Here, we show that POLRMT, in addition to its essential role in gene expression, is required for primer formation for initiation of mtDNA replication in vivo. Our data show that primer synthesis and initiation of mtDNA replication are prioritized over gene transcription because (i) there is a differential effect of POLRMT at the mtDNA promoters as illustrated in
Fig. 7. At lower POLRMT levels, transcription is primarily initiated at LSP, which helps to ensure that primer synthesis can be maintained. (ii) Under normal conditions, about 95% of all replication initiation events at O
H are prematurely terminated, forming the 7
S DNA. When POLRMT is depleted, 7
S DNA is no longer formed, suggesting that all residual replicative events continue to full-length replication. (iii) The increase in TWINKLE protein levels is likely a compensatory response aimed to promote productive mtDNA replication. The increased TWINKLE protein levels are most likely due to posttranscriptional regulation because the
Twinkle mRNA levels remain unchanged. Although the exact mechanisms of TWINKLE stabilization and regulatory role in mtDNA replication remain unknown, TWINKLE can be loaded at the 3′ end of 7
S DNA to promote full-length genomic replication (
58). It is possible that abortive mtDNA replication is favored by an antihelicase activity at the end of the D loop (
58,
59) and that increased TWINKLE levels will overcome this block. The increase in TWINKLE as a compensatory mechanism agrees with results from other mouse models with severe mtDNA depletion, suggesting an involvement of TWINKLE in the regulation of mtDNA replication in mammals. Surprisingly, the TFAM protein levels remain normal in the absence of POLRMT despite profound mtDNA depletion. The inability of this excess TFAM to bind mtDNA to form nucleoids results in an increased free pool of TFAM. There is strong experimental evidence that the amount of TFAM directly regulates mtDNA copy number and that mtDNA levels also reciprocally affect TFAM levels (
21,
56,
60). Disruption of
Tfam in mouse leads to loss of mtDNA, and overexpression of
TFAM leads to increased mtDNA copy number (
22,
45). Our data show that TFAM can be stable in the absence of mtDNA. The normal TFAM protein levels observed here could be a part of a compensatory response that attempts to maintain mtDNA. It has been suggested that TFAM turnover involves LONP, which degrades TFAM when it is not bound to mtDNA (
24,
25). Overexpression of
LONP was reported to result in reduced TFAM and mtDNA levels, whereas
LONP depletion was found to increase TFAM and mtDNA levels. The TFAM protein levels are normal despite increased LONP levels in the absence of POLRMT. TFAM is thus protected from LONP degradation. Moreover, the levels of
Tfam transcripts were increased in the
Polrmt knockout hearts, whereas all other investigated transcripts of nuclear genes involved in mtDNA expression were unaltered. These findings argue for a compensatory regulatory mechanism that controls TFAM levels in mitochondria. To our knowledge, this discrepancy between mtDNA and TFAM levels has not been found in other knockout mouse models with reduced mtDNA levels (
28), and it remains to be clarified whether this is specifically linked to the loss of POLRMT. LONP is known to play a role in degrading misfolded, unassembled, or oxidatively damaged proteins. Thus, the highly induced levels of LONP in the
Polrmt knockouts are likely due to an imbalance between mtDNA- and nucleus-encoded OXPHOS subunits, as already described in previous studies (
61). The mechanisms that regulate transcription levels in mammalian mitochondria remain to be clarified, and our study of the heterozygous
Polrmt knockout mice provides important insights. Despite a drop in POLRMT levels, the steady-state mitochondrial transcript levels in these mice are unchanged. We also observed increased TEFM protein levels, whereas all other nuclear-encoded factors involved in mtDNA expression, for example, LRPPRC, TFAM, and TWINKLE, were normal. This could be a compensatory response to ensure a normal transcription rate and mitochondrial steady-state transcript levels even when the ratio of POLRMT protein per mtDNA molecule is reduced in the heterozygous
Polrmt knockouts. In vitro studies recently reported that TEFM increases transcription processivity to allow mitochondria to increase transcription rates (
43,
44). In contrast to TEFM, neither TFAM nor TFB2M levels were affected in the heterozygous
