1. Author response


    September 9, 2021

    We would like to thank ASAPbio for selecting our preprint for review! We are excited to contribute to this new process and hope others will find it as helpful as we have. The comments generated by the “crowd” were detailed and thoughtful. Below we respond to the major discussion points and if there were specific reviewer comments relevant to the discussion point, we also included that statement. We also responded to each specific comment. We would love to continue this discussion, so we invite further feedback and responses! Thanks so much for your time.

    -Chelsea Kidwell, Joey Casalini, and Minna Roh-Johnson


    Major Discussion Point #1:One of the most important claims is that mitochondria are the organelles responsible for the activation of the signals of cell proliferation. However, a previous report by the last author reported that macrophages transfer cytoplasm to recipient cells. It cannot be excluded that other organelles or cellular fragments are transferred as well and contribute to the observed effects (ERK activity). Perhaps a good way to solve this would be the use of macrophages that are devoid of mitochondria. At least, this aspect should be discussed in the manuscript.

    🡪 We had first considered two approaches to test the requirement and sufficiency of macrophage mitochondria in cancer cell proliferation. The first was to generate rho-zero macrophages (mtDNA-deficient), as you mention in your comment, such that the macrophages did not have functional mitochondria. However, we use primary human macrophages for all of our studies, and these cells would not survive long enough to generate rho-zero cells (which requires that the cells be treated with low levels of ethidium bromide for weeks). The second is to biochemically purify mitochondria from macrophages and directly inject these mitochondrial preps into breast cancer cells. We actually did this experiment, and cancer cells injected with purified mitochondrial preps exhibited higher proliferation (by live timelapse microscopy) compared to control cells. However, we also found that the mitochondrial purifications were not clean, and contained other membranous components in the cytoplasm. We tried centrifugation-centric approaches, as well as IP-ing against a mitochondrially-localized tag, but in all cases, the mitochondrial preparations contained other cytoplasmic components. Therefore, we did not feel that this approach was an adequate way to test effects of specifically the mitochondria. We certainly wanted to discuss this aspect in the manuscript, but unfortunately, we were limited due to space. If folks have suggestions on how to best purify mitochondria, we’d love to know, so please reach out.

    However, in terms of the bigger question of whether the induced proliferation in cancer cells is specifically due to ROS accumulation in transferred macrophage mitochondria, we tried to address this question with the mito-KillerRed experiments, where we generate ROS using optogenetics, and ask whether this accumulation is sufficient to induce cancer cell proliferation (which we showed it was). We also showed that this same approach could induce Erk activity, and then in separate experiments, we show that macrophage mitochondrial transfer results in accumulation of ROS and increased Erk activity. We feel that these experiments support our conclusions, however, we’d love for a way to link it all together. Unfortunately, we are not convinced that such experiments are possible at this time.

    Major Discussion Point #2: Most of the positive examples of transferred mitochondria discussed appeared in a small clump. However, there also appears to be another population that was more diffuse and co-localizes with host mitochondria (e.g., Fig2B, bottom right panels). It would be helpful to show results of these sibling mitochondria for assays performed on their clumpy siblings. If they behave differently, it would be helpful to provide some explanation.

    Specific Comment: Figure 2 Majority (57%) of donated mitochondria do not colocalize with LysoTracker signal (N=24 cells, 4 donors) - Here the paper implies that some transferred mitochondria do co-localize with lysoTracker signal. More importantly, they co-localize with host mitochondria. It raises the question of whether they signal through ROS and ERK like their clumpy siblings who are in the limelight of most figures.

    🡪Yes, you are correct. There does appear to be a diffuse population of macrophage mitochondria. The majority of these mitochondria co-localize with lysotracker, suggesting that they are being actively degraded. We can’t say that they tend to co-localize with endogenous cancer cell mitochondria, however, it’s possible that this diffuse population is comprised of both mitochondria that are being degraded and mitochondria that are fusing with the endogenous network. We do not know if this population has a different effect on cancer cell behavior because we did not follow this population (mostly because once the mitochondria are degraded or fuse with the network, we can no longer follow those mitochondria!). However, we did follow cancer cells that contained punctate macrophage mitochondria. Often times this was the only population we could observe in the cell at that time, and this is the population in which we observe accumulated ROS.

    Major Discussion Point #3:The effects that are attributed to the transferred mitochondria are highly variable (figures 1F, 3A,E) and often due to a subpopulation of samples that show a few extreme values (e.g. figures 2D, 3E, S4B, S4D). This might be expected from effects that are caused by a single mitochondria (which has a small volume) that is transferred to a complete cell. This complicates the study of the transfer process and effects and should be discussed. Also, do the authors have ideas how to improve the system, to make it more robust and easier to study the effects?

    🡪The variability in the assays likely reflects the heterogeneity within the biology - Each experiment contains macrophages derived from primary monocytes that are harvested from different human blood donors! Due to the primary nature of these cells, we do expect a range of phenotypes as each donor would have a different genetic background and the monocytes were likely exposed to different environmental stimuli. In fact, even though working on this study was a giant pain due to the variability, we felt more confident about our findings because despite the heterogeneity in the system, we still observed consistent phenotypes. Below we indicate where we took a sample set and removed “outliers”, and ran the statistical tests again. The differences were still statistically significantly different, further suggesting robustness of our findings.

    However, we are always on the lookout for ways to make the system easier to study. One way that we will follow up on is using M2-like macrophages since they transfer mitochondria at a higher rate than unstimulated macrophages.

    Major Discussion Point #4: The authors conclude that the transfer of dysfunctional mitochondria generated a signal mediated by ROS that activates cell proliferation signals. The statement that "transferred mitochondria act as a signaling source that promotes cancer cell proliferation" is too strong. There is increased ROS production from mitochondria, yes, but an experiment in which ROS are decreased would be needed to properly sustain that conclusion. The title and abstract could be changed to better reflect the data.

