Tunable organelle extraction from live cells

To enable organelle manipulations, in particular their unconstrained flow through the probe, we manufactured FluidFM probes with a channel height of 1.7 μm and drilled apertures up to 1,100 nm × 1,100 nm (A = 1.2 μm²) with focused ion beam (FIB) compared to the diameter of mitochondria of 300 to 800 nm [25]) (Fig 1B). In addition, we designed and fabricated dedicated probes with a cylindrical tip to facilitate minimally invasive cell entry. These were sharpened by FIB milling to resemble a hollow needle to facilitate membrane insertion (Fig 1B, S1 Fig). These probes had an aperture area of A = 1.6 μm², further minimizing steric limitations and increasing the range of applicable hydrodynamic forces spanning from a few pN to over 100 nN, while showing great robustness toward mechanical stress (S2 Fig). The general workflow for the manipulation of intracellular membrane enclosed compartments involves positioning the FluidFM probe above a selected subcellular location and their insertion by AFM force spectroscopy, followed by either extraction of material from the cell by exerting negative pressure (Fig 1A, S1 Movie) or injection into the cell by positive pressure (Fig 1C). Exclusion of large organelles is achieved by fine-tuning the aperture size (A) and the strength of the applied negative pressure (−Δp). The extraction of cytoplasmic material is monitored in real time, and the extract is inspected inside the FluidFM channel by optical microscopy after relieving the pressure (Δp = 0) and retracting the probe (Fig 2A–2C, S1 Movie). Force feedback by the AFM enables flexible extraction and injection times (seconds to minutes) during which the probe can be maintained inside the cell cytoplasm without impairing cell viability, allowing extraction and delivery of solutions that represent large fraction of the manipulated cells’ own volume (up to 90%) [26]. FluidFM probes with a channel height of 1 μm can capture a total volume of 7 pL. The extraction process allows exclusion of organelle compartments that require larger apertures and higher hydrodynamic forces for extraction. However, when larger apertures are used, smaller and less strongly crosslinked organelles will be coextracted (Fig 1B).

The sampled material can be dispensed subsequently for downstream analyses or transplanted directly into a recipient cell (Fig 1). To examine the capabilities of the newly fabricated FluidFM probes for organelle sampling from single cells, we monitored the endoplasmic reticulum (ER) and mitochondria. We used human osteosarcoma epithelial (U2OS) cells and visualized in parallel the ER by expression of GFP fused to the resident protein Sec61β (sec61-GFP) and mitochondria by expression of BFP targeted to the mitochondrial matrix (su9-BFP). When utilizing pyramidal probes with an aperture size of A = 0.5 μm² and low pressure offsets, Δp < 20 mbar, we accumulated ER in the cantilever, which was accompanied by disappearance of GFP signal in the cell (Fig 2D, S2 Movie). During extraction, the ER was pulled toward the cantilever tip, and we observed a general conversion of cisternal to tubular ER, in both U2OS cells and a similarly labeled kidney cell line (COS7, S3 Fig, S2 and S3 Movies). Notably, under these conditions, the mitochondrial network remained unperturbed, and mitochondria were not extracted.

Next, we aimed at extracting mitochondria using pyramidal probes with a larger aperture size (A = 1.2 μm²) and newly developed, slanted cylindrical probes (A = 1.6 μm²) (Fig 1B). Tunable extraction of mitochondria was achieved using both kinds of probes, thus enabling aspiration of individual mitochondria or sampling of larger quantities of the mitochondrial network (Fig 2E and 2F, S4 and S5 Movies).

We examined cell viability upon subcellular manipulation of ER and mitochondria and did not find it compromised (>95% cell viability) (S4 Fig). To further ensure that our extraction protocol does not damage the cytoplasmic membrane upon probe insertion, we conducted a dedicated set of experiments and monitored potentially occurring Ca²⁺ influx from the cell culture medium using a fluorescent probe (mito-R-GECO1 [27]). Our experiments confirmed that there was no ion influx during and after manipulation, indicating integrity of the cytoplasmic membrane during organelle extraction (S7 and S8 Movies).

Monitoring mitochondrial extraction, we noticed that mitochondrial tubules exposed to tensile forces (negative pressure) underwent a shape transition reminiscent of a “pearls-on-a-string phenotype” [28] inside the cytoplasm of the target cells. This phenotype was characterized by discrete spheres of mitochondrial matrix, connected by thin and elongated membrane stretches (S5A and S5B Fig). These globular structures eventually pinched off upon further exertion of a pulling force and resulted in spherical shaped mitochondria in the cantilever (Fig 2E, S5A–S5C Fig). To date, it was not possible to exert hydrodynamic forces intracellularly, while distinguishing physical manipulation from other potential cellular triggers. The observed mitochondrial “pearls-on-a-string” phenotype was previously described to result from calcium overflow [28] or mitochondrial membrane rupture [29]. To ensure mitochondrial membrane integrity and thus functionality, we investigated whether both mitochondrial membranes remained intact during the process. To this aim, we used U2OS cells and performed time-lapse microscopy during extraction of mitochondria in which both the matrix (su9-BFP) and the outer mitochondrial membrane (OMM; Fis1TM-mCherry) were labeled. Consistent with the observations described above (Fig 2E), we observed the morphological change of mitochondrial tubules exclusively under direct fluid flow at the cantilever aperture and thus exertion of tensile forces, but not upon probe insertion without fluid flow (S5A and S5B Fig, S5 and S6 Movies). Our data further show that the force-induced shape transition propagated over tens of micrometers along the mitochondrial tubules in the millisecond to second range after negative pressure was applied with FluidFM (S5B and S5C Fig, S6 Movie). The shape transition of the matrix compartment propagated homogeneously along connected mitochondrial tubules, while the OMM between the matrix foci initially remained intact. When traction was maintained for a few seconds, the OMM separated at one or more constriction sites between previously formed “pearls”, which resulted in isolated spherical mitochondria, while the remainder of the tubular structure relaxed and recovered (S5A–S5C Fig, S6 Movie). Furthermore, we demonstrate that the observed scission process of “pearling” mitochondria (S5C Fig) succeeds recruitment of the fission machinery protein GTPase dynamin-related protein 1 (Drp1) [30] to the constricted sites (S1 Text, S5D Fig). Combining mitochondrial pulling experiments with mitochondria-localized calcium sensors, we were able to show that the shape transition toward the pearls-on-a-string phenotype and subsequent mitochondrial fission inside the cytoplasm was calcium-independent (S6 Fig, S7–S10 Movies) (S1 Text, S5 and S6 Figs, S7 and S8 Movies).

