Abstract
The mitochondrial genotype of heteroplasmic human cell lines containing the pathological np 3243 mtDNA mutation, plus or minus its suppressor at np 12300, has been followed over long periods in culture. Cell lines containing various different proportions of mutant mtDNA remained generally at a consistent, average heteroplasmy value over at least 30 wk of culture in nonselective media and exhibited minimal mitotic segregation, with a segregation number comparable with mtDNA copy number (≥1000). Growth in selective medium of cells at 99% np 3243 mutant mtDNA did, however, allow the isolation of clones with lower levels of the mutation, against a background of massive cell death. As a rare event, cell lines exhibited a sudden and dramatic diversification of heteroplasmy levels, accompanied by a shift in the average heteroplasmy level over a short period (<8 wk), indicating selection. One such episode was associated with a gain of chromosome 9. Analysis of respiratory phenotype and mitochondrial genotype of cell clones from such cultures revealed that stable heteroplasmy values were generally reestablished within a few weeks, in a reproducible but clone-specific fashion. This occurred independently of any straightforward phenotypic selection at the individual cell-clone level. Our findings are consistent with several alternate views of mtDNA organization in mammalian cells. One model that is supported by our data is that mtDNA is found in nucleoids containing many copies of the genome, which can themselves be heteroplasmic, and which are faithfully replicated. We interpret diversification and shifts of heteroplasmy level as resulting from a reorganization of such nucleoids, under nuclear genetic control. Abrupt remodeling of nucleoids in vivo would have major implications for understanding the developmental consequences of heteroplasmy, including mitochondrial disease phenotype and progression.
THE technique of cybridization to ρ0 cells that lack endogenous mtDNA has enabled the genetic analysis of human disease-associated, heteroplasmic mitochondrial mutations in a controlled nuclear background (King and Attardi 1989; Kinget al. 1992). One mutation intensively studied by this technique is the A to G point mutation at np 3243 (Gotoet al. 1990; Kobayashiet al. 1990), associated with several distinct pathological phenotypes (Van den Ouwelandet al.1992; Jean-Francoiset al. 1994; Mariottiet al. 1995), including ocular myopathy, diabetes plus deafness, and the MELAS syndrome (mitochondrial encephalopathy with lactic acidosis and stroke-like episodes). The mutation maps to a conserved residue in the gene for tRNA-leu(UUR) and is presumed to impair the metabolism of susceptible tissues by affecting mitochondrial protein synthesis.
Analysis of cybrid clones in two different nuclear backgrounds (the 143B osteosarcoma and A549 lung carcinoma cell lines) has demonstrated clearly the threshold effect of the np 3243 mutation on respiratory and biogenetic function (Chomynet al. 1992; Kinget al. 1992; Bentlage and Attardi 1996; Dunbaret al. 1996; El Mezianeet al. 1998a). Below 70% mutant mtDNA cells are apparently normal. Between 70 and 90% mutant mtDNA a mild complex I defect is evident. Above 90% mutant mtDNA there is a general impairment of respiration, but only at 98% or more mutant mtDNA is there a severe drop in mitochondrial protein synthesis. In lung carcinoma cybrids there is a drop in the steady-state level of tRNA-leu(UUR) proportionate to mutant load (El Mezianeet al. 1998a), and the mutant tRNA appears to be incapable of being aminoacylated (El Mezianeet al. 1998b), accounting for its effects on translation. Absolute mtDNA copy number also influences the exact phenotype (Bentlage and Attardi 1996), with cells at low copy number crossing these thresholds earliest.
This type of analysis has been made possible by the fact that cybridization results in the establishment of cell lines that remain stably heteroplasmic at various different levels of mutant mtDNA. However, not all cell clones arising from cybrid fusions exhibit a stable mitochondrial genotype. The direction in which unstable heteroplasmons segregate is dependent on nuclear background. In 143B osteosarcoma cybrids, segregation, if it occurs, is in the direction of an increasing level of mutant mtDNA (Yonedaet al. 1992; Dunbaret al. 1995; Bentlage and Attardi 1996). Conversely, in A549 lung carcinoma cells, the exact opposite is observed, with unstable heteroplasmons showing progressive loss of mutant (Dunbaret al. 1995). However, in both backgrounds the wide range of heteroplasmy values over which cybrid cell lines can remain stable for years indicates an effective absence of selection (Matthewset al. 1995; Shoubridge 1995). During cell culture 50,000–100,000 cells are typically sampled at each passage. Under these conditions, sampling error is extremely small, which accounts for the absence of significant drift in heteroplasmy values. A stable average heteroplasmy level could nevertheless mask diversification of mtDNA genotype by random mitotic segregation, as observed, for example, in the germ line of artificially heteroplasmic mice (Jenuthet al. 1996) and also in humans (Marchingtonet al. 1997). Earlier studies of 143B osteosarcoma cybrids heteroplasmic for the np 3243 mutation were consistent with such segregation (Shoubridge 1995).
Evidence that selection and/or drift are nevertheless possible in long-term cell culture comes from the observation that new mtDNA mutations can arise and become fixed. As regards the np 3243 mutation, probably the clearest example is the np 12300 suppressor mutation in the anticodon of tRNA-leu(CUN), which was discovered in a subculture of lung carcinoma cybrids carrying 99% np 3243 mutant mtDNA (El Mezianeet al. 1998a). Such cells, carrying the np 12300 mutation at the 11% level, are phenotypically restored to apparently wild-type levels of respiration, mitochondrial protein synthesis, and membrane assembly. The suppressor mutation is predicted to create a novel tRNA capable of decoding the codon group for which tRNA function is otherwise absent, due to the pathological mutation.
In an effort to determine the limits and parameters of mitotic segregation and to document episodes of selection and/or drift associated with transient instability of mitochondrial genotype, we carried out a longterm study of a panel of stably heteroplasmic lung carcinoma cybrids maintained in continuous culture. This study included cell lines carrying the np 3243 mutation at various levels as well as the subline carrying also the np 12300 suppressor mutation. The results are consistent with a norm of mitotic segregation, but with a high segregation number, much higher than the 200 inferred for the mouse germ line (Jenuthet al. 1996), and comparable with total mtDNA copy number in the cells. Intercellular selection can nevertheless be superimposed on this process. We were able to observe several episodes of mitochondrial genotypic instability, resulting in a concerted and rapid shift of mtDNA genotype. This almost certainly involved selection at the intercellular and perhaps also the intracellular level. These observations are consistent with several alternate views of mtDNA organization within mammalian cells, but lead us to propose a model in which faithfully replicated mitochondrial nucleoids are the unit of segregation, allowing the maintenance of stable, long-term heteroplasmy with very high apparent segregation numbers. The nucleoid model can also account for shifts in genotype if nucleoids are subject to transient destabilization, leading to an intracellular diversification of the mitochondrial genotype on which selection can act. We furthermore provide evidence that a nuclear genetic change, namely a gain of chromosome 9, is associated with one such episode that we have documented.
The frequency with which such events occur in vivo in different tissues could have major implications for the developmental evolution of heteroplasmy and for mitochondrial disease phenotype and progression. The nuclear genome may have an important role in the induction of such events.
MATERIALS AND METHODS
Computer modeling of mitotic segregation: Predicted changes with time in the distribution of mtDNA heteroplasmy values in the cell populations studied in the laboratory were derived using a customized computer program (available from K. M. I. Juhola). The program was based on a simple mathematical model (see appendix and results) modified from that used by Hoppensteadt and Peskin (1991) to predict the segregation of plasmids in bacteria.
