend a 3′ mismatch and, in addition, this enzyme does not have proof-reading activity. Nine primer sets showing specificity for C57Bl/6 or DBA/2 were established, as shown in Fig. 2. PCR results for each of the primer sets on each template DNA are shown in Fig. 3(A). C57Bl/6 and DBA/2 tail DNA samples were used as external controls and included with every PCR run. In addition, each sample contained an internal control. The internal control amplified a region from chromosome 6 on both C57Bl/6 and DBA/2 templates and resulted in a product of 722 bp. Each sample DNA was amplified with both C57Bl/6 and DBA/2 SNP-specific primers. Representative results, using chromosome 12 primers, from neurospheres derived from four young and four old mice are shown in Fig. 3B. Both alleles are present in all four samples from the young animal, whereas the C57BL/6 allele is missing in 3/4 samples and the DBA/2 allele is missing in 1/4 samples from the old animal. Similar experiments were performed, in duplicate, for each primer set using DNA from each neurosphere (a total of 756 assays plus controls).
Comparison of genetic changes in NSCs derived from young vs. old mice
Fifteen samples obtained from the 2-month-old mice, and 17 samples derived from the 2-year-old mice, were analysed for both alleles across nine chromosomes using SNP-specific PCR. The results of the SNP-specific PCR reactions are summarized in Fig. 4. No single instance of either LOH or a full deletion was seen on any of the nine chromosomes assessed in any of the samples derived from young animals. In contrast, all but one sample (no. 11) from the old mice showed at least one instance of LOH across the nine chromosomes assayed, and in most instances displayed multiple regions of LOH. Most samples showed events on three or more chromosomes.
The PCR assay also allows detection of full deletion events. Seven of 17 samples (nos. 4, 6–9, 12, 13) derived from the old animals showed at least one full deletion event, and four carried deletions on two or more chromosomes. A summary of the deletion events by chromosome is depicted in Fig. 5.
In cases where the SNP-specific primers indicated that neither the C57Bl/6 nor the DBA/2 allele was present, primers specific to a second region of the chromosome were utilized to determine whether both chromosomes of the homology pair were entirely deleted, or if a portion of at least one of each chromosome pair was present (Fig. 2). Unlike the SNP-specific primers, the proximal primers were not allele-specific. All but three of the samples that exhibited a deletion event with the SNP-specific primers generated a product using the proximal primers, indicating the presence of at least one allele. No product was amplified with the proximal primers for chromosomes 7 and 9 from sample 7 (Fig. 4).
Figure 6 depicts the data on strain specificity for instances of LOH in the neurospheres derived from the 2-year-old mice. The allele that was retained is displayed in the graph and the events are broken down by chromosome. Chromosomes 9 and 19 retained the C57BL/6 alleles in six out of six events. In each case, the probability of this occurring by chance is less than 5%, suggesting the possibility that allele-specific somatic selection occurs for one or more genes located on each of these chromosomes.
The potential roles of somatic mutation in both carcinogenesis and age-related tissue dysfunction have been recognized. However, previous attempts to measure directly the rates of mutation within somatic cells and tissues, or the overall levels to which mutations accumulate, have suggested that the number of point mutations within most tissues is insufficient to generate the 4–10 somatic mutations required for observed rates of carcinogenesis. The number of mutations required for disruption of cellular function leading to age-related tissue dysfunction has not been estimated, but could be even higher.
One method utilized to estimate gene mutation rates relies on selectable endogenous target genes. For example, selection for HPRT expression in T-cells recovered from young and old humans (King et al., 1994) and mice (Inamizu et al., 1986) demonstrates the presence of cells carrying HPRT mutations and shows that the number of such cells increases with age. Although HPRT mutant cells were detected at frequencies as high as ∼3 × 10−5 in older mice, if this rate is representative of the gene mutation rate across the genome, the total number of mutations would be negligible in relation to the frequency of either cancer or aging phenotypes.
