NCB-C33065-R3 Segregation of mitochondrial DNA heteroplasmy through a development genetic bottleneck in human embryos Vasileios I. Floros*,1,2 Angela Pyle*,3 Sabine Dietmann,4 Wei Wei,1,2 Walfred W.C. Tang,5 Naoko Irie,5 Brendan Payne,4 Antonio Capalbo,6,7 Laila Noli,8,9 Jonathan Coxhead,4 Gavin Hudson,4 Moira Crosier,10 Henrik Strahl,11 Yacoub Khalaf,8,9 Mitinori Saitou,12,13 Dusko Ilic,8,9 M. Azim Surani,5 Patrick F. Chinnery.1,2¶ 1MRC-Mitochondrial Biology Unit, University of Cambridge, Cambridge, CB2 0XY, UK. 2Department of Clinical Neurosciences, Cambridge Biomedical Campus, University of Cambridge, Cambridge, CB2 0QQ, UK. 3Wellcome Trust Centre for Mitochondrial Research, Institute of Genetic Medicine, Central Parkway, Newcastle University, Newcastle upon Tyne, NE1 3BZ, UK. 4Wellcome Trust-Medical Research Council Stem Cell Institute, Tennis Court Road, University of Cambridge, Cambridge CB2 3EG, UK. 5Wellcome Trust Cancer Research UK Gurdon Institute, Tennis Court Road, University of Cambridge, Cambridge, CB2 3EG, UK. 6GENERA, Centre for Reproductive Medicine, Clinica Valle Giulia, 00197 Rome, Italy. 7GENETYX, Reproductive Genetics Laboratory, Via Fermi 1, 36063 Marostica, Italy. 8Division of Women’s Health, Faculty of Life Sciences and Medicine, King's College London, London SE1 7EH, UK. 9Assisted Conception Unit, 11th Floor Tower Wing, Guy’s Hospital, Great Maze Pond, London SE1 9RT, UK. 10Human Developmental Biology Resource, Institute of Genetic Medicine, Central Parkway, Newcastle University, Newcastle upon Tyne, NE1 3BZ, UK. 11Centre for Bacterial Cell Biology, Institute for Cell and Molecular Biosciences, Baddiley-Clark Building, Richardson Road, Newcastle University, Newcastle upon Tyne, NE2 4AX, UK. 12Department of Anatomy and Cell Biology, Graduate School of Medicine, Kyoto University, Yoshida-Konoe-cho, Sakyo-ku, Kyoto 606-8501, Japan. 13JST, ERATO, Yoshida-Konoe-cho, Sakyo-ku, Kyoto 606-8501, Japan. *equal contribution ¶ Correspondence to: pfc25@cam.ac.uk Mitochondrial DNA (mtDNA) mutations cause inherited diseases and are implicated in the pathogenesis of common late-onset disorders, but how they arise is not clear1,2. Here we show mtDNA mutations are present in primordial germ cells (PGCs) within healthy female human embryos. Isolated PGCs which have a profound reduction in mtDNA content, with discrete mitochondria containing ~5 mtDNA molecules. Single cell deep mtDNA sequencing of in vivo human female PGCs showed rare variants reaching higher heteroplasmy levels in late PGCs, consistent with the observed genetic bottleneck. We also saw the signature of selection against non-synonymous protein-coding, tRNA gene and D-loop variants, concomitant with a progressive upregulation of genes involving mtDNA replication and transcription, and linked to a transition from glycolytic to oxidative metabolism. The associated metabolic shift would expose deleterious mutations to selection during early germ cell development, preventing the relentless accumulation of mtDNA mutations in the human population predicted by Muller’s ratchet. Mutations escaping this mechanism will show shifts in heteroplasmy levels within one human generation, explaining the extreme phenotypic variation seen in human pedigrees with inherited mtDNA disorders. The 16.5 Kb human mitochondrial genome (mtDNA) encodes 37 genes essential for the synthesis of 13 polypeptide subunits of the respiratory chain, the final common pathway of oxidative metabolism1. In humans, mtDNA is inherited exclusively down the maternal line. Diploid cells typically contain 1-10,000 mtDNA molecules partitioned between a network of fusing and budding mitochondria. Exposed to a potent source of oxygen free radicals and lacking protective histone proteins, mtDNA is particularly vulnerable to mutation, creating a mixed population of wild-type and mutated mtDNA within a cell (heteroplasmy)2. MtDNA mutations accumulate in in non-dividing (post mitotic) cells during life, with high levels observed in late onset human disorders including Alzheimer’s disease and Parkinson’s disease. Although the overall tissue mutation burden is low, very high levels within single cells lead to bioenergetic failure and ultimately cause cell death1. These findings raise the possibility that mtDNA mutations contribute to the pathogenesis of several common human diseases, and also play a part in the ageing process itself3, but the origin of these mutations has not been unclear. Age-associated mtDNA mutations were assumed to have arisen in somatic tissues during life, but deep sequencing of maternal relatives raises the possibility that low-level mtDNA heteroplasmic point mutations are inherited down the maternal line4. To determine whether this is the case, we developed methods to isolate pure primordial germ cells (PGCs) from first trimester