Cells and cell culture: A549 lung carcinoma cybrids containing different proportions of np 3243 and np 12300 mutant mtDNAs have been described previously (Dunbaret al. 1995; El Mezianeet al. 1998a). Except where stated, cells were routinely maintained in Dulbecco's modified Eagle's medium, supplemented with 150 μg/ml uridine plus 10% fetal calf serum, with weekly passaging. Cell cloning at limiting dilution was carried out in 96-well plates. For testing in Gal medium, glucose in the medium was replaced by galactose at 0.9 mg/ml, with omission of the uridine supplement. For testing the effects of dimethyl sulfoxide (DMSO), cells were cultured prior to cloning either for 1 wk in normal medium containing 0.2% DMSO, or for 2 hr in 1% DMSO, followed by a week in normal medium.
Oxygen consumption: Whole-cell oxygen consumption was measured as described previously (Dunbaret al. 1996).
DNA extraction: For PCR-based mitochondrial genotyping, DNA was extracted from cultured cells as described by Reid et al. (1994, 1997). For Southern blot analyses of relative mtDNA copy number, EcoRI-digested genomic DNA was prepared from cells as follows. One confluent 10-ml plate (~5 × 106 cells) was trypsinized, and cells were recovered by centrifugation for 3 min at 2000 × gmax. Cell pellets were resuspended in 50 μl of water, 5 μl of 10 mg/ml boiled RNase A were added, and samples were incubated for 30 min at 37°. After addition of 10 μl of proteinase K solution (18 mg/ml) and 350 μl of 6 m guanidine hydrochloride, 0.5 m sodium acetate pH 5.5, incubation was continued overnight at 56°. Crude DNA was precipitated by the addition of 800 μl of ice-cold isopropanol, and samples were stored at 4° before further processing. Crude DNA was recovered by centrifugation for 30 min at 7000 × gmax, and pellets, washed twice with 70% ethanol, were digested three times successively with EcoRI as follows. Pellets were resuspended by the successive addition of 89 μl of water, 10 μl of 10× EcoRI digestion buffer (New England Biolabs, Beverly, MA), and 1 μl of EcoRI (20 units/μl; New England Biolabs), and were incubated overnight at 37°. In the first digestion cycle an additional 1 μl of the enzyme was added after 5 hr. Samples were then phenol-chloroform extracted without vortexing, centrifuged at 7000 × gmax to separate phases, and ethanol precipitated. The thrice-digested DNA was finally resuspended in 30 μl of water.
Mitochondrial genotyping by last hot-cycle PCR: The amount of heteroplasmy for the A3243G and G12300A mutations was measured in DNA samples extracted from cells using the materials and methods described in El Meziane et al. (1998a). Briefly, the regions around the mutation sites were PCR amplified using mtDNA-specific primer pairs, with addition of [α-32P]dCTP in the last synthesis cycle. Labeled PCR products were then digested with the restriction enzymes diagnostic for the site polymorphisms created by the mutations (ApaI and AflII, respectively), and products of replicate digestions were analyzed by polyacrylamide gel electrophoresis and phosphorimaging.
Mitochondrial staining: Cells were grown to high density on 60-mm plates, washed with phosphate-buffered saline (PBS), and then stained for 15 min at 37° in 1 ml of serum-free Optimem medium (GIBCO, Grand Island, NY) containing 10 μg/ml JC-1 (5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolylcarbocyanine iodide; Molecular Probes, Eugene, OR). The cells were washed twice with Optimem, then twice with PBS, mounted in PBS containing 5% sucrose, and examined within 20 min under highest magnification on an Olympus BX-50 fluorescence microscope, using a U-MWB filter.
Southern blot hybridization: To estimate relative mtDNA copy number, DNA prepared as above was fractionated on a 0.7% agarose gel, capillary blotted under standard conditions to MagnaCharge nylon membrane (Micron Separation Inc., Westborough, MA) and UV linked. Blots were hybridized overnight at 65° in 1 mm EDTA, 7% SDS, 0.5 m sodium phosphate buffer pH 7.2 (Church and Gilbert 1984) with radiolabeled probes for human mtDNA (a PCR product derived from a cloned segment of the COXII gene) and rDNA, as described by Spelbrink et al. (1997). The rDNA probe was a PCR product for the first 500 bp of the 18S rRNA gene, obtained by amplification from genomic DNA of a control A549 cybrid (line B) using primers 18S-F (TACCTGGTTGATCCTGCCAG) and 18S-R (TCGGGAGTGGGTAATTTGC). Amplification conditions were as for mtDNA genotyping (El Mezianeet al. 1998a), and probe fragments were purified using the QIAquick PCR purification kit (QIAGEN, Chatsworth, CA) and labeled using the Oligolabeling kit (Pharmacia Biotech, Uppsala, Sweden) in the presence of [α-32P]dCTP (Amersham; 3000 Ci/mmol). After hybridization, filters were washed successively for 20 min each at 65° in 3× SSC, 0.1% SDS and 0.3× SSC, 0.1% SDS and then analyzed by autoradiography and quantitated using a PhosphorImager SI (Molecular Dynamics, Uppsala, Sweden) with Image-Quant software.
Comparative genomic hybridization: DNA (proteinase K-extracted, stored frozen) was RNase treated and then deproteinized by a standard phenol-chloroform extraction. Aliquots of DNA (4 μg) were nick translated in the presence of either TexasRed-6-dUTP or FITCH-12-dUTP (DuPont, Wilmington, DE) to provide the respective samples for comparison. Comparative genomic hybridization (CGH) analysis was performed as described previously (Kallioniemiet al. 1994; Karhuet al. 1997).
RESULTS
Mitotic segregation with high segregation number is the norm for stable np 3243 heteroplasmons: The genotypes of three stable cybrid cell lines, heteroplasmic for the np 3243 mutation (A3243G) at different levels, were followed over a >30-wk period of continuous culture. The np 3243 genotype of the bulk culture and of individual clones obtained at limiting dilution was determined by ApaI site restriction fragment length polymorphism (RFLP) analysis of PCR products radiolabeled during the last synthesis cycle. Up to eight replicate measurements were performed on each DNA sample, enabling reliable estimates of the amount of mutant mtDNA generally to within 1%. The bulk cultures of all three cell lines, maintained by weekly passaging of a minimum of 50,000 cells in uridine-supplemented medium, remained throughout the experiment at their initial heteroplasmy levels of 85% (line Y), 95% (line X), and 99% (line G), indicating an absence of selection.
Because of the large number of cells sampled at each passage, the genotypic stability of the bulk cultures could mask a considerable degree of mtDNA segregation, as predicted by a random partition model and as observed previously in osteosarcoma cybrids with the same mutation (Matthewset al. 1995; Shoubridge 1995). A set of ≥20 individual cell clones, derived from each of lines X and Y after >30 wk of continuous culture, were therefore analyzed, revealing in both cases a narrow distribution of genotypes (Figure 1, Table 1). The mean heteroplasmy value for each set of clones was similar to that of the corresponding bulk culture, although slightly lower in each case, possibly indicating the loss of a few clones that would have been homoplasmic or very high mutant.
