2.1 Plant Cells Show a Persistent Population of Mitochondria Without mtDNA

It has previously been shown that individual mitochondria in plant cells do not carry the full gene complement. Some do carry the full complement, while some carry a subgenomic molecule(s) with a few genes, and some have no mtDNA at all (Arimura et al. 2004; Gualberto et al. 2014; Kozik et al. 2019; Logan 2006; Preuten et al. 2010; Takanashi et al. 2006). We quantified the proportion of mitochondria with and without mtDNA in whole single cells across two cell types by fixing 5-day-old Arabidopsis thaliana seedlings encoding mitochondrially targeted mCherry (Kindly provided by Prof. Markus Schwarzländer (pBin atp-mCherry, Candat et al. 2014)). These were stained with SYBR green, an intercalating dye specific for double-stranded DNA (dsDNA), with a low affinity for RNA (Zipper et al. 2004; Arya et al. 2005; Dragan et al. 2012). Cells (8 hypocotyl cells and 11 leaf epidermal cells) were imaged as z-stacks using a confocal laser scanning microscope and segmented using IMARIS (see methods). The SYBR stain gives a signal for all dsDNA in the cell, including the nucleus and chloroplast, alongside the mtDNA signal of interest (Figure 1Ai,ii). Some organelles contain no mtDNA, while others have a range of weaker to stronger SYBR signals, all able to be detected by the segmentation thresholds implemented (Figure 1B). Only the SYBR signal that sits directly within mitochondrial volumes was quantified, avoiding other organellar DNA signals. This method revealed a persistent but varied proportion of mitochondria that do not carry genomic material (Figure 1C, see methods), confirming previous results (Arimura et al. 2004; Zhang et al. 2023).

The proportion and absolute number of mitochondria not carrying genomic material vary across and within cell types. As controls for the segmentation process, a Col-0 (wild-type, with no stable fluorescent marker) hypocotyl cell treated with SYBR and a mito-mCherry hypocotyl cell with no SYBR stain were also segmented, giving similar intensity and mitochondrial volumes, respectively (Figure S1). We also see a correlation between cell area and mitochondrial number, explaining the variation in total mitochondrial number across cells (Figure S2). This staining and segmentation offer a cell-by-cell quantification of mtDNA presence/absence, supporting previous observations of mitochondria without genomic material.

2.2 Defining mtDNA Presence or Absence in Moving Mitochondria

To investigate the hypothesized link between the physical and genetic dynamics of plant mitochondria, we characterized the movement of individual organelles with and without mtDNA. Laser scanning single-plane confocal time lapses of 5-day-old mito-mCherry hypocotyl cells stained with the dsDNA marker SYBR green were captured over ~4 min (Figure 2A, Video S1). Hypocotyl cells were chosen as they have been demonstrated to provide a quasi-2D field of view of the mitochondrial population, and the cellular geography allows clear tracking of individual organelles (Chustecki et al. 2021). Therefore, although mitochondria do move in 3D space, the cytoplasm is restricted to the edge of this cuboidal cell and can be up to only 100 nm thick at points in Arabidopsis epidermal cells (Erickson et al. 2024). This allows mitochondria to be imaged and tracked in this quasi-2D space above the vacuole, only disappearing around the edge of the cell when they leave the field of view, as modelled in (Chustecki et al. 2021) and discussed in Giannakis et al. (2022).

Tracking was performed using Trackmate (Tinevez et al. 2017), with manual verification. Tracking was performed only upon the mito-mCherry channel of the captured videos; therefore, the SYBR green channel did not influence the movement tracking. Following tracking, the SYBR green channel was used to define mtDNA presence or absence. Trackmate quantification results can be viewed as a trajectory map over time (Figure 2B).

Two approaches were tested to quantify the SYBR signal in the GFP channel. Firstly, Trackmate quantifies the signal in each dot encircling the tracked mitochondrion. From this, the Contrast statistic was used to define the presence or absence of mtDNA (Tinevez et al. 2017). Contrast is calculated as the Intensity (A.U) of pixels inside the spot radius (r) compared to outside as defined by 2r by (I_in − I_out)/(I_in + I_out) (based on Michelson 1995). This is a good indicator of the SYBR signal for punctate mitochondria (Figure S3A). However, due to the intensity considered for I_out, two distinct SYBR signals located within 2r of each other may interfere with the contrast calculation, leading to only one being accounted for. Therefore, we also use a second definition of absence/presence, inspired by the flexible parameter Prominence within mtFociCounter (Rey et al. 2023), implementing the Find Maxima function in ImageJ (Wagner 2016), which scans each pixel intensity value, and in analogy to geographic topography, finds the maxima in relation to those surrounding it. Find Maxima distinguishes two proximal SYBR signals more effectively than the previously defined Contrast value (Figures 2C and S3A) and was carried forward as the detection method.