karyotypically normal human female embryos (Fig.1). Anatomical regions containing migrating PGCs were dissected from fresh 4-week embryos, and genital ridges were isolated from 5-8 week embryos. Single cell suspensions were then labelled with fluorochrome-conjugated antibodies against putative human germ cell surface markers, and sorted using fluorescence-activated cell sorting (FACS). Tissue non-specific alkaline phosphatase (TNAP) histochemistry5, VASA immunocytochemistry6 (Supplementary Fig.1), and germ-cell specific transcripts (Supplementary Figs.1&2)7, were used to determine the proportion of PGCs after FACS sorting at appropriate developmental stages. Highly purified (>97%) PGCs were isolated in 4 week (Carnegie stage, CS12) embryos using a combination of TNAP, stage-specific embryonic antigen 4 (SSEA4) and the PGC-associated cell-surface marker CD38; and for weeks 5-8 (CS16/17 – CS20/21) we used TNAP and c-KIT antibodies (Fig.1b). After cell lysis, mtDNA was amplified in overlapping 2Kb fragments to minimise potential contamination from nuclear-encoded mitochondrial insertions (NumtS), before >5000-fold depth re-sequencing. NumtS contamination was further minimized by filtering-out all variants with a minor allele frequency (MAF) <1%. Sequencing of a cloned mtDNA template showed no variants present at ≥1% MAF (Supplementary Fig.3). However, we observed mtDNA single nucleotide variants (SNVs) at >1% heteroplasmy in PGCs isolated from all 12 embryos studied (Fig.2a), which did not map to known human NumtS8. This shows that low-level mtDNA heteroplasmy is present within the female germ line in healthy humans. The variants were predominantly nucleotide transitions, which are less prone to in vitro sequencing error and less likely to be due to oxidative damage9, and are consistent with errors during in vivo DNA synthesis by mtDNA polymerase ϒ or the deamination of cytidine and adenosine10. We also observed a change in the mutation spectrum from CS12 to CS20/21 (Fig.2a). For protein coding genes, there was a decrease in the proportion of non-synonymous variants (non-synonymous/synonymous ratio, P=0.02) and a corresponding change in the codon bias (Fig.2b&c). We also saw a decrease in the mutation rate / base pair for tRNA genes (Fig.2d), which are hallmarks of selection against potentially deleterious mtDNA variants. Similar patterns have been seen in mice11,12, but the timing and mechanism of the selection was not known. Based on our observations, this occurs at an early stage in the developing human female germ line during PGC specification and migration. We were surprised to see some evidence of selection against variants in the non-coding mtDNA D-loop (Fig.2d), but these did not appear to be evenly distributed. CS12 PGCs contained variants in conserved regions not present in CS20/21 PGCs (Fig.2e), consistent with selection against variants in regions involved in mtDNA transcription and replication13. Given that the majority of the mtDNA variants were present at low heteroplasmy levels (Fig.2a), below the conventional threshold required for a biochemical effect, it was not immediately apparent how the selection could occur. To explore the potential mechanism, we measured the absolute amount of mtDNA in the human PGCs at different stages of development in vivo, and then sequenced individual late-stage PGCs to look for evidence of variant segregation. First we studied human oocytes from three healthy donors, which contained a mean of 1.22 x106 mtDNA molecules (SD=189,742, n=3)14. This was in keeping with previous measurements15 and confirmed assay reliability. The amount of mtDNA in human PGCs from 14 embryos in vivo was ~1000-fold less than in oocytes (CS12 mtDNA copies: median 1425, mean 1446, SD=77) (Fig.2f), consistent with a human genetic bottleneck. The mtDNA levels in female and male PGCs were similar, as previously seen in mice16, and resembled new measurements made in Blimp1-mVenus and Stella-eCFP [BVSC] mice17 (Fig.2f, Mouse E9.5 corresponds to human CS12; and mouse E12.5 corresponds to human CS20/21). This validates previous findings in mice with different nuclear genetic backgrounds16,18,19, and shows a similar mechanism in both humans and mice. However, the lower amount of mtDNA we observed in human PGCs (human/mouse: CS12/E9.5, P<0.0001; CS16-7/E10.5, P<0.0001; CS20-21/E12.5, P<0.0001) predicts a ‘tighter’ genetic bottleneck in humans than in mice, and thus the more rapid segregation of heteroplasmy. This is in accordance with observations in human and mouse pedigrees where more dramatic shifts in heteroplasmy are seen in humans (see Fig.4a in Ref20). It is not technically possible to isolate human PGCs from earlier stages because the cells are collected from clinical samples. However, to gain further insight we also measured mtDNA levels in single cells