The 3243A→G MTTL1 gene mutation of mitochondrial DNA (mtDNA) is the most common heteroplasmic pathogenic mtDNA mutation and is found in approximately 1 in 6000 of the general population.¹ Although first described in mitochondrial encephalomyopathy with lactic acidosis and stroke-like episodes (MELAS), the phenotypic spectrum is extremely diverse, including isolated diabetes and deafness, hypertrophic cardiomyopathy and retinitis pigmentosa.² The clinical variability can be explained partly by tissue-specific differences in the percentage of mutated mtDNA.^3,4

Intriguingly, the percentage of mutated mtDNA is consistently lower in peripheral blood than in post-mitotic tissues such as skeletal muscle and brain.^3,5 Serial measurements in the same subject have shown that the percentage of the 3243A→G mutation in blood decreases over time,^6,7 but the reasons for this are not clear. One possibility is that vegetative segregation in rapidly proliferating leucocyte precursors leads to high percentages of mutated mtDNA in some cells. This causes a biochemical defect of the respiratory chain, which either impairs the further proliferation of that cell lineage or leads to cell death.⁷ This would ultimately lead to a decrease in the percentage of mutated mtDNA in the daughter cells present in the peripheral blood. However, it is currently not known whether the biochemical defect is primarily because of high amounts of mutated mtDNA,⁸ low amounts of wild-type mtDNA⁹ or a combination of both.

To advance our understanding of this process, we developed and validated a highly sensitive fluorescent assay to measure the changes in heteroplasmy over time, and also measured the absolute amount of mutated and wild-type mtDNA in 11 subjects known to harbour 3243A→G.

METHODS

Subjects and control samples

Sequential venous blood samples between 2.0 and 7.8 years apart were collected from 11 subjects known to harbour the 3243A→G mtDNA mutation (five men and six women; table 1), and single blood samples were taken from 10 healthy controls. For all the samples, the peripheral blood leucocyte and platelet counts were within the normal range, with a normal proportion of each leucocyte subset. Skeletal muscle biopsy specimens were available for 9 of the 11 subjects with the 3243A→G mutation and 6 controls for comparison. Total genomic DNA was extracted using a standard procedure.

Establishing an accurate fluorescent 3243A→G heteroplasmy assay

A range of heteroplasmic samples was generated by mixing mutated and wild-type cloned fragments containing 3243 nucleotides generated from two siblings, one with a high percentage of the 3243A→G mutation and one with a low percentage 3243A→G, but who otherwise had an identical mtDNA sequence. Genomic DNA from each sibling was amplified across the region containing the MTTL1 gene and cloned using the pGEM-T Easy vector system (Promega, Southampton, UK), according to the manufacturer’s instructions. Plasmids from transformed JM109-competent Escherichia coli cells were purified using the Qiaprep spin mini prep kit (Qiagen, Crawley, UK) and quantified by ultraviolet spectrophotometry (Eppendorf, Cambridge, UK). Mutated and wild-type cloned DNA were mixed to generate heteroplasmic samples of 0%, 5%, 10%, 25% and 50%.

The percentage of mutated mtDNA was determined in triplicate using the gold standard technique of last cycle hot polymerase chain reaction (PCR). A 154-bp region encompassing the MTTL1 gene was amplified using standard cycling conditions and the following primers: nucleotides 3200–3218 forward, nucleotides 3353–3334 reverse. After the addition of 5 μCi [α-³²P]dCTP (3000 Ci/mmol) to the last PCR cycle, products were precipitated and digested at 37°C with HaeIII (10 U; New England Biolabs, Ipswich, UK). Restriction fragments were separated on a 12% non-denaturing polyacrylamide gel, dried on to a support, and exposed to a PhosphorImager cassette. The level of mtDNA heteroplasmy was quantified using ImageQuant TL software (Amersham Biosciences, Buckinghamshire, UK).¹⁰

The last cycle hot PCR method was adapted to a fluorescent analyser (Beckman CEQ 8000; Beckman Coulter, Buckinghamshire, UK) using modified primers (nucleotides 3155–3171 forward, nucleotides 3353–3334 reverse) and optimised PCR conditions. Fluorescent-labelled forward primer was added in the last PCR cycle instead of the radioactive label, followed by HaeIII digestion (10 U). Serial dilutions (in water) of the last cycle fluorescent PCR restriction fragment length polymorphism (RFLP) products were analysed in triplicate using a Beckman Coulter CEQ 8000 fluorescent DNA fragment analyser and fragment analysis software (V.8.0).

