Research Article

ER-mitochondria contacts couple mtDNA synthesis with mitochondrial division in human cells

See allHide authors and affiliations

Science  15 Jul 2016:
Vol. 353, Issue 6296, aaf5549
DOI: 10.1126/science.aaf5549

How the ER manages mitochondrial division

It has been unclear how mitochondrial DNA (mtDNA) replication is spatially controlled in mammalian cells and how the mitochondrial nucleoid—the protein-DNA structure that is the unit of mtDNA inheritance—is distributed at the cellular level. Lewis et al. now show that homeostatic mtDNA synthesis in mitochondrial nucleoids in mammalian cells is spatially linked to a small subset of endoplasmic reticulum (ER)-mitochondria contact sites that are specifically destined for mitochondrial division. Successive events of mtDNA replication, mitochondrial division, and mitochondrial motility function together to ensure the accurate distribution of mtDNA in cells. Furthermore, ER-mitochondria contacts coordinate the licensing of mtDNA replication with division to distribute newly replicated nucleoids to daughter mitochondria.

Science, this issue p. 261

Structured Abstract


Mitochondria are endosymbiotic organelles that have their own genome and perform many essential functions in eukaryotic cells, including ATP synthesis via oxidative phosphorylation. Mitochondria evolved from bacteria, which stringently replicate and segregate their genome in a binary fission process. The residual circular ~16-kilobase human mitochondrial genome is essential, as it encodes mitochondrial ribosomal and transfer RNAs and respiratory chain complex proteins. Within human cells, hundreds to thousands of copies of mitochondrial DNA (mtDNA) are packaged into nucleoids, the unit of mtDNA inheritance, and distributed within dynamic mitochondrial networks. The nature of mtDNA transmission in mitochondrial syncytia is relaxed, with replication occurring asynchronously throughout the cell cycle and also in postmitotic cells. If and how nucleoids are actively chosen for mtDNA replication and distributed within mitochondrial networks is not understood. This question is highly relevant for understanding the basis of human metabolic diseases caused by mutations in mtDNA and in nuclear genes that affect mtDNA maintenance. In addition, aging and neurodegenerative disorders are also linked to defective mtDNA maintenance and mitochondrial dysfunction. In this study, we investigated the fundamental process of mtDNA transmission in mammalian cells.


To address the cellular mechanism for mtDNA transmission, we examined whether nucleoid and mitochondrial distribution were coupled. Mitochondrial distribution is determined by mitochondrial division, fusion, and motility events. Previous work established that the mitochondrial division site placement is not random and is instead spatially marked by regions of contact between the endoplasmic reticulum (ER) and mitochondria, in a process termed ER-associated mitochondrial division (ERMD). In biological systems, division events serve a fundamental role in the inheritance of genetic material. Thus, we addressed whether ERMD serves to facilitate the transmission of mtDNA by examining the behavior of mitochondria, ER, and nucleoids in mammalian cells via fluorescent microscopy. We marked the subset of nucleoids in cells actively engaged in mtDNA synthesis with a functional green fluorescent protein–tagged version of POLG2, the processivity subunit of the human mitochondrial DNA polymerase holoenzyme. Using this sensitive and highly specific marker for mtDNA synthesis in live cells, we asked whether replicating nucleoids were selectively linked to ERMD for the purpose of ensuring the segregation of nascent mtDNA to daughter mitochondria.


Our work revealed that nucleoids actively engaged in mtDNA synthesis in mammalian cells were spatially and temporally linked to a small subset of ER-mitochondria contacts destined for mitochondrial division. At division sites, mtDNA replication occurred upstream of mitochondrial constriction and assembly of the division machinery. Nucleoids containing nascent mtDNA localized to mitochondrial tips, and these products of division were preferentially distributed within cells as compared with nonreplicative nucleoids. Our observations also demonstrated that ER structure and mtDNA maintenance were intertwined; ER tubules proximal to nucleoids were necessary but not sufficient for mtDNA synthesis and also functioned in nucleoid distribution.


We propose that, at ER-mitochondria contacts destined for division, the consecutive events of mtDNA replication, mitochondrial division, and mitochondrial motility are connected together to ensure the accurate distribution of nucleoids within cells. Our findings suggest that ER-mitochondria contacts coordinate the licensing of mtDNA replication with downstream mitochondrial division events to distribute newly replicated mtDNA to daughter mitochondria. The connection revealed between ER structure and mtDNA replication and distribution has broad implications for understanding human cellular homeostasis and the cellular pathology underlying human diseases.

ER-mitochondria contacts coordinate mtDNA replication with mitochondrial division.

In human cells, a subset of ER-mitochondria contacts are spatially linked to mitochondrial nucleoids engaged in replication and are destined for mitochondrial division. (Left) Light image is of an osteosarcoma U2OS cell; (right) in the schematic depiction, colors are as on the labels to the left; and the replicating nucleoid is marked by POLG2 in green.


Mitochondrial DNA (mtDNA) encodes RNAs and proteins critical for cell function. In human cells, hundreds to thousands of mtDNA copies are replicated asynchronously, packaged into protein-DNA nucleoids, and distributed within a dynamic mitochondrial network. The mechanisms that govern how nucleoids are chosen for replication and distribution are not understood. Mitochondrial distribution depends on division, which occurs at endoplasmic reticulum (ER)–mitochondria contact sites. These sites were spatially linked to a subset of nucleoids selectively marked by mtDNA polymerase and engaged in mtDNA synthesis—events that occurred upstream of mitochondrial constriction and division machine assembly. Our data suggest that ER tubules proximal to nucleoids are necessary but not sufficient for mtDNA synthesis. Thus, ER-mitochondria contacts coordinate licensing of mtDNA synthesis with division to distribute newly replicated nucleoids to daughter mitochondria.