    Specific Comment:_ ‘Furthermore, treatment with an ERK inhibitor (ERKi) was sufficient to inhibit ERK activity ‘- curious as to whether antioxidant treatment would reverse any proliferative phenotypes?_

    🡪We wish we could quench the ROS at macrophage mitochondria. We really tried. We used a combination of ROS quenchers (NAC, mitoTempo, Tempo) and ROS readouts (mitoSOX, CellRox, DCFDA, and the 2 biosensors used in our study: Grx and Orp1), and treated cells for various amounts of time, and no matter what we tried, we could not reliably detect reduction of ROS levels in the host network or the transferred mitochondria (without killing the cells, that is). Another issue that we faced was that any pharmacological treatment would have a global effect on the mitochondrial network in the recipient cells and therefore it would not be possible to distinguish effects from global inhibition of ROS versus specifically at the site of the transferred mitochondria, and we certainly observed cell death upon treatment of ROS quenchers because of this fact. We talked to a couple of ROS experts, and they indicated that this issue is not unique to us, although we unfortunately did not have viable solutions, so if people have ideas or suggestions, please let us know!!

    However, despite our failed attempts at quenching ROS, the comment that "transferred mitochondria act as a signaling source that promotes cancer cell proliferation" is too strong of a statement… well, we don’t entirely agree given that we do perform sufficiency experiments in which we_induce_ ROS and observe both proliferation and ERK signaling, so we do feel reasonably justified to provide the title that we did. However, we will continue to mull over this comment. Thanks for sharing your thoughts.

    Major Discussion Point #5:The study may benefit from more direct evidence to support its conclusion of increased proliferation after mitochondrial transfer. While the RNA-seq, flow cytometry, counting of completion of cytokinesis and dry mass measurements provided in the present study do lend some support to the proliferation hypothesis, they all seem indirect. With the biomarkers labeling the mitochondria of donor and potential recipient cells, high content imaging and tracking of cells could be used to monitor cell division. A comparison of cell division rates of transfer-positive cells and transfer-negative cells will provide a more pertinent test of whether mitochondrial transfer promotes recipient cell proliferation.

    🡪We should probably do a better job at describing the dry mass measurements (QPI, quantitative phase imaging) because we view this quantification as one of the most direct measurement to monitor cell growth/division. The approach measures the changes in dry mass as the cells prepares for cell division. So not only do we get the final readout of division (complete cytokinesis), but we also get a measure of that growth rate (the cell getting ready to divide) before cytokinesis. This is why we are so tickled to collaborate with Tom Zangle’s lab because we could finally get a direct proliferation readout in real-time. We could also use this approach to follow thousands of cells at a time, a very critical aspect since mitochondrial transfer is rare event, and therefore, we need to follow many cells to have enough statistical power to quantify the growth rates. Check out some of the Zangle lab’s other papers (PMC5866559; PMC6917840; PMC4274116), and please let us know if you disagree with us!

    Major Discussion Point #6: The authors have used such a tracking-based approach on a very small scale (n=5) to measure daughter cell growth rate. However, the data do not show a statistically significant difference between the growth rates of daughters that inherited transferred mitochondria and those who did not (Fig S3). Increasing the case number via high content imaging would help obtain sufficient data points for a reliable statistical test. In addition, as suggested above, an accounting of the daughter cells' division rate for transfer positive and negative cells would provide another line of evidence to either prove or disprove the increased proliferation rate hypothesis. The same suggestion goes to the optically induced ERK activation experiments shown in Fig3F. It is also helpful to include references that studied how ERK signaling promotes proliferation and compare the evidence here with evidence or assays used in those studies as a benchmark.

    Specific Comment:Figure S3 - There is no statistical test to check for ‘increase in their rate of change of dry mass over time versus sister cells that did not inherit macrophage mitochondria’. What are the colours indicative of in S3B? Can this be reported in the figure legend.

    🡪You are right – the tracking-based approach on daughter cells is based on a small ‘n’. However, the tracking itself is performed on 1000s of cells. It’s just that in order to capture daughter cell data, we have to find a cancer cell with macrophage mitochondria (which is only ~1% of the population), and then follow that cell until it divides, and then follow BOTH daughter cells. So, even with the 1000s of cells that we followed, we could only capture a small number of daughter cells. The colors in S3B represent each individual triads – parent and 2 daughters. We will make this info clearer in the legend.

    In terms of the optically-induced ERK activation experiments, yes, it would be great to have a higher sampling. These experiments were performed at 63x so we could reliably draw small ROIs to mimic the size of a macrophage mitochondria. While we switched to lower magnification to follow cell division, we still were limited to only a few cells for the actual photoactivation. The technical aspects of this experiment were the reason for the low sampling. Despite these limitations though, we still observed increased cell division upon mito-killerred photoactivation, which we were honestly pretty surprised (and stoked) about.

    Other specific comments:

    -Figure S1A - The authors could perhaps use a more aggressive gating strategy here, clipping closer to the 231 population described in Fig S1A - picking only the center of the cluster in the upper left of the RFP vs CD11b plot would likely not affect results but make them more convincing by unequivocally excluding macrophages.

    -Figure 1D - Not sure about the 0.2% baseline assigned for the monoculture of cancer cells (that does not have the macrophages with the Emerald mitochondria). It is determined with cytometry - I am no expert on that topic, so maybe I missed something - but it looks weird to see some cells with transfer when there is a monoculture.

    🡪Due to the variable nature of the mito-mEm signal in the recipient cancer cells (i.e. transfer of one mitochondrion vs transfer of three), we found that an overlap of 0.2% set on a fully stained monoculture control was the most accurate way to gate for the recipient cancer cells. The final gating strategies used in our study were determined by FACS-isolating populations of interest based on several different gating strategies, and directly visualizing cancer cells with macrophage mitochondria without capturing macrophages or cancer cell/macrophage fusions (which is cool, but not what we wanted). To further clarify, there is no transfer occurring in the monoculture – the overlap of mEmerald signal into the transfer gate in that control sample is likely reflective of normally occurring autofluorescence. This is a very important point, so we will make this aspect clearer in the Methods section.