The next challenge was describing the presence/absence of mtDNA across time points of individual trajectories. The complexity and speed of mitochondrial dynamics in live plant cells pose a challenge for all tracking software. Correlating the SYBR signal at each time point for each mitochondrion, it was observed that organelles with no SYBR signal may occasionally have a single frame that crosses the threshold for the signal. This can be seen in trajectory schemes labelled with SYBR channel features—some trajectories have a consistent SYBR signal throughout time, and others have occasional frames with a higher/lower signal (Figure S3C,D). This could be due to an unspecific SYBR signal or crossing paths with a mitochondrion with a strong SYBR signal. It could also be that those trajectories with SYBR signal may have a few frames where the signal is below the threshold (either defined by Contrast or Maxima). It is, therefore, necessary to carefully define mtDNA presence/absence to classify this inherently noisy data.

We, therefore, implemented and evaluated three definitions for mtDNA ‘presence’ for trajectories—based on how many frames a mitochondrial trajectory needs to have an SYBR signal above a threshold for it to be categorised as having mtDNA. The first definition, which we will refer to as lenient, is that only one frame in the entire trajectory needs to be above a threshold value to define that mitochondrion as having mtDNA. The next is tolerant, where at least two adjacent frames must be above a threshold. The last definition is strict, where at least three adjacent frames must be above a threshold. This last definition necessarily excludes all trajectories that are only two frames long. Also, it should be noted these are definitions of mtDNA presence, aiming to find a balance between accurately defining presence while not allowing a leaky absent definition to be populated by trajectories with signal.

When quantifying the number of trajectories across each definition, the most lenient definition appears to allow too many false positives, while the strictest results in data loss (Figure 2D). Therefore, we settled upon the tolerant definition moving forward; however, we corroborated our findings across the definitions as they were examined (see Supporting Information). Paired with the Find Maxima SYBR detection method, false detection was limited while allowing flexibility to work around any mis-tracking through the software. This allowed us to confidently define and detect mtDNA per trajectory for the subsequent physical statistical analysis.

2.3 Mitochondria Without mtDNA Demonstrate Altered Physical Dynamics

Using positional data of mitochondrial movement and the previously described tolerant definition of mtDNA presence/absence, physical statistics of individual mitochondria with and without mtDNA were compared (see methods). First, mean speed was calculated per trajectory, and no difference was observed between mitochondria with and without mtDNA across 10 hypocotyl cells (Figure 3A). When comparing the mean area travelled per trajectory, a significantly smaller area was travelled by mitochondria without mtDNA (Figure 3B), despite no overall difference in speed here. Area travelled is necessarily dependent on speed, direction and, in this case, trajectory length. We see a difference in trajectory length between those with and without mtDNA, where mitochondria without mtDNA have shorter trajectory duration (Figure S4A). To investigate this further, we implemented the most extreme possible definition of mtDNA presence and absence: every single time point in a trajectory must pass the Maxima threshold to count as that mitochondrion having mtDNA, and every single time point in those without mtDNA must be below the Maxima threshold. This causes any trajectories that include both time points with and without mtDNA to be discarded. In this case, we see similar trajectory lengths but still a smaller area travelled by mitochondria without mtDNA, confirming that this relationship holds and rules out artefacts from presence/absence definitions (Figure S4B,C). We also see this across all definitions (Figure S8).

Returning to our tolerant definition of mtDNA presence/absence, when comparing intermitochondrial positioning, we see differences between mitochondria with and without mtDNA. Intermitochondrial distance is the minimum Euclidean distance between each mitochondrion in each trajectory, averaged per frame. Mitochondria without mtDNA show an increased intermitochondrial distance, meaning they are, on average, further away from every other mitochondrion in the system than those with mtDNA (Figure 3C). Following this, to test the propensity for sharing between mitochondria, colocalisation time was calculated as the number of frames mitochondria had been within 1.6 μm distance of each other (as in Chustecki et al. 2021). Mitochondria without mtDNA spend less time colocalising with any other mitochondria, compared to mitochondria with mtDNA (Figure 3D). Examination of the same physical statistics across the two foci definitions (Contrast/Maxima) and the three trajectory definitions (Tolerant/Lenient/Strict) reveals consistent conclusions to the Maxima and tolerant used to define mtDNA presence here (Figure S8). Overall, a difference in movement between those with mtDNA and those without mtDNA is observed across whole populations of mitochondria in single cells.