Polrmt knockout mice. Both of these factors are directly involved in the transcription initiation mechanism, but apparently, their levels are not limiting in promoting increased levels of transcription. In the nucleus, differential activation of the large number of protein-coding genes is not controlled by varying the amount of RNA polymerase II but is instead based on specific combinations of bound transcription factors that regulate promoter specificity. In contrast to the nucleus, where protein-coding genes typically are present only in two copies per cell, there are thousands of copies of mtDNA per cell. This means that increased transcription of mtDNA does not necessarily require increased transcription from each promoter, but it can instead be achieved by engaging more mtDNA templates in transcription. This mode of regulation will only be an efficient regulatory system if the number of target promoters is low, such as is the case for mtDNA that only has two promoters per genome. How the presence of TEFM may prevent early transcription termination and promote transcription beyond the immediate promoter region remains to be discovered. The genetic data we present here resolve the controversy concerning the role of POLRMT as a primase for mammalian mtDNA replication. This issue has been widely debated because nuclear DNA replication is dependent on specific primases, which are distinct from the RNA polymerases needed for transcription of nuclear genes. Our mouse knockout data show that in the absence of POLRMT, 7
S DNA is no longer formed, and there is a severe reduction of mtDNA levels. In vitro data have shown that POLRMT can generate the RNA primers needed for mtDNA synthesis at both O
H and O
L (
40), but the in vivo importance of these findings has not been established (
62). The 7
S RNA, a polyadenylated transcript that is terminated near the 5′ end of the nascent D loop strand, is undetectable in the absence of POLRMT. The function of the 7
S RNA is not clear, but it has been suggested that it is involved in primer formation for initiation of mtDNA synthesis at O
H (
36,
58). In mammalian mitochondria, PrimPol was reported to have DNA and RNA primase activities and play a role in mtDNA replication (
35). Our knockout data show that PrimPol or any other primase cannot complement for the absence of POLRMT when it comes to mtDNA replication initiation. Together, our study provides clear evidence that POLRMT functions as the primase for mtDNA replication in mammalian mitochondria in vivo. Deletion of
Polrmt in heart results in a severe decrease in all mitochondrial transcripts, but we also report that the mitochondrial transcripts derived from the L strand, in particular the
mt-Nd6 mRNA, are less reduced than those transcripts encoded on the H strand. This could be explained by different stabilities of the individual mitochondrial transcripts due to up-regulation/down-regulation of factors involved in mitochondrial transcript processing/stability. Previous studies reported that most L-strand transcripts, including
mt-Nd6, were unaffected when silencing or knocking out the
Lrpprc gene (
54,
63). LRPPRC levels decrease markedly once POLRMT levels drop below 50% of control levels to cause markedly decreased mitochondrial transcription. This finding is consistent with the previously observed reciprocal interdependency between levels of LRPPRC protein and mitochondrial mRNAs (
64). The strongly increased GRSF1 levels in
Polrmt knockouts are in line with the less reduced amounts of
mt-Nd6 mRNA in comparison with levels of H-strand transcripts, because GRSF1 has been suggested to be involved in regulating and interacting with
mt-Nd6 mRNA and its precursor strand (
51,
52). However, these circumstances cannot explain why tRNAs from the L strand are present at higher levels than tRNAs of the H strand. Hence, a likely explanation for the difference in transcript steady-state levels is the observed promoter-specific transcription initiation effects in heart mitochondria with low POLRMT levels. This is underscored by our in vitro findings that transcription initiation at LSP is better maintained than at HSP at low POLRMT levels. This difference in the strengths of the two promoters is in line with previous in vitro studies (
15). In conclusion, we found that POLRMT is essential for embryonic development, and that whole-body knockout results in embryonic lethality during midgestation, whereas a moderate reduction of
Polrmt expression does not cause any phenotype. We further provide the first in vivo evidence that POLRMT functions as the primase for mtDNA replication in mammalian mitochondria. At low POLRMT levels, we observed a significant discrepancy in the initiation of mitochondrial transcription between the HSP and LSP, suggesting that POLRMT is part of a mechanism that provides a switch between RNA primer formation for mtDNA replication and mtDNA expression. Coordinated replication and expression of the mitochondrial genome are essential for metabolically active mammalian cells. By using several mouse models, we characterized the molecular consequences of different expression levels of POLRMT, a key player in those processes and thus a good target for therapeutic strategies.