We compared the findings with the predictions of a random partition model (see appendix; K. M. I. Juhola, M. K. Juhola and H. T. Jacobs, unpublished results), in which it is assumed that each mtDNA molecule is replicated exactly once per cell cycle, that precisely 50% of the molecules are partitioned at random to each daughter at cell division, and that the fitness of cells of all possible mitochondrial genotypes is equal. Deviations from these assumptions in practice will tend to result in a greater or faster segregation toward homoplasmy than predicted by the model. We modeled outcomes based on several different numbers of independently segregating units: the 200 inferred for the mouse germ line by Jenuth et al. (1996), a lower number (100), and a higher number (1000), which approaches the copy number of mtDNA in typical somatic cells. For both line Y and line X the distribution of heteroplasmy levels, presented in bar chart form in Figure 1, was far from that predicted by the 200 segregating unit model. Homoplasmic cells were completely absent, despite the fact that they are predicted by random partition to be a major fraction of cells at 30 wk. Even if the absence of homoplasmic mutant cells is discounted as being due to negative selection, the overall distributions of genotype were much narrower than predicted for a model based on 100 or 200 segregating units and are consistent with segregation numbers of 1000 or more (compare, for example, the actual and predicted variances as shown in Table 1).
Heteroplasmy profiles of lines Y (85% mutant for np 3243) and X (95% mutant). (a and d) The observed profiles for lines Y and X, based on analysis of mtDNA genotypes of, respectively, 31 and 27 individual cell clones obtained from these lines after 30 wk of growth in culture. Bars representing the frequencies of clones in the different 10% intervals are plotted such that they include clones having values for the percentage of mutant mtDNA up to but not including the values on their right-hand margins (e.g., the 10–20% interval represents clones scored as having from 10% up to less than 20% mutant mtDNA, inclusive). Homoplasmic wild-type (hw) and mutant (hm) classes represent clones in which only one type of mtDNA was detectable. The unchanged values for the bulk cultures, as well as the mean of the clones analyzed, are shown by arrows. (b and c) The predicted heteroplasmy distributions for line Y, after 30 wk of culture under the conditions stated in the appendix, based on segregation numbers (i.e., number of independently segregating units, S) of 200 and 1000, respectively. Intervals are plotted as for the real clones analyzed. Hatched bars in b show the distribution that would result if homoplasmic mutant cells are excluded from consideration. (e and f) The equivalent predictions for line X.
Actual and predicted distributions of heteroplasmy levels in cybrid cell lines
In bulk culture, line G grew at the same rate as the other cell lines, but its cloning efficiency was very low, hence an insufficient number of clones could be obtained to enable a statistically meaningful analysis. However, we were able to measure the genotype of >20 clones of line GT, a derivative of line G heteroplasmic also for the np 12300 suppressor mutation (G12300A), which also retains the np 3243 mutation at the same level as line G (99%). A culture of line GT, containing 11% np 12300 mutant mtDNA, was maintained over 30 wk, at the end of which clones were obtained at limiting dilution and genotyped at both np 3243 and 12300. Genotyping at np 12300 employed the AflII site polymorphism created by the mutation. The results (Figure 2, a–c; Table 1) showed that the distribution of genotypes was again narrow, at both sites. A small number of clones (9 out of the 63 analyzed) were derived and grown in the absence of the usual uridine supplement in the medium, but the distribution of their heteroplasmy values was similar to that of the other 54 clones analyzed; hence the results from all 63 clones are pooled in Figure 2 and Table 1.
The implied segregation number, assuming random partition, was once again clearly in excess of 1000. The few clones obtained that were homoplasmic wild type at np 12300 generally contained appreciable amounts (3–8%) of mtDNA also wild type at np 3243, and it is possible that some of them represent cells in which the suppressor was never present, because line GT was originally obtained without cloning. Cells lacking the suppressor but still retaining 99% or more np 3243 mutant mtDNA would presumably have cloned only at very low efficiency, as was the case for line G, and this is probably reflected in the slightly higher average heteroplasmy level of the clones for the np 12300 mutation (13.6%) than of the bulk culture grown in parallel.
Intercellular selection can shift the mean genotype of np 3243 heteroplasmons: Mitotic segregation from line G is predicted to result in the continuous generation of a small number of cells at heteroplasmy values <99%. These would presumably have a growth advantage under conditions selective for respiratory function. The fact that long-term culture of line G did not show any shift in average genotype toward lower heteroplasmy values for the np 3243 mutation indicates that during culture in standard medium, either with or without uridine supplementation, no such selection is normally operating. To test whether such selection can be imposed, revealing the presence of cells at lower heteroplasmy values in the bulk culture, we grew G cells in Gal medium, in which glucose is replaced by galactose. The rate of metabolization of galactose via glycolysis is believed to be insufficient to maintain ATP levels in cells that cannot also use respiratory pathways to make ATP. Gal medium is therefore a standard selection system, used to distinguish cells on the basis of their capacity for oxidative phosphorylation (Whitfieldet al. 1981). As observed previously, G cells cultured in Gal medium for 72 hr undergo cell death (El Mezianeet al. 1998a). However, a small number of surviving clones can be recovered from such cultures.
In such an experiment we transferred ~6 × 106 G cells to Gal medium when they were 80% confluent and cultured them for 4 days. Surviving cells were allowed to recover in glucose-containing medium for 8 days and then recultured for 4 days in Gal medium. At this point >80 colonies were visible, of which 48 were subcultured in 24-well plates for 2 final days in glucose-containing medium. DNA was prepared from 34 surviving clones and genotyped at np 3243. As shown in Figure 3a, the proportion of mutant mtDNA was uniformly lower than that of the bulk culture of line G that had been grown continuously in glucose-containing medium (99.4 ± 0.2% mutant, genotyped alongside these clones) and ranged from 87.5 to 97.2% mutant mtDNA. The mean heteroplasmy value for the 34 clones was 93.2%, with a standard deviation of 2.4%. Replicate measurements for individual clones were generally within 0.5% of each other. The heteroplasmy values obtained are below the generally accepted threshold of 98% for severe impairment of mitochondrial protein synthesis, consistent with the idea that a small number of respiration-competent cells, arising in the bulk culture by mitotic segregation as inferred above, were selected by growth under these conditions.
Heteroplasmy profiles at np 12300 of line GT and of subculture GTS. (a) The observed distribution for line GT, based on analysis of mtDNA genotypes of 63 individual cell clones obtained after 30 wk of growth in culture. Frequency bars are plotted as in Figure 1. The percentage of suppressor mutant mtDNA of the bulk culture, as well as the mean of the clones analyzed, is shown by arrows. (b and c) The predicted heteroplasmy distributions for line GT after 30 wk of culture under the conditions stated in the appendix, based on segregation numbers (S) of 200 and 1000, respectively. Hatched bars in b show the predicted distribution that would result if homoplasmic wild-type clones (i.e., lacking the suppressor) are excluded from consideration, on the basis of the fact that they are also close to 100% mutant at np 3243. (d) The heteroplasmy distribution for subculture GTS, based on analysis of mtDNA genotypes of 33 individual cell clones from the culture. Twenty-one such clones were obtained immediately at the time the shift to the new average heteroplasmy level of 29% suppressor mutant was noted, and their distribution is represented by the filled portions of the frequency bars. The remaining 12 clones, whose distribution is represented by the shaded portions of the frequency bars, were obtained from the GTS culture 2 mon later. The percentage of suppressor mutant mtDNA of the bulk culture, as well as the mean of the clones analyzed, is shown by arrows.