from both the human inner cell mass (ICM) and trophectoderm (TE) cells (Fig.2g). Although we saw a wide range of mtDNA levels in ICM cells, the mean level (9794, SD=5661) was >5-fold greater than in PGCs. We then studied embryonic stem cells (hESCs) and human PGC-like cells (hPGCLCs)7,21 which are derived from the embryonic epiblast in vitro, thus allowing us to estimate the amount of mtDNA at an intermediate stage of human germ cell development (Fig.2h). hESCs (mean=3264, SD=378) contained less mtDNA than ICM cells, and hPGCLCs (mean=896, SD=368) contained less mtDNA than hESCs (P=8.9x10-33). These findings are in keeping with a post implantation mtDNA bottleneck. The female hPGCLCs (mean=533, SD=203) contained significantly less mtDNA than male hPGCLCs (mean=991, SD=342, P=1.5x10-5), and less mtDNA than both male (mean=1102, SD=426, P=2.2x10-16) and female (mean=972, SD=248, P=2.2x10-16) somatic cells isolated from the same in vitro system but not expressing germ-cell markers. These findings suggest a more stringent mtDNA bottleneck in the human female germ line. Returning to the PGCs isolated in vivo, next we analysed the sub-cellular distribution of mtDNA molecules within the developing human female germ line. We used the dissector method to measure the number of mitochondria in human PGCs on adjacent section electron micrographs (EM) (Fig.3a), validating our results with super-resolution whole cell imaging using Nikon Structured Illumination Microscopy in 3-dimensions (N-SIM, Supplementary Movies 1 & 2, Fig.3b). The number of mitochondria did not change between CS12 and CS20-21, allowing us, to estimate the number of mtDNA per mitochondrion at ~5 per organelle in human PGCs. The reduction in mtDNA content after fertilization of the oocyte (Fig.2f), and associated intracellular compartmentalisation of mtDNA (Figs.3a&b), has important implications for the distribution of mtDNA heteroplasmy between individual organelles. Assuming that mutated mtDNA molecules are randomly distributed between mitochondria, higher levels of heteroplasmy within a cell will be associated with a greater proportion of mitochondria containing very high levels (>80%) of a specific mutation (Fig.3c). For example, if the average level within a cell is 40-50% mutant, then 10-20% of the organelles will contain >80% mutant mtDNA, exceeding the threshold required to cause a biochemical defect for pathogenic mutations1,2. In this context, ~10-20% of the organelles would be dysfunctional and potentially vulnerable to removal by mitochondrial quality control mechanisms, removing mutated mtDNA through selection at the organellar level. To determine whether this is plausible, we performed deep mtDNA sequencing on individual late-stage in vivo PGCs (Fig.3d&e). This showed much higher percentage levels of mtDNA heteroplasmy within individual germ cells, reaching up to 47% mutant. Thus, following the bottleneck, high levels of mtDNA mutations are found in single PGCs, and are predicted to reach very high levels in individual mitochondria (Fig.3c). The most likely explanation for the observed change in codon bias seen between CS12 and CS20-21 (Fig.2b), in the is the rapid segregation of variants above and below the assay detection threshold (1% MAF) due to the genetic bottleneck (Fig.2a&c), rather than de novo mutation events per se. In this context, even a subtle bias against deleterious variants would tend to suppress them below the detection threshold, whilst allowing non-functional variants to segregate freely to high levels (Fig.3d). To cast light on the possible mechanism of selection, we performed RNA-seq on weeks 4-9 human PGCs. Unsupervised hierarchical clustering (UHC) of gene expression and principal component analysis (PCA) of transcriptome confirmed strong expression of germ-cell specific transcripts (Fig.4a-c). Importantly, we also observed a distinct transcript profile for the nuclear-encoded mitochondrial proteins (MitoCarta2.0)22 in the early stages of PGC development (Fig.4d). Principal component analysis (PCA) of both the nuclear-encoded and the mtDNA-encoded transcripts showed that the most dramatic changes were during the early stages of germ cell development, with mitochondrial genes forming a cluster distinct from late gonadal PGC genes, such as DAZL and DDX4 (Supplementary Fig.4). The mtDNA-encoded transcripts had the highest scores on PC1, accounting for 92% of the total variance (Fig.4e). This is consistent with the developmental progression of mtDNA gene expression during PGC development. In contrast, genes involved in the canonical glycolysis pathway had opposing trends on PC1 (Supplementary Fig.4&5). Although based solely on transcript analysis, these findings suggest that PGCs become less dependent on glycolytic metabolism and more dependent on oxidative metabolism as they develop. As