    -Figure S1B - Could perhaps be an interesting follow-up question for future works re: differences between cell lines and propensities to transfer mitochondria. Did the authors attempt to use other cell lines (ie, MDCK, HeLa, iPSCs, etc)?

    🡪Great question and something that we have also been thinking about. To date the only recipient cells we have used are 231, MCF10A, and PDxO cells. This would be a great avenue for future studies.

    -Figure S1B - Did the authors see an increase in growth rate in MCF10A line despite the lower growth rate?

    🡪We have not measured the growth rate in MCF10a recipient cells but something that would be great to follow up on in future studies.

    -‘physically separated from macrophages by a 0.4μM trans-well insert’ - should this read 0.4 micrometer?

    🡪Yes, great catch.

    -Figure S1F - The authors wrote that they used a two-way ANOVA analysis, could you report the factors used for that analysis in the Figure legend.

    🡪Noted!

    -Figure 1B - It is difficult to see the arrowheads in 1B, suggest moving them so they are not covering the magenta fluorescence, have them point from a different angle, and make them more brightly colored. Insets here would help the reader. A negative control image from a monoculture would also be helpful, to ensure the GFP signal is not an artifact of culture conditions.

    🡪Thank you for your feedback – we will take note of this.

    -Figure 1F - For graphs that do not show zero (as in 1F), the bar should be omitted. In these cases the length of the bar does not reflect the average of the data (as it does in 1D).

    -Figure 3C - Please omit bar, see comment on panel 1F.

    🡪 In the case of Fig 1F, we modified the y-axis to eliminate empty space. The bar is representative of mean of the data displayed in both 1D as well as 1F, but we can add a broken y-axis to help make this point.

    -Figure 1 - Given that these data are fractions of a population (ie. can be described via a contingency table), isn't something like a Fisher's exact test a better measure of significance here?

    🡪We think you are referring to Figure 1D? If so, we thought that we could not use Fisher’s exact test because that test assumed parametric distributions (which we do not observe). We have been working with a biostatistician for our statistics, but please do let us know if we have it wrong.

    -Single cell RNA- sequencing - In the methods section the authors mention doing a differential analysis between the cells that received the mitochondria and the cells that didn’t. It might be worth introducing a figure (a heatmap or a U-MAP) relating to this analysis. Single cell sequencing would not only affirm the heterogeneity between these two populations but also help in highlighting the novel cell surface markers associated with the two populations.

    🡪Yeah, good point – we can add a UMAP.

    -‘mito-mEm+ mitochondria remained distinct from the recipient host mitochondrial network, with no detectable loss of the fluorescent signal for over 15 hours’- It is surprising that the transferred mitochondria do (or cannot) fuse with the host 231 mitochondria.

    🡪We were also initially surprised to find that the transferred mitochondria do not fuse with the host 231 network! We think that the lack of fusion is due to the fact that the transferred mitochondria do not exhibit membrane potential (which is required for mitochondrial fusion). We also think that these results open interesting lines of questioning: Why are these depolarized mitochondria not degraded? Is this an active avoidance of the mitophagy pathway? How dynamic are these punctae? Many fun and interesting questions regarding the long-lived nature of these transferred mitochondria.

    -It is unclear in these images, but the 231 mitochondria appear fragmented too. Is it possible that the mitochondrial fusion machinery (Opa1 or Mfn1/2) are inactive?

    🡪231 cells are capable of fission and fusion (PMC7275541, PMC3911914, and in our own timelapse recordings), so we think that the machinery is functional. However, we don’t know whether the 231 mitochondrial machinery changes after receipt of macrophage mitochondria. Interestingly, the references above both investigate how mitochondrial dynamics promote tumor metastasis. A fascinating future direction could include an investigation to how macrophage mitochondrial transfer influences tumor cell mitochondrial dynamics.

    -Figure 2B - What does the MTDR staining of the macrophage mitochondria prior to transfer look like? Important to check this to confirm that only the transferred mitochondria had lower membrane potential.

    -‘significantly higher ratios of oxidized:reduced protein were associated with the transferred mitochondria versus the host network’-Here too, it would be important to check the mito-Grx1-roGFP2 readout of macrophage mitochondria prior to transfer.

    🡪The way that these comments are written is as if we already know that the mitochondria are dysfunctional_before_ transfer to cancer cells. But we actually do not know if that is the case. It’s also possible that macrophage mitochondria become dysfunctional once they are in the cancer cell, which would be equally cool. So, we are actively investigating this biology.

    -Figure 2A, 2BB and S1D - How were the colocalizations assessed? Was it just a visual assessment? Given the importance of these experiments for the whole story, having a quantification of the level of colocalization with each dye would be important.

    🡪This is a good point and it should be straightforward to include a Pearsons coefficient for these markers.

    -Figure S1D - The paper makes an argument about mitochondria transferred from Macrophages (marked green) having positive DNA stain (gray), but appearing depolarized (negative TMRM stain). The image in FigS1D is peculiar, as the majority of the 231 cells' mitochondria appear to not have any DNA stain but maintain membrane potential (positive in TMRM), while some (just above the green macrophage mitochondria) do have both DNA stain and membrane potential. The authors might want to clarify whether this is a typical scenario, and if so perhaps offer an explanation as to why the 231 mitochondria exhibit such heterogeneity.

    🡪The images in S1D are of a single z-plane image therefore the DNA signal in the endogenous network is more readily visible in planes that are not shown.

    -‘we confirmed that 91% of transferred mitochondria were not encapsulated by a membranous structure, thus excluding sequestration as a mechanism for explaining the lack of degradation or interaction with the endogenous mitochondrial network’ - This is based on co-staining with MemBrite 640/660, which is a dye that "covalently labels the surface of live cells", thus there is a concern as to whether this approach allows to study whether the mitochondrium is encapsulated by an endomembrane.