Mitochondrial transplantation into cultured cells

Our next goal was to demonstrate the functional delivery of mitochondria into new host cells and to achieve cell-to-cell organelle transplantation. In contrast to mitochondria extraction, for which both pyramidal probes and cylindrical probes could be used (Fig 1B), injection of mitochondria was possible only with the latter, newly developed probes. FluidFM offers 2 possibilities for mitochondrial transfer: transplanting mitochondria from a donor cell to a recipient cell by coupling mitochondrial extraction with reinjection of the extract into a new host cell or back-filling FluidFM probes with mitochondria purified by subcellular fractionation, followed by injection (Fig 3A–3D). Working with bulk-isolated mitochondria allows for a higher throughput of cells injected in series with one cantilever (>1 cell per minute). However, such a protocol is accompanied by reduced mitochondrial quality caused by the preceding purification process. We compared both approaches, the cell-to-cell approach (Fig 3A) and the injection of purified mitochondria (Fig 3C), with respect to the delivery of mitochondria into the cytoplasm of individual cultured HeLa cells.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Mitochondrial transplantation.

(A) Scheme of mitochondrial transplantation using the cell-to-cell transfer approach: Mitochondria are extracted via FluidFM aspiration. Subsequently, the cantilever holding the extract is moved to a recipient cell and the extract is injected. (B) Image of a FluidFM cantilever prefilled with perfluorooctane after mitochondrial extraction, mitochondria are labeled via su9-mCherry. Extracted volume approximately 0.8 pL. Scale bar: 10 μm. (C) Scheme of mitochondrial transplantation using mitochondria purified by a standard mitochondrial purification protocol. The purified mitochondria are resuspended in HEPES-2 buffer and filled directly into the fluidic probe. Cells are injected consecutively. (D) Image of a FluidFM cantilever filled with mitochondria isolated from bulk, labeled via su9-mCherry. Scale bar: 10 μm. (E) Images of a recipient cell post–mitochondrial transplantation via the cell-to-cell approach. The host cells’ mitochondrial network is labeled via su9-BFP, and the transplant is labeled via su9-mCherry. Scale bar: 10 μm. (F) Images of a recipient cell post–mitochondrial transplantation via the injection of isolated mitochondria approach, labels similar to c. Scale bar: 10 μm. (G) Evaluation of mitochondrial transplantation via the cell-to-cell approach upon optical inspection and the injection of isolated mitochondria approach. A total of 40 cells were evaluated per approach. (H) Absolute numbers of transplanted mitochondria of 22 individual cells evaluated for the cell to cell and the injection of isolated mitochondria approach. (I) Fusion states of transplanted mitochondria 30 post–cell-to-cell transplantation. Mitochondria are visualized using different fluorescent labels for the transplant (su9-mCherry) and for the host mitochondrial network (su9-BFP). Scale bar: 5 μm. (J) Fusion states of transplanted mitochondria 30 postinjection of purified mitochondria, similar labellng as in g. Scale bar: 5 μm. (K) Degradation of transplanted mitochondria, the transplant is split into multiple smaller fluorescent vesicles (su9-mCherry) showing no overlap of fluorescence with the labeled host cell mitochondrial network (su9-BFP). Scale bar: 5 μm. (L) Time-lapse image series of a single transplanted mitochondrion (su9-mCherry). The organelle donor was a HeLa cell, recipient cell is a U2OS-cell with a fluorescently labeled mitochondrial network (su9-BFP). Scale bar: 10 μm. The data underlying Fig 3G and 3H can be found in S1 Data.

https://doi.org/10.1371/journal.pbio.3001576.g003

To visualize the transfer of mitochondria, we used donor and acceptor cells with a differentially labeled mitochondrial matrix (Fig 3E and 3F; su9-mCherry and su9-BFP, respectively). When transplanting mitochondria directly from cell to cell using FluidFM, we achieved successful transfer of mitochondria into the cytosol of the recipient cells in 95% of all cases, while maintaining cell viability (Fig 3G, 39 out of 41 transplanted cells). Upon injection of purified mitochondria, we observed mitochondrial transfer and preserved cell viability in 46% of cases (19 of 41) (Fig 3G, S7 Fig). Quantification of the transplant showed that the number of transplanted mitochondria for these experiments varied from 3 to 15 mitochondria per cell (Fig 3H). The different success rates between the 2 alternative protocols can be explained by differences in mitochondrial condition. When evaluating mitochondrial extraction protocols, we observed that a fraction of the extracted mitochondria undergo rupture of their outer membranes (S7 Fig) [31]. Irreversible damage of mitochondria leads to degradation inside cells and potentially cell apoptosis due to cytochrome c leakage. While cell-to-cell transplantation of mitochondria reduces throughput, it has the advantage that the extracellular time is short (<1 minute) and that mitochondria sampled by FluidFM are maximally concentrated in native cytoplasmic fluid, bypassing the use of artificial buffers altogether. We ensured that the extract remained near the aperture during extraction by filling the probes with immiscible perfluorooctane before extraction and transplantation. Therefore, only small volumes (0.5 to 2 pL) are injected into the host cells (Fig 3B), up to the volume previously extracted from the donor cell (injection of larger volumes is automatically prevented due to inherent flow resistance properties of the prefilled fluorocarbon liquid).