a consequence, non-synonymous mtDNA mutations are progressively more likely to have phenotypic consequences during germ cell development. The proliferation of PGCs between CS12 and CS20/21 requires considerable mtDNA replication, inevitably introducing new mutations through ϒ polymerase errors. These new mutations are more likely to be synonymous (Fig.2) and have no functional consequence. The switch from glycolytic to oxidative metabolism provides an explanation for these findings, by exposing functional variants to the selective effects during this critical time period of PGC development. Germ-line mtDNA selection at the organelle level has been proposed in mammals12,23 and Drosophila24,25, but it is not known how this occurs. One possibility is that mitochondria depolarise when the proportion of mutated respiratory chain subunits is high, activating intracellular surveillance systems mediated by Pink1 and Parkin, and leading to the selective removal of damaged organelles containing mutant mtDNA by mitophagy26,27. Alternatively, impaired mtDNA replication in ATP-deficient organelles28 could leading to the preferential turnover of mitochondria containing wild-type genomes; or the transport of mitochondria may be compromised in deficient organelles, preventing their coalescence around the Balbiani body where intense mtDNA replication occurs18. In our data, specific genes that were upregulated >2.5-fold (Fig.4f) included POLRMT, C10orf2, SLC25A4, and MRPL28, which play a key role in mtDNA replication and gene expression (Supplementary Table 4), similarly seen in the PCA (Supplementary Fig.4). Thus, mtDNA variants influencing mtDNA replication and expression will also be more likely to have an effect as PGCs develop, providing a possible explanation for the signature of selection we observed against tRNA and D-loop variants between CS12 and CS21 (Figs.2d&e). We also saw low levels of mtDNA somatic cells in vitro (Fig.2h) and in vivo (Fig.5a), in keeping with a somatic cell bottleneck, as previously observed in mice19. We also detected mtDNA variants in the somatic cells in vivo at similar levels to PGCs (Fig.5b), but the segregation pattern appeared to be different. Some of the mtDNA variants were detected in both the PGCs and the somatic cells from the same embryo, but by the later stages (n=15 shared variants, CS16-17 & CS20-21), the somatic cells contained higher heteroplasmy levels (P 0.034, Fig.5c). Although based on limited data, this suggests that a less stringent mtDNA bottleneck is present in somatic cells within human embryos, and that mtDNA variants are less constrained by selection in somatic tissues than in the germ line. With rudimentary DNA repair mechanisms and negligible intermolecular recombination, uniparentally inherited genomes have limited means of removing de novo mutations. If left unchecked, the relentless accumulation of mtDNA mutations would eventually lead to ‘mutational meltdown’ through Muller’s ratchet, and extinction of the species29. The mtDNA genetic bottleneck has probably evolved to counteract this process, exposing potentially deleterious mutations to selective forces. Based on our findings, in humans this occurs early during germ-line development (Fig.5d). However, if a mutation is sufficiently mild to escape germ-line selection, the same genetic bottleneck can have disastrous consequences. First observed in Holstein cows30, dramatic shifts in heteroplasmy levels have been observed within one generation. In humans, healthy mothers harbouring a low level heteroplasmic mtDNA mutation can have offspring with high levels of heteroplasmy and a severe multisystem disease. Pedigree analyses are consistent with a human mtDNA bottleneck31, but there has been no direct evidence in humans. Here we show that a human mtDNA genetic bottleneck exists, and is narrower than the bottleneck in mice16,18 (Fig.2f). Reducing the amount of mtDNA during this critical time of germ cell development will expose variants to selection acting at either the cellular, organellar or mtDNA level, because this coincides with germ cells expressing transcripts for oxidative phorphorylation as they proliferate and migrate to form the gonad. Harnessing the underlying cell biology could help prevent human mtDNA disease which currently affect ~1 in 11,000. Moreover, given the potential role of mtDNA mutations in common age-related disorders32, preventing their transmission may have a wider impact by promoting healthy ageing in future generations. Supplementary Information is available in the online version of the paper. Acknowledgements PFC is a Wellcome Trust Senior Fellow in Clinical Science (101876/Z/13/Z), and a UK NIHR Senior Investigator, who receives support from the Medical Research Council Mitochondrial Biology Unit (MC_UP_1501/2), the Medical Research Council (UK) Centre for Translational Muscle Disease research (G0601943), and the National Institute for Health Research (NIHR) Biomedical Research Centre based at Cambridge University Hospitals NHS Foundation Trust and the University of Cambridge. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health. WCCT is supported by a Croucher Foundation studentship, and MAS by a Wellcome Investigator Award. Author Contributions V.I.F developed methods and isolated the in vivo human and mouse PGCs, and performed the microscopy; S.D. performed RNA-Seq bioinformatic analysis and contributed to the manuscript; A.P. and W.W.C.T. carried out the real-time PCR assays; W.W.C.T. performed the RNA-seq experiments and N.A. derived and isolated the hESCs, hPGCLCs, and in vitro somatic cells; A.C. and L.N. isolated the human inner cell mass and trophectoderm cells, overseen by D.I. and Y.K. J.C. carried out the library preparation and deep sequencing; M.C. helped with the human tissue dissection; H.S. helped with the super-resolution microscopy; B.P. preformed the technical validation of the deep sequencing protocol; M.S. provided the BVSC mouse; G.H. advised on the NGS data analysis; M.A.S. supervised the RNA-Seq experiments and real-time PCR expression assays and advised on the project; P.F.C. supervised the project, designed experiments, analysed the data, and wrote the paper. All authors contributed to the manuscript. Author Information The authors declare no competing financial interests. “Reprints and permissions information is available at www.nature.com/reprints”. Correspondence and requests for materials should be addressed to P.F.C. (pfc25@cam.ac.uk). References 1 Vafai, S. B. & Mootha, V. K. Mitochondrial disorders as windows into an ancient organelle. Nature 491, 374-383, doi:10.1038/nature11707 (2012). 2 Stewart, J. B. & Chinnery, P. F. The dynamics of mitochondrial DNA heteroplasmy: implications for human health and disease. Nature Reviews Genetics 16, 530-542 (2015). 3 Wallace, D. C. Colloquium paper: bioenergetics, the origins of complexity, and the ascent of man. Proc Natl Acad Sci U S A 107 Suppl 2, 8947-8953, doi:0914635107 [pii] 10.1073/pnas.0914635107 (2011). 4 Payne, B. A. et al. Universal heteroplasmy of human mitochondrial DNA. Hum Mol Genet 22, 384-390, doi:10.1093/hmg/dds435 (2013). 5 Hanley, N. A. et al. SRY, SOX9, and DAX1 expression patterns during human sex determination and gonadal development. Mech Dev 91, 403-407 (2000). 6 Anderson, R. A., Fulton, N., Cowan, G., Coutts, S. & Saunders, P. T. Conserved and divergent patterns of expression of DAZL, VASA and OCT4 in the germ cells of the human fetal ovary and testis. BMC developmental biology 7, 136, doi:10.1186/1471-213X-7-136 (2007). 7 Irie, N. et al. SOX17 is a critical specifier of human primordial germ cell fate. Cell 160, 253-268, doi:10.1016/j.cell.2014.12.013 (2015). 8 Dayama, G., Emery, S. B., Kidd, J. M. & Mills, R. E. The genomic landscape of polymorphic human nuclear mitochondrial insertions. Nucleic Acids Res 42, 12640-12649, doi:10.1093/nar/gku1038 (2014). 9 Kennedy, S. R., Salk, J. J., Schmitt, M. W. & Loeb, L. A. Ultra-sensitive sequencing reveals an age-related increase in somatic mitochondrial mutations that are inconsistent with oxidative damage. PLoS Genet 9, e1003794, doi:10.1371/journal.pgen.1003794 (2013). 10 Copeland, W. C. & Longley, M. J. Mitochondrial genome maintenance in health and disease. DNA Repair (Amst) 19, 190-198, doi:10.1016/j.dnarep.2014.03.010 (2014). 11 Stewart, J. B., Freyer, C., Elson, J. L. & Larsson, N. G. Purifying selection of mtDNA and its implications for understanding evolution and mitochondrial disease. Nat Rev Genet 9, 657-662, doi:nrg2396 [pii] 10.1038/nrg2396 (2008). 12 Freyer, C. et al. Variation in germline mtDNA heteroplasmy is determined prenatally but modified during subsequent transmission. Nat Genet 44, 1282-1285, doi:10.1038/ng.2427 (2012). 13 Falkenberg, M., Larsson, N. G. & Gustafsson, C. M. DNA replication and transcription in mammalian mitochondria. Annu Rev Biochem 76, 679-699, doi:10.1146/annurev.biochem.76.060305.152028 (2007). 14 Pyle, A. et al. Extreme-Depth Re-sequencing of Mitochondrial DNA Finds No Evidence of Paternal Transmission in Humans. PLoS Genet 11, e1005040, doi:10.1371/journal.pgen.1005040 (2015). 15 Craven, L. et al. Pronuclear transfer in human embryos to prevent transmission of mitochondrial DNA disease. Nature 465, 82-85, doi:nature08958 [pii] 10.1038/nature08958 (2010). 16 Cree, L. M. et al. A reduction of mitochondrial DNA molecules during embryogenesis explains the rapid segregation of genotypes. Nat Genet 40, 249-254, doi:ng.2007.63 [pii] 10.1038/ng.2007.63 (2008). 17 Ohinata, Y., Sano, M., Shigeta, M., Yamanaka, K. & Saitou, M. A comprehensive, non-invasive visualization of primordial germ cell development in mice by the Prdm1-mVenus and Dppa3-ECFP double transgenic reporter. Reproduction 136, 503-514, doi:10.1530/REP-08-0053 (2008). 