    🡪Thank you for your feedback. We actually do think that Membrite can label endomembrane in addition to the plasma membrane. This is from the published Membrite protocol: “MemBrite™ Fix dyes are designed to be fixed shortly after staining, when they primarily localize to the plasma membrane/cell surface. Cells also can be returned to growth medium and cultured after staining, however, dye localization in live cells changes over time. Labeled membranes become internalized, so staining gradually changes from cell surface to intracellular vesicles, usually becoming mostly intracellular after about 24 hours. Internalized MemBrite™ Fix dye is usually detectable for up to 48 hours after staining, though this may vary by cell type”.

    In our hands, we found that the dye started to become internalized and labeled vesicles within the cell within a few hours of staining. The images in the panels that you refer to came from time-lapse imaging experiments of between 10-15 hours, therefore the cells have internalized the MemBrite signal allowing for the visualization of internal vesicles. Also, in other studies not in the preprint, we perfused purified mitochondrial preparations onto 231 cells. The 231 cells took up the mitochondria from the environment, and all of these engulfed mitochondria were surrounded by a MemBrite positive membrane! These results further suggest that if the transferred mitochondria were encapsulated by a membrane, we would be able to visualize it.

    _-‘macrophage mitochondria are depolarized but remain in the recipient cancer cell’ -_Did the authors examine the extent of cancer cell death in their co-culture system (due to the activation of apoptosis by the depolarized mitochondria)?

    🡪We do not find any evidence of abnormal levels of cell death by both flow cytometry assays as well through our QPI image analysis.

    -Figure 2C–D - Like in Fig 2B, in the bottom left of panel of Fig 2C there are a lot of donor mitochondria not in highly oxidized state and the growth/proliferation phenotypes apply mostly to donor mitochondria that appear 'clumpy'.

    -Perhaps it is worth commenting on whether there is a link between donor mitochondrial morphology and the suspected proliferation-enhancing phenotype.

    🡪The images in Fig. 2C are of the same cell – a single recipient cancer cell which is expressing the Grx biosensor. The donor mitochondria are labeled with an arrowhead, the rest of the yellow/green signal (bottom right) is from the endogenous host network and therefore we do not expect it to be in a highly oxidized state (ie. more yellow than green).

    Regarding the mito morphology and proliferation – great question, and one that we are actively working on!

    -‘At 24 hours, we observed a similar trend, but no statistically significant difference (Fig. S4D). These results indicate ROS accumulates at the site of transferred mitochondria in recipient cancer cells’ - if a specific sensor fails to show a significant oxidation at 24 hours compared mito-Grx1-roGFP2 which reports on mitochondrial glutathione redox state, does that mean there are ROS independent ways to oxidize Glutathione? The authors did see cell growth phenotype both in 24 and 48 hours which suggests that something is happening in 24 hours despite no significant difference in ROS H2O2 sensor.

    🡪The additional biosensor that we used – mito-Orp1-roGFP2 - has been engineered to be a readout of one type of ROS – H2O2. The Grx probe is a surrogate for ROS of any type, of which there are many! To us, it is not completely unexpected that they would behave differently over time since they are readout for two separate things, and it generates an interesting possibility that different types of ROS accumulate over time. Given that the Grx probe shows an increase at 24 hours, which is when we observe the proliferation phenotype, we think we are on the right track. If you have ideas on robust ways to directly observe specific types of ROS, we would love to know!

    -The differences in ratio for the two sensors used are not very convincing. In Fig 2D and Fig S4B and D the “host” and “transfer” populations are very similar. The difference seems only due to the presence of a few outliers in the “transfer” populations. More importantly, sometimes it seems that these outliers come mostly from one donor rather than being present in all 3 donors. It could be good to show histograms of the two populations for each replicate/donor and maybe redo the stats excluding these outliers.

    🡪We think that the heterogeneity that is observed is due to the biology in the system – we are using primary macrophages derived from blood donors. However, for the data represented in Fig 2D, just as a test case, we took out the top four “outliers” in that data set and re-ran the Wilcoxon matched-pairs signed rank test and the p-value was 0.0010 (***), further suggesting that the ROS biosensors are revealing consistent and robust results.

    -Figure S5C - it seems like the percentage of cells that divided is the same for unstimulated cells and cells with stimulated mito-KillerRed. Isn't this contrary to the expectation? The figure shows that photobleaching cytoplasm decreased % cell division, which is puzzling.

    🡪The mean percent of cells that divided in unstimulated and mito bleach are very similar and was not significantly different. One point to be made that may not be well illustrated in our graphical representation is that if you look at the matched data (points connected are averaged per FOV for each condition in the same experiment) the trend shows that the mito bleach does seem to have an increase in cell division which is washed out with the average bar overlay. We should note that this experiment is very “noisy” and therefore we needed a lot of N to be able to detect significant changes. We are currently thinking about other ways to demonstrate sufficiency as it relates to cell proliferation – any experimental suggestions would be very welcome! Thanks for the feedback.

    -Figure 3A - In the 'cyto' condition 6 out of 13 fields have no cells that divide. Is that expected? What is the percentage of dividing cells for cells that were not illuminated at all (a control that is lacking)? There is large variation, ranging from 0% to 22%. The evidence that illumination of KillerRed leads to increased proliferation is rather weak. Also, since Cyto and Mito are different cells, is a "paired" statistical test the right kind of test to use here?

    🡪Additional data pertaining to Fig. 3A can be found in Fig. S5C, which includes the control for cells not illuminated at all. Having no cells that divide in a field of view is not surprising to us – the doubling time for these cells is ~35 hours, and we imaged for 18 hours. Also, for each field of view, our ‘n’ for each field of view was often 6-8 cells because we performed these experiments at 63X to allow for accurately drawn regions of interest for photoactivation. We also internally controlled every experiment (each experiment consisted of fields of view that had either mito activation, cyto activation, or no-activation controls, all of which were imaged overnight with multiple x/y positions). Cells that left the field of view over the 18 hours of imaging could not be quantified. It’s this sampling that caused the large variation in the graph. But again, as with many of our experiments, despite this variability, we still observe a significant difference in our experimental conditions over control cyto bleach. As for the statistical test, our understanding is that given each experiment is internally controlled, and we compare within each experiment, a paired statistical test is appropriate here. We will consult with our biostatistician to confirm, though.