2.4 Mitochondria Without mtDNA Are Less Well Connected to the Rest of the Population

To study mitochondrial connectivity across whole populations of individual organelles, a novel method borrowing from graph theory, analogous to social networks, has been developed (Chustecki et al. 2021). Defining individual mitochondria as nodes (here represented as circles), and close connectivity (within 1.6 μm distance) as a connecting edge (represented as sticks), we can successfully build historical networks describing and quantifying the encounters of individual organelles over time-lapse videos of motion in single cells. Onto this, we project for the first time, functional data—the presence or absence of mtDNA. We define each trajectory as with/without mtDNA (using the Maxima, and tolerant definitions) and build an interaction network over time from all mitochondria and all connections (Figure 4A).

Graph theory and social network analyses use a parameter called ‘degree’ to describe the connectivity between individuals in a network. A low degree means that an individual encountered a very small number of other individuals in the network during the time examined. For mitochondria, this would mean they travel around the cell without coming into close proximity with many other mitochondria. In contrast, a high degree means that a mitochondrion was in close proximity to many others over time. In order to compare connectivity between mitochondria with and without mtDNA, we focused on this node-specific value of degree, which is a measure of how many other mitochondria a single mitochondrion met over the time we tracked it within the cell, or in network graphical terms, how many sticks are connected to the circles. Strikingly, the node values for mitochondria with DNA reach much higher local connectivity levels than those without mtDNA (Figure 4). This means that mitochondria with mtDNA have more immediate neighbours (of either type) than those without.

By the final frame of all videos, we see mitochondria without mtDNA only reaching a maximum degree value of 15, while those with mtDNA reach a maximum degree value of 34 (Figure 4Biii). Examination of the degree parameter across the two foci definitions (Contrast/Maxima) and three trajectory definitions (Tolerant/Lenient/Strict) reveals the same significant relationship of higher degree values for those with mtDNA, regardless of the definition used (Figure S9). The combination of these 12 different definitions and analysis strongly suggests that our choices of how to define a mitochondrial image and how to decide whether it has mtDNA in it or not do not change the interpretation, and the results are robust to possible artefacts of the definitions used during image analysis.

Examination of the network illustrations (Figure 4Ai–iii) suggests that those without mtDNA are enriched in those with degree 0 (i.e., no close encounters with other mitochondria) and by those with degree 1 (i.e., at the periphery of the network). This is confirmed by quantifying the number of nodes with a certain degree value in proportion to the number of nodes of each status at representative time points across all cells analysed (Figure 5). For those nodes of degree 0 or 1, we see a higher proportion of nodes without mtDNA than those with (Figure 5A,B). When compared to those of degree 3, that is, with three immediate neighbours (Figure 5D), which are well represented in both those with and without mtDNA (Figure 4B), we see no particular mtDNA status as prominent over frames or cells, in contrast to the representation of those without mtDNA in the lowest degree nodes of 0, 1 or 2 (Figure 5A,B). At higher degree values, above 3 (Figure 5E–I), we see a higher proportion of nodes with mtDNA compared with every other node of the same mtDNA status. Above degree 8, comparison between nodes with and without mtDNA becomes difficult due to a lack of nodes without mtDNA and with high degree (Figure 4B). When this analysis is extended to all frames of all cells, the same pattern is observed (Figure S7). Thus, mitochondria without mtDNA do not have as many immediate neighbours as those with mtDNA and are more likely to be nodes with only zero or one immediate neighbour(s), while those with higher degree values are more likely to be mitochondria with mtDNA.

2.5 Mt-HI-NESS Nucleoid Binding Provides Improved Visualisation of mtDNA Nucleoids in Plants

Until now, mtDNA imaging in plants has relied on exogenous stains that face the impermeability barrier of the plant cell wall or fluorescent proteins that are target-specific and do not bind mtDNA ubiquitously. Live rates of mtDNA exchange across mitochondrial populations in plants are also yet to be established. Therefore, a photoconvertible ubiquitous mtDNA marker was sought. We modified the mt-HI-NESS (Mitochondrial HNS-based Indicator for Nucleic Acid Stainings) construct developed in human cells for stable transformation and ubiquitous mitochondrial targeting in plants (Deng et al. 2023). This construct relies on the H-NS bacterial nucleoid binding protein attached via a linker to the photoconvertible Kaede protein. We codon-optimised for Arabidopsis and chose two plant-specific mitochondrial targeting sequences to test, IVD and AOX (see methods). As a further test, we also used the original human Cox8 presequence construct to see whether it would function in a plant system.

