Specialized high-capacity mitochondria fuel cell invasion.

Caenorhabditis elegans ( C. elegans ) were maintained at 20°C on nematode growth media (NGM) plates and fed Escherichia coli strain OP50. In text and figures, wild-type refers to N2 Bristol strain C. elegans, and standard C. elegans nomenclature is used to convey genotypes of strains. Promoter-driven transgenes are denoted by the italicized gene name followed by “p”, “::” signifies an adjoining protein, and italicized mutant alleles are enclosed within parentheses following the gene name. As previously described, anchor cell (AC) invasion was scored in L3 hermaphrodites and staged in reference to the number of vulval precursor cells (VPCs) ( Figure 1 ) 32 . Synchronization of L1 animals was performed using standard hypochlorite treatment for all RNAi and rotenone experiments 113 . All endogenously tagged strains in this study were generated by injecting into the gonads of wild-type (N2) C. elegans . Endogenous strains that were homozygous infertile were either maintained by picking animals with the fluorescent signal of interest or crossed with a balancer strain (detailed in Key Resources Table ). Measurements of fluorescence intensity in all strains were done in homozygous animals (See Table S3 for details on the health of endogenous tagged strains). All animals were well-fed (3+ generations without starvation) prior to any experimentation or imaging to control for nutrient-dependent phenotypes. Genotypes of all strains were verified by sequencing and/or plate-level phenotype. All strains used in this study are listed in the key resources table . Human orthology to C. elegans proteins annotated throughout the manuscript was identified using Alliance of Genome Resources 114 . In cases where multiple human orthologs were present, the ortholog with the highest sequence similarity to the C. elegans protein was selected. CRISPR-Cas9 mediated genome editing with a self-excising hygromycin selection cassette (SEC) was used to generate endogenously tagged strains 115 . In brief, the optimal location to insert the fluorescent protein was identified based on the tertiary and quaternary structure of the protein and any protein cleavage sites were accounted for (such as mitochondrial localization sequences that get cleaved). To generate the SEC repair plasmids, N2 gDNA was used as a template to amplify ~2KB homology arms upstream and downstream of the PAM site. To ensure the Cas9 didn’t subsequently cut the PAM site after the initial insertion cut, the mNeonGreen (mNG) was inserted in between the guide sequence and PAM site or silent mutations to the guide sequence were made in the homology arms. The homology arms were inserted into the vector mNG-C1ˆSECˆbothlink SEC repair template 116 and correct assembly was confirmed by colony PCR and sequencing. PureLink ™ HiPure Plasmid Miniprep (Invitrogen #K210002) was used to isolate high purity sgRNA plasmid DNA and SEC repair plasmid DNA for injection. https://crispor.gi.ucsc.edu/ was used to identify short guide sequences where the fluorescent protein could be inserted and PAM sites where the Cas9 protein cleaves. The short guide RNA (sgRNA) plasmid was generated by cutting the Cas9 guide plasmid, pDD122 (Addgene, #47550), with Nhel and EcoRV, and HiFI assembly (New England Biolabs, #E2621L) was used to insert the respective sgRNA sequences into the plasmid. For each strain the germline of 10–30 young adult N2 hermaphrodites were injected with a mixture containing 100ng/μL of the SEC repair plasmid DNA, 50ng/μL of the sgRNA plasmid DNA, and 2.5 ng/ul of co-injection markers (pCFJ90 myo-2p ::mCherry, pCFJ104 myo-3p ::mCherry). After 3–4 days at 20°C, F1 progeny of singled-out injected animals were treated with 500μl of 2mg/ml Hygromycin B (Sigma-Aldrich #H3274). Candidate knockin animals exhibited rolling (due to sqt-1(e1350) within the SEC), survived Hygromycin B treatment, and lacked the red fluorescent co-injection markers. F2 rolling ( sqt-1(e1350) +) animals were singled, confirmed homozygous knockin through roller phenotypes and presence of consistent fluorescence signals. To excise the SEC, we heat shocked about 6 L3/L4 homozygous rollers at 34°C in a water bath for 4 hours. After 3–4 days, adult non-rolling animals were singled to check for homozygous excision by confirming loss of rolling in all progeny. Successful genome editing was verified with fluorescence and PCR genotyping. See Table S5 for list of oligonucleotides used in strain generation and genotyping and See Key Resources Table for all strains. Promoter-driven transgenic strains were generated through the Mos single copy insertion (MosSCI) on Chromosome I or Chromosome II or extrachromosomal array integration. Promoters and fusion proteins were either amplified from N2 gDNA ( nuo-1 promoter, nduv-2 promoter, ucp-4 , tomm-20 ), another plasmid ( eef-1A.1 promoter, lin-29 promoter, GFP, mNeonGreen, mKate2, PercevalHR, HYlight, HYlight-RA ), or a C. elegans codon-optimized gene block ( iATPSnFR1.0 ). The Cas9 guide plasmid, pAP082, which is targeted near the ttTi4348 Mos insertion site, was used to generate transgenes via the Mos single copy insertion (MosSCI) on chromosome I 117 . The Cas9 guide plasmid, pDD122, which is targeted near the ttTi5605 Mos insertion site, was used to generate transgenes via the Mos single copy insertion (MosSCI) on chromosome II 118 , 119 . The SEC repair plasmids for MosSCI chromosome I (pCFJ352) and MosSCI chromosome II (pCFJ150) 119 were cut with restriction enzymes, NheI and NotI, then the respective promoter and fusion protein were cloned into the cut SEC repair plasmid backbone using NEBuilder HiFi DNA Assembly master mix (NEB, #E2621L). pCFJ352 was a gift from Erik Jorgensen (Addgene plasmid # 30539; http://n2t.net/addgene:30539 ; RRID:Addgene_30539). Injection, selection, excision, and genotyping were performed as described above. The following strains were generated by MosSCI: the biosensors (iATPSnFR1.0 and HYlight) and their controls (cytosolicGFP and HYlight-RA) all driven by the eef-1A.1 ubiquitous promoter, ubiquitous mitochondrially-localized mKate2 to visualize all mitochondria independent of endogenous fluorescence intensity (driven by rpl-28 ubiquitous promoter), and AC-specific overexpression of UCP-4::SL2::mKate2 (driven by the AC promoter lin-29 ). See Key Resources Table for all strains. To generate strains using extrachromosomal array integration, unc-119(ed4) hermaphrodites were injected with promoter-driven fusion protein plasmids, 50 ng/ml unc-119+ rescue DNA, 50ng/ml pBsSK, and 25 ng/ml EcoRI cut salmon sperm DNA. Stable extrachromosomal lines were established and then integrated using gamma radiation 40 . To reduce the possibility of background mutations resulting from radiation, integrated lines were backcrossed three times with N2 animals. See Key Resources Table for all strains. RNA interference (RNAi) was performed by feeding animals Escherichia coli strain HT115 containing the L4440 RNAi vector targeting genes of interest or the empty L4440 vector. RNAi clones were sourced from either the Vidal or Ahringer feeding RNAi libraries, then streaked out on LB plates containing tetracycline and ampicillin. A single RNAi colony was inoculated in LB with 100 μL/mL Ampicillin (Sigma-Aldrich #A0166) and grown overnight (12–16 hours) at 37°C in an incubator with a shaker. To induce double-stranded RNA expression, 1 μL/mL isopropyl β-d-1-thiogalactopyranoside (IPGT, Sigma-Aldrich #I6758) was added to RNAi cultures and placed back in the 37°C shaker for 1 h. After initial dsRNA induction, RNAi cultures were seeded on NGM plates containing topically applied 1mM IPTG and 100-mg/ml ampicillin and allowed to dry at room temperature overnight for further induction. All RNAi clones were sequenced to confirm correct target gene. Synchronized L1 animals were plated on RNAi NGM plates, fed for 39 hours for imaging P6.p 2-cell stage or 42 hours to score for invasion at the P6.p late 4-cell stage. For every RNAi experiment, a negative control (empty L4440 RNAi vector) and either knockdown efficiency measurement ( Table S6 ) or positive control ( fos-1a RNAi that causes a penetrant invasion defect) was used to ascertain RNAi activity. Rotenone (EMD Millipore #557368) was reconstituted in DMSO to make a 20mM stock solution and stored at −20°C. Synchronized L1 animals were plated on NGM plates seeded with E. coli OP50 and allowed to grow for 34 hours at 20°. Early L3 larvae were washed off from NGM plates with M9 solution and collected into 1.5 mL canonical tubes. 