Rapid, concerted shifts in mtDNA genotype occur as a rare event in np 3243 cybrids: The above experiment represents the deliberate selection of a small proportion of cells against a background of massive cell death. Line G has otherwise been grown continuously in the laboratory for periods of up to 1 year without any perceptible change in genotype or phenotype. On several occasions, however, we have observed an abrupt shift in genotype of subcultures of line G, grown continuously in glucose-containing medium and without any apparent cell death, change in growth rate, or any other observable parameter. In two such subcultures, designated G5S and G6S, the average genotype shifted toward increased levels of wild-type mtDNA over at most a few weeks, with the average heteroplasmy value restabilizing in these cultures at 92 and 89% mutant, respectively. The cultures were grown continuously for a further period of at least 28 wk, cloned at limiting dilution, and genotyped. The heteroplasmy values of 22 clones obtained from subculture G5S (Figure 3b) were now once again narrowly distributed, with an overall diversity similar to that of other lines exhibiting mitotic segregation with high segregation number. The mean heteroplasmy value of the clones (90.6%) was slightly less than that of the bulk culture (92.4%), probably reflecting once again the low cloning efficiency of cells at very high level of mutant that were therefore lost from the analysis. The 5 clones of subculture G6S that were analyzed also had heteroplasmy levels narrowly distributed about the mean of the bulk culture (data not shown).
Changes in genotype of line G under selection and as a result of a spontaneous shift. Genotype at np 3243 (percentage mutant mtDNA) of (a) 34 cell clones obtained from line G, selected by growth in Gal medium. Arrows show the mean of the distribution (93.2%) and the value for the bulk culture of line G grown continuously, in parallel, in glucose-containing medium (b) Twenty-two cell clones obtained from subculture G5S, which had undergone a spontaneous shift to a new mean heteroplasmy level of 92% that was noted 30 wk before cloning. Arrows show the genotype of the bulk culture of G5S and the mean heteroplasmy levels of the clones analyzed. Frequency bars are plotted as in Figures 1 and 2.
The events that gave rise to the shifts of genotype in subcultures G5S and G6S presumably involve a mechanism of selection similar to that inferred previously to be operating in unstable heteroplasmic cybrids (Dunbaret al. 1995). One possible explanation for the behavior of these cultures is that the limitations on mitotic segregation were transiently relaxed, leading to the rapid emergence of truly homoplasmic mutant cells and of many cells with significantly lower levels of mutant, upon which selection acted progressively to shift the average genotype. The kinetics of these events are also not fixed, in the sense that the duration and “genotypic end point” of such shifts can vary. In a further such subculture of line G, shifting continued progressively over a period of several months and resulted in the eventual complete loss of mutant mtDNA, as was observed previously in cybrid clones that were unstable from the time they were established (Dunbaret al. 1995).
Concerted shifts in mtDNA genotype are associated with transient diversification of genotype, followed by reestablishment of stable heteroplasmy, in np 3243 cybrids: Finally, we obtained a subculture of line GT (designated GTS) that underwent a shift at some point within an 8-wk period, from 11 to ~29% mutant at np 12300, while remaining 99% mutant at np 3243. Thereafter, the bulk culture has remained stable for 10 mon at the new heteroplasmy value. Twenty-one individual clones obtained at limiting dilution at an early stage in the culture following the shift in average heteroplasmy level exhibited a wide variety of genotypes (Figure 2d), when analyzed 2 wk after cloning. These ranged from 6 to 87% mutant at np 12300. A further 12 clones, derived 2 mon later from the shifted cells, showed a slightly narrower distribution of heteroplasmy values (Figure 2d), but still wider than in any sampling of the stable lines.
All clones obtained at the early time point from the GTS culture were grown for a further 12 wk and regenotyped (Figure 4a), revealing that while about half had remained stable within the accuracy of measurement (estimated at ±2% for the np 12300 mutation), many had continued to shift to new heteroplasmy values both lower and higher than those originally measured 2 wk after cloning. In addition, 6 clones had apparently become homoplasmic for the np 3243 mutation, within the detection limits (<1 copy per cell). Aliquots of each clone had been frozen 2 wk after cloning, and these were now recovered from the freezer and genotyped again at various times during growth over 3 mon, to determine whether the continuing shifts in genotype at np 12300 observed in the first experiment were random or reproducible and to evaluate the kinetics of these shifts. As shown in Figure 4b, there was a strong positive correlation between the change in heteroplasmy level over 3 mon observed for each of 20 GTS clones in the first experiment and the change exhibited over the same time period by the same clone during the second experiment. In other words, clones that remained stable the first time generally did so again, and those that showed increases or decreases in the amount of suppressor mutant mtDNA over 3 mon did so a second time, generally to a very similar extent. Only 1 of the 20 clones behaved differently in the second experiment, namely clone GTS-20, which was stable the first time at 25% mutant, but in the second experiment declined progressively to 9% mutant.
Temporal changes in heteroplasmy in GTS clones. (a) The percentage of np 12300 mutant mtDNA was measured in each clone at 2 and 14 wk after cloning, and the values obtained at the two time points are plotted against each other. Solid circles represent clones that remained at an unchanged heteroplasmy level, within the limits of accuracy of measurement (±2% np 12300 mutant). Open circles represent clones whose heteroplasmy levels were significantly different at the two time points. The dotted line represents the diagonal on which unchanged clones should lie. (b) The change in percentage heteroplasmy of each clone was compared in two separate experiments: (1) after 14 wk of growth since cloning and (2) in a second period of growth for a total of 15 wk since cloning, after having been frozen down at 2 wk postcloning. Solid circles represent clones whose direction of change, if any, was the same in the two experiments. Open circles represent clones that shifted in opposite directions in the two experiments. The dotted lines represent the diagonal along which clones exhibiting perfectly correlated behavior should lie and demarcate the quadrants of positive and negative correlation. (c) Measurements of heteroplasmy values (percentage np 12300 mutant mtDNA) for six individual clones, over time. Solid circles, first experiment over 14 wk; open circles, second experiment over a total of 15 wk since cloning.
The kinetics with which these clones shifted to new heteroplasmy values differed, but in general, the values reached within 7 wk changed little thereafter. Some examples of the behavior of individual clones showing up- or downshifts, or that remained stable, are shown in Figure 4c. The extent or direction of change was not related in a simple way to the starting proportion of mutant. For example, clones GTS-12, GTS-13, and GTS-19 all had ~50% mutant mtDNA at 2 wk yet after 3 mon showed, respectively, a modest decrease, a large decrease, and a significant increase in the amount of mutant. Stable clones also exhibited a wide variety of heteroplasmy levels, ranging from ~15% to >70% mutant.
Reestablishment of stable heteroplasmy values is independent of obvious phenotypic selection: The fact that the change in genotype of each clone of the shifted GTS subculture was reproducible suggests that the reestablishment of stable heteroplasmy values involved intra- or intercellular selection. The simplest hypothesis is that this was driven by respiratory constraints, favoring the growth of actively respiring cells. We therefore measured the oxygen consumption of each GTS clone, both at the earliest time point from which we had sufficient cells to carry out the assay (3 wk postcloning) and at 15 wk of growth, when heteroplasmy levels had generally restabilized.
The results, depicted in Figure 5, do not support the view that the shifts observed in individual clones were due to respiratory selection. Furthermore, they indicate that the relationship between respiratory phenotype and mitochondrial genotype at np 12300 is complex. Although overall heteroplasmy level for np 12300 is clearly not the sole predictor of respiratory phenotype in these cells, a general pattern is evident, wherein clones with rather extreme heteroplasmy levels (<10 or >70%) were the worst respirers. Oxygen consumption values for these clones were in the same range as the 95% np 3243 mutant line X, which lacks the suppressor mutation completely. In contrast, the best respirers were generally in the range of 20–30% heteroplasmy for np 12300, and many clones appeared to converge into this zone during growth, with concomitant improvement in respiration. Nevertheless, a number of clones clearly shifted in the opposite direction: away from optimal heteroplasmy, and with accompanying phenotypic “deterioration.” The overall conclusion is that the shifts in mitochondrial genotype that we observed during restabilization are not obligatorily associated with selection driven by respiratory constraints, but are instead attributable, at least in part, either to drift or else to some other kind of selection that we have not identified. The notion of drift seems incompatible with the reproducible behavior of these clones and with the general narrowing of the distribution of heteroplasmy values with time: however, in the discussion we present a hypothesis to reconcile these ideas.