Initial tests were carried out using the Nicotiana benthamiana/Agrobacterium tumefaciens transient transfection system (see methods), with IVD and AOX mt-HI-NESS specifically localising within the individual mitochondria of epidermal cells (Figure S5D,E), using a red fluorescent IVD-FP611 construct as a mitochondrial marker. Compared to the same tissue using IVD-FP611 and stained with SYBR green, we see a similar nucleoid signal, showing that mt-HI-NESS is ubiquitously staining mtDNA nucleoids (Figure S5B,D,E). Using Cox8-mt-HI-NESS, we observed non-specific chloroplast and nuclear DNA targeting (Figure S5C), presumably because the human Cox8 mitochondrial targeting sequence does not function in plants.

Following successful testing in N. benthamiana, the AOX/IVD-mt-HI-NESS constructs were moved into Arabidopsis plants. Floral dip of mito-mCherry and Col-0 plants was performed (see methods), and positive transformants were isolated. Each combination demonstrated mitochondrial nucleoid-specific binding, and here we show mito-mCherry with AOX-mt-HI-NESS (Figure 6A) and Col-0 with IVD-mt-HI-NESS (Figure 6C). This binding was identical to the same tissue stained with SYBR green (Figure 6B), although the exogenously applied SYBR is also visible around the outside of the cell. We did not find any differences in localisation or expression between the two mitochondrial targeting sequences, IVD and AOX mt-HINESS (Videos S2 and S3). To demonstrate the accuracy and ubiquitous mtDNA staining provided by the mt-HINESS marker, we used a different nucleic acid marker, Syto Blue 40 in mito-mCherry mt-HINESS. We used this new stain because Kaede and SYBR Green have the same emission wavelengths. Unfortunately, Syto Blue 40 fails to stain hypocotyl cells, so we used root tissue for this assay. Syto Blue 40 and unconverted mt-HINESS Kaede colocalise within the mCherry marked mitochondria (Figure S14). It is not known whether spot brightness or size is related to the amount of mtDNA.

We tested the photoconvertible properties of the Kaede protein under UV laser (405 nm) and demonstrated complete green-to-red switching of each mitochondrial nucleoid in a specified region of interest (Video S4).

A series of physiological experiments were conducted to investigate if the IVD/AOX-mt-HI-NESS lines impact mitochondrial function through their nucleoid binding. The rosette areas of transformed Arabidopsis did not differ from WT (Figure S6A,B). We initially saw a difference in root length, a key phenotype for mitochondrial function, when testing Col-0 against mito-mCherry AOX or IVD-mt-HI-NESS (Figure S6Ci,ii). However, subsequent testing of the mito-mCherry line (into which these constructs had been transformed) against Col-0 revealed the same shorter root phenotype (Figure S6Ciii,E). Therefore, the difference in root length is due to the mito-mCherry background and not the mt-HI-NESS nucleoid binding protein. This was confirmed by the root lengths in Col-0 background mt-HI-NESS lines and Col-0 showing no significant difference (Figure S6Civ,F). All statistical comparisons in the physical and connectivity analyses were conducted for samples using the mito-mCherry fluorescent marker. We therefore conclude any alterations to dynamics and connectivity within the context of the mito-mCherry background. Next, oxygen consumption assays as an indicator of mitochondrial respiration were performed on leaf discs in the dark on the three genotypes, and no difference in oxygen consumption rate was observed between them (Figure S6D).

2.6 Mitochondria Without mtDNA Demonstrate Altered Motility and Connectivity in a Stable Nucleoid Marker Line

Upon validating this new stable fluorescent mitochondrial nucleoid marker line, we performed live imaging, tracking and differential motility analysis on 11 mito-mCherry AOX-mt-HI-NESS hypocotyl cells (Figure 7). This system offers a precise and ubiquitous marking of mtDNA in plant cells, and together with the mito-mCherry marker, allows visualisation of mitochondria with and without mtDNA in a cell with very little background signal, as was often seen in the exogenous SYBR stained cells (Figures 2A and 7A). The clear signal in the stable genetic line is amenable to precise thresholding using the Find Maxima function used to define mtDNA puncta in the time-lapse images (Figure 7Aiii). The same tracking methodology and mtDNA presence analysis were performed as with the SYBR green lines. Again, only the mito-mCherry (red) channel was used for mitochondrial tracking, and the mtDNA presence definitions were applied after tracking was performed.