Mutations in mitochondrial DNA (mtDNA) and in nuclear genes that control mtDNA maintenance cause mitochondrial dysfunction and are linked to human disease and aging, which affirms the functional importance of mtDNA (14). The units of mitochondrial inheritance are mtDNA-protein complexes called nucleoids (5). Accurate maintenance of mtDNA requires replication, repair, packaging, and distribution of the mitochondrial nucleoid at the cellular level. In mammalian cells, mitochondrial DNA replication is mediated by a nuclear-encoded replisome composed of a polymerase gamma holoenzyme containing a catalytic subunit (POLG1) and a processivity subunit (POLG2) (610). In addition to the polymerase complex, the replisome contains the helicase Twinkle and a mitochondria-specific single-stranded DNA binding protein, which together facilitate the formation of a single-stranded DNA replication template (11, 12). Within cells, mtDNA is packaged into a nucleoid by TFAM, a nuclear-encoded DNA-bending protein, which also plays a role in mtDNA replication and transcription (1318).

Although molecular players involved in mtDNA replication and packaging have been described, the mechanisms underlying the spatial regulation of mtDNA replication and intracellular nucleoid distribution have been elusive. This is partly because mtDNA is present in cells in multiple copies, and the spatiotemporal regulation of its replication and distribution is relaxed in comparison with the nuclear genome (19). Indeed, mtDNA replication occurs asynchronously with the cell cycle and within postmitotic tissues, such as the brain and muscle (20, 21). Nucleoids are evenly distributed within mitochondria and constrained in their motility (5, 22). Mitochondrial distribution is in large part dependent on cytoskeletal-based motility and on mitochondrial division (23), mediated in mammalian cells by DRP1, a cytosolic dynamin-related guanosine triphosphatase (GTPase) that forms assemblies around mitochondria to facilitate membrane scission (24, 25). DRP1 recruitment and assembly occur at sites of endoplasmic reticulum (ER)–mitochondrial contact, where mitochondrial constriction is also observed (26). Perturbation of mitochondrial division in both yeast and mammalian cells causes nucleoid aggregation, mtDNA deletions, and mtDNA depletion, which suggests a fundamental and functional link between mitochondrial division and mtDNA maintenance (2729). In yeast, ER-linked division sites, marked by the fungus-specific ER-mitochondria encounter structure (ERMES) complex, are spatially linked to nucleoids, which further suggests a role for ER-mitochondria contacts in mtDNA maintenance (30). Here, we asked whether ER-mitochondria contact sites function to couple mtDNA replication with mitochondrial division for the purpose of distributing newly replicated mtDNA in human cells.

ER-mitochondria contacts and ER-associated mitochondrial division are spatially linked to nucleoids in mammalian cells

We asked if ER-associated mitochondrial division (ERMD) events are spatially linked to mitochondrial nucleoids in human cells by simultaneously imaging mitochondria, nucleoids, and the ER network at high spatial and temporal resolution using spinning disk confocal microscopy. Osteosarcoma cells (U2OS) were transiently transfected with green fluorescent protein (GFP)–tagged TFAM (TFAM-GFP), a well-characterized marker of the total nucleoid population, mitochondrial matrix–targeted blue fluorescent protein (mito-BFP), and ER-targeted mCherry or mRuby (Sec61b-mCherry or mRuby-KDEL). TFAM-GFP–labeled foci were evenly spaced within mitochondria, as previously described for nucleoids in other cell types (Fig. 1A) (5, 16, 18). TFAM-GFP–labeled nucleoids were also localized adjacent to points where ER tubules crossed over mitochondria in a perpendicular fashion (Fig. 1A, right), and a subset of nucleoids remained stably linked to ER-mitochondria contacts over time, despite ER network remodeling and mitochondrial motility (Fig. 1B, arrowheads). We further assessed the spatial link between ER-mitochondria contacts and nucleoids by determining the Pearson correlation coefficient of mRuby-KDEL and TFAM-GFP fluorescence intensity along line scans of mitochondria imaged in live U2OS cells (n = 58). Consistent with our observations, this analysis indicated a highly significant enrichment of ER signal specifically within 17 pixels (~1 μm) laterally adjacent to nucleoids (Pearsons’ R = 0.59) (fig. S1A). Thus, in general, nucleoids are spatially linked to ER-mitochondria contacts in human cells.

Fig. 1 Mitochondrial DNA nucleoids are spatially linked to mitochondria-ER contacts in human cells.