20mM rotenone (40 μM final working concentration) was added to the collecting tube and placed on a rocker or rotator (to avoid hypoxia) for 2 hours at room temperature. The same volume of DMSO in M9 was used on control animals for the equivalent amount of time. Before imaging, animals were recovered from the incubation onto fresh OP50 NGM plates. All experiments were performed with bleach synchronized L3 C. elegans from wild-type N2 lines and 8 endogenously tagged ETC component lines ( nuo-1::mNG , nduv-2::mNG , sdhb-1::mNG , mev-1::mNG , ucr-2.1::mNG , cox-4::mNG , cox-6A::mNG , cox-10::mNG ). L3 C. elegans were washed from OP50 plates and transferred into each well of a 96-well Seahorse plate. The Seahorse Extracellular Flux Bioanalyzer was used to quantify oxygen consumption rate (OCR) as described in previous studies 65 . The number of individual animals per well were counted by imaging the plate using a Keyence BZ-X710 microscope to normalize the OCR measurements per individual worm. For each of the 9 lines, 20–40 L3 animals per strain were used per well, four technical replicates were done for each strain, and a minimum of three biological replicates were performed. Results were normalized to the OCR per worm of N2 animals and are presented as % N2 OCR. Statistical analysis was performed in GraphPad Prism 10.2.1. A one-way ANOVA followed by Dunnet’s post-hoc was used to determine statistical significance. Assessment of anchor cell (AC) invasion was performed as previously described 32 . In brief, synchronized L3 animals at the VPC P6.p late 4-cell stage were mounted onto 5% agar with 1% sodium azide to anesthetize the worms, which were then imaged on a Ziess upright compound microscope equipped with 488 and 561 nm filters and a Nomarski prism for differential interference contrast (DIC). The assessment of AC invasion was based on two indicators: loss of the DIC phase dense BM line and absence of the fluorescence BM marker underneath the AC. Complete removal of the BM resulting in a gap of the width of anchor cell nucleus is scored as normal invasion. Incomplete or blocked invasion is scored if the size of the BM gap is less than the width of the AC nucleus or if the BM is fully intact, respectively. The sensitized strain referred to in Table S1 is null for matrix metalloproteases (MMPs) zmp-1, 3, 4, 5, 6 . Loss of MMPs delays invasion and more ATP is required by the AC to breach the BM 37 . Candidate genes for the mitochondrial import and cristae screen were compiled by finding C.elegans orthologs of mammalian genes annotated as having a role in mitochondrial protein import and cristae formation 58 . All ortholog genes were tested if RNAi clones were available and sequenced correctly. RNAi plates were made according to the RNAi protocol described above. Screening was conducted with the strain (NK2657) with co-labeled endogenous NUO-1::mNG (mNeonGreen) and AC-specific marker ( cdh-3p :moeABD::mCherry). Animals were synchronized by hypochlorite treatment, plated on RNAi or L4440 control, and grown until the P6.p 2-cell stage. Screening was performed on a Ziess upright compound microscope equipped with 488 and 561 nm filters. Animals were scored for the loss of NUO-1::mNG fluorescence enrichment in the AC compared to the uterine cell within the same animal. Systemic effects of the RNAi to AC health were evaluated by comparing signal of the promoter driven AC-specific mCherry ( cdh-3p :moeABD::mCherry) to the control. RNAi treatment that led to a decrease in the mCherry signal would suggest a nonspecific effect and would not be included in the analysis. All genes that decreased the enrichment of NUO-1::mNG in the AC after RNAi mediated knockdown were reported as affecting the generation of high-capacity mitochondria ( Table S6 ) and each of these genes was repeated via RNAi three times to validate the effect. The plasmid containing the AC-specific overexpression of UCP-4 also contained a membrane-localized mKate2 (2xmKate2::PLC∂P). When imaging the effects of UCP-4 on the ATP:ADP ratio, we confirmed expression of mKate2. Confocal images were acquired on a Ziess Axioimager microscope equipped with either a Yokogawa CSU-10 or CSU- W1 spinning disk confocal controlled by Micromanager Software 120 vv1.4.23 or v2.0.1 using a Zeiss 100x Plan-Apochromat 1.4NA oil immersion objective and with either a Hamamatsu ORCA-Fusion sCMOS camera, Hamamatsu ORCA-Quest qCMOS camera, or ImageEM EMCCD camera. Figure 5B and S5B confocal images were acquired using a Yokogawa CSU-W1 spinning disk confocal on a Nikon Ti2E Motorized PFS Microscope with an ORCA-Quest qCMOS camera and deconvolved using Nikon Elements Software. The ImageEM EMCCD camera was used to boost signal to noise, such as in Figures 2D and 5B to compare all endogenously tagged components even those that were present at low levels in the AC and in 4A to detect the reduced fluorescence of NUO-1::mNG and UCR-2.1::mNG in the egl-43 RNAi knockdown experiments. Animals were mounted on 5% noble agar pads and anesthetized with either 0.01M sodium azide (Sigma-Aldrich #S2002) or 5mM Levamisole (Millipore Sigma #L9756). Time-lapse imaging of AC invasion was performed as previously described on a Ziess Axioimager microscope with a Yokogawa CSU-W1 spinning disk and Hamamatsu ORCA-Fusion sCMOS camera 121 . Fluorescence recovery after photobleaching (FRAP) was done on a Ziess Axioimager microscope equipped with an iLas targeted laser system from BioVision using an Omicron Lux 60mW 405nm laser, Yokogawa CSU- W1 spinning disk confocal controlled by Metamorph imaging software, a Zeiss 100x Plan-Apochromat 1.4NA oil immersion objective and Hamamatsu ORCA-Fusion sCMOS camera. All quantitative mitochondrial protein measurements of endogenously tagged strains ( e.g., Figures 2D and 3E ) were imaged using identical acquisition settings (488nm laser at 0.6 laser power, 150ms exposure, 128eGain, relative z-stack from −3.5μmto 3.5um, z-step size 0.37um) along with a control strain (NUO-1::mNG) to confirm reproducibility across multiple imaging sessions 122 . PercevalHR was imaged as previously described 41 . Briefly, optimal acquisition settings for both the ATP (488nm excitation and 525nm emission) and ADP (405nm excitation and 525nm emission) channels were determined to be 488nm laser at 2.5 laser power with 500ms exposure and 405nm laser at 3.0 laser power with 1000ms exposure. To prevent photobleaching from the 405nm laser affecting 488nm excitation, multidimensional acquisition parameters were set to acquire all z-slices of the ATP channel before the ADP channel. HYlight was imaged as previously described 48 . Using the published physiological ratio for the Hylight-RA control (FBP-bound/FBP-unbound ratio of ~0.4) 48 , appropriate acquisition settings were determined as 488nm laser at 2.5 laser power with 800ms exposure and 405nm laser at 3.0 laser power with 800ms exposure. For all experiments measuring glycolysis, both HYlight and HYlight-RA were imaged with the same acquisition settings. Due to the photosensitivity of the sensor HYlight(-RA) one z-plane was imaged per animal. Quantitative imaging of iATPSnFr1.0, GFP, and ETC promoter driven fluorescence expression, was acquired with standardized acquisition settings to ensure intensiometric measurements were reproducible across developmental time and experimental condition. Comparisions between the uterine cells and anchor cell were made in the same animal for all comparisions, except for the HYlight(-RA), due to photobleaching. Comparisions between the apical and basal regions of the anchor cell were made within the same cell. Tetramethylrhodamine, Ethyl Ester, Perchlorate (TMRE, ThermoFisher Scientific #T669) was reconstituted in DMSO and diluted to 1μM in M9 buffer. A mixed population of animals were added to fresh OP50 NGM plates with 400μL of 1μM TMRE. Animals were stained and allowed to continue to grow overnight in the dark at room temperature before imaging. Nonyl Acridine Orange (NAO, ThermoFisher Scientific #A1372) was reconstituted in M9 then diluted to 5μM in M9 buffer. As with the TMRE protocol, 100μL of 5μM NAO was added to OP50 NGM plates with a mixed population of animals. The NAO treated animals were grown overnight at room temperature before imaging. To account for variability in the uptake of dyes, AC mitochondrial measurements of TMRE and NAO were internally controlled by either measuring neighboring uterine cells or measuring both apical and basal mitochondria within the same cell. MitoCarta is a compendium of human and mouse mitochondrial genes coding for proteins in the 149 annotated mitochondrial pathways 58 . The current version of MitoCarta3.0 includes 1136 human mitochondrial genes. Using OrthoList2.0, a comparative orthologs prediction tool between C. elegans and human