Upon analysis of the physical dynamics of mitochondrial motion between mitochondria with and without mtDNA in this new system, the significant relationships remain as seen in the SYBR system. Mitochondria without mtDNA travel around a significantly smaller area in the cell on average than those with mtDNA (Figure 7C). Those without mtDNA also demonstrate an increased intermitochondrial distance between each individual without mtDNA and the rest of the mitochondrial population (Figure 7D). This, in turn, is matched by a decreased colocalisation time for mitochondria without mtDNA, in that each mitochondrion without mtDNA does not spend as many frames associated with other mitochondria as those with mtDNA do (Figure 7E). For each of these conclusions, we tested the relationship across the two definitions for mtDNA foci, Contrast and Maxima and the three definitions for trajectories (tolerant/lenient/strict) with and without mtDNA (Figure S10). These relationships are consistent across all definitions tested, except for tolerant and strict definitions for the contrast-detected foci in the intermitochondrial mean data, although the variation seen in those mitochondria without mtDNA remains consistent (Figure S10Biii,Ciii) and matches that seen in the SYBR system (Figure S8Biii,Ciii).

Interestingly, we see that mitochondria without mtDNA have a reduced speed in this mt-HI-NESS system (Figure 7B), and this conclusion is consistent across the definitions of foci and trajectories (Figure S10A–Fi). This difference was not observed in most comparisons across definitions in the SYBR system, although some did indeed pass the threshold of significance (0.001) (Figure 3A, S9A–Fi). We do not see the opposite conclusion anywhere of mitochondria without mtDNA having an increased speed. The slower speeds in mitochondria without mtDNA may also be reflected in the smaller area travelled by mitochondria without mtDNA (Figure 7C). Using the mt-HI-NESS system, we see significantly slower mitochondrial motion in those without mtDNA.

Upon network analysis of local connectivity between organelles in the population, mitochondria marked as having nucleoids by mt-HI-NESS again have much higher connectivity as measured by degree values than those without mtDNA (Figure 7F,G). We also see that mitochondria without mtDNA are more likely in nodes with lower degree values (degree 0, 1, 2) than higher degree values (3–8) (Figure S12). This confirms the relationship seen in the SYBR-treated cells and is seen consistently across all definitions of foci and trajectories (Figure S11). Furthermore, preliminary results using TMRM as a membrane potential marker show that mitochondria with and without mtDNA maintain a membrane potential (Figure S13).

Overall, the new mt-HI-NESS system provides a more specific marker of mtDNA with less background, enabling a clear-cut analysis of the presence or absence of mitochondria nucleoids per organelle. Upon analysis of physical and social dynamics in the SYBR and mt-HI-NESS systems, the same conclusions of differential motility and connectivity can be drawn, with the addition of a significant difference in speed.

4.1 Plant Growth

Seeds of Col-0, mito-mCherry, mtGFP, mito-mCherry mt-HI-NESS and Col0 mt-HI-NESS lines were surface sterilised using 50% (v/v) household bleach solution with continual inversion before washing three times with sterile ddH₂O, and plating onto ½ MS agar with 5 μg/mL Nystatin at pH 5.7 before stratification at 4°C for 2 days in the dark. Seedlings were either grown to 4–5 days old in ½ MS before use or transplanted to Berger BM2 germinating soil with Turface MVP. Any plants grown on soil were grown at 22°C in 8/16 h dark/light.

N. benthiamiana plants were grown in greenhouse conditions with supplemented light 12 h a day and temperature ranges at 27°C–24°C daytime and 21°C–19°C night time from seed until ~3–4 weeks old.

4.2 Plant Tissue Fixation Staining and Imaging

5-day old Arabidopsis seedlings were fixed in 3.5% paraformaldehyde (EMS, #15710) in PBS buffer (Gibco, #10010023) for 30 min, ensuring complete submersion by applying gentle vacuum pressure to the solution. Seedlings were washed with PBS before further staining. Images of fixed samples were taken as z-stacks. Due to the highly motile nature of plant mitochondria, every other image and time-lapse videos were taken as planar (2D) images, not as z-stacks.

SYBR green (invitrogen, #S7563) was used at a 1:1000 dilution in ddH₂O, and 5-days-old Arabidopsis seedlings were stained for 45 min then washed in PBS or H₂O before use.

All laser scanning confocal micrographs, with the exception of the 14 mito-mCherry AOX-mtHINESS time lapse videos, were taken on an upright Nikon A1-NiE system (RRID:SCR_020318) using NIS-Elements Confocal image acquisition and analysis software, using a 60X water objective. Images were taken using a sequential channel series setting to minimise cross-channel signal, and the channels used were: GFP, Ex: 488 nm, Em: 525 (500–550), TxRed, Ex: 561, Em: 595 (570–620), Cy5, Ex: 640 nm, Em: 700 (663–738).