Labeling mitochondria of the recipient cell (su9-BFP) in addition to labeling donor cell mitochondria (su9-mCherry) allowed us to survey the state of the mitochondrial network in the transplanted cell. In both transfer approaches described above (transplantation and purification followed by injection), the tubular, interconnected phenotype of the host mitochondrial network remained unaltered by the injection process. In addition, labeling allowed us to monitor the fate of the transplanted mitochondria. We observed mitochondrial acceptance, defined by fusion of the transplant with the host mitochondria network, and thus overlap of both fluorescent signals and mitochondrial degradation, marked by further fragmentation of the transplant and segregation into presumptive mitophagosomal structures (Fig 3I–3K). These processes were observed irrespective of the transfer method, cell to cell (Fig 3I), or injection of purified mitochondria (Fig 3J). We followed the fate of the transplant over time in 22 cells: 18 cells showed full mitochondrial fusion of the transplant, and 4 cells showed mitochondrial degradation. Fusion events were first observed within 30 minutes posttransplantation in most cases (14 of 18 cells).

As indicated above, the high cell-to-cell transplant efficiency allows for direct observation of the fate of individually transplanted mitochondria. To showcase this, we transplanted labeled mitochondria (su9-mCherry) from HeLa cells into differentially labeled U2OS cells (su9-BFP), regularly used for studies of dynamic mitochondrial behavior. The strong label together with a highly sensitive microscope camera enables tracing of individual mitochondria within a recipient cell over time (Fig 3I, S11 Movie). In this case, we observed rapid spread of the fluorescent mitochondrial matrix label after the initial fusion event to the network 23 minutes posttransplantation.

In summary, we established 2 methods for mitochondria transfer into single cultured cells. One involves bulk purification of mitochondria and their injection into recipient cells. The injection protocol is rather rapid but inevitably compromises mitochondrial and cellular function. The second consists in cell-to-cell transplantation. Evaluation of cell viability and transplant fate showed an efficient protocol that allows observation of the dynamic behavior of transplanted mitochondria after transfer.

Fate of transplanted mitochondria in primary cells

Having developed an efficient protocol for cell-to-cell transplantation of mitochondria, we sought to test whether primary cells show similar uptake behavior as the tested cancer cells and, if so, what are the dynamics of integration of foreign mitochondria. We considered these particular experiments important because quality control mechanisms are impaired in cancer cell lines [32] and to demonstrate the broad applicability of the established protocol. In addition, several studies link naturally occurring mitochondrial transfer events with short-term benefits for individual cells and tissues, for example in osteocytes [33], adipose tissue [34], or in neurons [35]. However, to the best of our knowledge, the fate of mitochondria or dose-response relationships have not been studied, and appropriate technologies of mitochondrial transfer that preserve cell viability have been lacking.

We used primary human endothelial keratinocytes (HEKa), a skin cell type that is generally susceptible to radiation damage and aging [36]. In standard culture conditions, the mitochondrial network of HEKa cells is mostly tubular, forming a large connected network (S8 Fig) similar to HeLa cells studied above, indicating an active mitochondrial fusion machinery [37]. Notably, cell-to-cell transplantation of mitochondria into HEKa cells showed no impact on their viability, allowing analysis of all injected cells. We conducted time-lapse experiments of transplanted labeled mitochondria (su9-mCherry) from HeLa cells into HEKa cells, focusing on the fluorescent signal of the transplant and its dynamic behavior over time (Fig 4A–4C, S12 Movie).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Mitochondrial fate in HEKa cells.

(A) Time series of cell-to-cell transplantation from HeLa (su9-mCherry) to HEKa, fusion and degradation to the unlabeled host cell network. Scale bar: 10 μm. (B) Surface-render (total su9-mCherry signal) of the time-lapse series of mitochondrial fusion and degradation within the host cell shown in a. The transplant is depicted in red, host cell mitochondrial network carrying fluorescent traces of the transplant is shown in gray, presumptive mitophagosomal structure is shown in light blue. (C) Enlarged section of the 2 hours time point surface render from b. (D) Scatter plot of size and normalized fluorescence intensity distribution of objects detected at the 2-hour time point. (E) Zoom in on low-size fraction of the scatter plot shown in c. (F) Total number of objects classified as primary transplant or secondary particle over time. (G) Absolute number of primary and secondary particles detected directly after transplantation (0 hour) and after 3 hours for 12 cells. Paired t test, p = 0.0095. (H) Fusion and degradation behavior of drug-compromised mitochondrial transplants in HEKa cells. Each condition was tested with 22 cells. (I) Distribution of mitochondrial quantity transplanted per cell from all conditions tested in e, total number of cells: 132. Dashed line indicates mean value. Total number of transplanted mitochondria: 1,117. (J) Histogram plot correlating the amount of transplanted mitochondria with uptake outcome; bin width is 5. Correlation of the absolute number of transplanted mitochondria per cell with the cell response of either mitochondrial fusion or degradation across all transplant conditions in percent. Total cell number: n = 135. The data underlying Fig 4D–4J can be found in S1 Data. HEKa, human endothelial keratinocyte.