It has been argued that mutation levels based on the HPRT test underestimate the mutational load due to selection against mutant cells (Deubel et al., 1996) or the possibility that the region of the genome at which HPRT is located is not representative. Alternative methods based on transgenic mice carrying lacI- (Stuart & Glickman, 2000; Stuart et al., 2000) or lacZ-expressing (Martus et al., 1995; Ono et al., 1995) bacteriophage lambda or plasmid-based recovery systems have also been used to estimate mutation frequencies. For the majority of transgenes used in such studies, the observed mutation frequency is near to the range defined using HPRT (Vijg, 2000). Exceptions include a line in which a lacZ-containing bacteriophage lambda vector was integrated near the pseudoautosomal region of the X-chromosome in which an approximately 100-fold higher mutation rate was observed, and a similar transgene located on chromosome 7 where an approximately 20-fold higher rate was observed (Gossen et al., 1991). These observations may indicate that specific regions of the genome undergo mutation at a higher frequency, although a contribution of the transgenic integration itself to instability cannot be eliminated. Mutation frequencies based on lamda phage or plasmid rescue experiments, as described above, require retention of the vector sequence and cannot detect large deletions or chromosomal loss.
In the present study we have surveyed genomic changes occurring on nine autosomes of the mouse genome in neural stem cells using a PCR-based approach. Results from these studies suggest that, whereas the mutational load in NSCs derived from 2-month-old animals is low, by 2 years of age virtually all NSCs carry multiple genomic changes in the form of deletions or loss of heterozygosity. Both the strong age dependency and the high frequency with which mutations are detected here differ from previously reported estimates. At least two aspects of the present approach could account for these differences, relative to previous studies. First, because there is no requirement for retention of vector sequences in the present study, large deletions or chromosomal losses are observable events. Second, bacteriophage or plasmid recovery-based techniques assay changes occurring within total tissue. Depending on the rate and proportion of tissue replacement, the mutation frequency observed in vector rescue experiments will reflect a composite of cells that were born at various times during the lifetime of the animal. Here, however, the genomic DNAs assessed are derived specifically from the neural stem cell/proliferating progenitor population following a brief expansion in vitro and are likely to reflect the mutation frequency within the mitotically active component of the brain at the time of isolation.
The spectrum of defects observed within individual neurospheres from old animals has several implications. First, the distribution of events between different neurospheres is dissimilar, suggesting that the events observed are independent. Second, deletion of both alleles was observed at least once for each of the chromosomal regions assayed, suggesting that essential genes are not located near these locations. However, in all but two of the instances where a deletion of both alleles occurred, an additional marker located more proximal to the centromere was present, suggesting that complete chromosomal loss is less frequent. Finally, the observation that for both chromosomes 9 and 19 the C57Bl/6 allele was retained in 6/6 cases in which LOH occurred suggests the possibility that selection for a C57Bl/6, or against a DBA/2, allele of one or more genes carried on these chromosomes is occurring in the NSC compartment. If mitotic recombination occurs at a sufficiently high frequency, this observation raises the possibility that genes affecting NSC survival could be localized through genetic mapping using clonally derived neurosphere samples from old individuals.
Both the age dependency and frequency with which genomic changes are found in neurospheres derived from old animals are consistent with a primary role for genomic instability in age-related carcinogenesis and tissue dysfunction. Indeed, the finding that only one of 17 (6%) of the neurospheres derived from 2-year-old mice exhibit a normal genotype for the nine chromosomes assayed here suggests that very few genotypically normal cells are present in the NSC compartment of these animals. It is likely that this degree of genomic rearrangement affects the function of these cells, and transgenic models which exhibit elevated rates of genomic instability support this possibility. For example, even young mice lacking methyl CpG binding protein 1 show elevated rates of genomic instability, which is correlated with deficits in neurogenesis and hippocampal function (Zhao et al., 2003).
The present findings may be relevant to the possibility that somatic neural stem cells can be exploited in therapies for neurodegenerative diseases. Although there are potential caveats to extrapolating these observations to humans, e.g. polymorphisms within essential genes could suppress the accumulation of cells exhibiting LOH in outbred populations, if similar levels of genomic rearrangement occur in human neural stem cells to those found here for the mouse it could limit the utility of these cells when taken from older individuals.
Three 2-month-old and three 20-month-old hybrid B6D2F1 (C57BL/6 × DBA/2) mice were obtained from the NIA Aged Rodent Colonies. All procedures involving animals were approved by the IACUC.