18 Wai, T., Teoli, D. & Shoubridge, E. A. The mitochondrial DNA genetic bottleneck results from replication of a subpopulation of genomes. Nat Genet 40, 1484-1488, doi:ng.258 [pii] 10.1038/ng.258 (2008). 19 Cao, L. et al. The mitochondrial bottleneck occurs without reduction of mtDNA content in female mouse germ cells. Nat Genet 39, 386-390 (2007). 20 Wonnapinij, P., Chinnery, P. F. & Samuels, D. C. Previous estimates of mitochondrial DNA mutation level variance did not account for sampling error: comparing the mtDNA genetic bottleneck in mice and humans. Am J Hum Genet 86, 540-550, doi:S0002-9297(10)00139-4 [pii] 10.1016/j.ajhg.2010.02.023 (2010). 21 Kobayashi, T. et al. Principles of early human development and germ cell program from conserved model systems. Nature 546, 416-420, doi:10.1038/nature22812 (2017). 22 Calvo, S. E., Clauser, K. R. & Mootha, V. K. MitoCarta2.0: an updated inventory of mammalian mitochondrial proteins. Nucleic Acids Res 44, D1251-1257, doi:10.1093/nar/gkv1003 (2016). 23 Stewart, J. B. et al. Strong purifying selection in transmission of mammalian mitochondrial DNA. PLoS Biol 6, e10, doi:07-PLBI-RA-2319 [pii] 10.1371/journal.pbio.0060010 (2008). 24 Ma, H., Xu, H. & O'Farrell, P. H. Transmission of mitochondrial mutations and action of purifying selection in Drosophila melanogaster. Nat Genet 46, 393-397, doi:10.1038/ng.2919 (2014). 25 Hill, J. H., Chen, Z. & Xu, H. Selective propagation of functional mitochondrial DNA during oogenesis restricts the transmission of a deleterious mitochondrial variant. Nat Genet 46, 389-392, doi:10.1038/ng.2920 (2014). 26 Chen, Y. & Dorn, G. W., 2nd. PINK1-phosphorylated mitofusin 2 is a Parkin receptor for culling damaged mitochondria. Science 340, 471-475, doi:10.1126/science.1231031 (2013). 27 Narendra, D. P. et al. PINK1 is selectively stabilized on impaired mitochondria to activate Parkin. PLoS Biol 8, e1000298, doi:10.1371/journal.pbio.1000298 (2010). 28 Fukuoh, A. et al. Screen for mitochondrial DNA copy number maintenance genes reveals essential role for ATP synthase. Molecular systems biology 10, 734, doi:10.15252/msb.20145117 (2014). 29 Muller, H. J. The relation of recombination to mutational advance. Mutat Res 1, 2-9 (1964). 30 Upholt, W. B. & Dawid, I. B. Mapping of mitochondrial DNA of individual sheep and goats: rapid evolution in the D loop region. Cell 11, 571-583 (1977). 31 Chinnery, P. F. et al. The inheritance of mtDNA heteroplasmy: random drift, selection or both? Trends Genet 16, 500-505 (2000). 32 Keogh, M. & Chinnery, P. F. Hereditary mtDNA heteroplasmy: a baseline for aging? Cell metabolism 18, 463-464, doi:10.1016/j.cmet.2013.09.015 (2013). 33 Funkuda, T. Ultrastructure of primordial germ cells in human embryo. Virchows Arch B Cell Pathol 20, 85-89 (1976). 34 Villesen, P. & Fredsted, T. Fast and non-invasive PCR sexing of primates: apes, Old World monkeys, New World monkeys and Strepsirrhines. BMC Ecol 6, 8, doi:10.1186/1472-6785-6-8 (2006). 35 Noli, L., Capalbo, A., Ogilvie, C., Khalaf, Y. & Ilic, D. Discordant Growth of Monozygotic Twins Starts at the Blastocyst Stage: A Case Study. Stem cell reports 5, 946-953, doi:10.1016/j.stemcr.2015.10.006 (2015). 36 Sterio, D. C. The unbiased estimation of number and sizes of arbitrary particles using the disector. J Microsc 134, 127-136 (1984). 37 Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754-1760, doi:10.1093/bioinformatics/btp324 (2009). 38 Li, H. et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 2078-2079, doi:10.1093/bioinformatics/btp352 (2009). 39 Koboldt, D. C. et al. VarScan: variant detection in massively parallel sequencing of individual and pooled samples. Bioinformatics 25, 2283-2285, doi:10.1093/bioinformatics/btp373 (2009). 40 Wang, K., Li, M. & Hakonarson, H. ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data. Nucleic Acids Res 38, e164, doi:10.1093/nar/gkq603 (2010). Figure 1. Isolation of a pure population of human primordial germ cells (PGCs). (a) Human and mouse developmental time points used in this study. Upper panel - Prdm1 (Blimp1)-mVenus and Dppa3 (Stella)-eCFP [BVSC] mouse embryos (E9.5 and E10.5) and gonads (E12.5) 17. Arrows show the fluorescent mouse PGCs subsequently isolated by fluorescent sorting. Upper panel scale bars from left to right: 0.5 mm, 1 mm, 0.5 mm. Lower panel – fresh human female embryos showing lower core-body dissected area (red line drawing) for 4 Wk (Carnegie stages, CS, 12) embryos (left), and dissected female embryonic gonads for 5.5 - 6 Wk (CS16/17) and 8 Wk (CS20/21) embryos (middle & right) Lower panel scale bar = 1 mm. (b) Isolation of human PGCs by fluorescent sorting using: for week 4 (CS12) dissected embryos - a triple positive combination for TNAP, stage-specific