Real-time PCR

The total amount of mtDNA (referred to as the mtDNA copy number) in homogenate DNA samples from patients and controls was determined by real-time PCR using iQ Sybr Green on the BioRad ICycler (BioRad, Hercules, California, USA) to a target template spanning from nucleotide 3459 to nucleotide 3569 of the mtDNA ND1 gene. This was normalised using a nuclear-encoded template for the glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene spanning from nucleotide 804 to nucleotide 903 (NCBI, NM_002046.2). Relative copy number was calculated from the threshold cycle value, ΔCt value, where the mean amount of mtDNA/cell = 2(2^−ΔCt), to account for the two copies of GAPDH in each cell nucleus. Each reaction was optimised and confirmed to be linear over an appropriate concentration range using genomic DNA standards. Samples were analysed in triplicate for both assays, enabling calculation of the average mtDNA to nDNA ratio. As the overwhelming majority of blood mtDNA is present in the leucocyte population, this assay provides a good estimation of the average amount of mtDNA within the blood leucocyte population.

Statistical analysis

Statistical analyses were carried out using Minitab (V.14), including correlations analysis with Spearman’s correlation coefficient (R²), and two-sample t tests for the comparison of the amount of mtDNA.

RESULTS

Comparison of the radioactive and fluorescent last cycle assays

The last cycle hot PCR and last cycle fluorescent PCR methods were compared directly by analysing the mixed cloned DNA. Each assay was performed in triplicate from the mixed clone templates. Fluorescent-labelled products can be analysed in two ways: measuring the area under the curve or measuring the peak height amplitude. Both measurements are likely to be reliable when the signal to noise ratio is high, but the point at which they become unreliable has not been formally studied. We therefore compared both measurements over a range of dilutions with the current gold standard technique of the last cycle hot PCR-RFLP (fig 1).

In each case, there was a strong correlation between the different measures (R²>0.95), but the undiluted and 1:2 dilution of the last cycle fluorescent-labelled assay had the strongest correlation with the last cycle hot assay (fig 1). For greater dilutions, the last cycle fluorescent assay showed the most variation at lower percentage levels of heteroplasmy. We found no obvious difference between the area under the curve or peak height amplitude methods.

Sequential blood heteroplasmy quantification

The percentage of mutated mtDNA was determined in each sample from the 11 subjects harbouring the 3243A→G mutation using the last cycle fluorescent PCR-RFLP. Measurements were made from the original genomic DNA sample on three independent occasions. The last cycle fluorescent PCR-labelled HaeIII digestion products were not diluted. In each subject, the percentage of 3243A→G decreased over time (fig 2). The overall mean change was a decrease of 0.6% per year (standard deviation (SD) 0.5).

mtDNA copy number in leucocytes

The mean mtDNA copy number in control blood leucocytes was 370.1 (SD 61.9) (fig 3, table 1). All the samples from subjects with the 3243A→G mutation had low levels of mtDNA (mean for sample set 1 = 69.8 (SD 25.4), two-sample t test comparison with controls, t = −14.28, p<0.001; mean for sample set 2 = 98.0 (SD 52.8), two-sample t test comparison with controls t = −10.79, p<0.001). In eight of the subjects with the 3243A→G mutation, the mean amount of mtDNA per leucocyte increased over time. In three subjects, the mean amount of mtDNA per leucocyte decreased over time. The overall mean change was an increase in the mean amount of mtDNA per leucocyte (4.4 molecules/cell/year (SD 7.6); fig 4). We found no clear correlation between the total amount of mtDNA and the age of the subjects, the initial percentage of mutated mtDNA or the absolute amounts of mutated and wild-type mtDNA.