All time-series images taken on the Nikon system have a frame time of 2.21 s.

For imaging of live hypocotyl cells, and fixed hypocotyl and leaf epidermal cells, 5-day-old seedlings grown on plates were used. For mt-HI-NESS leaf epidermal images, ~2/3-week-old plants were used. Tissues were mounted in water on glass slides, with double-sided sticky tape and a glass cover slip.

For photoconversion of the Kaede fluorescent protein of the AOX/IVD-mt-HI-NESS lines, the A1 and ND Stimulation features of the NIS-Elements were used to target 4% of the total field of view for 5 s at 10% of 20 mW LD 405 nm laser power. Frame intervals were 2.15 s and the total time imaged was 95 s.

The 14 mito-mCherry AOX-mtHINESS time-lapse videos analysed were of single hypocotyl cells of 5-day seedlings, taken on an inverted Zeiss LSM 980 with ZEN blue image acquisition and analysis software, using a 63X oil immersion lens. Time lapses were taken in two tracks, GFP, Ex: 488 nm, Em: 509 nm (490-561 nm), mCherry, also capturing chloroplast autofluorescence, Ex: 587, Em: 610 nm (596-685 nm). This collection of time series has a frame time of 2.52 s.

TMRM staining of 7-day-old seedlings involved a 20-min submersion in 20 nM TMRM (Fisher Scientific Cat No. T668) in ½ MS solution, then washing with ddH₂O.

Syto Blue staining of 5-day-old mito-mCherry AOX/IVD-mtHINESS seedlings used 2.5 μM Syto 40 Blue (Invitrogen Cat no. S11351) for 20 min, and root cells were imaged with the upright Nikon A1-NiE system equipped with a 60X water-immersion lens (NA 1.2). Images of root cells were captured sequentially through 561, 488 then 405 nm.

4.3 Cell Segmentation With IMARIS

Z-stack files of fixed cells were imported into IMARIS (version 10.1.0). The cell feature was used to define thresholds for segmenting vesicles (mtDNA puncta) and nuclei (mitochondria). The cell outline segmentation was not implemented, as a reliable cell wall/membrane stain could not be implemented. To work around this, manual cropping of cell outlines based on SYBR stain at the cell wall was performed in ImageJ2 (2.14.0/1.54f) before import. Further cropping was done using cropping planes in IMARIS, until a clear view of a single cell in each case was achieved. In most cases, the same thresholding parameters were used, briefly: Nucleus Smooth Filter Width = 0.2 μm, Nucleus Background Subtraction Width = 0.25–0.4 μm, Nucleus Manual Threshold Range: 77.3–261.3, Vesicle Estimated Diameter = 0.600 μm, Vesicle ‘Quality’ above range: 63.5–207. The hypocotyl hyp1 sample also had an extra filter of ‘Nucleus Number of Voxels’ above 2.05.

During analysis of exported data, any segmented ‘mitochondrion’ < 0.1 μm³ was removed. SYBR signals from the nucleus and chloroplasts were avoided by quantifying only mtDNA puncta within segmented mitochondria.

4.4 Video Analysis

Video analysis closely follows (Chustecki et al. 2021). Single-plane time-lapse videos of SYBR-stained mito-mCherry hypocotyl cells of 5-day-old Arabidopsis thaliana were cropped to single cells using an extracellular SYBR signal. Eight videos were taken, from which 10 unique cells were analysed (n = 10). Ensuring resolution (4.83 pixels/μm) and time steps (2.21 s) were consistent across the videos analysed, Trackmate v7.11.1 (Tinevez et al. 2017) in ImageJ v2.14.0 was used to track movement in the mito-mCherry channel only. Therefore, SYBR or mt-HINESS signals do not influence the movement tracking itself. Typical settings for SYBR-stained cells were using the LoG Detector filter with a blob diameter of 1.1 μm, a threshold of 76–100, and a quality filter on spots, if necessary, of between 87 and 180. Then, the Simple LAP Tracker was applied with a linking max distance of 2.5, a gap-closing distance of 3.5 and 2 frames used for gap-closing.

The same process was undertaken for live mito-mCherry AOX-mtHINESS cells, for 11 cells from 5 different seedlings (n = 11). Again, resolution (3.8 pixels/μm) and time steps (2.52 s) were consistent across the videos analysed. Tracking used slightly different parameters of blob diameter of 0.8 μm, a threshold of 9.8–29.8, and a quality filter on spots, if necessary, of between 16.5 and 28.8, and the LAP Tracker linking max distance of 2.5, a gap-closing distance of 3.5, and 1 frame used for gap-closing. For both video collections, manual editing of tracks was undertaken if deemed necessary.