https://doi.org/10.1371/journal.pbio.3001576.g004

The analysis revealed the rate of transplant fusion with the host mitochondrial network and its movement inside the cytosol. In contrast to the findings described above for HeLa-to-HeLa transplantations, i.e., host cells either accepting or degrading mitochondria (S9 Fig), we observed a third, intermediate scenario in which single cells both fused and degraded parts of the transplant. To further define the transplant uptake mechanisms in HEKa cells, we classified the transplanted mitochondria as primary particles—mitochondria that retain their original shape and fluorescence intensity, indicating that they have neither fused with the host network, nor been targeted to degradation—secondary particles—mitochondria that retain their fluorescence but are distinctively smaller in size, suggesting fragmentation for degradation—and tertiary particles—mitochondria fused with the host network, showing a characteristic decrease of transplant specific fluorescence (>10-fold), due to dilution into the host mitochondrial network (Fig 4D and 4E, S12 Movie). Shape transition of the transplant from discrete spheres toward tubules was observed exclusively in the context of fusion of the transplant to the existing host network. Emergence of secondary particles hinted toward rearrangement of the transplant into parts recognized as “viable” mitochondria, designated for fusion to the network, while another part was targeted for degradation. Tracing this process within an individual cell, we counted the number of “primary” and “secondary” particles over 14 hours (Fig 4F). In the first 3 hours postinjection, we observed an increase in “primary” and “secondary particles”, which subsequently lost their high fluorescent signature while the fraction of fluorescent mitochondrial network increased over time (Fig 4B–4E). To test whether restructuring of the transplant into subparticles was common in cells that showed emergence of secondary particles, we counted the number of primary and secondary particles immediately after transplantation and 3 hours after in cells that showed both degradation and integration of the transplant (Fig 4G; n = 10). The results indicated variance in the absolute extent of rearrangement; however, all cells showed an increased particle number by a factor of 1.7 ± 0.4.

In this first set of transplantation experiments from HeLa cells to HEKa cells, 12 cells showed complete uptake of the transplant, 6 showed both uptake and degradation of the transplant, and 4 fully degraded the transplant (Fig 4H). The degradation of defined amounts of transplanted mitochondria can be traced over time in single cells, as exemplarily shown in S10 Fig.

In the experiments described above, the first fusion or degradation events occurred 20 minutes posttransplantation and continued for more than 16 hours. Remarkably, the speed at which the primary transplant was processed was independent of transplant sizes. Fusion or degradation of large transplants, >40 mitochondria per cell, advanced at similar rate as smaller transplants, <8 mitochondria per cell (S12 Movie, S11 Fig).

As outlined above, the established cell-to-cell transplantation protocol is minimally invasive regarding the integrity of the mitochondria themselves when using “healthy” donor cells, and cells receiving transplants from unperturbed donor cells showed mostly uptake of the transplant (Fig 4H). Next, we wondered whether the acceptance of mitochondria by host cells was altered, if the quality of the transplant was impaired by prior drug treatment of the donor cell. Such treatments are commonly used to study the pathways that control the maintenance of the mitochondrial network. However, drugs have been applied to the entire cell so far, likely interfering with regulatory processes of cells as a whole. In consequence, the here established cell-to-cell transplantation of mitochondria provided an opportunity to follow the fate of damaged mitochondrial subpopulations in the context of an otherwise intact cell. In vitro, the chemical triggers used to study mitophagy are uncoupling agents, inhibitors of the respiratory chain or combinations thereof, robustly causing changes of membrane potential and simulating low mitochondrial quality [38]. To investigate how otherwise healthy cells respond to impaired mitochondrial subpopulations, we treated transplant donor cells with the proton ionophore CCCP (S8 Fig) and transplanted the depolarized mitochondria into HEKa cells. Most cells, 14 of 22, showed fusion of initially depolarized mitochondria to the network, while 8 cells degraded the transplant (Fig 4H). This reaction was similar to the previous condition tested, the transplantation of untreated mitochondria, which was unexpected, because literature suggests rapid mitochondrial degradation of mitochondria having lost their membrane potential [39,40]. However, the membrane potential can potentially recover quickly, if a functional ATPase is present, as expected in our study. Next, we tested the influence of doxycycline, an antibiotic, inhibiting protein synthesis of the mitochondrial ribosome, which induces fractionation of the mitochondrial network in HeLa cells after 24 hours (S8 Fig), putatively due to nonfunctional mitochondrial OXPHOS protein complexes [41]. The reaction of the host cells was similar as toward depolarized and untreated mitochondria: 15 show full fusion, 4 show both fusion and degradation, and 3 show full degradation of the transplant. We then treated donor cells with H₂O₂, which damages proteins and induces double strand breaks in mtDNA. The fraction of mtDNA containing double strand breaks in H₂O₂ treated cells was reported to remain high for several hours posttreatment [42], and the donor cells showed a fragmented mitochondrial network after 3 hours (S8 Fig). However, even upon H₂O₂ treatment, and even upon application of all drugs simultaneously to donor cells, mitochondrial acceptance in healthy recipient background was still comparable to all other conditions, indicating the potential of cells to cope with highly damaged mitochondria when occurring as isolated events (Fig 4H, S12 Fig).