(A) (Left) Panels show a merged image of a live U2OS cell expressing mito-BFP, TFAM-GFP, and Sec61b-mCherry (ER). (Right) The pixel intensity of mito-BFP, TFAM-GFP, and Sec61b-mCherry from a line scan drawn along the mitochondrial tubule (dashed line), arrows indicate nucleoid positions. (B) Time-lapse images of a U2OS cell expressing mito-BFP, TFAM-GFP, and mRuby-KDEL (ER); a single plane is shown. Arrowheads indicate a site of persistent colocalization between a TFAM-GFP–labeled nucleoid and an ER-mitochondria contact. (C) Number of mitochondrial divisions in U2OS cells spatially linked to TFAM-GFP–labeled nucleoids, from 43 cells. (D) Time-lapse images of mitochondrial division (marked by arrowhead) spatially linked to a TFAM-labeled nucleoid focus in a U2OS cell. (E) The displacement of PicoGreen-labeled nucleoids in live COS-7 cells over 12.5 min as a function of their intramitochondrial position. Data are means ± SD. (F) (Left) Merged image of a live U2OS cell expressing mito-BFP and Sec61b-mCherry (ER). (Right) Examples of mitochondrial constrictions colocalized with ER tubules (arrowheads). (G) The percentage of persistent mitochondrial-ER colocalizations that become sites of mitochondrial constriction or division over 5 min in live U2OS cells. Scale bars: (A), (B), (D), and inset in (F), 2 μm; (F), 10 μm.

Accordingly, we observed that a majority of ERMD events (82%, n = 62) were spatially linked to nucleoids (within a 1-μm distance), which resulted in their localization at mitochondrial tips after division (Fig. 1, C and arrowhead in D). We also performed time-lapse imaging of nucleoids and mitochondria in COS-7 primate cells labeled with the selective vital dyes PicoGreen DNA stain and MitoTracker Red, respectively. Retrospective nucleoid tracking over time revealed that nucleoids at mitochondrial tips had been displaced a greater distance on average (3 times as great over a time period of 12.5 min) than intramitochondrial nucleoids, which suggested that the subset of nucleoids linked to ERMD sites are preferentially distributed within cells (Fig. 1E).

ER-mitochondria contacts are not rate-limiting for mitochondrial division

The high density of ER and mitochondrial networks in mammalian cells suggested that ER-mitochondria contact sites may not be rate-limiting for ERMD. To test this, we defined and quantified persistent ER-mitochondria contacts in cells by time-lapse imaging of mitochondria and ER for 5 min at 15-s intervals using mito-BFP and mRuby-KDEL/Sec61b-mCherry, respectively, in U2OS cells. ER and mitochondria were segmented in time-lapse images by thresholding, and regions of overlap between the organelles were identified and tracked over 5 min (see fig. S1B, arrowheads). Most regions of ER-mitochondria colocalization were transient, but ~100 distinct regions per cell were identified in which the ER and mitochondria persistently colocalized over the duration of imaging (fig. S1C). We followed the fate of persistent ER-mitochondria colocalized regions in relation to mitochondrial division over the duration of imaging. At a small fraction of these regions, mitochondrial constrictions were observed and/or developed (arrowheads in Fig. 1F; also Fig. 1G). An even smaller fraction of ER-mitochondria colocalized regions was linked to division events, consistent with published observations indicating that ER-linked mitochondrial constriction precedes division (Fig. 1G) (26). Thus, although there are many ER-mitochondria contacts, only a subset are destined to be linked to ERMD. Given that ER-mitochondria contacts are also linked spatially to nucleoids in general (fig. S1A), we considered whether a functionally specialized subset of nucleoids marks ERMD events in cells.

Nucleoids engaged in mtDNA synthesis mark nascent mitochondrial division sites at ER-mitochondria contacts

It has been proposed that functional subpopulations of nucleoids coexist in mammalian cells, distinguished on the basis of mtDNA replication and/or transcription status (31, 32). In yeast, a subpopulation of nucleoids is spatially linked to ERMES foci, which in turn mark a fraction of mitochondrial division sites (30, 33). Thus, we tested whether in mammalian cells mtDNA synthesis is specifically coupled to ERMD, which might help to ensure the segregation of nascent mtDNA to daughter mitochondria.

To visualize nucleoids engaged in mtDNA synthesis in cells, we used POLG2-GFP, a fluorescently tagged version of the processivity subunit of the human mitochondrial DNA polymerase holoenzyme, whose enzymatic properties and affinity for mtDNA are indistinguishable from untagged POLG2 in vitro (34). Sixteen hours after transfection of live U2OS cells, POLG2-GFP labeled a subset of the total nucleoid population within mitochondria, which was revealed by the addition of PicoGreen DNA stain and subsequent reimaging (Fig. 2A and fig. S2A). Consistent with this, comparison of the total number of nucleoids labeled with TFAM-GFP to the number of POLG2-GFP–labeled nucleoids in live U2OS cells indicated that there were significantly fewer POLG2-GFP foci per cell (9.4% of total nucleoids) (Fig. 2B). To test whether the POLG2-labeled nucleoids were selectively engaged in mtDNA synthesis, live cells expressing POLG2-GFP and labeled with MitoTracker Red, were incubated with the nucleotide thymidine analog 5-ethynyl-2-deoxyuridine (EdU) for 1 hour, which was subsequently visualized in fixed cells with AlexaFluor647 using copper click chemistry (35). The vast majority of POLG2-GFP foci (as detected via α-GFP–AlexaFluor488) colocalized with detectable EdU incorporation at nucleoids within mitochondria (96% from 15 cells) (Fig. 2, C and D). Consistently, there was no significant difference between the number of EdU or POLG2-GFP foci per cell in either live or fixed cells (fig. S2B). We also assessed whether expression of POLG2-GFP perturbed mtDNA maintenance by examining mtDNA copy number by quantitative polymerase chain reaction (qPCR). We detected no difference in mtDNA levels between cells expressing POLG2-GFP versus mock-transfected cells (fig. S2C). In contrast, and as previously shown, a significant increase in mtDNA copy number was observed in cells overexpressing TFAM-GFP, which indicated that U2OS cells were capable of modulating mtDNA synthesis during our experiments (13, 36). Thus, exogenously expressed POLG2-GFP is recruited to the endogenous mtDNA replisome and provides a sensitive and highly specific marker for monitoring mtDNA synthesis in live cells.