Relationship of changes in heteroplasmy for the np 12300 suppressor mutation to respiratory activity. Oxygen consumption was determined for each GTS clone studied, at 3 and 15 wk after cloning, and plotted against heteroplasmy level (percentage of mutant at np 12300) at the two time points. Solid circles represent values measured at 3 wk, open circles represent those measured at 15 wk. Solid lines join the two values for each given clone, to indicate the extent and direction of change of both variables over time.
Shifts in mitochondrial genotype are not correlated with mitochondrial morphology or copy number: We next considered two other simple hypotheses to account for transient instability of mitochondrial genotype: alterations in the morphology or intracellular distribution of mitochondria and differences in mtDNA copy number. In both cases we examined the stable heteroplasmic lines X, Y, G, and GT, plus all of the clones from the GTS subculture, some of which were still shifting in genotype at the time of the experiment. Mitochondrial morphology was analyzed by staining of cells with the mitochondrial membrane potential-sensitive dye JC1 (Figure 6). This revealed an identical pattern of staining in all cell lines, with the sole exception of the respiration-deficient line G (99% mutant at np 3243). All of the respiration-competent lines, including those that were shifting in mtDNA genotype, contained threadlike mitochondria, with many intracellular foci of intense orange-yellow staining indicative of high membrane potential. The mitochondrial network extended over the entire cell, but its exact intracellular distribution varied between cells of a given culture. Lines with different heteroplasmy values, with or without the np 12300 suppressor, whether genotypically stable or shifting, and regardless of oxygen consumption values, all showed this typical pattern. Line G gave a completely different pattern, with diffuse green staining and only a few yellow spots indicating high redox activity. We interpret this as a feature of the np 3243 mutant phenotype not relevant to the stability of heteroplasmy as such. A very low proportion of cells of line G (estimated at <0.1% of the total) gave a normal mitochondrial staining pattern. These cells are probably equivalent to those selected by growth in Gal medium, which had higher levels of mtDNA wild type at np 3243.
Cellular distribution of mitochondria and mitochondrial redox activity, by fluorescence microscopy of JC-1 stained cells. Top, a cell from clone GTS-8, which is typical of all the cell lines and clones analyzed except for line G, shown at the bottom.
Relative mtDNA copy number was estimated by Southern cohybridization, using probes for mtDNA and nuclear rDNA (illustrated in Figure 7). The relative signals for nuclear and mitochondrial DNA were arbitrarily normalized to those for one of the clones (GTS-8). Replicate measurements were made on each clone at 11 and 15 wk of growth, giving small differences (generally within 50%) that may reflect growth state or inherent limitations in the accuracy of the method, rather than real differences in copy number. To improve accuracy the two measurements were averaged and then compared with the shift in heteroplasmy over 3 mon for each clone (the value taken was the average heteroplasmy shift that occurred for each clone, in the two experiments depicted in Figure 4b). The relative copy numbers were all within a fourfold range, and generally much closer. No correlation was obtained between mtDNA copy number and any of the parameters investigated (average heteroplasmy level, stability vs. instability, rate of restabilization, or oxygen consumption). The analysis is illustrated in Table 2, in which the clones are rank ordered according to the amount of shift in heteroplasmy that was measured over 3 mon. Clones that exhibited genotypic instability included those with both higher and lower relative copy numbers. The clones showing the greatest disparity in copy number measurements at 11 and 15 wk (GTS-5, GTS-7, and GTS-22) were also among the most stable clones genotypically. We conclude that shifts in mtDNA genotype can take place without significant alterations in mtDNA copy number. We cannot, however, exclude the possibility that a transient change in copy number occurred at the time of the original event that led to mitotic segregation in the GTS culture, although copy number measurements in the bulk cultures revealed no major differences between GT, GTS, and other cybrids studied.
Example of mitochondrial DNA copy number estimation in GTS clones. Genomic DNA digested with EcoRI (see materials and methods) was fractionated by 0.7% agarose gel electrophoresis and Southern blots were cohybridized with probes for human mtDNA (COXII region) and rDNA (5′ end of 18S rRNA). The autoradiograph shows the signals obtained from a panel of GTS clones at 15 wk of growth after cloning. For estimating the values quoted in Table 2, phosphorimager signals from a series of such blots on DNA samples extracted at both 11 and 15 wk were used to derive ratios of mtDNA/rDNA hybridization that were then arbitrarily normalized to the value obtained for clone GTS-8. The fragments detected are as predicted from the relevant database entries for mtDNA (GenBank accession no. J01415) and rDNA (GenBank accession no. U13369).
Cytoskeletal disaggregation by DMSO treatment does not induce systematic mtDNA segregation: Cells of line GT that underwent the dramatic diversification in heteroplasmy level had been subjected to very low levels of the cytoskeleton-altering drug DMSO (Nishidaet al. 1987), subsequent to an aliquot of the cells being frozen down in a DMSO-containing medium at passage. To test whether such exposure to this drug could account for shifts in mtDNA genotype, we treated cells of two otherwise stable clones, GTS-52 and GTS-53, with low levels of DMSO under comparable conditions to those imposed earlier on the GT culture that shifted, as well as at a higher dose for a shorter time (see materials and methods). The results, summarized in Table 3, show clearly that low-level DMSO treatment did not induce any systematic shift or diversification of heteroplasmy levels in these cultures, despite the minor differences observed. Specifically, clone GTS-52 showed a modest increase in variance after treatment, but no change in mean heteroplasmy, while clone GTS-53 showed a minor change in average heteroplasmy level but no systematic increase in variance. These variations are almost certainly attributable to the small number of subclones analyzed. Moreover, in those cases where variance increased, the increases were quantitatively minor compared with those that occurred during the dramatic shift of line GTS.
Relationship between mtDNA copy number and stability of heteroplasmy in GTS clones
Shifts in mtDNA genotype can be associated with a nuclear genetic change: One hypothesis to account for the episodes of genotypic shift and diversification that we observed is that they were induced by a nuclear genetic change. To test this idea, we carried out CGH on DNA samples from paired cell lines before and after such episodes. DNA from line G (>99% np 3243 mutant) was compared with its derivative G5S, which had shifted to 92% mutant, revealing no detectable differences. CGH can detect chromosomal deletions or amplifications down to the 1 Mb level, but obviously this negative result does not exclude point mutations or smaller scale aberrations. By contrast, comparisons of DNA from line GT (stable and homogeneous at 11% np 12300) with its derivative GTS (genotypically heterogeneous and shifted to 29% np average heteroplasmy) showed clear evidence of a chromosomal difference, namely a gain of most or all of chromosome 9 (Figure 8a). To confirm this result, CGH was carried out to compare DNA from each of lines GT and GTS with normal human DNA, indicating that line GT is indeed deficient for chromosome 9, whereas the complement of chromosome 9 DNA in line GTS is as normal (Figure 8b). CGH further revealed that the chromosome 9 content of line GTS is also increased relative to that of another of the A549 lung carcinoma cybrids of the series, line Y (Figure 8c). The shift in mitochondrial genotype, in this instance, is therefore correlated with a gross change in nuclear genotype, although this does not prove causality.
Effects of DMSO on mtDNA heteroplasmy levels
DISCUSSION
The experiments described in this paper document three quite distinct types of behavior of heteroplasmic cells in culture: (1) stable heteroplasmy with high segregation number, (2) genotypic shift effected by an externally applied selection, and (3) genotypic shift associated with an apparent, transient relaxation of partition constraints.