Since the tested conditions for drug-impaired mitochondria did not show an impact on mitochondrial uptake behavior, we wondered whether the amount of injected mitochondria might have an impact on transplantation outcome. Across all conditions, we transplanted 1,117 mitochondria (Fig 4I) and followed their fate in more than 100 individual primary cells after transplantation of 1 up to 53 mitochondria (S12 Fig). Pooling uptake data of cross-tested conditions (n = 135 cells), we plotted the frequency of cells showing transplant fusion and transplant degradation versus the number of individually transplanted mitochondria and could not find any correlation between the number of transferred mitochondria per cell and their uptake behavior (Fig 4J).

Overall, transplantations were successful in all experiments and cells remained viable, irrespective of the amount of transplant received. We demonstrated that primary HEKa cells incorporate a majority of transplanted mitochondria into their network via mitochondrial fusion. A subset of transplanted cells showed individualized responses, with one fraction of mitochondria fusing with the mitochondrial network and another subpopulation undergoing rearrangement associated with particle formation, while a third fraction was subjected to degradation. This behavior highlights the presence of quality control mechanisms that are in place to sort each mitochondrion and exemplifies the potential of the FluidFM approach to transplant mitochondria into new host cells to study their fate and host cell response with single mitochondria resolution.

Mitochondria transplantation and transfer of mitochondrial genomes

After demonstrating the short-term response of cells to mitochondrial transplantation, we focused on the long-term effects of mitochondrial transfer over generations of host cells. Mitochondria differ from other membrane compartments in that they carry their own genome, which is propagated within the cell’s inherent mitochondrial pool. It has been shown that mtDNA can be transferred into somatic cells via miniaturized probes under selective pressure, either by transfer into cells artificially rendered free of mitochondrial DNA (rho zero cells) [19,22] (2% to 25% efficiency) or by selection using antibiotics (<0.01% efficiency) [9]. Therefore, the introduction of new mtDNA sequences into functional somatic cells remains a challenge [20,43]. Before demonstrating the transfer of mtDNA into host cells upon transplantation, we first wondered whether the FluidFM-extracted mitochondria contain mitochondrial DNA, which is organized in discrete complexes termed mitochondrial nucleoids, because the extraction process leads to rapid fragmentation of the network. To visualize the behavior of mitochondrial nucleoids during mitochondrial extraction using FluidFM, we expressed a fluorescently tagged version of p55 (p55-GFP), a polymerase-γ subunit that colocalizes with mitochondrial nucleoids and appears in discrete speckles scattered throughout the mitochondrial matrix [44] (Fig 5A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Mitochondria transplantation and transfer of mitochondrial genomes.

(A) Time-lapse evaluation of mitochondria extraction visualizing mitochondrial nucleoids. An overlay of mitochondrial matrix shown in blue (su9-BFP) and mitochondrial nucleoids shown in red (p55-GFP). Yellow box indicates position of the cantilever. (B) Dynamic localization of labeled nucleoids within a fractionating mitochondrial network. Plotted are the total amounts of mitochondria (su9-BFP) that show overlap with a labeled nucleoid (p55-GFP) in red and idle mitochondria showing no fluorescent trace of p55-GFP in blue over time. See also S13 Fig. Scale bar: 5 μm. (C) Strategy for the quantification of mtDNA maintenance after mitochondrial transplantation. (D) Quantification of retained U2OS mtDNA in transplanted HeLa cells of all approaches tested. Bars indicate mean value. Difference between condition i and ii, p = 0.0138 and between condition i and iii, p = 0.0034. The data underlying Fig 5B and 5D can be found in S1 Data. mtDNA, mitochondrial DNA.

https://doi.org/10.1371/journal.pbio.3001576.g005

We evaluated time-lapse experiments upon mitochondrial extraction with labeled nucleoids and counted mitochondrial fragments either containing labeled nucleoids or being devoid of their fluorescent signal (Fig 5B, S13 Fig). We observed p55-GFP in >90% of formed mitochondrial spheres (n = 18), indicating that most transplanted mitochondria contained mtDNA.

Next, we applied genomics to quantify uptake and maintenance of mtDNA after mitochondria transplantation (Fig 5C). We compared 2 methods of mitochondrial transfer, using U2OS cells as mitochondria donors and HeLa cells as acceptors: direct cell-to-cell transplantation of mitochondria with FluidFM and FluidFM injection of purified mitochondria. In addition, to control for nonspecific uptake of mitochondria or otherwise unspecific mtDNA carryover, we seeded cells into a culture dish and mixed them with extracted mitochondria from donor cells. All experiments were executed in biological triplicates and cells were passaged onto a fresh culture dish; for a more detailed description of the experimental procedure and conditions, see Methods section. Subsequently, we extracted DNA, then amplified and sequenced part of the variable D-loop region of the mitochondrial genome. Four single nucleotide polymorphisms [45] allowed the detection of transplanted mtDNA. The percentage of transplant mtDNA propagation was highest in samples from cell-to-cell mitochondria transplantation (mean 2%), followed by FluidFM injection of purified mitochondria (mean 0.5%), whereas mtDNA transfer by nonspecific uptake was below the detection limit (Fig 5D, S1 Table). The obtained amount of mtDNA was well in line with our expectations considering the amount of mtDNA copies per cell (approximately 1,100) [46] and mtDNA copy number per nucleoid (1.4 to 3.2) [47,48], and the quantity of mitochondria we transplanted within this experimental series (Fig 3H, median = 6).