embryonic antigen 4 (SSEA4) and CD38, and for weeks 5-8 (CS16/17 – CS20/21) gonadal tissue double positive for TNAP and c-KIT. Boxes in the FACS plots show the sorted cell population. (c) Purity of isolated hPGCs was validated by alkaline phosphatase activity-staining showing >97% positive cells. Scale bar = 100 μm. See also Fig.4a-c and Supplementary Figs. 1&2. Observations for b & c were independently replicated three times at each developmental stage. Figure 2. Mitochondrial DNA variants in human primordial germ cells. (a-e) Mitochondrial DNA (mtDNA) variants exceeding the assay detection threshold (minor allele frequency, MAF) of >1% during human primordial germ cell (PGC) development. Carnegie stage, CS12 = blue, CS16/17 = red, CS20/21 = orange. Data were derived from n=80 single cells sorted from two CS12 embryos, n=240 cells sorted from four CS16/17 embryos, and n=360 cells sorted from six CS20/21 embryos. (a) Frequency distribution of mtDNA variants at sequential stages of germ line development. (b) Codon positions affected by the coding region mutations (No codon bias for CS12, P = NS, significant codon bias for CS20/21, P = 0.005, Fisher’s exact text, 2-sided). (c) Proportion of non-synonymous (shaded) variants and synonymous (open) (dN/dS, P = 0.02, Fisher’s exact text, 2-sided). (d) Evidence of selection against mtDNA variants in tRNA genes (P=0.03, Fisher’s exact text, 2-sided) and the non-coding Displacement (D)-loop (P = 0.003) between CS12 and CS20/21. (e) Position of the variants detected in the non-coding mtDNA D-loop region. (f) MtDNA content of single in vivo PGCs from human (left, n = 117) and Blimp1-mVenus and Stella-eCFP [BVSC] mouse (right, n = 68) embryos. Each data point represents the mtDNA content in a single cell corresponding to the mean value from the independent qPCR measurements. Source data available in Supplementary Table 6. (g) MtDNA content of single human inner cell mass (ICM, n = 20), trophectoderm (TE, n = 12) cells from five embryos in vivo. Each data point represents the mtDNA content in a single cell corresponding to the mean value from the independent qPCR measurements. Source data available in Supplementary Table 6. (h) MtDNA content of single human embryonic stem cells (hESCs, n = 18), somatic cells (n = 30), and human primordial germ-cell like cells (hPGCLC, n = 48) isolated in vitro7,21. Two independent PGCLC inductions were performed for both the male (WIS2) and female (H9) ESC lines, with cells analysed at day 5/6. Source data available in Supplementary Table 6. Each data point represents the mtDNA content in a single cell corresponding to the mean value from the independent qPCR measurements. Horizontal lines in (f-h) represent the mean values. Figure 3. Morphology and deep sequencing of single human primordial germ cells in vivo. (a) Upper, N-SIM image of a human PGC labelled with anti-TOM20 (Supplementary movie 1) with mitochondria identified and counted by IMARIS (Supplementary movie 2). Scale bar = 2 μm. Lower, transmission electron micrographs (TEM) characteristic of human PGCs33. White fat arrow = nucleolus; white thin arrow = chromatin, electron dense glycogen particles, and lipid droplets; pink arrow = pseudopodia in the migratory stage (CS12) PGCs (right image). Round discrete mitochondria around the nucleus (orange arrows). Scale bar = 2 μm. Similar observations made from three independent experiments. (b) Number of mitochondria in human primordial germ cells. Left (EM, n=24 cells): mitochondrial count per cell using the dissector method and serial section EMs. Right (N-SIM, n=12 cells): super-resolution microscopy. Each data point represents a measurement from a single PGC. EM: mean=275.9, median=272, SD 87.76. N-SIM: mean=288.8, median=272, SD=72.12. The horizontal bar represents the mean value. (c) Probability of an individual mitochondrion containing mutated mtDNA molecules as a function of the overall heteroplasmy level within the cell. Binomial distributions assuming 5 molecules/mitochondrion that are randomly distributed between organelles. For cells containing 40-50% mutant mtDNA, many mitochondria are likely to contain 4 or more mutated molecules (>80%). If deleterious these high levels will potentially depolarise the cell triggering mitophagy, or inhibit mtDNA replication within that organelle, potentially leading to selection against mutations at the organellar level. (d) Minor allele frequency (MAF) of mtDNA variants in 20 single human PGCs from three CS20/21 embryos detected by deep sequencing (left panel). Right panel shows the same data with the Y-axis scaled from 1-10% MAF. In both panels the detection threshold was set at MAF >1%. (e) Position of the variants from single human PGCs within the mitochondrial genome (left panel). Right panel shows the same data with the Y-axis scaled from 