Fig. 2 POLG2-GFP is specifically recruited to replicating nucleoids in live cells.

(A) A live U2OS cell expressing mito-BFP and POLG2-GFP was imaged (left), then stained with PicoGreen DNA dye and reimaged 10 min later (right). (Insets) 488-channel signal intensity in an example mitochondrion (left), and the same organelle after PicoGreen staining (right). Magnified 2×. (B) The number of mitochondrial POLG2-GFP (n = total 441 foci from 10 cells) or TFAM-GFP foci (n = 5182 foci from 10 cells) per U2OS cell (***P < 0.001, two-tailed t test). Data are means ± SD. (C) Representative image of a U2OS cell expressing POLG2-GFP and pulse-labeled with 50 μM EdU, fixed, and stained with 4′,6′-diamidino-2-phenylindole (DAPI) (DNA, blue); MitoTracker (mitochondria, red); anti-GFP–AlexaFluor488 conjugate antibody (POLG2-GFP, green); and Click-iT EdU-AlexaFluor647 (nascent DNA, magenta). Arrowheads indicate colocalization. (D) Observed colocalization between mitochondrial POLG2-GFP and EdU foci in fixed U2OS cells labeled as in (C), from 15 cells. Scale bars: (A) and (C), 10 μm; inset in (C), 5 μm.

With this tool in hand, we asked whether nucleoids engaged in mtDNA synthesis, labeled faithfully by POLG2-GFP, were spatially linked to ERMD in a selective manner. As shown in a representative time-lapse series (Fig. 3A), POLG2-GFP–labeled nucleoids were indeed spatially linked to mitochondrial division at ER-mitochondria contact sites in live U2OS cells, as labeled by mRuby-KDEL and mito-BFP markers, respectively. A majority of ERMD events were linked to POLG2-labeled nucleoids (within 1 μm) and and ERMD occurred at a rate more than 3 times that expected from random chance, at persistent ER-mitochondria contacts (Fig. 3B). In addition, a comparable proportion of ERMD events were linked to nucleoids in cells expressing either the general nucleoid marker TFAM-GFP or the replication-specific marker POLG2-GFP (82 versus 73%, P = 0.71) (Figs. 1C and 3B). Thus, nucleoid-linked ERMD events seem to occur predominantly at nucleoids engaged in mtDNA synthesis.

Fig. 3 Replicating nucleoids mark sites of ER-mediated division.

(A) Representative time-lapse images of a U2OS cell expressing mito-BFP, mRuby-KDEL (ER), and POLG2-GFP, demonstrating mitochondrial division at a mitochondrial-ER contact site spatially linked to POLG2-labeled nucleoid (arrowheads indicate division site). (B) Percentage of ERMD events (marked by mRuby-KDEL and mito-BFP as in Fig. 1B) in live U2OS cells that occurred within 1 μm of a POLG2-GFP–labeled nucleoid (n = 15 events from seven cells, **P < 0.01, Fisher’s exact test). Significantly more division events occurred at preexisting POLG2-GFP foci (73.3%) than expected by random chance at stable ER-mitochondria contacts (22.1%). Scale bar, 2 μm.

To further test the idea that nucleoid-linked ERMD events occur predominantly at nucleoids actively engaged in mtDNA synthesis, we exploited the observation that POLG2-GFP nucleoids localized at the tips of daughter mitochondria as a consequence of ERMD (Fig. 3A). Steady-state analysis of nucleoid position in U2OS cells revealed a highly significant spatial enrichment of POLG2-GFP labeled nucleoids within 1 μm of mitochondrial tips, as compared to the total TFAM-GFP labeled nucleoid population (Fig. 4, A and B), despite the nearly 10 times as great number of TFAM-GFP–labeled nucleoids detected per cell (Fig. 2B). To validate that our observations were generally reflective of mammalian mtDNA segregation, we analyzed two additional cell lines: noncancerous ARPE19 retinal epithelial cells and COS-7 primate fibroblasts. As in U2OS cells, POLG2-GFP–labeled nucleoids were enriched at mitochondrial tips in these additional lines, where they were also colocalized with focal EdU labeling in fixed cells (fig. S3, A and B). In addition, we reasoned that if mtDNA replication was selectively linked to ERMD, the number of POLG2-GFP– and/or EdU-labeled nucleoids per mitochondrion would be constrained to the number of mitochondrial tips created by mitochondrial division, as opposed to scaling to the total length of the organelle. Thus, we examined the distribution of replicating and total nucleoids per mitochondrion in live U2OS cells (labeled with TFAM-GFP or POLG2-GFP) and in fixed ARPE19 cells (labeled with PicoGreen and EdU) (Fig. 4C, left and right, respectively). Indeed, consistent with the relatively uniform distribution of nucleoids within mitochondria (Fig. 1A), individual organelles contained numerous TFAM-GFP– or PicoGreen-labeled nucleoids, and the total number of nucleoids per mitochondrion was highly correlated with the length of the organelle (Fig. 4C). In contrast, the number of POLG2-GFP– or EdU-labeled nucleoids per organelle was not well correlated with mitochondrial length, and in a majority of instances, there was a clear constraint on the number of EdU- or POLG2-GFP–labeled nucleoids to two or less per organelle, regardless of organelle length (Fig. 4C). Closer examination of outlier EdU- or POLG2-labeled nucleoids in mitochondria that contained greater than the median number of nucleoids, revealed that these organelles were either branched and, consequently, had a greater number of tips with associated EdU- or POLG2-labeled nucleoids or were unbranched and contained an additional internally localized pair of EdU foci (fig. S4). Internal EdU foci pairs were closely spaced, suggestive of a segregation intermediate. Thus, in mammalian cells, a majority of ERMD events are spatially linked to the subset of nucleoids that are actively engaged in mtDNA replication, consistent with a role for ERMD in the coordinated segregation of nascent mitochondrial genomes.