In summary, we showed efficient transplantation of mitochondrial DNA into somatic cells without the need for selective pressure. These results indicate that HeLa cells show little or no selection for mitochondrial DNA variants of another cell line, here U2OS cells, under standard culture conditions.

Manufacturing process of FluidFM probes with a cylindrical tip

The process starts with a standard 4” silicon wafer in the (100) crystallographic orientation. A 400-nm thick silicon-rich nitride (SiRN) layer is deposited by low pressure chemical vapor deposition (LPCVD) on the selected wafer. The thickness of the SiRN layer determines the thickness of the bottom wall of the microfluidic channel. A silicon oxide layer is deposited by LPCVD from TEOS on the top of the SiRN layer. Then, a circular opening is patterned by reactive ion etching (RIE) on the deposited multilayer. By using deep reactive ion etching (DRIE), a cylindrical pattern is subsequently formed in silicon. The depth of the cylindrical mold defines the height of the cylindrical tip. After the formation of the cylindrical mold, a 150-nm thick SiRN layer is deposited by LPCVD. The thickness of the second SiRN layer determines the thickness of the walls in the cylindrical tip. After the tip molding, a blanket etch (RIE without an etching mask) RIE is performed to remove silicon nitride from the top surface of the wafer and from the bottom of the cylindrical mold. Due to the directionality of the RIE process, the material on the sidewalls is preserved. The silicon oxide layer, which protects the underlying SIRN layer during the RIE etching, is removed. Subsequently, a 1,500-nm thick polysilicon layer is deposited by LOCVD. Polysilicon is used as a sacrificial material to form a microfluidic channel. The thickness of the polysilicon layer determines the height of the microfluidic channel. The layout of the microchannel is patterned by RIE of polysilicon. A SiRN layer with a thickness of 400 nm is deposited by LPCVD after the patterning of polysilicon. This SiRN layer forms the top wall of the fluidic microchannel. The layout of the cantilever and the inlet of the microchannel are defined by RIE of silicon nitride. Next, a reflective metal layer is deposited and patterned on the cantilever. The silicon wafer is anodically bonded to a prediced glass wafer in which the fluidic inlets are premanufactured by powder blasting. After the anodic bonding, the sacrificial polysilicon layer is removed by tetramethylammonium hydroxide (TMAH) etching in order to empty the microchannel and form a hollow cantilever. The bulk silicon is also removed during the TMAH etching to completely release the cantilever. A small part of silicon remains underneath the microfluidic inlet to provide mechanical strength for the channel walls.

FluidFM probe processing and FIB-SEM imaging and milling

The FluidFM probes were mounted into a custom probe holder and coated with an 18-nm carbon layer using a CCU-010 Carbon Coater (Safematic, Switzerland) before milling by a FIB-SEM Nvision 40 device (Zeiss, Germany) using the Atlas software (Zeiss).

Pyramidal probes used for extraction of organelles and hydrodynamic pulling experiments: The milled face of the pyramidal probe was aligned perpendicularly to the FIB beam equipped with a gallium ion source. Subsequently, the probes were milled with an acceleration voltage of 30 kV at 10 pA until the aperture was milled, controlled by optical observation (approximately to an electric charge of 10 nC per μm²). The pyramidal probes have a 100-nm thick silicon nitride layer at the aperture region.

Cylindrical probes used for mitochondrial transplantation experiments were “sharpened” by milling the probe apex at a 50° angle alongside the cylinder at an acceleration voltage of 30 kV at 40 pA. The probes were optically controlled after the milling procedure.

Then, the probes were glued onto a cytoclip holder by Cytosurge (Cytosurge AG, Switzerland). Before each experiment, the cantilevers were cleaned by a 90-second plasma treatment (Plasma Cleaner PDG-32G, Harrick Plasma, USA) before coating overnight with vapor phase SL2 Sigmacote (Sigma-Aldrich, USA) in a vacuum desiccator. The siliconized probe was oven dried at 100 °C for 1 hour. The cantilever spring constant was measured using software-implemented scripts (cylindrical probes: 2 ± 0.4 Nm⁻¹, pyramidal probes: 5 ± 1 Nm⁻¹).

FluidFM setup and microscopy

The FluidFM setup is composed of a FlexAFM 5-NIR scan head controlled by a C3000 controller (Nanosurf, Switzerland), a digital pressure controller (ranging from −800 mbar to +1,000 mbar) and Microfluidic Probes (Cytosurge). The scan head is mounted on an inverted AxioObserver microscope equipped with a temperature-controlled incubation chamber (Zeiss). The microscope is coupled to a spinning disk confocal microscope (Visitron, Germany) with a Yokogawa (Japan) CSU-W1 scan head and an EMCCD camera system (Andor, UK). For all images and videos, a 63× oil objective with 1.4 numerical aperture and a 2× lens switcher was used (without lens switcher: 4.85 pixel/micron and 9.69 pixel per micron with lens switcher); images are in 16 bit format. Image acquisition was controlled using the VisiView software (Visitron); linear adjustments and video editing were made with Fiji [67]; additionally, images and videos were noise-filtered using the wiener noise filtering function (wiener2; 3 by 3 neighborhood size) in MATLAB R2018a (MathWorks, USA). Movies were created using a self-written MATLAB script in order to visualize several sections or channels within the same movie. Colormaps originate from Thyng and colleagues [68]. Images of cantilevers containing extracts (Fig 2) were created summing up the slices of a Z-stack via Fiji and reconverting the image to 16 bit format.