    -‘ROS induces several downstream signaling pathways’ - We would not expect the authors to investigate every signaling pathway, but wonder if the PI3K pathway was explored? It seems to be the other major cancer/proliferative pathway induced by ROS.

    🡪Yes, this is a very good point! We actually assessed three different pathways at first – ERK, PI3K-AKT, and NLRP3/inflammasome. While analyzing these 3 pathways simultaneously, we discovered that ERK inhibitors resulted in decreased proliferation in cancer cells with macrophage mitochondria. As a result, we then focused on the ERK pathway. We still do not know if PI3K-AKT or NLRP3/inflammasome pathways play a role in this biology because we have not gone back and revisited these experiments yet, however in figure 3F, ERKi treated recipient cells exhibit a partial ‘rescue’ of baseline proliferation. This suggests that other pathways may indeed be involved and we plan to investigate this possibility.

    -‘Recipient 231 cells had significantly higher cytoplasmic to nuclear (C/N) ERK-KTR ratios compared to cells that did not receive transfer’-Since two different quantification styles with opposite fraction values were used, is it possible to please specify which one was used here.

    🡪Will do!

    -Figure 3B - Please show the outlines of the nuclei and that of the cell.

    🡪That would be helpful, wouldn’t it? Will do!

    -Figure 3D - it is peculiar that ERK-KTR in Fig 3D is so strongly cytosolic while in Fig 3B it is almost exclusively nuclear. If this sensor behaves differently in different situations, the authors may want to comment on how that would affect their conclusions.

    🡪The panels in Fig. 3B were taken with the ImageStream flow cytometer which takes a lower resolution image of a single plane of a cell in suspension in the flow stream. In Fig. 3D, those images are from confocal spinning disk microscopy which allows for higher resolution, z-stack images of adherent cells on glass. Therefore, we think the differences that you point out are likely due to the fact that the two images come from very different imaging systems.

    -Figure 3E - The effect of 'opto-induced' ERK activity is weak. The initial ERK-KTR is 1 at time point zero (as the data is normalized to this timepoint) and around 1 for both the cyto and mito condition. A statistical difference is observed, but the effect is minor and it is unclear whether it is biologically meaningful. The 'cyto' condition shows an average below 1 and the mito condition remains 1, suggesting that ERK activity remains constant when ROS are produced in the mitochondria.

    -Also from S8C and 3E it appears cyto actually shows a decrease rather than mito showing an increase, could the authors comment on this?

    🡪We have a few thoughts on this. The first is that we don’t expect a dramatic change in ERK signaling because the ROS accumulation is localized to a small region in the recipient cell. This is not a situation where we would expect a large-scale change because we are adding a growth factor. We can understand that the change in ERK activity may appear to be minor, but our data suggest that these subtle changes in kinase signaling translate into significant changes in downstream behavior – proliferation. The way that we interpret differences as “biological meaningful” is whether they exhibit a functional response, and in our study, we show that inhibiting the induction of ERK activity in cancer cells with macrophage mitos inhibits proliferation. What is most interesting to us is that cancer cells that do not have macrophage mitochondria have an unchanged fraction of cells in G2/M phase of the cell cycle in response to the concentration of ERK inhibitor we used, suggesting that the ERK inhibition specifically blocks macrophage mitochondria-induced proliferation.

    In Fig. S8C, bleaching a region of cytoplasm does seem to cause a decrease in ERK activity over time. We really can’t explain this result. However, we do think that ERK activity is higher in mito-bleached cells because mt-ROS is generating an increase in ERK activity which compensates for the decrease in activity that occurs when the cytoplasmic region of interest is photobleached. It’s still a head scratcher, though, but we did perform internal controls for every experiment (as we describe above), and the mito-bleach, cyto-bleach, and no-bleach conditions were run side-by-side such that we can make apples-to-apples comparisons.

    -‘patient-derived xenografts (PDxOs)’ - As a control it would be relevant to include a normal mammary organoid model perhaps from the same patient to demonstrate that the transfer of mitochondria specifically to the cancer cells is more beneficial.

    🡪Using a normal mammary organoid cells as a control to compare efficiency of transfer and downstream phenotypes would be very interesting. Due to the fact that these are patient-derived organoids, we are unable to acquire non-malignant cells from the same patient. Expanding our studies in the MCF10A cell line that we utilized in this paper would be an alternative to what you propose and would also expand our understanding of general biology underlying mitochondrial transfer.

    -‘macrophages to both HCI-037 and HCI-038 PDxO cells (Fig. 4G)’ - Why is M0 able to transfer efficiently to HCL-037 tumour when its mitochondrial network is less fragmented as M2?

    🡪These results really stood out to us. It was quite surprising that in HCI-037, both M0 and M2 macrophages were able to transfer their mitochondria at similar efficiencies, but in HCI-038, M2 macrophages were more efficient at transfer. HCI-037 is a primary tumor, and HCI-038 is a metastases from the same patient, so there are some exciting avenues of study to examine how macrophage mitochondria transfer differs at the primary versus metastatic site. There is still very little known about how donor cell dynamics influence mitochondrial transfer!

    -Are mito transfer from M0 depolarised and accumulate ROS or show increased ERK activity or increased cell proliferation?

    🡪Yes – all studies, except studies pertinent to fig 4 (where we assessed macrophage differentiation states), were done with M0 macrophages.

    -‘M2-like macrophages preferentially transferred mitochondria to the bone metastasis PDxO cells (HCI-038) when compared to primary breast tumor PDxO cells (HCI-037)’ -The authors may want to check this statement here as it is in consistent with their data plot. In Fig. 4G, M2/PDxO transfer percentages for HCI-037 and HCI-038 are about the same, unless the authors provide statistical tests to prove otherwise. Instead, M0 appears to transfer mitochondria to HCI-037 much more efficiently than it does HCI-038.