Fig. 4 Nascent mtDNA is segregated to daughter mitochondria by division.

(A) Representative images of live U2OS cells expressing mito-BFP and TFAM-GFP (top) or POLG2-GFP (bottom). Scale bar, 5 μm.(B) Significant spatial enrichment of the total population of POLG2-GFP foci within 1 μm of mitochondrial tips, as compared with TFAM-GFP foci (in dark gray) in live U2OS cells as labeled and imaged in (A). (Data are means +/- SD, ***P < 0.001, two-tailed t test.) (C) (Left) In live U2OS cells, the number of TFAM-GFP foci per mitochondrion, but not POLG2-GFP foci, is correlated with mitochondrion length. Linear regression with best-fit line is shown. For TFAM-GFP, adjusted R2 = 0.86, Pearson’s R = 0.92 (P < 0.0001). For POLG2-GFP, adjusted R2 = 0.15, Pearson’s R = 0.41 (n.s.). (Right) In fixed ARPE19 cells, the number of PicoGreen foci per mitochondrion, but not EdU foci, is correlated with mitochondrion length. Linear regression as on the left. For PicoGreen, adjusted R2 = 0.68, Pearson’s R = 0.82 (P < 0.0001). For EdU, adjusted R2 = 0.07, Pearson’s R = 0.31 (n.s.). (D) (Top) EdU pulse-chase experiments in ARPE19 cells. (Bottom) Empirical cumulative distribution analysis of EdU focus position along mitochondria, demonstrating depletion (D) of pulse-labeled nascent mtDNA from mitochondrial tips over time, toward a simulated random distribution (***P < 0.001, *P < 0.05, Kolmogorov-Smirnov test).

To rigorously test this model and to determine the fate of replicating nucleoids linked to ERMD, we performed an EdU pulse-chase analysis of mtDNA in cells under native conditions, in the absence of POLG2-GFP expression (Fig. 4D, top). We pulse-labeled ARPE19 cells with EdU for 1 hour, under conditions where the replication of nuclear DNA was inhibited; chased for 1, 24, or 48 hours; and subsequently analyzed the position of EdU-labeled nucleoids relative to mitochondrial tips after fixation. Consistent with our previous observations (Figs. 3, A and B, and 4, A and B), EdU nucleoids detected during the pulse and after the 1-hour chase were highly enriched within 1 μm of mitochondrial tips (Fig. 4D and fig. S5, A to C, representative images). Quantification revealed that the positioning of EdU-labeled nucleoids relative to tips progressively decreased after the 24- and 48-hour chases, toward a random distribution (Fig. 4D). Thus, replicating nucleoids are indeed both spatially and temporally linked to mitochondrial division, and after the completion of mtDNA replication, nascent mtDNA is segregated into daughter mitochondria and subsequently distributed away from the preceding division sites.

The mtDNA replisome is an early marker of nascent mitochondrial division sites

To gain further insight into the relationship between ERMD and replicating nucleoids, we further examined the temporal relationship of active mtDNA synthesis to known mitochondrial division events: ER-linked mitochondrial constriction and recruitment of the mitochondrial division dynamin, DRP1 (26). We performed time-lapse microscopy to assess the relationship of POLG2-GFP–labeled nucleoids to mitochondrial division events in U2OS cells coexpressing mito-BFP and mCherry-DRP1. Consistent with our previous observations, a majority of mitochondrial division events marked by DRP1 assemblies occurred within 1 μm of a POLG2-GFP focus (Fig. 5, A and B). To further define the spatial link between replicating nucleoids and DRP1-marked division sites in live cells, we quantified the distance from the center of each POLG2-GFP focus to the position of matrix marker discontinuity, for 26 division events. We observed that division sites were enriched within a zone greater than 200 nm but less than 400 nm from a POLG2-GFP focus (fig. S4C). In every division event marked by mCherry-DRP1 (100%, n = 26), POLG2-GFP–labeled nucleoids marked a future division site before mitochondrial constriction and also preceded the recruitment of mCherry-DRP1 to mitochondrial constrictions (Fig. 5B). Indeed, for some division events, POLG2-GFP was detected at a future division site more than 15 min before the recruitment of mCherry-DRP1. Thus, replication of mtDNA precedes known events linked to mitochondrial division.

Fig. 5 Replicating nucleoids mark division sites before mitochondrial constriction or DRP1 recruitment.