Full .xml files were exported and imported into RStudio (v2024.04.0+735) using the TrackmateR package (v0.3.10) (Royle 2024), to include all spot statistics. To define trajectories as having or not having mtDNA, we use the second channel of the SYBR/mt-HI-NESS signal and two definitions –Contrast and Maxima. Contrast is calculated as the Intensity (A.U) of pixels inside the spot radius (r) compared to outside as defined by 2r by (I_in − I_out)/(I_in + I_out), and for each cell, the contrast cutoff for what counted as a candidate SYBR/mt-HI-NESS signal was manually selected using the visualisation tool and spot pseudo-colouring in Trackmate. Cutoffs ranged from 0.05–0.22 contrast values, where anything below was without mtDNA, and anything above was with mtDNA.

For the Maxima definition of mtDNA presence, the FindMaxima (Wagner 2016) function in ImageJ was run through a macro across three frames (1,50,100) of each video and looped over prominence values in increments of 50 from 150 to 500, with the ‘strict’ setting applied. Final prominence values used across the videos ranged from 200 to 400 for SYBR videos and 10 or 11 for mt-HI-NESS videos. Images with maximum points as crosses were viewed to choose the representative best prominence value for the true SYBR/mt-HI-NESS signal, and this chosen prominence value was used for each video to generate a list of pixel coordinates from puncta that were SYBR/mt-HI-NESS signal. After conversion to microns to match Trackmate output, the R point.in.polygon function from sp package v2.1-4 was used to assess whether or not the maxima puncta sat inside a Trackmate spot from the mito-mCherry signal. If it did, the mitochondrion at that frame has mtDNA according to our Maxima definition, and if not, it was without mtDNA.

For both Maxima and Contrast definitions, three strictness definitions were then defined. Each video had its own contrast and prominence value. We define the first as Lenient in that only one frame in the entire trajectory needs to be above the threshold value to pass (as this mitochondrion has mtDNA). The next is tolerant, where at least two adjacent frames must be above the threshold to pass. The last definition is strict, where at least three adjacent frames must be above the threshold to pass.

Physical statistical analysis includes speed (μm s⁻¹), taken as the distance moved per frame per trajectory. Area (μm³) is taken as the region inside a polygon plotted using the furthest points of a trajectory of frame length ≥ 3 (convex hull area). The minimum intermitochondrial distance (μm) is the minimum Euclidean distance between each mitochondrion and any other mitochondrion in the cell per frame. Any log values represent values at log base e.

We characterise encounter networks as undirected edges between nodes representing mitochondria. Edges are built when mitochondria are within 1.6 μm of each other and do not represent fusion events. Encounter networks build as frame times progress, and network build-up is historical. Node degree is then calculated as the number of immediate neighbours to which each node is connected. This analysis aimed to investigate connectivity values for nodes with and without a characteristic; therefore, we did not implement global values of connectivity for entire networks, as we analysed connectivity between individual nodes instead of between whole networks and avoided splitting the network into subgraphs. R packages iGraph (v2.0.3) and brainGraph (v3.1.0) were used to calculate network statistics and visualise networks.

4.5 Correlation Analysis

Spearman correlation was calculated using the cor.test function of the stats (version 3.6.2) R package, using Spearman's rho statistic for a rank-based association between values, and which uses p values computed using algorithm AS 89, set out in (Best and Roberts 1975).

4.6 mtHINESS Plasmid Construction

The mt-Kaede-HI-NESS construct was kindly provided by Professor Timothy Shutt, University of Calgary, and developed for use in U2OS human cells, using Cox8 as the MTS (mitochondrial targeting sequence). This construct was transformed into TakaraBio Stellar E. coli cells (TakaraBio, #636763) and grown on LBA plates with Ampicillin (100 μg mL⁻¹).

Plasmid DNA was extracted using a column-free plasmid prep method (Green and Sambrook 2012), and mt-Kaede-HI-NESS-HA was amplified from the plasmid (linearised using AvrII restriction enzyme (NEB, # R0174S) using RP (5′-CAAGAAAGCTGGGTTTAAGCG-3′) and FP2 primers (5′-CACCGCCACCATGTCCGT-3′)) resulting in only one Cox8 sequence being carried through, and a CACC sequence for directional TOPO cloning being introduced. This 1003 bp PCR product was extracted from a 1% agarose gel using the TakaraBio gel extraction kit (TakaraBio, #740609.50) and introduced to the pENTR TOPO cloning vector (Kan^R) (Invitrogen, #K240020), which contains the attL1 and attL2 sites, LR clonase reactions and transformed into OneShot Top10 competent E. coli (Invitrogen, #C404003).