    🡪Upon re-reading our sentence again, we now realize that it’s actually quite poorly written, so we can understand the confusion! What we meant to articulate is that M2-like macrophages are better at transferring mitochondria to HCI-038 than M0 macrophages. Whereas in HCI-037, we do not observe the same preferential transfer (ie. M0 and M2 can transfer at the same efficiency). We will certainly clarify this language in the manuscript.

    -‘M2-like macrophages exhibit mitochondrial fragmentation’ - Is there a correlation between the status of the mitochondrial network in the donor and the % of transfer to the recipient? If so, this would be a correlation that would support the conclusions.

    🡪Yes, please see Fig. 4C for transfer rates with different donor subtypes and Fig. 4H for a general working model on how we think these data fit into the larger picture.

    -‘accumulate ROS, leading to increased ERK activity’ - Did the authors obtain similar results with the PDXOs? It would be an interesting observation if the primary samples also exhibit a mechanism similar to established cell lines wherein there are more accumulated genetic changes.

    🡪Our main limitation with PDxOs is overcoming the technical hurdles related to our downstream assays. These include introducing relevant reporters and generating stable lines in the PDxOs, and imaging at high-resolution when the PDxOs are cultured in 3D. However, we are very interested in this question as well, and are actively thinking about ways to overcome these hurdles.

    -It would also be interesting to examine whether there is any difference in the ROS-ERK mechanism for primary and metastatic tumour.

    🡪We agree and this is an active avenue of investigation for us. We agree and are currently pursing models to understand how our findings fit into the larger picture of tumorigenesis and metastatic potential. We had spent months pursuing an_in vivo_ approach using a murine Cre/LoxP system to genetically label mouse macrophage mitochondria with GFP. We crossed mice which express Cre under a monocyte-specific promoter (Jax, SN: 004781) and mice with germline expression of Lox-Stop-Lox-3xHA-EGFP-OMP25 (Jax, SN: 032290) with the expectation of seeing Cre-based excision of the stop cassette – thus resulting in offspring with macrophages expressing mitochondrial-localized GFP. However, the macrophages of the resulting offspring do not express GFP (by flow cytometry, imaging, and western blot analysis), despite the PCR-verified presence of both transgenes and the excision of the stop cassette. Needless to say, this was quite frustrating! We are currently in the process of developing a newly available MitoTag model which has been optimized for visualization purposes (Jax, SN: 032675). If you have any suggestions or advice on this matter we would much appreciate your thoughts!

    -‘in cancer cells that receive exogenous mitochondria’ - Since these macrophages also transfer mitochondria to non-malignant cells, such as MCF10A cells shown in Fig S1B, perhaps the authors could comment on whether this is part of a physiological process that would also promote normal cell growth?

    🡪 There are so many questions regarding when and why macrophages might transfer mitochondria. In general, mitochondrial transfer is observed in stressed cells. Our data suggest that transfer happens to MCF10A cells although at a much lower rate than their malignant counterparts, 231 cells, but we do not know whether similar downstream mechanisms and phenotypes are also occurring in the non-malignant cells. Thanks for your feedback – more to come here!

    Read the original source
    Was this evaluation helpful?
  2. This review reflects comments and contributions by Vaishnavi Ananthanarayanan, Xianrui Cheng, Joachim Goedhart, Arthur Molines, Eric Peterman, Sanjana Pillay, Pablo Ranea-Robles, Mugdha Sathe and Zara Weinberg

    The study by Kidwell et al. reports that lateral transfer of mitochondria from macrophages to breast cancer cells activates signaling pathways that promote cell proliferation. The study uses single-cell imaging strategies to show transfer of fluorescently labeled mitochondria; the transfer is rare, but it is convincingly demonstrated. The study reports that transferred mitochondria were depolarized and accumulate ROS, which signaled through ERK and promoted proliferation of the recipient cells as well as their daughter cells who also inherited the transferred mitochondria.

    The report sheds light on the mechanism by which macrophages stimulate cancer cell proliferation in the tumor microenvironment, which has strong implications in the cancer field. One of the novelties of this study is the fact that the transferred mitochondria are rather dysfunctional (depolarized), in contrast with most of the previous reports on mitochondrial transfer, which was thought to be a way to "rejuvenate" the mitochondrial pool of a cell.

    The authors make some interesting observations and the methods are described in detail. With the current data, there are questions as to whether all effects are due to biological changes, and there may be a need to re-evaluate some of the data and nuance the interpretation of the results.

    General comments are outlined below followed by annotations/specific comments on manuscript content (in order of appearance).

    • One of the most important claims is that mitochondria are the organelles responsible for the activation of the signals of cell proliferation. However, a previous report by the last author reported that macrophages transfer cytoplasm to recipient cells. It cannot be excluded that other organelles or cellular fragments are transferred as well and contribute to the observed effects (ERK activity). Perhaps a good way to solve this would be the use of macrophages that are devoid of mitochondria. At least, this aspect should be discussed in the manuscript.
    • Most of the positive examples of transferred mitochondria discussed appeared in a small clump. However, there also appears to be another population that was more diffuse and co-localizes with host mitochondria (e.g., Fig2B, bottom right panels). It would be helpful to show results of these sibling mitochondria for assays performed on their clumpy siblings. If they behave differently, it would be helpful to provide some explanation.
    • The effects that are attributed to the transferred mitochondria are highly variable (figures 1F, 3A,E) and often due to a subpopulation of samples that show a few extreme values (e.g. figures 2D, 3E, S4B, S4D). This might be expected from effects that are caused by a single mitochondria (which has a small volume) that is transferred to a complete cell. This complicates the study of the transfer process and effects and should be discussed. Also, do the authors have ideas how to improve the system, to make it more robust and easier to study the effects?
    • The authors conclude that the transfer of dysfunctional mitochondria generated a signal mediated by ROS that activates cell proliferation signals. The statement that "transferred mitochondria act as a signaling source that promotes cancer cell proliferation" is too strong. There is increased ROS production from mitochondria, yes, but an experiment in which ROS are decreased would be needed to properly sustain that conclusion. The title and abstract could be changed to better reflect the data.
    • The study may benefit from more direct evidence to support its conclusion of increased proliferation after mitochondrial transfer. While the RNA-seq, flow cytometry, counting of completion of cytokinesis and dry mass measurements provided in the present study do lend some support to the proliferation hypothesis, they all seem indirect. With the biomarkers labeling the mitochondria of donor and potential recipient cells, high content imaging and tracking of cells could be used to monitor cell division. A comparison of cell division rates of transfer-positive cells and transfer-negative cells will provide a more pertinent test of whether mitochondrial transfer promotes recipient cell proliferation.