(A) (Top) Time-lapse images of a U2OS cell expressing mito-BFP, mCherry-DRP1, and POLG2-GFP. Arrowhead indicates site of division. (Bottom) Line scan drawn along the mitochondrial tubule to show relative fluorescence intensity of mitochondria, DRP1 division machinery, and POLG2 signal for time points t = 0 s (preconstriction), t = 45 s (postconstriction), and t = 90 s (postdivision). Scale bar, 1 μm. (B) The percentage of mitochondrial divisions [marked by mCherry-DRP1 as in (A)] that occur within 1 μm of a POLG2-GFP focus in live U2OS cells (n = 26 events from 22 cells, **P < 0.01, Fisher’s exact test).

ER structure is required for mtDNA replication licensing and nucleoid distribution

The spatiotemporal relationship between replicating nucleoids and nascent ERMD sites prompted us to further examine the role of the ER in mtDNA replication and nucleoid distribution in cells. In addition to the nuclear envelope, the peripheral ER forms a dense and dynamic network of interconnected tubules and sheetlike structures (3739). Members of the conserved reticulon protein family (RTN1, RTN2, RTN3, and RTN4/Nogo) are thought to be key factors in determining the ratio of ER tubules and sheetlike structures by localizing to and stabilizing highly curved ER tubules and edges of ER sheetlike structures (40, 41). The coiled-coil membrane protein CLIMP63 is enriched in sheetlike ER structures where it is thought to maintain the lumenal distance between ER membrane bilayers (4143). Overexpression of the ER-shaping proteins RTN4 or CLIMP63 in mammalian cells shifts the proportion of tubules to sheetlike structures or sheetlike structures to tubules, respectively (40, 41, 44). Thus, we used transient overexpression of CLIMP63 or RTN4A in COS-7 cells to manipulate ER structure and to examine its potential role in nucleoid replication and distribution. As expected, acute overexpression of fluorescently tagged RTN4A or CLIMP63 caused a dramatic increase in the proportion of tubular ER or sheetlike ER structures, respectively, as compared with control cells expressing Sec61b-mCherry, an ER membrane marker with no known role in membrane morphogenesis (Fig. 4A) (40). In contrast, and consistent with previous observations, ER morphology in cells simultaneously overexpressing CLIMP63-GFP and RTN4A-GFP was similar to control cells, consistent with a dynamic balance between the tubules and sheetlike structures within ER networks (Fig. 6A) (41).

Fig. 6 ER tubules license mtDNA synthesis and are required for nucleoid distribution.

(A) ER network morphologies in representative COS-7 cells expressing fluorescently tagged ER membrane proteins: Sec61b-GFP (top left), RTN4A-GFP (top right), CLIMP63-GFP overexpression (OX) (bottom left), and RTN4A-GFP and CLIMP63-mCherry double overexpression (bottom right). Inset regions magnified 5×. (B) Quantification of the number of mitochondrial EdU foci per fixed COS-7 cell after a 4-hour pulse of 50 μM EdU, in cells labeled with MitoTracker Red and indicated ER markers, n = 15+ cells per condition (***P < 0.001, **P < 0.01, two-tailed t test). (C) Quantification of fluorescence intensity of mitochondrial EdU foci, n = 600+ foci from 15+ cells per condition (**P < 0.01; *P < 0.05, two-tailed t test). (D) Image of fixed COS-7 cell overexpressing CLIMP63-mCherry (ER) after a 4-hour pulse of 50 μM EdU in cells labeled with MitoTracker Red. (E) Image of a live COS-7 cell expressing mito-BFP (mitochondria), TFAM-GFP (nucleoids), and overexpressing CLIMP63-mCherry (ER). (Left) Full field view of entire cell; (right) examples of aggregated nucleoids associated with sheetlike ER (top) and distributed nucleoids associated with reticular ER (bottom). Arrowheads indicate colocalization. Scale bars: (A), (D), and (E), 10 μm; insets in (D) and (E), 2 μm.

In COS-7 labeled with MitoTracker Red and overexpressing CLIMP63, RTN4, or both, EdU pulse-labeling of mtDNA was used to assess the number of nucleoids engaged in mtDNA synthesis within a 4-hour window. After fixation, EdU-labeled nucleoids were detected in cells using AlexaFluor647, and total nucleoids and the ER were visualized using PicoGreen staining and indirect immunofluorescence with a Calreticulin antibody, respectively. Relative to control cells, there was a significant reduction in the number of EdU-labeled nucleoids in the cell population overexpressing CLIMP63-mCherry but not RTN4A-GFP (Fig. 6B). In addition, the fluorescence intensity of AlexaFluor647 at EdU-labeled nucleoids was significantly reduced in CLIMP63-overexpressing cells depleted of tubular ER in comparison with control and RTN4A-GFP–overexpressing cells (Fig. 6C). This observation indicates that the reduced number of apparent EdU foci in CLIMP63-overexpressing cells was not due to aggregation of replicating nucleoids. Moreover, in cells simultaneously overexpressing CLIMP63-mCherry and RTN4A-GFP, the number and fluorescence intensity of EdU-labeled nucleoids and ER morphology were similar to control cells, which suggested that the nucleoid phenotypes in CLIMP63-overexpressing cells were a consequence of changes in ER structure and not CLIMP63 expression per se (Fig. 6, B and C). Analysis of mtDNA copy number by qPCR indicated that there was no significant difference in mtDNA content in cells transiently overexpressing CLIMP63-mCherry, RTN4A-GFP, or both, relative to control cells (fig. S6A). This observation suggests that the transient reduction of the proportion of ER tubules in cells acutely disrupted mtDNA synthesis, as opposed to causing mtDNA loss or mtDNA EdU-labeling resistance (Fig. 6B). ER morphological phenotypes were not completely penetrant within the total population of cells overexpressing CLIMP63-GFP. We observed some cells with mosaic intracellular phenotypes in which the severity of ER tubule depletion varied spatially within the cytoplasm (Fig. 6, D and E). In this context, rare EdU-labeled nucleoids were detected even in cells highly overexpressing CLIMP63 and, in every case, were spatially linked to tubular ER colocalized with mitochondria (Fig. 6D, full field views in fig. S6B). Consistently, in live COS-7 cells overexpressing CLIMP63-mCherry and expressing POLG2-GFP and mito-BFP, POLG2-GFP–labeled nucleoids were observed adjacent to residual ER tubules colocalized with mitochondria (fig. S6C); these sites subsequently marked mitochondrial constriction and, ultimately, division events. In addition, we observed a reduction in the number of POLG2-GFP foci in cells overexpressing CLIMP63, as compared with control cells expressing the ER marker Sec61b-mCherry (fig. S6C). Thus, ER structure and, in particular, tubular ER-mitochondria contacts are necessary but not sufficient for homeostatic mtDNA replication.