Skeletal muscle mtDNA

In each patient, the percentage of 3243A→G was greater in muscle than in blood. The amount of mtDNA per nucleus was greater in the skeletal muscle of 3243A→G subjects than of the controls (supplementary table A, available online at www.jmg.bmjjournals.com/supplemental).

DISCUSSION

Several conclusions can be drawn from this work. On the technical side, we have shown that the addition of a fluorescent-labelled oligonucleotide primer in the last PCR cycle enables the accurate quantification of mtDNA heteroplasmy. Although the strongest correlation was between the undiluted and 1:2 diluted fluorescent products and the last cycle hot PCR, all the assays correlated well with an R²>0.95. We were surprised to find that the serial dilution of the fluorescent PCR products did not have a dramatic effect on the assay. This probably reflects the automated electrostatic injection methods used by most capillary-based DNA analysers, which load a fixed amount of labelled DNA rather than a fixed volume of solution. It is possible that different results are obtained for manually loaded gel-based systems. Arguably, the reproducibility of the fluorescent assay exceeds that of last cycle hot PCR (see the approximately 50% 3243A→G data in fig 1), and, given the inherent risks of using radiochemicals, last cycle fluorescent analysis should now be the benchmark for measuring mtDNA heteroplasmy.

Using the last cycle fluorescent PCR, we confirmed that the percentage of 3243A→G decreases in blood over time. The mean percentage decrease of 0.6% per year corresponds well to the largest previous study (0.69% per year; SD 0.61).⁶ In our study (fig 2), the greatest rate of decrease in the percentage of mutated mtDNA was seen in the younger subjects, consistent with an exponential loss of 3243A→G from the peripheral blood. With this in mind, reporting mean values for the percentage decrease could be misleading, and explains why the mean value in our adult study was significantly less than that found in one study of six subjects (1.12% per year; SD 0.65),⁷ which predominantly involved observations made over childhood (two-sample t test comparing the data in table 1 with the data in Rahman et al⁷: t = 3.53, p = 0.004).

The most striking novel finding is the profound depletion of mtDNA in blood leucocytes in patients harbouring this mutation. Our control data for the mean mtDNA copy number/leucocyte are remarkably similar to other published values based on real-time PCR and other techniques,^11,12 confirming that we were using a reliable assay. How can we therefore explain these observations? Given that the peripheral blood count was normal in each case, a change in the number of cell nuclei or cell types probably does not account for the profound depletion we have observed. From first principles, the depletion of mtDNA could be due to decreased rates of mtDNA synthesis relative to the intense cellular proliferation in leucocyte precursors. An alternative (and non-exclusive) explanation is an increased loss of mtDNA from the precursors of the mature cells.

The peripheral leucocyte population is continuously recycled. Neutrophils are the largest single component (up to 80% of circulating leucocytes), maturing in the bone marrow for 11–12 days before being released into the circulation, where they remain for 6–8 h.¹³ The entire neutrophil population in the peripheral blood (approximately 5×10¹⁰ cells) is therefore replenished on a daily basis, requiring a massive amount of mtDNA replication to maintain normal mtDNA levels. This is sustained by a limited number of stem cells (1 in 10⁴ to 1 in 10⁵ marrow cells), which turn over at a low rate and repopulate the blood through a hierarchical cascade of cell divisions.¹⁴ It is therefore far more probable that any effect on mtDNA replication would manifest during the massive expansion of daughter cells leading to the formation of mature leucocytes, rather than in the stem-cell population itself. Even a subtle decrease in the replication rate could lead to the dramatic depletion that we have observed in the peripheral blood. For example, if the mtDNA replication rate was a fraction f of the normal replication rate, then after a series of n cell divisions the leucocyte precursors would be depleted by a factor fⁿ. Twenty cell divisions (the number required for a million-fold increase in cells) would give our observed depletion to 23% of the normal level with f = 0.93.