genome 123 , we identified 824 C. elegans orthologs of the human MitoCarta3.0 genes. Out of the 312 human genes that Ortholist2.0 did not annotate, orthologs for an additional 131 genes were retrieved manually from the Alliance of Genome Resources 114 . In total, 84% of the human MitoCarta genes had orthologous genes in C. elegans (955 out of 1136 genes) and 16% of the human genes did not have known C. elegans orthology (181 out of 1136). Ortholist2.0 and Alliance of Genome Resources cross reference 6 and 9 different gene ortholog databases, respectively, accounting for the discrepancies in ortholog annotations. The number of databases confirming orthologs for each gene is reported in Table S2 . A total of 1094 (including duplicates of any C. elegans orthologs) C. elegans mitochondrial genes were compiled ( Table S2 ). The C. elegans MitoCarta (CMC) genes were cross-referenced with a recently generated AC transcriptome 11 to determine the transcriptional enrichment of each gene in the AC compared to the whole animal (as measured by Log2-fold change). Gene Set Enrichment Analysis (GSEA) was conducted based on the data set of the transcriptional enrichment (log2-fold change) of each CMC genes and the gene set of 149 mitochondria pathways 124 – 126 . The GSEA algorithm filters the gene set size and removes all mitochondrial pathways where the number of genes is less than 2 126 . A list of 145 gene sets/mitochondrial pathways was analyzed and sorted based on the normalized enrichment score (NES). The significance of the enrichment was interpreted with a false discovery rate (FDR) Q-value that is less than 0.05. Plotting for the GSEA ( Figure S2 ) was conducted in R studio (version 4.4.0). The C. elegans Anchor Cell Mitochondrial Gene Transcriptome Explorer is accessible at https://sherwoodlab.shinyapps.io/AC_MitoGenes/ without login requirement, or can be downloaded and accessed locally at https://github.com/DavidSherwoodLab/ACmitogenes . All acquired images of endogenously-tagged mitochondrial proteins were processed using ImageJ/Fiji 127 . Mitochondria signal within the cell of interest (either AC or UC) was analyzed using 3-slice sum projections (0.2 μm z-step size) confocal z-stacks (with background subtraction, 6.7 pixel rolling ball radius). A mask for the mitochondria in the cell of interest was generated using the adaptive thresholding plugin 128 . This mask was used as a region of interest and the mean fluorescence intensity was measured to determine the “total” fluorescence per mitochondrial area for all mitochondria in the cell (e.g, Figure 2D and 3E ). For measurements of apical and basal mitochondria ( e.g., Figure 5C ), the cell was manually subdivided into apical and basal regions and the mean fluorescence intensity per mitochondrial area was measured in each region. Polarity measurements were calculated using the following ratio (e.g., Figure 5C ): polarity = basal mean fluorescence intensity per mitochondrial area apical mean fluorescence intensity per mitochondrial area Using a mitochondrially localized mKate ( rpl-28p::tomm-20::mKate2 ), a confocal z-stack was acquired through the entire uterus (51 slices, 0.2 μm step size) and the images were background subtracted using 6.7 pixel rolling ball radius. To generate an accurate projection of all mitochondria in each analyzed cell (both AC and uterine cell), regions of interest (ROI) around the boundary of each cell were manually drawn and any signal outside of this ROI was removed for each z-slice. The new compiled stack for each cell of interest was subsequently analyzed using Imaris Version 9.9.1 to generate a 3D isosurface rendering. 3D isosurface renderings relied on adaptive thresholding as described above and were used to determine the volume of the mitochondria within each cell. Z-stacks of all the mitochondria within the cell of interest were generated as described above. Using the ImageJ plugin “Mitochondrial Analyzer 128 ”, each z-stack was thresholded and mitochondrial sphericity and branch number were measured. Using animals with endogenously tagged HMG-5 ( hmg-5::GFP ) 54 , a confocal z-stack was collected through the entire uterus (51 slices, 0.20 μm step size). 4 central slices of the AC or UC were sum z-projected and the number of puncta were manually counted in each cell. Using transmission electron microscopy (TEM) images from a depositary of images from previously published work 34 , boundaries around individual mitochondria were manually drawn and annotated as either AC or UC. Those images were then assigned random numbers to blind the analysis. For each blinded image the freehand selection tool was used to circle and measure the area of both the outer perimeter of the mitochondria (total mitochondrial area) and the clear space lacking membrane (mitochondrial matrix area). To estimate cristae density we used the following ratio ( Figure S3A ): cristae density = ( Total Mitochondrial Area − Mitochondrial Matrix Area ) Total Mitochondrial Area Ratiometric biosensors PercevalHR and HYlight(-RA) were used to determine the ATD:ADP ratio and FBP-bound:FBP-unbound ratio, respectively. As described above (See Microscopy and Image Acquisition ), the imaging data for these biosensors were acquired as two-channel images. To transform the two-channel images into ratiometric images, the “Imaging Calculator > Divide” function in Fiji was used to divide the signal of the 488nm-excitation channel (ATP for PercevalHR and FBP-bound for HYlight(-RA)) by the 405nm-excitation channel (ADP for PercevalHR and FBP-unbound for HYlight(-RA)). The divided ratiometric images were used for quantification. Representative images for figures were displayed as the spectral intensity map using the “Fire” look-up table in Fiji (LUT). DHB::2xmKate2 ratios were quantified using 2-slice sum projection (0.2 μm z-step size) confocal z-stacks at the center of the AC (with background substraction). Two ROIs were drawn by hand around the nucleus and around the cytoplasm excluding the nucleus and avoiding pixels belonging to the cytoplasm of neighboring cells. A measurement of the mean fluorescence intensity was obtained and a cytoplasmic:nuclear ratio was calculated. NDUV-2::mNG fluorescence intensity per mitochondrial area was analyzed in the same cell as decribed previously. Spectral intensity maps were used to emphasize differences in fluorescent intensity. Using the “Fire” look-up table (LUT) in Fiji, the spectral intensity map and corresponding calibration bar were applied to representative images shown in figures. In instances where all representative images within the panel were constrained to the same minimum and maximum pixel intensity, exact values were displayed at the bottom and top of the calibration bar, respectively. Calibration bars were annotated with “hi” and “lo” (represent high and low intensity pixel values) for representative images that were displayed with different minimum and maximum pixel intensity either due to variation in the image acquisition settings or normalization of ATP:ADP ratios. For all experiments, the sample size ( n ) of animals, cells, or mitochondria measured were listed in the figure legend or table, along with the statistical tests used and the p-values. All statistical analyses and graph generation were done in GraphPad Prism (Version 10). Normality of each experimental data set was evaluated using the Shapiro-Wilk test. Comparisons across multiple time-points were performed using a one-way ANOVA with Tukey’s post hoc test for multiple comparisons. Comparisons of fluorescence intensity between two cells in the same animal (AC vs UC) or between two mitochondrial populations within the same cell (apical vs basal) were conducted using paired t-test or mixed-effects ANOVA with Geisser-Greenhouse correction and uncorrected Fisher’s LSD to allow for comparison between designated pairs. Comparisons of fluorescence intensity between a control and treatment were made using a two-tailed unpaired t-test with Welch’s correction if the variance was unequal between the two groups. Comparisons of fluorescence intensity between a single control and two or more treatments were performed by one-way ANOVA with Dunnett’s test for multiple comparisons. When assessing AC invasion defects based on invasion scoring, a Fisher’s exact 2×2 test was used to compare the treatment versus the control. When determining if there was a statistically significant relationship between the lag-2 expression ratio and the mitochondrial component fluorescence intensity per mitochondrial area ratio between the α1 and α2 cells, measure of non-zero slope was used to interpret the results of the simple linear regression analysis.