Given the significant spatial link observed between ER-mitochondria contacts and nucleoid position (fig. S1A), we also examined the overall distribution of the total nucleoid population in cells under conditions of perturbed ER structure. In a majority of cells, overexpression of CLIMP63-mCherry, but not RTN4A-GFP, caused a significant increase in the area of resolvable TFAM-GFP foci, consistent with nucleoid aggregation and hence disturbed distribution (Fig. 6E and fig. S7, A and B). Further examination of a subset of CLIMP63-overexpressing cells that contained some normally distributed nucleoids revealed that these cells contained ER tubules colocalized with mitochondria at positions adjacent to TFAM-GFP–labeled nucleoids (fig. S7B). In addition, at the subcellular level, depletion of ER tubules was highly correlated with nucleoid aggregation (Fig. 6E). In cells simultaneously overexpressing CLIMP63-mCherry and RTN4A-GFP, the ER morphology and nucleoid distribution were similar to those in control cells (fig. S7A). Thus, normal ER structure is required for intracellular nucleoid distribution, which suggests that tubular ER-mitochondria contacts play a role in nucleoid segregation within the mitochondrial network of mammalian cells. These findings demonstrate critical functional interdependencies between mitochondrial and ER dynamics and mitochondrial genome maintenance.


Our data indicate that within mitochondrial nucleoids in mammalian cells, homeostatic mtDNA synthesis is spatially linked to a small subset of ER-mitochondria contacts that are selectively coupled to mitochondrial division. We also observed that nucleoids at mitochondrial tips, which are the products of mitochondrial division, exhibit preferential motility within cells as compared with intramitochondrial nucleoids. We propose that the successive events of mtDNA replication, mitochondrial division, and mitochondrial motility are intimately linked and function together as part of a programmed process that ensures the accurate distribution of mtDNA within cells. Such a process may be especially important for highly polarized cells, such as neurons, whose long axonal processes likely depend on the transport and amplification of mitochondria and associated mtDNA derived from the cell body.

It will be important to determine the fundamental molecular mechanisms linking mtDNA replication initiation to mitochondrial division. Our analyses suggest that contacts between the ER and mitochondria are required to license mtDNA replication. Increasing evidence implicates interorganellar membrane contacts in the formation of membrane microdomains with specialized lipid and protein composition (45). In this capacity, ER-mitochondria contacts could function to facilitate the creation of a spatially defined platform within and on mitochondria that selectively recruits components required for the initiation of mtDNA replication, such as POLG1, POLG2, or other components of the mtDNA replisome. Supporting this idea are recent biochemical observations suggesting that mtDNA and mtDNA replisome proteins associate with cholesterol-rich membrane structures that would be predicted to have raft-like properties (46). Our findings also raise the question of how mtDNA replication, which occurs inside mitochondria, is coordinated with division events associated with the outer surface of the organelle. Perhaps, in addition to contact sites between mitochondria and the ER, there are intramitochondrial spatial determinants that contribute to division-site placement.

Our findings connect ER structure with mtDNA maintenance. This connection has implications for understanding the cellular pathology underlying human diseases and suggests that, for human diseases linked to defects in ER morphogenesis, pathogenesis could be a consequence of mitochondrial dysfunction (47, 48).



All fluorescent protein constructs have been previously described. Mito-BFP and mCherry-DRP1 (26), mCherry-Sec61b (49), mCherry-CLIMP63 (50), GFP-CLIMP63, and RTN4A-GFP (44) were gifts from G. Voeltz. mRuby-KDEL was a gift from J. Wiedenmann (51). Human POLG2-GFP was a gift from W. Copeland (34). Human TFAM-GFP was a gift from M. Alexeyev (52).

Mammalian cell growth and transfection

U2OS, COS-7, and ARPE19 cells (ATCC) were grown in high-glucose Dulbecco’s Modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin. Cells were seeded at ~0.5 × 105 cells per ml in 35-mm glass-bottom dishes (MatTek) 24 hours before transient transfection and 40 hours before imaging. Plasmid transfections were performed for 4 hours in serum- and antibiotic-free DMEM with 3 μl FuGENE6 reagent (Millipore) per dish. Sixteen hours later, cells were imaged in Fluorobrite DMEM (ThermoFisher) containing 10% FBS.