    The authors have used such a tracking-based approach on a very small scale (n=5) to measure daughter cell growth rate. However, the data do not show a statistically significant difference between the growth rates of daughters that inherited transferred mitochondria and those who did not (Fig S3). Increasing the case number via high content imaging would help obtain sufficient data points for a reliable statistical test. In addition, as suggested above, an accounting of the daughter cells' division rate for transfer positive and negative cells would provide another line of evidence to either prove or disprove the increased proliferation rate hypothesis. The same suggestion goes to the optically induced ERK activation experiments shown in Fig3F. It is also helpful to include references that studied how ERK signaling promotes proliferation and compare the evidence here with evidence or assays used in those studies as a benchmark.

    Specific comments

    Figure S1A - The authors could perhaps use a more aggressive gating strategy here, clipping closer to the 231 population described in Fig S1A - picking only the center of the cluster in the upper left of the RFP vs CD11b plot would likely not affect results but make them more convincing by unequivocally excluding macrophages.

    Figure S1B - Could perhaps be an interesting follow-up question for future works re: differences between cell lines and propensities to transfer mitochondria. Did the authors attempt to use other cell lines (ie, MDCK, HeLa, iPSCs, etc)?

    Figure S1B - Did the authors see an increase in growth rate in MCF10A line despite the lower growth rate?

    ‘physically separated from macrophages by a 0.4μM trans-well insert’ - should this read 0.4 micrometer?

    Figure S1F - The authors wrote that they used a two-way ANOVA analysis, could you report the factors used for that analysis in the Figure legend.

    Figure 1B - It is difficult to see the arrowheads in 1B, suggest moving them so they are not covering the magenta fluorescence, have them point from a different angle, and make them more brightly colored. Insets here would help the reader. A negative control image from a monoculture would also be helpful, to ensure the GFP signal is not an artifact of culture conditions. Figure 1D - Not sure about the 0.2% baseline assigned for the monoculture of cancer cells (that does not have the macrophages with the Emerald mitochondria). It is determined with cytometry - I am no expert on that topic, so maybe I missed something - but it looks weird to see some cells with transfer when there is a monoculture.

    Figure 1F - For graphs that do not show zero (as in 1F), the bar should be omitted. In these cases the length of the bar does not reflect the average of the data (as it does in 1D).

    Figure 1 - ​​Given that these data are fractions of a population (i.e. can be described via a contingency table), isn't something like a Fisher's exact test a better measure of significance here?

    Single cell RNA- sequencing - In the methods section the authors mention doing a differential analysis between the cells that received the mitochondria and the cells that didn’t. It might be worth introducing a figure (a heatmap or a U-MAP) relating to this analysis. Single cell sequencing would not only affirm the heterogeneity between these two populations but also help in highlighting the novel cell surface markers associated with the two populations.

    Figure S3 - There is no statistical test to check for ‘increase in their rate of change of dry mass over time versus sister cells that did not inherit macrophage mitochondria’. What are the colours indicative of in S3B? Can this be reported in the figure legend. ‘mito-mEm+ mitochondria remained distinct from the recipient host mitochondrial network, with no detectable loss of the fluorescent signal for over 15 hours’ - It is surprising that the transferred mitochondria do (or cannot) fuse with the host 231 mitochondria. It is unclear in these images, but the 231 mitochondria appear fragmented too. Is it possible that the mitochondrial fusion machinery (Opa1 or Mfn1/2) are inactive?

    Figure 2B - What does the MTDR staining of the macrophage mitochondria prior to transfer look like? Important to check this to confirm that only the transferred mitochondria had lower membrane potential.

    Figure 2A, 2BB and S1D - How were the colocalizations assessed? Was it just a visual assessment? Given the importance of these experiments for the whole story, having a quantification of the level of colocalization with each dye would be important.

    Figure S1D - The paper makes an argument about mitochondria transferred from Macrophages (marked green) having positive DNA stain (gray), but appearing depolarized (negative TMRM stain). The image in FigS1D is peculiar, as the majority of the 231 cells' mitochondria appear to not have any DNA stain but maintain membrane potential (positive in TMRM), while some (just above the green macrophage mitochondria) do have both DNA stain and membrane potential. The authors might want to clarify whether this is a typical scenario, and if so perhaps offer an explanation as to why the 231 mitochondria exhibit such heterogeneity.

    ‘we confirmed that 91% of transferred mitochondria were not encapsulated by a membranous structure, thus excluding sequestration as a mechanism for explaining the lack of degradation or interaction with the endogenous mitochondrial network’ - This is based on co-staining with MemBrite 640/660, which is a dye that "covalently labels the surface of live cells", thus there is a concern as to whether this approach allows to study whether the mitochondrium is encapsulated by an endomembrane.

    Figure 2 ‘Majority (57%) of donated mitochondria do not colocalize with LysoTracker signal (N=24 cells, 4 donors) - Here the paper implies that some transferred mitochondria do co-localize with lysoTracker signal. More importantly, they co-localize with host mitochondria. It raises the question of whether they signal through ROS and ERK like their clumpy siblings who are in the limelight of most figures.