Maintenance of stable heteroplasmy: The ability of cells to maintain stable heteroplasmy with a high effective segregation number differs from what was previously observed in the mammalian germ line (Jenuthet al. 1996), as well as in yeast (Treat and Birky 1980; Birky 1983). In both of the latter cases, segregation number is at least an order of magnitude below the mtDNA copy number per cell. Copy number in somatic cells is generally in the range 103–104, comparable with the segregation numbers of >>1000 we inferred here. Extant culture records do not allow us to estimate the total time during which each of these lines was grown since its original cloning, which in most cases is certainly far longer than 30 wk. Therefore, this segregation number we have observed in tumor-derived cell lines must be regarded as a minimum. In principle we can suggest three mechanisms by which high segregation number might be achieved, based on radically different views of the intracellular organization of mtDNA molecules. The simplest is that copy number and segregation number are synonymous, but other interpretations are possible.
In this first model, therefore, we assume that every mtDNA molecule has an equal chance of being partitioned to either daughter cell, regardless of where in the cell it is physically located. This might be achieved if, like mitochondria themselves, mtDNA was highly dynamic within the cell, so that the physical locations of mtDNA molecules become randomized during each cell cycle. This model assumes that mtDNA is dispersed in the cell and that single mtDNA molecules are the unit of inheritance. Almost nothing is known about the degree to which mtDNA molecules move about the cell in the course of the mammalian cell cycle. In yeast mating, when mtDNA molecules are supposedly programmed to meet and exchange information, mtDNA remains much more rigidly fixed in the cytoplasm than mitochondrial proteins/membranes, during at least the early phase of the mating process (Nunnariet al. 1997).
The idea of mtDNA as a dynamic molecule is supported by observations in ρ0 HeLa cell cybrids, in which introduced mtDNA rapidly spread throughout the cell (Hayashiet al. 1994) and in which complementation was observed between partially deleted mtDNA and mtDNA carrying a chloramphenicol-resistance mutation (Takaiet al. 1997). In contrast to this, in 143B osteosarcoma cell cybrids Yoneda et al. (1994) failed to observe complementation between pathological mtDNA point mutations in different tRNA genes, suggesting that mtDNA molecules may have spatially limited spheres of influence within the mitochondrial network.
The second model is almost the exact opposite of the first, in that it proposes that the direction of partition of every mtDNA molecule is not random, but absolutely constrained. At each round of mtDNA replication the two daughter molecules are proposed to be marked in some way that determines their eventual partition in opposite directions at cell division. This would imply the existence of some kind of mtDNA kinetochore, perhaps analogous to the structure recently discovered in bacteria (Glaseret al. 1997; Lewis and Errington 1997; Wheeler and Shapiro 1997), and would furthermore imply hard-wired connections between mtDNA and the cytoskeleton, through the mitochondrial membrane system, at least during cytokinesis. Such a system is indeed suggested by observations of organelle partition in unicellular eukaryotes (Kuroiwaet al. 1998), and disruption of the cytoskeleton is known to impair the normal patterns of mtDNA segregation in yeast (Berger and Yaffe 1996).
Comparative genomic hybridization. CGH signals are shown only for chromosomes 9 and 10, averaged from a minimum of 5 replicate chromosomes in each hybridization. For each chromosome the average spectral difference is plotted along the length of the chromosome on a scale that indicates the margins of error for signal equivalence. Signal to the right indicates gain relative to the reference DNA; signal to the left indicates loss. Solid bars alongside the plot indicate the chromosome regions of significant signal difference. (a) Comparison of DNA from line GTS (shifted) against line GT (unshifted) as reference. The signal for chromosome 10 indicates no significant differences. The signal for chromosome 9 indicates gain of DNA in line GTS along the entire length of the chromosome, except perhaps for the tip of the long arm, as denoted by the solid bar to the right of the plot. Repeat experiments (not shown) gave the same results. (b) Comparisons of DNA from, respectively, lines GT (top) and GTS (bottom) against normal human genomic DNA as reference. The two lines exhibit a similar profile for chromosome 10, with a possible underrepresentation of a short region toward the telomere of the long arm. For chromosome 9 the two cell lines give different profiles, consistent with the results shown in a. Line GTS gives signal within the normal range, except at the tip of the long arm, where there is a probable amplification. By contrast, line GT gives signal indicating a relative underrepresentation of almost the entire chromosome, except for the tip of the long arm, as denoted by the solid bars to the left of the plot. (c) Comparison of line GTS with line Y as reference. The profile for chromosome 10 is very similar to that seen in b for each cell line compared with normal DNA. The profile for chromosome 9 shows a uniform overrepresentation, straddling the limit of significance along the whole length of the chromosome. This is consistent with a general deficiency of chromosome 9 in A549 cybrid lines, with a gain of this chromosome uniquely in line GTS.
In mammalian cells mtDNA is generally believed to exist in a proteinaceous complex associated with the inner membrane, as in plants, yeast, and lower eukaryotes (Kuroiwaet al. 1994; Miyakawaet al. 1995; Newmanet al. 1996; Oldenburg and Bendich 1998). Connections between the two membranes and between mitochondria and the cytoskeleton are well documented, hence the physical components of such a machinery are not unprecedented, although the nature of any mtDNA kinetochore is unknown. The effectiveness of such a system for constraining partition would be dependent on all or almost all mtDNA molecules replicating during each cell cycle. However, long-standing observations indicate that, at least in tumor cells in culture, not every mtDNA molecule replicates exactly once during each round of cell division (reviewed in Clayton 1982).
The third model, which we favor, is that mtDNA molecules are grouped, as in yeast, plants, and slime molds (Oldenburg and Bendich 1998; Sasakiet al. 1998), into essentially permanent nucleoid structures, each containing a number of physically associated, protein-bound copies of the genome that usually are faithfully replicated as a unit. In the stably heteroplasmic cell, this model assumes that each nucleoid is itself heteroplasmic (i.e., contains molecules of the two different types), but that the heteroplasmy levels of the different nucleoids within the cell are roughly homogeneous. Thus, the composition of the mtDNA population is maintained at cell division, because the units of segregation are essentially identical. Any segregation that is actually observed would then be attributable to the fact that the exact number of mtDNA molecules in each nucleoid might vary, and that especially where one mtDNA genotype is extremely prevalent (e.g., 95% or more), some heterogeneity in the degree of heteroplasmy of different nucleoids must exist. In addition there may be a degree of “leakiness” of the system that faithfully replicates such nucleoids.
This model has the advantage of being compatible with the noncomplementation of pathological mtDNA mutations reported by Yoneda et al. (1994), i.e., that genetically distinct nucleoids in the cell would continue to replicate faithfully and each have a distinct sphere of genetic influence. Furthermore, it does not require stable interactions between mtDNA and the cytoskeleton to be invoked, although it does still imply the existence of some kind of mitochondrial kinetochore, and what might be termed mitokinesis, as proposed by Kuroiwa et al. (1994), and as inferred cytologically in a number of organisms. In an earlier study of the (short-term) evolution of heteroplasmy in fibroblast clones carrying the np 3243 mutation, and also in cybridization experiments using one such clone as a donor, Matthews et al. (1995) also concluded that the unit of mitochondrial inheritance in heteroplasmic cells was itself heteroplasmic. However, there is no physical evidence to support the view that mtDNA is organized thus in mammalian cells.
