107
6 DISCUSSION
For nearly the last three decades it was thought that the mechanism of mammalian mitochondrial DNA replication had been elucidated thoroughly and only the identification of the relevant replicative proteins and their characterization remained. While the investigation of plant and yeast mtDNA maintenance has been hampered by the evident complexity of the processes involved, much of the mammalian mtDNA research has been biased by the assumptions of mechanistical simplicity and failure to examine tissue- or species-specific features. Most of the research has been concentrated on cultured cells or on only one model mammal – the mouse. The work by Kajander et al. (2001) was one of the first indications that the extrapolated view is inaccurrate, because it described for the first time abundant recombination as a completely new feature of mammalian mtDNA. The observation was obviously not just anecdotal, as the relative quantities of junctional molecules were much higher than those reported in other systems with well-documented active recombination. It was obvious that these molecules represented some central aspects of mtDNA maintenance in healthy human heart, although the finding has been largely ignored by the field.
In this series of studies, I conducted extensive investigations of mammalian mtDNA replication under various conditions in cell culture as well in different tissues of the organism. I found new functions for genes known to be involved in mitochondrial transcription or mtDNA maintenance, characterized replication phenotypes of catalytically-defective replisome proteins, revealed physiologically-significant tissue-specific differences in mtDNA replication, and showed that these features could be manipulated in a transgenic model organism. Finally, by applying several different analytical methods in concert, I revealed evidence for a novel mechanism of mtDNA replication in human heart. Taken together, the work published in original communications I-IV has helped to deepen our understanding of the maintenance of mammalian mtDNA.
6.1 Mitochondrial DNA replication in mammalian tissues and cultured
108
2000, Yang et al. 2002, Yasukawa et al. 2006]). A paper by Brown et al. (2005) questioned these 2DNAGE findings, suggesting that the apparently-duplex mammalian mtDNA replication intermediates most likely arise from artifactual branch migration of strand displacement replication intermediates. Based on AFM evidence, they furthermore proposed that there are additional light strand origins that would also explain some of the 2DNAGE findings. This notion is not new. In fact, Wolstenholme et al. (1973) as well as Pikó and Matsumoto (1977) reported at least two double-stranded regions on the displaced lagging-strand of mouse mtDNA.
I have shown by TEM that mtDNA extracted by our method from sucrose gradient- purified mitochondria contains mostly fully double-stranded theta-type replication intermediates, which can be converted into partially single-stranded species by RNase H digestion, as predicted by the 2DNAGE results (unpublished data of Pohjoismäki et al. 2007; submitted manuscript not included in this thesis). Furthermore, these TEM results are in agreement with the work done by Kirschner et al. (1968) and Wolstenholme et al. (1973), who conducted their studies on partially purified mtDNA without sucrose or CsCl gradient centrifugation steps. All replication intermediates seen by Kirschner et al. (1968) from rat liver were fully duplex, and 10-60% of hepatoma cell line mtDNA replication intermediates were fully duplex theta molecules reaching up to 85% of the genome (Wolstenholme et al. 1973). The partially single-stranded molecules seen by Wolstenholme et al. (1973) could result from degradation of RNA in crude mitochondrial preps, as reported by Yang et al. (2002).
Our experiments show that carefully isolated mtDNA contains RITOLS replication intermediates, although there is, as yet, no evidence as to how they arise. Several possibilities remain open. In the light of the SDM, RNA incorporation would be a good alternative to protect the displaced lagging-strand from degradation instead of having long stretches coated by SSB. The RNA could be synthesized de novo or incorporated from preformed RNA at the replication fork from 3´ to 5´ by an RNA helicase activity. One candidate for such a helicase might be hSUV3, a poorly-characterized mitochondrial RNA-DNA helicase–like protein (Minczuk et al. 2002, 2005).
An obvious candidate for de novo RNA synthesis would be the mitochondrial RNA polymerase MTRPOL. If MTRPOL is coupled to the replication fork, the synthesis of the RNA lagging-strand would be discontinuous, just like in the case of Okazaki fragments. These RNA-Okazaki fragments could then function as elongated RNA primers for delayed lagging-strand synthesis. If MTRPOL could be shown to initiate promoter independent-transcription from a single-stranded DNA template without the presence of any transcription factors, it might provide alternative means for RNA-DNA hybrid formation. As mentioned earlier, the E. coli RNA polymerase is capable in synthesizing persistent RNA-DNA hybrid on single-stranded DNA template (Chamberlin & Berg 1964).