    ‘macrophage mitochondria are depolarized but remain in the recipient cancer cell’ - Did the authors examine the extent of cancer cell death in their co-culture system (due to the activation of apoptosis by the depolarized mitochondria)?

    ‘significantly higher ratios of oxidized:reduced protein were associated with the transferred mitochondria versus the host network’ - Here too, it would be important to check the mito-Grx1-roGFP2 readout of macrophage mitochondria prior to transfer.

    Figure 2C–D - Like in Fig 2B, in the bottom left of panel of Fig 2C there are a lot of donor mitochondria not in highly oxidized state and the growth/proliferation phenotypes apply mostly to donor mitochondria that appear 'clumpy'. Perhaps it is worth commenting on whether there is a link between donor mitochondrial morphology and the suspected proliferation-enhancing phenotype.

    ‘At 24 hours, we observed a similar trend, but no statistically significant difference (Fig. S4D). These results indicate ROS accumulates at the site of transferred mitochondria in recipient cancer cells’ - if a specific sensor fails to show a significant oxidation at 24 hours compared mito-Grx1-roGFP2 which reports on mitochondrial glutathione redox state, does that mean there are ROS independent ways to oxidize Glutathione? The authors did see cell growth phenotype both in 24 and 48 hours which suggests that something is happening in 24 hours despite no significant difference in ROS H2O2 sensor.

    The differences in ratio for the two sensors used are not very convincing. In Fig 2D and Fig S4B and D the “host” and “transfer” populations are very similar. The difference seems only due to the presence of a few outliers in the “transfer” populations. More importantly, sometimes it seems that these outliers come mostly from one donor rather than being present in all 3 donors. It could be good to show histograms of the two populations for each replicate/donor and maybe redo the stats excluding these outliers.

    Figure S5C - it seems like the percentage of cells that divided is the same for unstimulated cells and cells with stimulated mito-KillerRed. Isn't this contrary to the expectation? The figure shows that photobleaching cytoplasm decreased % cell division, which is puzzling.

    ‘ROS induces several downstream signaling pathways’ - We would not expect the authors to investigate every signaling pathway, but wonder if the PI3K pathway was explored? It seems to be the other major cancer/proliferative pathway induced by ROS.

    ‘Recipient 231 cells had significantly higher cytoplasmic to nuclear (C/N) ERK-KTR ratios compared to cells that did not receive transfer’ - Since two different quantification styles with opposite fraction values were used, is it possible to please specify which one was used here.

    Figure 3A - ​​In the 'cyto' condition 6 out of 13 fields have no cells that divide. Is that expected? What is the percentage of dividing cells for cells that were not illuminated at all (a control that is lacking)? There is large variation, ranging from 0% to 22%. The evidence that illumination of KillerRed leads to increased proliferation is rather weak. Also, since Cyto and Mito are different cells, is a "paired" statistical test the right kind of test to use here?

    Figure 3B - Please show the outlines of the nuclei and that of the cell.

    Figure 3C - Please omit bar, see comment on panel 1F.

    Figure 3D - it is peculiar that ERK-KTR in Fig 3D is so strongly cytosolic while in Fig 3B it is almost exclusively nuclear. If this sensor behaves differently in different situations, the authors may want to comment on how that would affect their conclusions.

    Figure 3E - The effect of 'opto-induced' ERK activity is weak. The initial ERK-KTR is 1 at time point zero (as the data is normalized to this timepoint) and around 1 for both the cyto and mito condition. A statistical difference is observed, but the effect is minor and it is unclear whether it is biologically meaningful. The 'cyto' condition shows an average below 1 and the mito condition remains 1, suggesting that ERK activity remains constant when ROS are produced in the mitochondria. Also from S8C and 3E it appears cyto actually shows a decrease rather than mito showing an increase, could the authors comment on this?

    ‘Furthermore, treatment with an ERK inhibitor (ERKi) was sufficient to inhibit ERK activity ‘- curious as to whether antioxidant treatment would reverse any proliferative phenotypes?

    ‘patient-derived xenografts (PDxOs)’ - As a control it would be relevant to include a normal mammary organoid model perhaps from the same patient to demonstrate that the transfer of mitochondria specifically to the cancer cells is more beneficial.

    ‘macrophages to both HCI-037 and HCI-038 PDxO cells (Fig. 4G)’ - Why is M0 able to transfer efficiently to HCL-037 tumour when its mitochondrial network is less fragmented as M2? Are mito transfer from M0 depolarised and accumulate ROS or show increased ERK activity or increased cell proliferation?

    ‘M2-like macrophages preferentially transferred mitochondria to the bone metastasis PDxO cells (HCI-038) when compared to primary breast tumor PDxO cells (HCI-037)’ -The authors may want to check this statement here as it is in consistent with their data plot. In Fig. 4G, M2/PDxO transfer percentages for HCI-037 and HCI-038 are about the same, unless the authors provide statistical tests to prove otherwise. Instead, M0 appears to transfer mitochondria to HCI-037 much more efficiently than it does HCI-038.

    ‘M2-like macrophages exhibit mitochondrial fragmentation’ - Is there a correlation between the status of the mitochondrial network in the donor and the % of transfer to the recipient? If so, this would be a correlation that would support the conclusions.

    ‘accumulate ROS, leading to increased ERK activity’ - Did the authors obtain similar results with the PDXOs? It would be an interesting observation if the primary samples also exhibit a mechanism similar to established cell lines wherein there are more accumulated genetic changes. It would also be interesting to examine whether there is any difference in the ROS-ERK mechanism for primary and metastatic tumour.

    ‘in cancer cells that receive exogenous mitochondria’ - Since these macrophages also transfer mitochondria to non-malignant cells, such as MCF10A cells shown in Fig S1B, perhaps the authors could comment on whether this is part of a physiological process that would also promote normal cell growth?

    Read the original source
    Was this evaluation helpful?