Anchor cell (AC) invasion is a stereotyped basement membrane (BM) transmigration that can be staged with the underlying 1° fated P6.p vulval precursor cell (VPC) divisions 31 , 32 . The AC is a specialized uterine cell specified between the L2/L3 larval stages (early P6.p 1-cell stage, Figure 1A ) 39 . During L3, the AC grows and translates pro-invasive proteins (P6.p 1-cell stage) 11 , 33 , 40 . At the P6.p early 2-cell stage, an F-actin rich invadosome penetrates the BM, triggering lysosome exocytosis and focused F-actin generation to form a single large protrusion 9 , 10 , 35 . The protrusion degrades and displaces BM, allowing the AC to initiate direct uterine-vulval contact ( Figure 1A ). Mitochondria at the AC invasive front generate localized ATP to fuel invasion 37 , 38 , 41 . We previously used an AC-expressed ratiometric ATP:ADP biosensor, PercevalHR 38 , 42 , which revealed a dramatic increase in the ATP:ADP ratio at the site of mitochondria enrichment during BM breaching and clearance 38 . This suggested that ATP generation might be distinctly regulated in the AC. To compare AC ATP metabolism with neighboring non-invasive uterine cells, we ubiquitously expressed genetically encoded biosensors, PercevalHR and iATPSnFR1.0 ( eef-1a.1 promoter) 43 . PercevalHR measures the ATP:ADP ratio, which reflects the free energy of ATP hydrolysis available for driving energy demanding processes 42 . iATPSnFR1.0 is formed from circularly permuted superfolder GFP inserted between the ATP-binding helices of the ε-subunit of a bacterial F0-F1 ATPase and is responsive to physiological cytoplasmic ATP levels 44 . There was an increase in the ATP:ADP ratio in the AC compared to neighboring uterine cells at the P6.p 1-cell stage (n = 20/20 animals) several hours prior to invasion at the time when the AC is growing and translating pro-invasive proteins ( Figure 1B ). This ratio peaked and polarized toward the invasive front during BM breaching ( Figure 1B , 2-cell n = 27/27 and 2–4-cell stage, n = 21/21 ATP polarized). Similarly, fluorescence levels of iATPSnFR1.0 were elevated at the P6.p 1-cell stage and increased ~50% during BM breaching ( Figure 1C ). Given the known dynamic response range of iATPSnFR1.0 in vitro , this likely represents over a 3-fold increase in ATP in the AC 44 . We also found a ~40% increase in AC iATPSnFR1.0 fluorescence compared to neighboring uterine cells ( Figure 1D and S1A ) 43 . Unlike PercevalHR, which has rapid fluorescence kinetics with exposure to ATP and ADP 42 , iATPSnFr1.0 takes up to 10 seconds to return to baseline fluorescence after ATP exposure 44 . This likely accounts for its uniform cytosolic and nuclear localization. iATPSnFR1.0 is a GFP based sensor and, like GFP, is sensitive to pH 45 . Importantly, GFP does not show pH sensitivity in the AC 38 . We conclude that the AC uniquely produces high ATP levels prior to invasion that peak during BM transmigration. Cellular ATP is produced via mitochondrial oxidative phosphorylation (OXPHOS) and glycolysis 46 . To assess the contribution of glycolysis, we used the glycolytic ratiometric biosensor HYlight 47 . The fluorescence ratio produced by HYlight serves as a proxy for glycolytic activity, as it measures the glycolytic metabolite fructose 1, 6-biphosphate (FBP), which is the glycolysis commitment step 47 , 48 . HYlight showed similar glycolytic acitivity in the AC and uterine cells ( Figure S1B ). Rotenone treatment, a mitochondrial toxin that reduces OXPHOS 49 , significantly decreased iATPSnFR1.0 fluorescence (~40%), similar to the reduction in the ATP:ADP ratio we previously reported using PercevalHR 38 , and caused invasion defects ( Figures 1E and 1F ; Table S1 ). AC-specific overexpression of mitochondrial uncoupling protein UCP-4 ( lin-29p::UCP-4::SL2::mKate2::PH ), which dissipates the mitochondrial protein gradient, also reduced the AC ATP:ADP ratio ( Figure S1C ) 50 , 51 . Impaired mitochondria respiration shifts neuronal metabolism towards glycolysis 48 . Rotenone treatment increased glycolysis (30% increase in HYlight ratio, Figure 1G , HYlight-Reduced Affinity (-RA) control Figure S1D ), potentially explaining the residual ATP production in the AC and moderate invasion defect after OXPHOS inhibition ( Figures 1E and 1F ). Together, these results implicate mitochondrial OXPHOS in the increased ATP production necessary for AC invasion. Increased mitochondrial biogenesis, increased mitochondrial volume, and altered mitochondrial morphology can enhance ATP production 52 , 53 . However, we found neighboring uterine cell mitochondria have a similar morphology to the AC and an overall higher mitochondrial volume ( Figures S1E and S1F ). Further, we examined the localization of mitochondrial transcription factor-A puncta (TFAM, C. elegans HMG-5::GFP) 54 , which marks mitochondrial nucleoids 54 – 56 . There was no difference in the number of TFAM puncta in the AC compared to neighboring uterine cells ( Figure S1G ), strongly suggesting a lack of enhanced AC mitochondrial biogenesis. These results indicate that the AC does not require greater mitochondria number, volume, or morphology to drive higher OXPHOS dependent ATP levels. We next determined if AC mitochondria have a distinct molecular composition to facilitate high ATP production. Mitochondria have two phospholipid bilayers: an outer membrane (OMM) separating mitochondria from cytoplasm, and a folded inner membrane (IMM) enclosing the mitochondrial matrix ( Figure 2A ) 57 . These mitochondrial compartments house proteins that perform many functions, including OXPHOS, apoptosis, and calcium regulation 57 . To determine if mitochondrial molecular functions are augmented in the AC, we referenced human MitoCarta3.0, a dataset of 1136 mitochondrially-associated genes with mitochondrial pathway annotations 58 , and identified 1076 known or predicted C. elegans orthologs ( Table S2 ). Using an AC transcriptome, we cross-referenced these C. elegans mitochondrial genes to determine which are AC enriched ( Figure S2A ; Table S2 ) 11 . We generated an RShiny app for identifying mitochondrial gene orthologs ( C. elegans and human genes) and for plotting differentially expressed genes within mitochondrial pathways ( https://sherwoodlab.shinyapps.io/AC_MitoGenes/ ). Of the top 10 enriched pathways, 8 encoded electron transport chain (ETC) and OXPHOS proteins ( Figure S2A ), suggesting increased ETC levels might contribute to elevated AC ATP production. The ETC comprises five multi-protein complexes (CI-V) localized to the IMM that generate an electrochemical proton gradient driving ATP synthase ( Figure 2A ) 59 . RNAi-mediated reduction of transcriptionally-enriched ETC components from complexes I, II and IV, but not complex III, caused invasion defects ( Figures 2B and S2B ; Table S1 ). As components of OXPHOS complexes might have modulatory roles in ATP production 60 , we also targeted ant-1.1 (Human SLC25A4 / 5 / 6 ), the dominant C. elegans adenine nucleotide translocase (ANT) that transports ATP out of the mitochondria 61 . RNAi-mediated depletion of ANT-1.1 resulted in a strong invasion defect (60%) ( Figures 2B and S2B , Table S1 ). Additionally, RNAi-mediated depletion of the ETC components NUO-1 (Human NDUFV1) and UCR-2.1 (Human UQCRC2) proteins, but not NDUV-2 (Human NDUFV2 ), decreased the ATP:ADP ratio in the AC ( Figures 2C and S2C , Methods ). As with rotenone treatment, RNAi-mediated reduction of NUO-1 increased AC glycolysis ( Figure S2D ). To determine whether transcriptional ETC enrichment correlated with protein abundance, we used genome-editing to insert mNeonGreen (mNG) at the C-terminus of 15 ETC proteins: 4 Complex I (CI) (C. elegans/ Human : nduv-2 / NDUFV2, nduf-7/NDUFS7, nuo-1/NDUFV1, nuo-6/NDUFB4), 2 Complex II (CII) ( sdhb-1/SDHB, mev-1/SDHC ), 3 Complex III (CIII) ( ucr-2.1/UQCRC2, isp-1/UQCRFS1, cyc-1/CYC1 ), 5 Complex IV (CIV) ( cox-10/COX10, cox-5a/COX5A, cox-5b/COX5B, cox-6aCOX6A1/2, cox-4/COX4I1/2 ) proteins, and 1 Complex V (CV) ( atp-4/ATP5PF ) protein. Of the 15 tagged strains, 4 ( nuo-6, isp-1, cyc-1, atp-4 ) were homozygous sterile but viable as heterozygotes (see Table S3 for viability and health of knock-in strains; Methods ), suggesting a sensitive mitochondrial germline requirement—a tissue where mitochondria are highly active 62 , 63 . Seven of the 11 homozygous strains had growth rates comparable to wild-type animals, while 4 were slower growing ( nuo-1 , sdhb-1 , mev-1 , cox-5a ; Table S3 ). Seahorse analysis for mitochondrial respiratory capacity of 8 homozygous viable strains revealed normal mitochondria health, except for COX-4::mNG and SDHB-1::mNG. Only COX-4::mNG showed impaired basal function ( Figure S2E ) 64 , 65 . In all 15 tagged strains, homozygous animals displayed normal AC invasion, suggesting ETC function in the AC was largely unaffected ( Table S1 ). We next measured fluorescence intensity of each mNG tagged ETC component in the AC compared to the non-invasive neighboring uterine cell mitochondria at the P6.p 2-cell stage ( Figures 2D , 2E , and S2F ). To control for mitochondrial density, we utilized adaptive thresholding and quantified ETC component mean intensity per mitochondrial area ( Methods ). Strikingly, 14 of the 15 ETC proteins were enriched (~1.2 to 2.1-fold) in AC mitochondria compared to uterine cell mitochondria ( Figures 2D ). Transcriptional reporters of nuo-1 and nduv-2 ( nuo-1 p::mNG; nduv-2 p:mNG) also showed increased expression in the AC ( Figure S2G ). Since the ETC generates the mitochondrial membrane potential to produce ATP, we used a mitochondrial membrane potential sensitive dye, Tetramethylrhodamine Ethyl Ester (TMRE). While a higher membrane potential is not always associated with increased oxidative capacity or ATP production 66 , we found that the AC has a higher membrane potential than uterine cells ( Figure S2H ). We conclude that the AC has ETC-enriched mitochondria that generate an increased membrane potential and produce higher ATP levels. We hereafter refer to these as high-capacity mitochondria. We next investigated the structure of the high-capacity AC mitochondria. The ETC localizes to inner mitochondrial membrane (IMM) folds, or cristae, which increase surface area and allow for dense packing of ETC proteins 67 . Using transmission electron microscopy (TEM), we measured areas of mitochondrial matrix (less dense, clear space and lacking membrane) relative to the total mitochondrial area to indirectly assess cristae density (the higher the density of cristae, the lower the matrix area). AC basal mitochondria had less mitochondrial matrix area relative to total mitochondrial area compared to non-invasive neighboring uterine cell mitochondria implying that the AC basal mitochondria contain a higher cristae density ( Figure S3A ). We next asked if molecular differences support high ETC-enriched AC mitochondria. We reasoned that high-capacity mitochondria might require more import machinery to support denser cristae. We thus examined mitochondrial protein translocases ( C. elegans TOMM and TIMM complexes) that facilitate protein import from the cytoplasm, mitochondrial contact site and cristae organizing system (MICOS) components that stabilize cristae, and the cardiolipin synthase (CRLS) enzyme that synthesizes cardiolipin, which supports cristae folding ( Figure 3A ) 68 , 69 . An RNAi screen targeting these genes revealed that knockdown of the majority decreased AC mitochondria enrichment of ETC components NUO-1::mNG (CI), NDUV-1::mNG (CI) and UCR-2.1::mNG (CIII) ( Figures 3B , S3B , and S3C ; Table S4 ). Furthermore, RNAi knockdown of tomm-20 (TOMM complex; Human TOMM20 ), immt-1 (MICOS component, Human IMMT ), and crls-1 (cardiolipin synthase, Human CRSL1 ), disrupted AC invasion, but only tomm-20 and crls-1 knockdown significantly decreased the AC ATP:ADP ratio ( Figures 3C and 3D ; Table S1 ). Endogenous mNG knock-ins of TOMM-20, IMMT-1 (MICOS component), and CRLS-1 were enriched in AC mitochondria compared to neighboring uterine cells (1.3 to 1.6-fold, Figures 3E and 3F ). Similarly, nonyl acridine orange (NAO), which stains cardiolipin 70 , showed ~2-fold more cardiolipin accumulation in AC versus uterine mitochondria ( Figure S3D ). Thus, AC high-capacity mitochondria are built with denser cristae and an increased import system to harbor ETC-enriched mitochondria that increase ATP production for invasion. We next wanted to determine if the ETC-enriched mitochondria are specified by the AC pro-invasive transcriptional program 71 . EGL-43 (MECOM oncogene, Human MECOM) is crucial to specifying AC invasive fate and independently promoting cell cycle arrest 33 . RNAi targeting egl-43 dramatically reduced AC mitochondrial ETC enrichment of NUO-1::mNG, UCR-2.1::mNG, and NDUV-2::mNG ( Figures 4A , S4A , and S4B ). Additionally, egl-43 RNAi significantly reduced nuo-1 and nduv-2 transcriptional reporter expression ( nuo-1 p::mNG; nduv-2 p:mNG) ( Figure 4B ). Knockdown of egl-43 can cause the AC, which is typically arrested in the G1 phase, to enter the cell cycle and give rise to multiple ACs 72 . To ensure decreased ETC component enrichment following egl-43 knockdown was not due to a change in the cell cycle state, only worms with one AC were analyzed. We monitored the cell cycle using a CDK activity sensor based on human DNA helicase B (DHB), DHB::2xmKate2 73 . In a normal post-mitotic AC, DHB::2xmKate2 localizes primarily to the nucleus and does not diffuse into the cytoplasm (low cytoplasmic-to-nuclear ratio, empty vector control, Figure S4B ). Following egl-43 RNAi knockdown, regardless of the cell cycle state of the AC (low or high DHB::2xmKate2 cytoplasmic-to-nuclear ratio), NDUV-2::mNG enrichment decreased ( Figure S4B ). To determine when AC high-capacity mitochondria form, we examined molecular enrichment of ETC components, cristae components, and mitochondrial import machinery. The AC and neighboring ventral uterine cells (VU) are stochastically specified from two proto-uterine cells (α1 and α2, Figure 4B ). The earliest AC fate marker is LIN-12 (Notch, Human NOTCH4 ) ligand LAG-2 upregulation ( Figure 4C ) 74 . We examined AC mitochondrial components required for ETC enrichment (TOMM-20::mNG, IMMT-1::mNG, CRLS-1::mNG), an additional protein translocase (TOMM-70::GFP) 75 , and ETC components (NDUV-2::mNG and UCR-2.1::mNG) in combination with lag-2 p::2xmKate2::PH ( Figures 4D and S4C ). We correlated the lag-2p:: 2xmKate2::PH α1:α2 signal ratio, with the mitochondrial component α1:α2 ratio in proto-uterine cells at the time of AC/VU specification. The non-zero slope, a measure of regression, was only significant for the relationship between both TOMM-20::mNG and TOMM-70::GFP and the lag-2 driven signal ( i.e . they both increased in the early specified AC at the same time) ( Figure 4D ). This indicates that, of the AC mitochondria components tested, TOMM-20 and TOMM-70 are the first to become enriched and this occurs during AC specification. We next quantified levels of TOMM-20::mNG, IMMT-1::mNG, CRLS-1::mNG, NDUV-2::mNG, and UCR-2.1::mNG per mitochondrial area after AC specification (P6.p early 1-cell stage) and leading up to (1- and 2-cell) and after (4-cell) invasion. All components at the early 1-cell stage were elevated compared to neighboring uterine cells. Further, TOMM-20 remained stable across time, IMMT-1 peaked at the early 1-cell, CRLS-1 increased over time, and both ETC proteins peaked between 1-cell and 2-cell stage ( Figures 4E and S4D ). Thus, high-capacity AC mitochondria are specified by the pro-invasive transcriptional network and begin forming early during AC invasive differentiation. We previously reported that mitochondria localize towards the AC invasive front 38 . To better quantify enrichment, we examined mitochondrial volume and found that ~70% was basally localized at the 2-cell stage, while neighboring uterine cells showed equal distribution ( Figure S5A ). Additionally, basal mitochondria had higher mitochondrial membrane potential (TMRE staining) compared to apical mitochondria ( Figure S5C ), suggesting that basal AC mitochondria might be high-capacity and distinct. To