Cell fixation, antibodies, and immunofluorescence

Cells were seeded as described above; 24 hours later, cells were stained with 500 nM MitoTracker Red chloromethyl-X-rosamine (CMX-Ros) (ThermoFisher), and where indicated, a 1:1000 dilution of PicoGreen DNA stain (ThermoFisher) for 15 min at 37°C. Cells were rinsed once in complete medium and twice in warm phosphate-buffered saline (PBS) and were fixed in 4% paraformaldehyde in PBS pH 7.4 for 20 min at room temperature. Dishes were washed twice in PBS and permeabilized in 0.1% TritonX-100 for 20 min. Dishes were blocked in 3% bovine serum albumin (BSA) PBS solution for 1 hour at room temperature. Primary antibodies were added at 1:1000 dilution in PBST (PBS pH 7.4, 1% BSA, 0.1% Tween-20) overnight at 4°C, rinsed twice in PBS, and incubated with Alexa-Fluor–conjugated secondary antibodies at 1:2000 dilution in PBST for 1 hour. Antibodies used: mouse anti-GFP AlexaFluor 488 conjugate (ThermoFisher), donkey anti-rabbit AlexaFluor 405 conjugate (ThermoFisher), anti-GFP (clone N86/8, Neuromabs), rabbit anti-Calreticulin (2907, Abcam).

EdU incorporation was detected via Click-iT EdU AlexaFluor 647 labeling kit (C10640, ThermoFisher) according to the manufacturer’s instructions with minor deviations. Briefly, cells were incubated in 7 μM aphidicolin (A4487, Sigma) for 4 hours in complete medium before a pulse of 50 μM EdU. The EdU pulse was followed by a chase in EdU-free complete DMEM for 1, 24, or 48 hours as described in the main text. For all EdU-labeling experiments, cells were fixed and stained while subconfluent, during logarithmic growth.

Spinning-disk confocal microscopy

Live cell imaging was performed using the spinning-disk module of an inverted objective fluorescence microscope [Marianas spinning-disk confocal (SDC) real-time 3D Confocal-TIRF (total internal reflection) microscope; Intelligent Imaging Innovations] with 100×, 1.46 numerical aperture objective. Either a Photometrics QuantiEM electron multiplying charge-coupled device or Hamamatsu Orca Flash 4.0 scientific complementary metal–oxide–semiconductor (sCMOS) camera was used, depending on the experiment. Images were captured with Slidebook (Intelligent Imaging Innovations); if necessary, linear adjustments were made with ImageJ (NIH). Morphological and quantification analyses were performed in Nikon Elements-Advanced Research (Nikon) as described below. Scale bars were generated using Slidebook.

Mitochondrial division proximity analysis

To predict the frequency that mitochondrial division could occur at POLG2-GFP–marked ER-mitochondrial contacts by random chance, we counted the total number of persistent contacts from each mitochondrion containing a POLG2-GFP focus in the frames leading up to each division event, considering each to be a potential future division site. We then compared the number of ERMD events that occurred within 1 μm of the POLG2-GFP to the total number of persistent contacts and averaged over all events. For example, a mitochondrion with three persistent contacts and one division event would have a per-contact expected frequency of one-third.

MtDNA copy number and qPCR analyses

Total DNA was isolated from U2OS cells by using the DNeasy Blood and Tissue Kit (Qiagen). Quantitative PCR was carried out using SsoAdvanced universal probes supermix (Biorad). Mitochondrial DNA copy number of control, and TFAM-, CLIMP63-, RTN4A- and double-overexpression osteosarcoma cells was performed as in (11) by using primer sequences described therein. Briefly, mtDNA copy number was normalized to nuclear DNA by amplifying an ~132-nucleotide fragment of cytochrome b as compared with a similarly sized fragment of the single-copy nuclear gene, APP. The delta delta Ct method (the ratio of our target gene in our treated sample relative to our untreated sample change in measured cycle thresholds) was used in calculations of fold change.

Statistical analyses, plotting, and modeling

All statistical analyses and plotting were performed in R version 3.2.0 within the RStudio development environment, version 0.99.441. To model the distribution of EdU foci in fixed ARPE19 cells, we used the empirical cumulative distribution function ecdf(). Random simulated data were generated with the runif() function by using 83 observations and a range of 0 to 7, consistent with our real data from the first biological replicate of the 1-hour EdU pulse with 1-hour chase. To recreate the random data set shown in Fig. 4D, use the set.seed() function as follows:

>set.seed(100)#random number generator, initial state

>randomDistToTip<-runif(83, min = 0, max = 7)#generate random dataset

>R<-ecdf(randomDistToTip)#calculate ecdf

>plot(R)#plot random dataset

Supplementary Materials

References and Notes

Acknowledgments: We thank members of the Nunnari lab and W. Copeland for discussion, and M. Paddy and the University of California (UC), Davis, Department of Molecular and Cellular Biology Microscopy Facility for helpful suggestions. This work is supported by NIH grants R01GM106019 and R01GM097432 (to J.N.), an NIH Ruth L. Kirschstein Postdoctoral Fellowship F32GM113388 (to S.C.L.), and a UC Davis Provost’s Undergraduate Research Fellowship (to L.F.U.). J.N. is on the Advisory Board of Mitobridge and declares no financial interest related to this work. The other authors declare that no competing interests exist. All data are included in the main manuscript and supplementary materials.
View Abstract

Stay Connected to Science

Navigate This Article