Cell culture

U2OS, COS7, and HeLa cells were maintained in Dulbecco’s Modified Eagle Medium containing 1% penicillin-streptomycin (Thermo Fisher Scientific, USA) and 10% fetal bovine serum (Thermo Fisher Scientific) culture medium at 37 °C and 5% CO₂ in a humidified incubator. Primary adult HEKa cells were purchased from Thermo Fisher Scientific. HEKa cells were cultivated in EpiLife Medium, with 60 μM calcium, with Human Keratinocyte Growth Supplement, 1%, (Thermo Fisher Scientific) and Gentamicin/Amphotericin Solution (Thermo Fisher Scientific). Cells were seeded 48 hours preceding the experiments onto 50-mm tissue culture treated low μ-dishes (ibidi, Germany) inside 2-well culture inserts (ibidi) or, for experiments coupling mitochondrial extraction via FluidFM to injection of mitochondria, inside 4-well microinserts (ibidi). For the experiments, the culture media was replaced with CO₂-independent growth medium containing 10% FBS (Thermo Fisher Scientific) and 1% penicillin-streptomycin (Thermo Fisher Scientific). Cell lines stably expressing fluorescent proteins markers were created via lentiviral transduction; the constructs were previously described in Helle and colleagues [52].

Transient transfection

U2OS-cells were seeded 72 hours preceding the experiment into 2-well culture inserts (ibidi) inside 50-mm tissue culture treated low μ-dishes (ibidi) in a total volume of 100 μL per well. Transfections were performed overnight 36 hours preceding the experiment using 0.2 μg plasmid DNA and 0.2 μL Lipofectamine P3000 solution following the manufacturer’s instructions (Thermo Fisher Scientific). The media was exchanged to standard culture medium 12 hours posttransfection. The plasmid used for calcium imaging CMV-mito-R-GECO1 was a gift from Robert Campbell (Addgene plasmid #46021; http://n2t.net/addgene: 46021; RRID: Addgene_46021).

Mitochondrial pulling and extraction experiments

All experiments were executed at 37 °C. The cells for extraction/transplantation were selected by optical inspection using light microscopy. Z-stacks were taken before and after the manipulation step to document the workflow. FluidFM probes were prepared as specified above and prefilled with octadecafluorooctane (perfluorooctane) (Sigma-Aldrich). Subsequently, the FluidFM probe was moved over a targeted area in the cytosol of a selected cell, usually close to the nucleus or mitochondrial tubes in the cell periphery. The cantilever was then inserted at the specified location driven by a forward force spectroscopy routine in contact mode, until the set point of 400 nN was reached. The probe was then kept at this position (in the X-Y dimension) at the given force offset. Then, negative pressure in the range between −10 and −150 mbar was applied to aspirate cellular content. Before retracting the probe at the end of the aspiration process, the pressure was set back to 0 mbar. The force set point was adjusted by analyzing force distance curves from neighboring cells within the same experiment; the force value at which the curve takes a linear shape (in this case 80 nN) was estimated. This force value marks the point at which the probe makes contact with the glass bottom below the cell; consequently, this value was chosen as a set point for the extraction. Extraction of the ER fraction: The experiments were performed using pyramidal cantilevers featuring an aperture area of 0.5 μm² (see Fig 1B) at −20 mbar using U2OS or COS7 cells. Experiments for both cell lines were repeated twice on different days. Extraction of mitochondria and mitochondria pulling experiments were performed using pyramidal cantilevers featuring an aperture area of 1 μm² or cantilevers with a cylindrical apex featuring an aperture area of 1.5 μm² (see Fig 1B) at −20 to −100 mbar using U2OS or HeLa cells. Experiments including extraction of mitochondria were executed in 32 individual experimental setups on 32 individual days. Experiments showing recruitment of Drp1 were repeated 4 times on 4 different days with at least 5 cells per experiment in U2OS cells. Mitochondrial pulling experiments connected with calcium imaging and thapsigargin treatment (Fig 5) were executed 3 times, each time on an individual day with at least 7 cells per experiment.

For probe insertion with minimal Ca²⁺ influx as shown in S7 Movie, newly coated (Sigmacote; see above) FluidFM cantilevers were used. To deliberately disturb the cell membrane (S9 Movie), the probe was driven into the cell using the same set point, then the optical table (Newport, USA) was gently flicked with the index finger to cause the probe to shift slightly, effectively disturbing the cell membrane. Cell viability was controlled using the LIVE/DEAD cell imaging kit (Thermo Fisher Scientific).

Mitochondrial transplantation experiments

FluidFM injection of bulk-purified mitochondria: Mitochondria were purified from approximately 2*10⁶ cultured HeLa cells continuously expressing the mitochondrial matrix marker su9-mCherry using the Qproteome Mitochondria Isolation Kit (Qiagen, Germany) following the manufacturer’s instructions. After purification, the mitochondria were washed 2 times in injection buffer (see above) and finally resuspended in 40-μl injection buffer before being loaded into FluidFM probes having a sharpened cylindrical apex.