determine if the basal mitochondria are high-capacity, we quantified ETC component intensity per mitochondrial area of endogenously tagged ETC components (as in Figure 2 ) and TOMM-20, IMMT-1, and CRSL-1 ( Figure 3 ). All but one were significantly enriched (1.5–2.0-fold) in the basal mitochondria ( Figures 5A – 5C and S5C ). Higher resolution imaging of NUO-1::mNG and NDUV-2::mNG (single slices,~150nm resolution) showed higher average fluorescence signal in the basal mitochondria relative to the apical. Further, there were bright puncta within the basal population, suggesting heterogeneity in enrichment within the invasive mitochondria ( Figure 5B and S5B ). We also note that, GFP::MIRO-1 (Human RHOT1/2), a mitochondrial adaptor protein, exhibited equal apical and basal mitochondria distribution ( Figure S7B ). Additionally, mitochondria were initially uniformly distributed shortly after AC specification (early-1-cell stage), but segregated into two populations at the late 1-cell stage ( Figure 5D ). To ascertain if these two mitochondrial populations intermix, we performed fluorescence recovery after photobleaching (FRAP) on NUO-1::mNG mitochondria at the initiation of invasion (2-cell stage). A portion of basal invasive mitochondria were photobleached, then the change in fluorescence intensity in the bleached and non-photobleached apical and neighboring basal mitochondria was measured to determine if, and from where, mitochondria moved into the bleached region ( Figure 5E ). Following photobleaching, the mitochondrial signal in the bleached basal mitochondria recovered. Notably, the signal in the non-photobleached adjacent basal region decreased, while the apical signal was unchanged ( Figures 5E , 5F , and S5D ). Following photobleaching of AC apical mitochondria, the mitochondrial signal in the bleached apical mitochondria recovered and the signal in non-photobleached adjacent apical mitochondria decreased, while the basal mitochondria signal remained unchanged ( Figure S5E ). In contrast, photobleaching in neighboring uterine cells revealed exchange between all mitochondria ( Figures 5F , S5D , and S5E ). Therefore, the AC contains two spatially and compositionally distinct mitochondrial populations: a specialized high-capacity basal mitochondria population isolated from an apical population with lower ETC levels. We next examined high-capacity mitochondrial dynamics during invasion. Using endogenously-tagged type IV collagen to mark the BM (EMB-9::mRuby) 76 and NUO-1::mNG to image mitochondria, we performed time-lapse imaging of BM breaching and protrusion formation. In all cases we observed high-capacity NUO-1::mNG enriched mitochondria at the breach site and concentrated within the emerging invasive protrusion (n = 6/6; Figures 6A , S6A , and S6B ; Video S1 ). Netrin ( C. elegans UNC-6), secreted from the underlying 1° fated P6.p VPCs, polarizes invadosomes, the invasive protrusion, and prenylation enzymes towards the invasive plasma membrane 9 , 10 , 35 , 38 . To determine if netrin signaling also polarizes high-capacity mitochondria, we examined mitochondria (marked with NUO-1::mNG) in an unc-6 (ev400) null mutant. Loss of netrin disrupted the basal enrichment of high-capacity mitochondria and the ATP:ADP ratio ( Figures 6B and 6C ). Thus, netrin signaling directs high-capacity mitochondria to the invasive front, where they concentrate at the BM breach site and enter the invasive protrusion. Microtubules transport mitochondria to areas of high ATP demand in neurons and cultured migrating cells 22 , 23 , 28 , 77 . Netrin regulates microtubule dynamics, stability, and elongation to attract and direct growing neuronal axons and microtubules are enriched at the AC’s invasive front 38 . We visualized AC microtubules using the microtubule-binding domain of ensconsin ( lin-29p ::EMTB::GFP) 78 and found that microtubule basal enrichment was greatly reduced in netrin mutants ( Figure 7A ). Microtubules have two functionally distinct ends—the plus-end with β-tubulin exposed that polymerizes and depolymerizes faster than the minus-end with α-tubulin exposed 79 . We examined the endogenously tagged plus-end protein (EBP-2::GFP; Human MAPRE1/2/3 ) and minus-end protein (GFP::GIP-1) 80 , and found that microtubules are oriented with the plus-end at the basal region of the AC and the minus-end at the apical ( Figure 7B ). Thus, trafficking along polarized microtubules could direct high-capacity mitochondria to the invasive front. Microtubules serve as tracks for motor proteins and their adaptors to transport cargo. Notably, genes encoding mitochondrial trafficking proteins were among the upregulated MitoCarta AC transcriptome pathways ( Table S2 ). Of these, we examined membrane adaptor complexes that mediate the movement of mitochondria in plus-end directed mitochondrial trafficking 81 , 82 . The two C. elegans metaxin homologs, MTX-1 and MTX-2 (Human MTX-1/3 and MTX-2 , respectively), are OMM mitochondrial adaptors with functionally distinct roles in trafficking. MTX-2 works with microtubule adaptor, MIRO-1 (Human RHOT1/2 ), for both microtubule plus-end and minus-end transport, whereas MTX-1 only serves as an adaptor for plus-end transport 82 . RNAi-mediated loss of mtx-1 and mtx-2 , but not miro-1 , significantly reduced the basal polarity of high-capacity mitochondria ( Figures 7C , S7A , and S7C ). Endogenously tagged mtx-1 and mtx-2 with mNG were basally enriched in the high-capacity mitochondria ( Figure 7D ), whereas GFP::MIRO-1 82 was equally localized ( Figure S7B ). Previous work indicates that MTX-2 localizes to mitochondria independently of MIRO-1 and can compensate for the loss of MIRO-1. Thus, MIRO-1 might be compensated for or does not play a role in AC mitochondrial localization. Next we examined AC high-capacity mitochondria (marked with NUO-1::mNG) following RNAi knockdown of kinesins, the molecular motor proteins primarily responsible for plus-end directed movement. We identified and targeted kinesins expressed at high levels in the AC transcriptome 11 . RNAi knockdown of klp-4 (Human KIF13A/B ) and klp-19 (Human KIF4A/B ), but not klc-2 (Human KLC1 ; n = 10/10 animals), klp-7 (Human KIF2A/B/C ; n = 10/10), or klp-13 (Human KIF19 ; n = 13/13), resulted in a significant loss of NUO-1::mNG basal enrichment ( Figure S7C ). Furthermore, metaxin localization was netrin dependent and loss of netrin disrupted basal metaxin enrichment ( Figure S7D ). We conclude that high-capacity mitochondria enrichment at the site of invasion is netrin dependent and trafficked on plus-end microtubules using kinesins and the metaxin adaptor complex. We next determined what links the netrin (UNC-6) directional cue to polarized microtubule transport. Src family kinases couple netrin signaling to microtubule dynamics through phosphorylation of β-tubulin 79 , which regulates microtubule dynamics 79 , 83 . C. elegans harbor two src genes , src-1 (Human YES1) and src-2 (Human FRK) , and src-1 is highly AC expressed 11 . Endogenously-tagged SRC-1::GFP showed basal AC enrichment in an unc-6 dependent manner ( Figure 7E ). Further, the basal enrichment of both high-capacity mitochondria and microtubules was reduced in a src-1(lq185) mutant background 84 ( Figures 7F and 7G ), indicating that SRC-1 polarizes microtubules and mitochondria to the basal invasive front. AMPK, an energy sensor and metabolic regulator, promotes mitochondria localization to the leading edge of lamellipodia in ovarian cancer cells 23 . RNAi knockdown of four AMPK components, including the the catalytic α-subunit of AMPK, aak-2 (Human PRKAA1/2) , did not disrupt mitochondrial localization ( Figure S7E ). Interestingly, previous studies in neurons and several cancer cell lines have shown ATP production loss reduces microtubule mitochondrial trafficking 22 , 23 , 85 . This suggests that ATP output might preferentially traffic high-capacity AC mitochondria along polarized microtubules to the invasive front. Consistent with this, both pharmacological (rotenone treatment) and genetic ( nuo-1 RNAi) perturbation of ETC function resulted in a loss of basal enrichment of high-capacity mitochondria ( Figures 7H and S7F ). Taken together, our results suggest netrin (UNC-6) signaling through SRC-1 polarizes microtubules towards the invasive front, which facilitates trafficking of high-capacity mitochondria harboring the metaxin adaptor complex to supply high levels of ATP for BM breaching.