1-10% MAF. In both panels the detection threshold was set at MAF >1%. Figure 4. Transcriptome analysis of human primordial germ cells (PGCs). Analysis of n = 13 transcriptomes. (a) Unsupervised hierarchical clustering (UHC) of RNA-seq expression profiles in H9 Embryonic stem cells (ESCs), Wk4, Wk7, and Wk9 female (F) human PGCs and Wk5.5 male (M) hPGCs, and gonadal somatic cells (Soma) showing clustering of gene expression consistent with the appropriate developmental stage (see also Supplementary Fig.4). Results from a single analysis of the entire transcriptome data set. (b) Principal component analysis (PCA) of RNA-seq gene expression in hPGC samples. Arrow line indicates developmental progression along PC1. (c) Heat map showing expression of key PGC-associated transcripts (early and late) and of pluripotency, mesoderm, endoderm, and gonadal somatic (Soma) markers. (d) Correlation of transcription profiles of human PGCs of various stages for nuclear-encoded mitochondrial transcripts predicted by MitoCarta2.0 (see Methods, Pearson’s correlation, two-sided) and mtDNA-encoded transcripts. Only genes with normalized reads counts with (log2 (normalized counts) > 2.5) were used for analysis. Expression levels were averaged over biological replicates. The number of biological replicates at each developmental stage is shown in the cluster diagram shown in Fig.4a generated from a total of 13 separate transcriptomes. This shows a progressive change in transcriptome profiles for mitochondrial genes in a sequential manner during PGC development. (e) Two-dimensional principal components analysis (PCA) of RNA-seq gene expression (PC2 against PC1) for mitochondrial nuclear-encoded and mtDNA-encoded genes (MitoCarta2.0) at the different stages of hPGCs (left panel), with a corresponding plot of the PC scores on PC2 and PC1 (right panel). Gene names are color-coded to illustrate up-regulated expression in hPGCs-wk9 (purple), and down-regulated expression in hPGs-wk9 (red) when compared with hPGCs-wk4 sample (see also Fig. 4f, Supplementary Fig.4&5 and Supplementary Table 4). Arrow line indicates potential mitochondrial-related-transcript progression for each developmental stage from migratory hPGC-wk4 to gonadal hPGC-wk9. (f) Volcano plot showing differentially expressed genes (from MitoCarta2.0 list) between week 9 (n = 2 independent replicates) and week 4 (this measurement was performed once) hPGCs [log2(fold change)>2.5, p < 0.05, DEseq two-sided, see Methods]. See also Supplementary Fig.4&5 and Supplementary Table 4. Note that the week 9 PGCs included in the RNAseq analysis were from a slightly later developmental stage than CS20/21, reflecting the clinical samples that were available. Figure 5. Mitochondrial DNA analysis of somatic cells in vivo & mechanisms of selection. (a) MtDNA content of single human somatic cells in vivo at the same stage as the female primordial germ cells (PGCs) shown in Figs.1&2 (CS12 n=36, CS16-17 n=36; CS20-21 n=60). Each data point represents the mtDNA content in a single cell corresponding to the mean value from the independent qPCR measurements from 12 embryos. The horizontal bar represents the mean value. Source data available in Supplementary Table 6. (b) Mutation frequency for human female PGCs and corresponding somatic cells (soma) in developing human embryos (CS12-20/21) in vivo. Data derived from 12 embryos. (c) Segregation of the same heteroplasmic mtDNA variants detected in somatic cells and PGCs from the same embryo at CS16-17 & CS20-21 (n = 15 variants) in vivo. Variants in the somatic cells had a higher heteroplasmy level than the PGCs (P=0.03, Fisher’s exact test, two-sided), which was not the case at CS12. Source data available in Supplementary Table 6. (d) In vivo PGCs contain ~ 300 discrete mitochondria which each containing 5 mtDNA molecules. In vitro data suggests that the amount of mtDNA may be even lower in female germ cells at an earlier stage. Some of the mtDNA molecules are mutated with either neutral (green) or deleterious (red) mutations. Between Carnegie Stage (CS) 12 and CS 20/21 the number of cells increases from ~3000 to ~80,000 accompanied by a massive proliferation of mtDNA molecules within the migrating PGCs. As the mitochondria replicate, the mutations will be copied. The random segregation of mtDNA molecules during mitochondrial fission will generate daughter mitochondria with either high or low levels of the original mutation. If the mutation is deleterious, then it will be selected against. This could be because it compromises ATP synthesis and activates surveillance systems that remove defective mitochondria, or because it compromises mtDNA replication, and thus prevents that particular mitochondrion from being copied. Both of these mechanisms can act in concert at the cellular level and at the mitochondrial level. 26 28