Coupling mitochondrial extraction with transplantation from individual cell to cell: Mitochondria were aspired as described above, using FluidFM probes having a sharpened cylindrical apex from HeLa-cells that were cocultured on the same dish, previously seeded within another quadrant of the 4-well micro insets. Subsequently, the probe was moved to a region containing cells targeted for transplantation. For injection of mitochondria, the probe was positioned above a target cell as described above for the extraction of organelles. The cantilever containing the organelles was then inserted into the cell in contact mode (set point 100 nN). The injection process was controlled by observing the process in brightfield in real time, at first, a positive pressure of 20 mbar was applied and the separation line between extract and perfluorooctane was monitored, the pressure was increased until the line was moving toward the aperture (maximal value of 200 mbar), effectively pushing the cell extract into the recipient cell. When the extract was totally or in large parts transferred to the recipient cell, the pressure was then set to 0 mbar, and the probe was retracted. Z-stacks were taken to control for successfully transferred mitochondria. Cell viability was controlled using the LIVE/DEAD cell imaging kit (Thermo Fisher Scientific) following the manufacturer’s instruction. Both experimental approaches were performed in 6 individual experiments, each on individual days.

Quantification of genetic incorporation of transplanted mitochondria

All experiments were conducted in biological triplicates. (A) FluidFM injection of purified mitochondria: Single HeLa cells continuously expressing the mitochondrial matrix marker su9-BFP were seeded into a single quadrant of a 4-well microinsert (ibidi) 72 hours before the experiment, at the day of the experiment, each seeded well contained 4 to 6 cells. All cells were injected as described above. (B) Coupling mitochondrial extraction with transplantation from individual cell to cell: One quadrant of a 4-well microinsert (ibidi) was seeded with recipient HeLa (su9-BFP) cells as described above. A second quadrant was seeded with U2OS cells continuously expressing su9-mCherry as donors. The transplantation was executed as described above. Subsequently, the U2OS cells were removed, again using a FluidFM probe filled with 1% sodium dodecyl sulfate (Merck Group, Germany). Before sequencing, the whole cell population was controlled for any signal mCherry positive cells via fluorescence activated cell sorting, but no unwanted carryover of remaining U2OS cells (mCherry+) could be detected. Control for nonspecific mitochondrial carryover: Approximately 100,000 Hela cells expressing the mitochondrial matrix marker su9-BFP were seeded onto a culture dish. They were mixed with purified mitochondria from approximately 2*10⁶ U2OS cells expressing su9-mCherry by pipetting the extracts gently on top of the cultured cells. Cells obtained after treatments A and B were grown for 8 days, transferred into a new culture dish and grown for another 6 days. The control cells were grown for 3 days, transferred into a new culture dish and grown for another 3 days. Subsequently, the total DNA of all samples was purified using the MasterPure Complete DNA and RNA Purification Kit (epicenter) following the manufacturer’s instructions. Part of the D-loop region was amplified via PCR using primers 1 and 2 (S2 Table). To create a “null hypothesis” for the following sequencing workflow, 3 test samples were created consisting of 0.1%/0.5%/1% U2OS amplicons, mixed with HeLa amplicons. Subsequently, all samples were sequenced using the a Pacbio Sequel SMRT cell, subsequently analyzed as described in Russo and colleagues [69] The reads were assigned to either U2OS mtDNA or Hela mtDNA using 4 conserved mutations of U2OS cells within this region previously identified via Sanger sequencing (15959G>T, 16069C>T, 16108C>T, 16126T>C; see S1 Table).

Statistics and reproducibility

All representative experiments were repeated at least 3 times independently with similar results. Absolute counts of cells or mitochondrial particles are included in the main text or in the figure legends. Analysis for single nucleotide polymorphisms was performed as described in Russo and colleagues [69]. Performed statistical tests and significance levels are indicated in the figures and respective figure legends. Statistical analyses can be found below the plotted data in S1 Data.

Analysis of mitochondrial quality control mechanisms with mitochondrial transplantation in primary HEKa cells

Cells were seeded into as described above, into low μ-dishes (ibidi) inside 2-well culture inserts (ibidi). The special separation allows for drug-treatments of the donor cell population before the experiment, while the host cell population of HEKa cells remains under standard culture conditions. At the beginning of the experiment, the cell media was replaced with CO₂-independent growth medium containing 10% FBS (Thermo Fisher Scientific) and 1% penicillin-streptomycin (Thermo Fisher Scientific). A total of 4 μM oligomycin was added when indicated. Treatment conditions: 10 μM CCCP for 3 hours, 8 μM doxycycline for 48 hours, 750 μM H₂O₂ for 3 hours. Mitochondria were extracted from Hela cells and transplanted into HEKa cells. A Z-stack image series of the transplant and the host cell was taken directly after the transplantation process with a 63× oil objective with 1.4 numerical aperture and a 2× lens switcher in 500-nm steps. Further Z-stacks were taken at other time points depending on the individual experiment. For the endpoint at 18 to 22 hours posttransplantation, the mitochondrial network was visualized using MitoTracker Green FM (Thermo Fisher Scientific) following the manufacturer’s instructions. The overlap between the transplant and the host mitochondrial network was controlled. For the quantitative analysis, data were analyzed with self-written MATLAB (R2018a) scripts.

Split-kinesin experiments

pKIF5C-HA-FRB was a kind gift from Prof. Sean Munro. Expresses fusion of Rat kinesin minus tail to FRB; pFKBP-mCh-Fis1TM was subcloned from pFKBP-GFP-myc-GRIP (Sean Munro) by swapping the GFP-myc-GRIP fragment with mCherry fused to the TM domain of yeast Fis1 for OMM targeting. U2OS cells stably expressing mtBFP and a shRNA against DRP1 [52] were reverse transfected with the 2 constructs above. Two days later, cells were imaged using a spinning disk microscope. Rapamycin was diluted in growth medium to a final concentration of 1 nM. A microfluidic imaging device was used to allow (rapamycin containing) medium replacement while imaging. Experiments were performed twice on different days.