Could the bioenergetic defect be responsible for a relative decrease in rate of mtDNA synthesis in the leucocyte precursors? Although plausible, this explanation is difficult to reconcile with the increased amount of mtDNA seen in post-mitotic tissues of subjects harbouring 3243A→G.¹⁵ In addition, if the loss of mtDNA were directly related to a defect of ATP synthesis, we would expect a closer relationship between the percentage level of mutated mtDNA and the degree of depletion. This was clearly not the case—for example, low amounts of mtDNA were found in a young subject with approximately 50% 3243A→G in blood and an older subject with approximately 5% 3243A→G in blood (figs 2 and 3). Recent in vitro studies on human NT2 teratocarcinoma cybrids harbouring 3243A→G (NT2.3243 cybrids) cast light on this issue.¹⁶ High percentages of 3243A→G were associated with a profound loss of mtDNA from the NT2.3243 cybrids. This seemed to be mutation specific, as the same depletion was not seen in NT2 cybrids homoplasmic for the 1555A→G mtDNA mutation. Moreover, the presence of a known suppressor mutation (12300G→A), which ameliorates the biochemical effects of 3243A→G at high levels, did not prevent the loss of mtDNA from the cybrid cell lines, indicating that the biochemical defect is not the principal factor behind the depletion. Remarkably, not all the cell lines with >99% 3243A→G lost their mtDNA, proving that the percentage of mutated mtDNA is not an absolute determinant of the depletion. Further work showed slow rates of segregation of the 3243A→G mutation in NT2.3243 cybrids, providing indirect evidence that functional or physical partitioning of mtDNA molecules contributes to the segregation process. These observations highlight the complex interplay between intercellular and intracellular signals that are involved in the mitotic segregation of the 3243A→G mutation.¹⁶ Similar mechanisms may be involved in rapidly dividing tissues in vivo, including the leucocyte population. It is intriguing that, as the percentage of mutated mtDNA in blood decreases in subjects harbouring the 3243A→G mutation (fig 2), there is an overall increase in the amount of mtDNA in blood, particularly wild-type mtDNA (fig 4). This suggests a degree of recovery from the depletion as time passes, although the amount never reaches normal values.

An alternative explanation for the loss of one mtDNA genotype in blood is through immune surveillance mediated through an abnormal cell surface epitope. This hypothesis is supported by the description of a complex I (ND) gene variant in mice that forms a maternally transmitted major histocompatibility antigen.¹⁷ However, recent work on heteroplasmic mice has shown that immune-mediated cell loss does not occur in heteroplasmic mice with the C57B/BALB genotype, and nuclear genes may be important.¹⁸

Finally, the depletion may be a secondary phenomenon, as in other mitochondrial disorders. Mutations in OPA1 cause the most common form of autosomal dominant optic atrophy. In a recent study, subjects with OPA1 mutations had markedly lower levels of mtDNA in leucocytes than controls.¹¹ Given the central role of OPA1 in maintaining the structural integrity of mitochondria and mtDNA, it was suggested that the primary genetic defect led to the loss of both mitochondria and mitochondrial genomes. A similar structural disintegration of the mitochondrial network has been seen in subjects harbouring primary mtDNA defects, potentially explaining the observations we report here. By contrast, one study¹⁹ described a mild increase in the amount of mtDNA in leucocytes from subjects harbouring mtDNA mutations that cause Leber hereditary optic neuropathy. However, in contrast with the 3243A→G mutation, the biochemical defect in patients with Leber hereditary optic neuropathy is mild and sometimes undetectable.^20,21 Further work will determine whether depletion of mtDNA is seen in other mtDNA disorders.

If the 3243A→G mutation is causing depletion through an effect on the structural integrity of mitochondria in leucocytes, why does this not occur in non-dividing tissues such as skeletal muscle, where there tends to be a proliferation of mitochondria and mtDNA (supplementary table A online at www.jmg.bmjjournals.com/supplemental)? At present, we can only speculate, but recent work in our laboratory has shown that, although initially the amount of skeletal muscle mtDNA is high in subjects with mtDNA mutations, there is a progressive loss of mtDNA over time.²² This could be partly due to muscle deconditioning, but it may be due to a more general process related to mitochondrial fragility. These findings have important implications for the molecular diagnosis of mitochondrial disorders—the presence of depletion may be a secondary phenomenon and the primary molecular defect may be a mtDNA mutation or a nuclear gene mutation not directly involved in the synthesis of mtDNA.