Animals were killed by cerebro-spinal dislocation. Brains were isolated and the tissue dissociated using 0.25% trypsin solution in DMEM. One-quarter of the recovered cells from each brain were seeded into two 96-well microtitre plates (non-tissue culture grade; ICN Biomedicals, Aurora, OH, USA) containing culture media. The culture medium was DMEM/F-12 (Invitrogen, Gaithersburg, MD, USA) supplemented with glutamine (2 mm; Invitrogen), gentamycin (50 µg mL−1; Sigma, St Louis, MO, USA), B27 supplement (Invitrogen), epidermal growth factor (EGF, 20 ng mL−1; Peprotech, Rocky Hill, NJ, USA), and fibroblast growth factor-2 (FGF-2, 20 ng mL−1; Sigma). The dissociated cells gave rise to individual neurospheres in less than one-fifth of the microtitre wells, consistent with a clonal origin of each neurosphere. Medium was changed on days 2, 4 and 7 after seeding. On day 9 wells were scored for the presence of neurospheres and were harvested on day 10 and lysed for genomic DNA isolation.
Multiple displacement amplification
Multiple displacement amplification (MDA) was performed using the GenomiPhi DNA Amplification Kit (Amersham Biosciences, Piscataway, NJ, USA). MDA reactions were performed with 1 ng genomic DNA in 1 µL, which was added to 9 µL of sample buffer (50 mm Tris/HCl pH 8.2, 0.5 mm EDTA containing random hexamer primers), and denatured at 95 °C for 3 min. One microlitre of 29 DNA polymerase mix including additional random hexamers was mixed on ice with 9 µL of reaction buffer containing dNTPs, and the mixture was added to the denatured sample. The MDA reaction was allowed to proceed for 18 h at 30 °C. The enzyme was deactivated by heating to 65 °C for 10 min. The success of the MDA reaction and the absence of product in negative control samples were assessed by agarose gel electrophoresis.
SNPs and primer design
Mouse SNPs specific to C57BL/6 and DBA/2 were chosen, from the Whitehead Institute/MIT Center for Genome Research database, for each chromosome. Locations of polymorphisms and PCR primers are given in Fig. 2. PCR primers were chosen so that the SNP occupied the 3′ most position of the primer. Primers were designed to have a Tm near 58 °C, a G + C content of 40–60%, and a length of 18–25 bases with a product size of 200–500 bases. All primers were purchased as custom synthesis products from Integrated DNA Technologies (Coralville, IA, USA).
Allele-specific PCR amplifications
All reactions were performed in a volume of 25 µL containing 10 ng total mouse genomic DNA template, 10× PCR reaction buffer, 2.5 mm MgCl2, 250 µm of each dNTP, 0.2 µm of each primer, and 1.5 U Platinum Taq DNA polymerase (Invitrogen). Amplification was performed using a Bio-Rad MyCycler with an initial denaturation at 95 °C for 2 min, followed by 30 cycles of denaturation at 95 °C for 30 s, primer annealing at 59 °C for 30 s and extension at 72 °C for 45 s. An internal control primer set was included with every PCR reaction. The internal control primers amplified a segment off a proximal region of chromosome 6 (Fig. 2). Tail DNA from C57BL/6 and DBA/2 mice, included with every PCR run to ensure that allele-specific amplification occurred, were used as external controls. The internal control was also included in the tail-specific control reactions. The ratio of internal control product and allele-specific product was used to assess the relative amounts of each product in the sample reactions. Lack of a PCR product with a set of allele-specific primers determined the absence of that specific allele segment. In cases where both alleles appeared to be deleted (i.e. no PCR product produced with either set of primers) a more proximal region of the corresponding chromosome was amplified to determine the extent of the deletion (Fig. 2).
Analysis of PCR products
An aliquot (10 µL) of each PCR product was loaded with 10× loading buffer onto a 1% agarose gel and run in 1× TAE buffer at 15 mA constant current. Amplification products were visualized by UV transillumination after staining with ethidium bromide. The agarose gels were photographed using ChemiImager software.
Single factor analysis of variance (anova) was used to determine the significance between the results obtained from young and old animals. Two-factor anova was employed to estimate the significance of the effect of chromosome number and strain on the resulting variance (presence or absence of tested alleles).
This work was supported by NIH grants AG020946 and AG019863 to S.C.P. and utilized animal facilities supported by an NCI CCSG grant.