Genotypic shift under external selection at the cellular level: Selection against cells with the highest levels of np 3243 mutant mtDNA was achieved in the experiments described here in two different ways. Cells of line G, other than the tiny minority in which mitotic segregation had raised the level of wild-type mtDNA above the threshold of 2–3%, were unable to grow in the respiration-selective Gal medium, nor were they able to grow from single, isolated cells in rich medium, even though their survival and growth rate when passaged at higher cell density was indistinguishable from that of control cybrids. The unclonability of high mutant cells may reflect a defect in the ability of an individual cell to retain bioenergetic or ionic balance throughout the cell cycle, for which neighboring cells, perhaps in gap junctional communication, can compensate. The physiological relevance of this is unclear, because cells do not normally find themselves isolated from their neighbors. However, the imposition of external nutritional constraints may provide a good model for the additional demands placed upon a cell's bioenergetic capacity during differentiation of stem cells to perform specialized cell functions (as nerve, muscle, secretory tissues, and so on). Selection in vivo against cells with high levels of the np 3243 mutation has already been inferred from analysis of patient-derived fibroblast cell clones (Matthewset al. 1995). Stem cells must survive a prolonged lifetime, with many rounds of division under conditions of little or no selection for respiratory function. Segregation toward homoplasmy for deleterious mutations such as A3243G during this period would be catastrophic, but only once the cell's descendants are required to differentiate. A system of constrained segregation, such as we observed in our A549 lung carcinoma cybrids, would minimize the risk that deleterious mutations arising during the lifetime of a stem cell can accumulate to a damaging level. Tumor cells are widely regarded as dedifferentiated representatives of the stem cell compartments from which their tissues of origin are derived. Mitochondrial DNA is arguably located in a highly mutagenic environment, due to the generation of oxygen radicals as a by-product of respiration (Miquel 1992); hence this system may have evolved to compensate for a hypothetical high somatic mutation rate, preserving the fitness of the stem cell population throughout life.
Genotypic shift arising from relaxation of partition constraints: In cultures of two different cybrid cell lines we observed, as a rare event, a concerted shift in average heteroplasmy level without evidence of massive cell death or interrupted growth. In the one case that we were able to analyze in detail shortly after the primary event, there was clear evidence for a dramatic diversification of mitochondrial genotype in a culture of cell line that has otherwise behaved in the standard fashion, i.e., with highly constrained segregation. We therefore interpret concerted genotypic shifts, such as occurred in both lines G and GT, as being the result of a sudden relaxation of partition constraints, followed by selection at the cellular level against cells that segregate, as a result, toward homoplasmy for either the np 3243 or np 12300 mutation.
The relaxation on partition constraints that we infer to have occurred in both cell lines appears to have been transient. Some 28 wk after its initial shift in genotype was noticed, line G5S had reestablished a relatively narrow distribution of heteroplasmy values, and this was also the case for 2 GTS clones analyzed in detail (Table 3). In addition, the 10 clones picked from line GTS 2 mon after the shift was recorded were more narrowly distributed about the mean heteroplasmy value than the clones picked at the earlier time point (see Figure 2d). The individual GTS clones generally reestablished a new stable heteroplasmy level within weeks, but with shifts in both directions. Only one clone, GTS-20, was clearly continuing to shift in genotype at 3 mon, and this was also the clone that behaved differently when grown the second time, which possibly indicates that it had initiated a shift de novo.
However, considering the amount of suppressor mutant in the set of GTS clones analyzed at the two time points (2 and 14 wk after cloning), only a small overall decrease in variance was observed (416 at 14 wk as opposed to 471 at 2 wk of growth). In other words, although the bulk culture seemed gradually to return to a more homogeneous state following the shift, this is not reflected in the distribution of mtDNA genotypes of cell clones drawn from the population, which remained wide. In addition, we observed that cell clones with extreme heteroplasmy levels for np 12300 (<10 or >70%) seemed to have a respiratory impairment, and it is thus reasonable to assume that they may also have a mild growth defect. The behavior of the bulk culture following the diversification of genotypes may therefore involve a degree of selection at the cellular level against cells with very high or very low amounts of suppressor mutant.
The mechanism by which partition constraints are relaxed obviously depends on what is the mechanism constraining segregation in the first place. One obvious model, but which is unfortunately difficult to test, is that a catastrophic event caused a transient but major reduction in mtDNA copy number in the whole cell population, i.e., a bottleneck phenomenon. Such a transient reduction in copy number would entrain a corresponding reduction in segregation number of exactly the kind we observed. Arguing against this is the fact that we observed no abnormality of cell growth during the process, which might be expected to accompany a sudden, major loss of mtDNA. If the drop in copy number was milder but more prolonged, then we might expect to have detected at least an echo of it in cells still shifting in genotype during the recovery phase, which we did not. On balance, while we cannot entirely reject this hypothesis, we think it unlikely, and now focus on alternative possible explanations.
Proceeding from the same three models presented earlier, we can propose the following. According to the dynamic mtDNA model, mitotic segregation would be induced if mtDNA movement around the cell were blocked; hence sister molecules would remain close together following replication. In a rather small number of cell divisions this would lead to the emergence of cytoplasmic domains of a defined genotype, resulting in a significant drop in the apparent segregation number. Restabilization of heteroplasmy would then be viewed as a resumption of dynamic mtDNA movement. An alternative version of this is to invoke the formation of mitochondrial nucleoids in a cell that previously (or normally) does not possess them. If the thousands of mtDNA molecules previously segregating as independent units simply coalesced into a few hundreds or tens of mtDNA aggregates, then the segregation number would drop. This would especially be the case if such nucleoids formed mainly from mtDNA molecules of one or another genotype.
In the mitochondrial mitosis model, in which daughter mtDNA molecules are constrained to segregate in opposite directions at cell division, mitotic segregation would result if the system simply broke down, such that mtDNA molecules now moved to daughter cells only according to their positions in the cell, in a scenario rather similar to that envisaged above. If mitochondrial nucleoids of different genotypes were already present, then this would accelerate segregation still further, upon rupture of the links between mtDNA and the apparatus of cytokinesis. Another possibility is that mtDNA replication might suddenly become highly nonrandom, in the sense that a small number of molecules may replicate many times in one cell cycle, rather than all or most molecules replicating once. This would have the same effect as a transient drop in mtDNA copy number, leading to a founder effect. The possibility that mtDNA replication in mammalian cells might, under some circumstances, switch to a rolling circle mode as in yeast (Maleszkaet al. 1991) and plants (Backertet al. 1996) should be considered, because this would entail selection of a relatively small number of molecules as replication templates.
Finally, if the nucleoid/mitokinesis model is considered, mitotic segregation would be predicted to result from a transient disaggregation and reaggregation of mtDNA nucleoids, especially if several rounds of DNA replication took place in the interim, generating a highly nonuniform pattern of intracellular heteroplasmy. We strongly favor this model, not only because it is the most consistent with other published data, but because it best accounts for the progressive reestablishment of stable heteroplasmy that we observed in the GTS clones, with reproducible, clone-specific shifts in genotype apparently independent of phenotypic selection. How we envisage this is depicted in Figure 9.
We propose that at the time of the primary event, the previously homogeneous nucleoids are replaced with a collection of nucleoids of very different internal heteroplasmy levels, some of which may be homoplasmic or near homoplasmic for one or another genotype. During subsequent cell growth, these nucleoids remain in their new configurations, but are subject to mitotic segregation. The implied number of segregating units would be rather small (at most, the number of nucleoids physically present), leading rapidly and progressively to the reestablishment of stable heteroplasmy. The new level of heteroplasmy reached in each cell clone is that of the one remaining type of nucleoid in that cell after mitotic segregation is complete. The reproducible behavior of each cell clone is understandable if we assume that already at the time of cloning, or at the first time point at which we were able to sample the mtDNA population 2 wk later, there is already a predominant nucleoid type present in each cell.