109
If preformed RNA is incorporated by an active mechanism at the replication fork, one would expect to see overhanging RNA molecules that are drawn in at the replication fork like shoelaces (Figure 6.1.), hence the term bootlace model (Yasukawa et al. 2006). The 2DNAGE data is very suggestive in the favour of this model. It is possible to clean up a 2D image using the single- strand RNA digesting enzyme RNase If: the heterogeneous high molecular-weight intermediates vanish and better defined RITOLS intermediates appear (Figure 6.1.). For the TEM experiments the samples also had to be treated with RNase If to remove massive amounts of heterogeneous single- stranded RNA in order to visualize any DNA. If any overhanging RNA was present on the DNA molecules, it was lost in the procedure. A careful examination of replication forks seen in TEM show that the single-strandedness at or near the fork is usually very short, indistinguishable from standard strand-coupled replication. If the RNA on the lagging-strand resulted from an artefactual hybridization of RNA on ssDNA during mtDNA purification, some discontinuities would be expected. RNA tends to form secondary structures and these would need to be opened before such an even hybridization on long stretches of ssDNA can be achieved.
Figure 6.1. The bootlace model intermediates on 2DNAGE. Pre-formed RNA is hybridized on the lagging-strand at the replication fork resulting in heterogeneous replication intermediates with varying lengths of overhanging RNA. RNase If cleaves the overhanging RNA resulting in better defined RITOLS intermediates.
In TFAM-overexpressing cells the increase in the dsDNA RIs coincides with transcriptional depletion (Article I). As TFAM overexpression causes several changes in mtDNA, it is hard to draw conclusions of the exact causal relationships. The loss of heterogeneous RNA intermediates could be due to a lack of available transcripts close to the site of replication. This observation is supported by the fact that the increased steady-state levels of RNA per template
110
mtDNA in TFAM RNAi cells result in a corresponding increase of heterogeneous RNA intermediates (Figure 5.11). This can be due to the reduced compaction of the DNA template, resulting in better access for RNA polymerase.
However, the effect of TFAM overexpression in cultured cells on mtDNA replication is similar to that caused by ddC and catalytic mutants of PolG and Twinkle, although the effect of TFAM is more specific, being especially pronounced in the rDNA regions. As TFAM overexpression causes compaction of mtDNA (Kaufman et al. 2007), the decreased rate of fork progression is likely to be due to the replisome having difficulties in progressing through the coiled molecule. As in other cases of replication stalling (I, II), TFAM overexpression results in mtDNA depletion over time.
The catalytic mutants of PolG and Twinkle provide some important insight into the nature of lagging-strand synthesis in RITOLS replication. While the severe PolG mutants “freeze”
all replication intermediates, the severe Twinkle mutants selectively enhance dsDNA replication intermediates. Coinciding with mtDNA copy number depletion, the accumulation of RIs in both cases is a hallmark of severe replication stalling. In the case of Twinkle variants, this results from the fact that the mutant helicase is unable to unwind the DNA duplex and replication fork progression is halted. Delayed lagging-strand synthesis catches up with the retarded replication fork, resulting in fully dsDNA replication intermediates. More importantly, replication intermediates of all sizes, including those with forks close to OH, are converted to dsDNA RIs, indicating that lagging-strand synthesis can initiate at any region of the genome (e.g. Figure 5.13). Unless SDM with random initiation of the lagging-strand synthesis is suggested, which in any case would be close to the conventional (COSCOFA) model, the data strongly indicates that the RNA of the lagging-strand can be matured to DNA from any given point. When PolG is inhibited by ddC in addition to the presence of the catalytically-inactive mutant Twinkle, which inhibits fork progression inducing stalling, the RITOLS intermediates reappear (Figure 5.14). This ddC treatment has little further effect on the progression of the leading-strand fork, which is already extremely slowed. Since ddC has a drastic effect on the rate of lagging-strand synthesis, it supports the idea that PolG is responsible for the DNA maturation and the RITOLS intermediates are genuine precursors of the fully-dsDNA species detected in control cells. Their maturation presumably requires processing of the long hybridized RNAs into shorter RNA primers for the DNA polymerase to employ.
As a further objection to the SDM, OH appears to function as a definite replication terminus for both leading- and lagging-strand synthesis, as demonstrated by single-cleavage 2DNAGEs and TEM, meaning that the first lagging-strand origin is at or close to the OH. It may
111
well be that the sister molecules separate before lagging-strand maturation is complete, resulting in nicked molecules seen on 2DNAGE. The end product of replication might be catenated molecules, as proposed earlier (Clayton 1982). Unlike trypanosome minicircles that need to be decatenated prior to replication, it seems that this is not the case with human mtDNA (Figure 5.1B).
In conclusion, RITOLS replication appears to be the main replication mode of mtDNA in mammalian cultured cells, liver, kidney and skeletal muscle. The “bootlace” maturation model is supported by a variety of experiments in cultured cells.