Further information and requests for resources and reagents should be directed to and will be fulfilled by David R. Sherwood ( [email protected] ). All worm strains generated in this study will be available from the Caenorhabditis Genomics Center (CGC, gc.umn.edu ). All data reported in this paper will be shared by the lead author upon request. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

There is an emerging evidence that mitochondria specialize to meet the functional needs of different cells and tissues 29 , 86 . Tissue-level biochemical studies provided the first evidence of specialized mitochondria and showed distinct mitochondrial proteomes in the mouse brain, liver, heart, and kidney 87 . Subsequent findings identified distinct mitochondria within different cell types, including oocyte mitochondria lacking assembled ETC complex I to mitigate damaging ROS production 88 , fibroblast mitochondria lacking ATP synthase specialized to produce proline and ornithine 89 , mitochondria with increased Tomm20 clusters that correlate with increased membrane potential 90 , and osteosarcoma cell mitochondria with decreased matrix proteins and mitochondrial DNA and an upregulated Ca 2+ uniporter that regulate filipodia length 91 . Here we show the C. elegans AC has specialized mitochondria that produce high ATP levels to fuel BM invasion. Transcriptomic analysis revealed broad ETC component enrichment and examination of 15 endogenous mNeonGreen (mNG) tagged ETC proteins identified a distinct mitochondrial subpopulation localized to the AC invasive front. In these high-capacity mitochondria, increased TOMM complex proteins and dense cristae facilitate import and housing of greater ETC component levels that drive high ATP production. These mitochondria form shortly after AC specification, coinciding with increased ATP levels ~5 hours before AC invasion, which likely supports increased translation and lipid synthesis 10 – 13 . Further, we found mitochondrially-produced ATP peaked during BM breaching when energy consuming invasive protrusion dynamics occur 38 . Heterogeneity in the upregulation of different ETC components within the high-capacity mitochondria, both within ETC complexes and between complexes may indicate further specialization of ETC function. ETC-enriched mitochondria might be a common strategy to support energetically demanding processes. Proteomic studies have shown heart muscle and synapse mitochondria, which are energy intensive, contain higher levels of some ETC components 92 , 93 , and synaptic mitochondria contain dense cristae and elevated cytochrome C (complex III) 94 . Adipocyte mitochondria associated with lipid droplets also have dense cristae, elevated ETC complex IV component Cox4 and elevated respiratory capacity, which is thought to support ATP-dependent triacylglyceride synthesis 30 . In addition, high ETC levels might be a common feature of invasive cells, as aggressive ovarian, breast, pancreatic, and lymphoma cancers show upregulation of ETC components and complexes 95 – 99 . Mitochondria traffic on microtubules to areas of high energy demand, including to the leading edge of invasive and migrating cells and to neuronal synapses and dendrites 21 , 23 , 27 , 85 , 100 , 101 . However, the cues and mechanisms that direct trafficking are largely unknown. Our studies indicate that netrin signaling is key for high-capacity mitochondria trafficking to the BM breach site through microtubule polarization to the invasive front. Netrin signaling may polarize and redirect microtubules towards sources of netrin by balancing microtubule dynamics and stabilization 79 , 83 , potentially through Src family kinase mediated tubulin phosphorylation 83 . Consistent with this, we discovered that SRC-1 is upregulated in the AC and localized to the invasive front in a netrin dependent manner. Further, loss of src-1 disrupted microtubule and high-capacity mitochondria polarization. We also show that high-capacity mitochondrial trafficking is dependent on metaxin adaptors. Evidence from C. elegans , Drosophila , and human neurons suggests that MTX-2 is a core mitochondrial transport adaptor complex component that directs plus-end or minus-end mitochondrial movements with MTX-1 and TRAK-1, respectively 81 , 82 . Our work extends metaxin function beyond neurons to invasive cells. In the AC, microtubule plus-ends are oriented towards the site of BM invasion. Consistent with studies in neurons, loss of mtx-1 (plus-end adaptor) and mtx-2 (core adaptor) dramatically perturbed basal (plus-end directed) enrichment of high-capacity mitochondria. Further, we previously found that loss of TRAK-1 does not alter basal mitochondria enrichment 38 , suggesting a minor or absent function in minus-end directed movement of high-capacity mitochondria. An open question is how high-capacity mitochondria preferentially localize to the invasive front over mitochondria that have reduced ETC components and ATP production. AMPK, an energy sensor, promotes mitochondrial localization to the leading edge of lamellopodia in ovarian cancer cells, however, AMPK does not appear to regulate mitochondrial trafficking in the AC 23 . In contrast, reduction of the ETC component NUO-1 and OXPHOS inhibition with rotenone, which both reduced ATP production, perturbed high-capacity mitochondrial enrichment to the invasive front. As ATP diffusion is limited in cells 102 and microtubule trafficking is energy intensive 23 , 103 , one possible explanation is a competitive self-organizing system where high-capacity mitochondria with MTX-1 (plus-ended directed adaptor) and increased ATP output are more competitive in moving along microtubules, thus preferentially trafficked to microtubule ends at the invasive front. Consistent with this model, disruption of ATP production in mitochondria limits their trafficking in cancer cells and neurons 22 , 85 . Molecular competition of shared limited resources governs a number of biological processes, such as mRNA competition for ribsome translation, sigma factor competition for RNA polymerase binding, and competition between RNA binding proteins that drives localized phase separation in C. elegans embryos 103 – 105 . Mitochondria are responsible for cellular processes beyond ATP production, including fatty acid synthesis, stress response, cell-cell signaling, and calcium homeostasis 106 , 107 . Thus, specialized mitochondria might be a common, but hidden cellular feature. A key impediment in assessing mitochondrial diversity is detecting differences in individual mitochondrial molecular composition within and between cells. By endogenously editing 20 genes encoding mitochondrial proteins with mNG tags—15 ETC, two cristae, two metaxins and a TOMM complex component—our studies have greatly expanded the live-cell reagent toolkit to examine mitochondrial diversity. Our endogenously tagged strains revealed ETC-enriched high-capacity mitochondria in the AC and led to insights into single cell mitochondrial heterogeneity, specialization, specification, and trafficking. As C. elegans cell lineages and differentiation programs are known in detail 39 , this toolkit can be used and expanded to investigate mitochondrial specialization in other cell types, how mitochondria are regulated in cell division, apoptosis, and oogenesis, how mitochondria respond to environmental changes, and how mitochondria are altered during aging. Our Mitocarta-based curation indicates approximately 1000 C. elegans mitochondrial proteins. Ultimately, it should be possible to endogenously tag most mitochondrial proteins, which offers to reveal the breadth of mitochondrial diversity and lead to insights into unknown mitochondrial functions.