The cell line in which we observed these events was initially 99% mutant at np 3243 mtDNA and 11% mutant at np 12300, and homoplasmy for either mutation would presumably be deleterious, because mitochondrial protein synthesis would be severely impaired due to a specific decoding deficiency. During the final resolution period, the shift in average genotype of each clone can be understood as the loss of homoplasmic nucleoids, by virtue of their segregation into cells that hence become counterselected at the cellular level.
A model for changes in mtDNA genotype in GT cells, based on the heteroplasmic nucleoid model. In the stable heteroplasmic cells (a) it is proposed that all nucleoids (small circles) have the same relative composition of np 12300 suppressor mutant (black) and wild-type (white) mtDNAs, with all or almost all molecules also mutant at np 3243. During successive rounds of cell division the mtDNA genotype of individual cells in the culture (b) remains unchanged, because nucleoids are proposed to replicate and divide faithfully, thus preserving their uniformity. In response to an unknown stimulus, cells in the culture can enter a different state (c), in which their nucleoids become transiently disaggregated, or in which their faithful replication is temporarily abrogated. The result is that individual nucleoids in the cell, when they reform, can have a great variety of different internal heteroplasmy levels. During subsequent growth of the culture, mitotic segregation progressively purifies the nucleoid population in each cell descendant, such that after many rounds of division only one type of nucleoid remains in each cell. Cells containing only homoplasmic mutant nucleoids (f), and cells whose nucleoids contain only np 3243 mutant mtDNA without the suppressor (g), are proposed to be eliminated, progressively, by negative selection, leaving a diverse collection of stably heteroplasmic cells (d and e), whose heteroplasmy levels are now different and much more diverse than the original culture. The average heteroplasmy level of the remaining cells may differ from that of the original culture as a result of the different selective values of very high or very low levels of the suppressor mutant.
Induction of altered mtDNA partition behavior: In an attempt to determine how the altered partition behavior that we observed in our cultures may have come about, we tested two hypotheses. The first postulated that the inducing mechanism may be environmental, related to the action of DMSO, a drug that is known to affect cytoskeletal organization. The second proposed that mtDNA segregation may have been induced by a gross change in the nuclear genotype of the cells. The DMSO hypothesis was tested directly, with negative results. Small differences in the mean and variance of heteroplasmy were observed in cell clones after treatment with the drug, but these were not systematic. We conclude that if transient DMSO treatment was in any way involved in the shift of mtDNA genotype it must have been in combination with something else that we were not able to test.
The possible involvement of the nuclear genome was studied retrospectively, by comparative genomic hybridization. This indicated that while the shift observed in line G5S was not accompanied by any detectable nuclear genetic change, that of line GTS was associated with a gain of chromosome 9. The present results do not establish causality. However, the simplest explanation is that the gain of chromosome 9 induced the abrupt but transient diversification of mtDNA genotype that we observed. We postulate that a gain of chromosome 9 in an otherwise aneuploid cell line resulted in a modest growth advantage, most probably unrelated to mitochondria, that rapidly replaced the previous cell population with the new variant. Because this is, by definition, a cloning event, the diversification of mtDNA genotype that appears to have accompanied it would seem even more dramatic than if it occurred in a population of cells. We suggest that enhanced expression of one or more genes on chromosome 9 resulted in a significant, but transient, modification of mtDNA organization within the cell, whether by the disaggregation and reformation of nucleoids, the induction of rolling circle replication, the disruption of mtDNA/cytoskeleton interactions, or some other mechanism. The event then manifested as mitotic segregation over a period of weeks, such as already illustrated in Figure 9, leading finally to the reestablishment of a stable and homogeneous heteroplasmy level. The shift observed in line G5S may, however, have involved another mechanism. Given the clear selective advantage of lower heteroplasmy values for np 3243 mutant mtDNA in the A549 background, we cannot rule out a simple selective mechanism in this case.
To establish which of the proposed mechanisms accounts for the segregation behavior of mtDNA during the long-term culture of heteroplasmic mammalian cells and to establish the exact role of the nuclear genome in this process, it will be necessary to build a detailed understanding of how mtDNA is organized in cells, the proteins with which it interacts, and its physical behavior during the cell cycle. This knowledge is currently lacking. Nevertheless, the phenomena that we have documented indicate that the mtDNA population of heteroplasmic cells can remain stable over long periods, but can also be abruptly remodeled, in response to selection, altered expression of one or more nuclear genes, or some other stimulus. Elucidating the physical basis of that stability and of the stimuli that can transiently abrogate it is clearly of great importance to our eventual understanding of the developmental consequences of heteroplasmy, for example, in organogenesis, where systematic changes in heteroplasmy levels have been observed in mice (Jenuthet al. 1997), as well as in somatic aging and mitochondrial disease.
Acknowledgments
We thank Anja Rovio and Arja Alkula for technical assistance and our many colleagues within the EU MBDD and NMI Networks, as well as Pekka Lappalainen, for relevant discussions. We also thank journal editor Kathy Newton and an anonymous reviewer for insightful and helpful suggestions to improve our manuscript. This work was supported by funding from the Academy of Finland, Tampere University Hospital Medical Research Fund, the Muscular Dystrophy Group of Great Britain, the Juselius Foundation, and the European Union. I.J.H. is a Royal Society University Research Fellow.
APPENDIX: COMPUTATION OF MITOTIC SEGREGATION IN HETEROPLASMIC CELL LINES
The problem is defined with the following assumptions. The cell lines under consideration consist of copy number N copies of mtDNA per cell after every cell division. Each molecule replicates exactly once per cell cycle, and half of the cell's molecules are partitioned randomly to each daughter cell. The cells are maintained in culture for a period of v weeks. The cell doubling time (cell cycle) is 24 hr. The cultured cells are maintained by dilution every u days. At each passage c cells are seeded and allowed to grow to confluence.
We assume that at the beginning the cell line used is homogeneous and that s% of the molecules of each cell are mutant.
At first, we consider the probability distribution of the process after v weeks or d = 7v cell cycles.
After replication there are 2N molecules that are randomly partitioned to two daughter cells. It was assumed that each of them contains exactly N mtDNA molecules. Now a problem arises: What is the probability pij that one daughter cell includes j mutants if the mother cell had i mutants (and at the same time N − j and N − i wild type, respectively)? After replication the number of 2i mutant molecules can be randomly partitioned to two daughters—under condition that one daughter will have j molecules—in
Let us denote the first cell generation by vector
Note that the last formula (4) is deterministic. Once the first generation q(0) is known, all vectors q(d), i.e., the later generations, are uniquely evolved by the formula. The crucial issue is which type, percentage s (the corresponding qi is 1), of the first generation is used.
At the onset of the process, we are interested in cases in which exactly one component in the vector (2) is equal to 1 and the others are equal to 0. In other words, all cells are then homogeneous.
Up to this point, we have considered an arbitrary cell starting from the first generation. Nevertheless, because we presupposed all cells to be identical at the beginning, all their descendant cells have similar probabilities to arise, concerning the number of mutant and wild-type molecules they contain. Of course, the condition for this is that there are a great number of cells at the beginning as well as that a large number of them are randomly chosen after every week (u = 7) to continue the process. Because c = 50,000 cells were taken in this way, the condition was well satisfied.
Footnotes
-
Communicating editor: K. J. Newton
- Received December 20, 1998.
- Accepted September 21, 1999.
- Copyright © 2000 by the Genetics Society of America