In parallel, the Cox8 MTS was switched for the plant-specific AOX or IVD MTS, and AOX/IVD-mt-Kaede-HINESS-HA constructs were synthesised by ThermoFisher GeneArt (https://www.thermofisher.com/us/en/home/life-science/cloning/gene-synthesis/geneart-gene-synthesis.html), following codon optimisation for Arabidopsis using the VectorBuilder Optimisation tool (https://en.vectorbuilder.com/tool/codon-optimization.html), and with attL1 and attL2 sites introduced for the LR clonase reaction into Agrobacterium tumefaciens. These were shipped in pMA-RQ plasmids (Amp^R) and transformed into TakaraBio Stellar E. coli cells before plasmid DNA extraction.

The three plasmid types, Cox8/AOX/IVD-mt-Kaede-HINESS-HA, were inserted into separate pUB-DEST vectors via the LR clonase reaction (Spec^R) (Invitrogen, #11791020), transformed into Top10 competent E. coli, sequenced and verified. This puts the inserted construct under the control of the UBQ10 promoter. After one final plasmid prep, they were all transformed into A. tumefaciens strain C58C1 for use in infiltration and floral dipping. At each step, plasmids were sequenced using Eurofins Genomics whole plasmid sequencing service.

4.7 Transient Infiltration of N. benthiamiana

C58C1 A. tumefaciens lines were grown in YEBS + RGS for 2 days alongside a mitochondrial-targeted red fluorescent protein IVD-FP611 (in pZP212). Following the protocol previously described (Mangano et al. 2014), ~4-week-old N. benthiamiana leaves were syringe infiltrated on the abaxial surface with bacterial preparations (OD = 0.35 of each line). Plants were left to recover in a growth chamber for 48 h before imaging.

4.8 Stable Transformation of A. thaliana

Approximately, 4-week-old Col-0 and mito-mCherry plants were transformed via floral dip with A. tumefaciens carrying either AOX or IVD-mt-Kaede-HINESS-HA constructs. A previously published protocol was followed (Zhang et al. 2006), using a 5% (wt/vol) sucrose solution at pH 5.6 to resuspend cultured bacteria, and a second dip was performed 7 days after the first. Seeds were left to develop before screening with BASTA on soil.

4.9 BASTA Treatment for Transformed Arabidopsis Screening

Glufosinate Ammonium (BASTA, plantmedia.com) was diluted to 300 μM in a spray bottle with ddH₂O and sprayed upon seeds before stratification, then also to newly emerged seedlings and 1–2-weekweek-old plants, removed from growth chambers. Col-0 (susceptible) and pUBdest transformed (resistant) plants were grown alongside to monitor the effectiveness of treatment. Once sprayed, trays were left with plastic domed lids on for 8 h before returning to the growth chamber.

4.10 Root and Rosette Assays

Seeds of each genotype were bleach sterilized and sown onto square ½ MS plates after stratification at 4°C for 2 days, placed at a 70° angle in a growth cabinet and grown for 9 days without disturbance. Images were taken using the brightfield setting on a GelDoc Imager. For Col-0 background root comparison, the same method was used, but seedlings were grown on 1/2 MS with sucrose and imaged using a Nikon camera. For all, root lengths were hand traced from the cotyledon base to the root tip with the flexible line tool in ImageJ, after setting the scale based on the plate size, lengths were exported and further analysed in R.

For rosettes, seeds were grown on upright plates for 11 days before sowing on soil and growing for another 19 days. Rosette images were traced using freehand draw in ImageJ and the area was taken after setting the scale for pot size, exported and analysed in R.

4.11 Oxygen Consumption Measurements

Briefly, ~60 (±1) mg of leaf discs (7 mm width) were cut from the middle tip region of 25-day-old Arabidopsis leaves and placed in respiratory activity incubation medium (10 mM HEPES, 10 mM MES, 2 mM CaCl₂, pH 7.2) for 30 min in the dark. Oxygen consumption measurements were taken using the Oxygraph+ oxygen electrode system (Hansatech Instruments), following (Sew et al. 2013). Calibration by removal of all oxygen in 1 mL ddH₂O was performed using an excess of Sodium dithionate. Oxygen concentration was read over 15mins in the dark in 2 mL of respiratory activity incubation medium, and oxygen consumption rate was calculated as nmol O₂ per minute per gram of fresh weight (FW). Jithesh Vijayan kindly provided dry ice as needed.