Our studies indicate that the AC has ETC-enriched mitochondria that generate a higher membrane potential and ATP to drive cell invasion. The ETC is composed of over 100 components with unique functions subdivided into five complexes 53 . Testing the function of the entire upregulated ETC complex was not possible. Instead, we relied on targeting individual ETC components, chemically inhibiting complex I, reducing mitochondrial membrane potential, and decreasing ATP release from mitochondria. While our data demonstrate that ETC upregulation generates more ATP, we cannot rule out that upregulation enhances or alters other ETC functions, such as ROS signaling, metabolite transport, Ca 2+ regulation, and regeneration of electron carriers 106 , 107 . Additionally, our targeting methods for inhibiting ETC-mediated ATP generation have caveats. For example, while rotenone treatment inhibits ETC-dependent ATP production, it can also upregulate ROS production and oxidative stress, damaging mitochondrial DNA and disrupting microtubule assembly 108 – 112 . Also, we could weaken, but not eliminate, ETC upregulation. Our RNAi mediated knockdown of individual ETC components might target non-essential proteins or lead to ETC compensation or AC adaptation, like we see with increased glycolysis. This may explain the variable and mild phenotypes from individual component knockdown. Similarly, reducing membrane potential using the uncoupler UCP-4 and targeting of the ATP transporter ant-1.1 , do not completely reduce the ability of the upregulated ETC to produce ATP, preventing assessment of its full contribution to breaching the BM and cell invasion.

Cell invasion through basement membrane (BM), a dense sheet-like extracellular matrix (ECM) surrounding and separating tissues 1 , is crucial for development and immune cell trafficking. Although a formidable barrier, many cells acquire a specialized ability to invade through BMs, including trophoblasts, neural crest cells, neuronal axons, and immune cells 2 – 5 . Dysregulation of BM invasion underlies many human diseases, such as rheumatoid arthritis, pre-eclampsia, and endometriosis 6 , 7 . Most notably, the acquisition of invasive behavior initiates metastasis, which is the primary cause of cancer lethality 8 . Despite its importance, the mechanisms that drive this specialized behavior are not fully understood. Invasive cells form F-actin-rich invadosomes, membrane-associated protrusions that harbor and secrete matrix metalloproteinases (MMPs) to breakdown and physically displace BM 9 . To develop invasive capabilities, cells upregulate ribosome biogenesis, increase translation of pro-invasive proteins, and undergo de novo lipid synthesis to form dynamic invadosomes 10 – 13 . Protein translation, lipid biogenesis, membrane trafficking, and F-actin turnover, require significant energy in the form of ATP 14 – 18 . Thus, invasive cells posess distinct metabolic mechanisms to fuel BM breakdown 19 . Metastatic cancers largely depend on mitochondrial oxidative phosphorylation (OXPHOS) to power invasion 19 , 20 . Mitochondria localize to the leading edge of many cancer cells and migrating fibroblasts through kinesin-mediated trafficking along microtubules 21 – 26 . Mitochondrial enrichment is thought to ensure localized ATP production to fuel F-actin polymerization, membrane dynamics, actomyosin contractility, and focal adhesion turnover required for cell movement and invasion 19 , 27 . In pancreatic ductal carcinoma cells, mitochondria fuse, increase in size, and generate more ATP within protrusions invading through artificial matrices 28 . Emerging evidence indicates that mitochondria are uniquely tailored for distinct functions 29 , 30 . Whether mitochondria at the leading edge of invasive cells have other attributes that facilitate high ATP production is unknown. Anchor cell (AC) invasion through BM in C. elegans is a visually accessible, stereotyped, and genetically tractable model of invasion 31 , 32 . The AC is a specialized uterine cell that invades the underlying linked gonadal and ventral epithelial BMs to initiate uterine-vulval attachment 33 , 34 . During AC invasion, a netrin ( C. elegans UNC-6) cue secreted from the vulval cells polarizes F-actin-rich invadosomes to the AC’s invasive front, where, through dynamic assembly and disassembly, they depress the BM until one breaches the BM 9 , 35 . At the breach, a large invasive protrusion forms via lysosome exocytosis to expand the BM opening 9 , 10 . The AC’s pro-invasive transcriptional program, including the proto-oncogenic Fos and MECOM transcription factors ( C. elegans FOS-1 and EGL-43, respectively), and BM breaching machinery, including actin regulators Arp2/3, cofilin, and matrix-degrading MMPs, are shared with metastatic cancer cells 33 , 36 . The AC harbors a robust energy acquisition and delivery system, including glucose transporters, glycolytic enzymes, and mitochondria that enrich at the invasive front and provide ATP to fuel the invasive machinery 37 , 38 . The experimental accessibility of AC invasion, combined with an AC transcriptome 33 , provide a powerful model to establish the cellular and molecular underpinnings of mitochondria formation and composition within invasive cells. Here, using whole-body ATP biosensors, we show the AC harbors elevated ATP levels during its differentiation that peaks and is enriched at the invasive front during BM breaching. AC mitochondrial gene expression analysis revealed high expression of electron transport chain (ETC) complex components. We endogenously tagged 15 ETC components across all five complexes and 5 mitochondrial proteins involved in import and cristae formation with mNeonGreen (mNG). These strains, combined with electron microscopy, mitochondrial transmembrane potential and lipid composition dyes, and targeted disruptions, revealed a population of specialized, ETC-enriched, transport enhanced, and cristae-dense, mitochondria that are preferentially localized to the invasive front and required for the high ATP levels for BM invasion. We show that high-capacity mitochondria are specified early during AC invasive differentiation by the proto-oncogenic transcription factor network and that netrin signaling guides their trafficking to the BM breaching site through mitochondrial metaxin adaptor complexes, microtubules, kinesins, and the Src family kinase, SRC-1. Together, we present an endogenously tagged mitochondrial protein toolkit and discover a specialized subset of high-capacity mitochondria that are preferentially trafficked fuel BM invasion.

Document S1. Figures S1 – S7 , Tables S1 , S3 , S4 , and S6 . Table S2. C. elegans mitochondrial genes with human orthologs, AC transcriptome data, and mitochondrial pathways. Related to Figure S2 and STAR Methods . Table S5. Oligonucleotides used in strain generation. Related to Figure 1 – 7 and STAR Methods . Video S1. High-capacity mitochondria localize to the BM breach site. Related to Figure 6 and S6 . Time-lapse of mitochondria (NUO-1::mNG, gray) prior to and during AC invasion through the BM (EMB-9::mRuby2, magenta) and in the emerging protrusion. (Top, related to Figure 6 ) Animals were imaged every 1 min for 31 min. (Middle, related to Figure S6 ) Animals were imaged every 1 min for 25 min. (Bottom, related to Figure S6 ) Animals were imaged every 1:30 min for 51 min. Scale bar, 5 μm.