fMetastatic potential in clonal melanoma cells is driven by a rare, early-invading subpopulation

  1. Amanpreet Kaur
  2. Luciann Cuenca
  3. Karun Kiani
  4. Gianna T. Busch
  5. Dylan Fingerman
  6. Margaret C. Dunagin
  7. Jingxin Li
  8. Ian Dardani
  9. Eric M. Sanford
  10. Jordan Pemberton
  11. Yogesh Goyal
  12. Ashani T. Weeraratna
  13. Meenhard Herlyn
  14. Arjun Raj

Reviewed by

Read the full article

Listed in

Log in to save this article

Abstract

Metastasis occurs when tumor cells leave the primary tumor site and disseminate to distal organs. Even though most cells remain in the primary tumor, the circumstances by which a small fraction of them disseminate remain unclear. Here, we show that a rare, highly invasive subpopulation of melanoma cells can be detected within clonal cell lines due to non-genetic fluctuations in gene expression. The highly invasive phenotype was intrinsic to the cells, independent of their environment, and was marked by transiently high levels of SEMA3C expression, as revealed by RNA-sequencing analysis. Furthermore, the invasive subpopulation drove the bulk dissemination of tumor cells to distal locations in a mouse model of melanoma. The transcription factor NKX2.2 regulated the proportion of invasive cells in the melanoma 1205Lu cell line. Furthermore, an overall tradeoff between proliferation and invasion in single cells was observed. Our results suggest that phenotypes like metastasis may arise from intrinsic differences stemming from non-genetic fluctuations between single cells.

Article activity feed

  1. Version published to 10.1101/2022.04.17.488591v6 on bioRxiv
  2. Version published to 10.1101/2022.04.17.488591v5 on bioRxiv
  3. Version published to 10.1101/2022.04.17.488591v4 on bioRxiv
  4. Review Commons

    Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Reply to the reviewers

    We would like to thank the reviewers for their thorough and positive assessment of our work. We also thank them for their careful review of our manuscript. Our responses to their specific comments are provided in the lines below.

    __Reviewer #1 (Evidence, reproducibility and clarity (Required)): __

    Summary:

    The manuscript entitled „Metastatic potential in clonal melanoma cells is driven by a rare, early-invading subpopulation" by Kaur and colleagues provides a phenotypical analysis of the invasive potential of established melanoma cell lines on single cell level. The aim of the study was to answer the question if even homologous tumor cells bear the intrinsic potential to give rise to cells with high invasive (and therefore potentially metastatic) capacity in absence of selection pressure from the tumor microenvironment.

    The authors used clones from two different melanoma cell lines (to prevent the accumulation of random (epi)genetic changes during cultivation) and performed invasion assays with Matrigel-coated transwell inlays to differentiate between cells that were able to invade early (up to 8 h, approx. 1% of the total cell population) or late (8-24 h; approx. 3% of the total cell population) after plating. Comparative RNA sequencing of early invaders and non-invaders populations revealed a high expression of SEMA3C in early invaders, which was then established as marker in the used cell lines. Interestingly, in vivo models using NSG mice injected with a mixture of early and late invading melanoma cells revealed that both contributed similarly to the primary tumor, while metastatic cells in the lung consisted almost exclusively of early invaders. Subsequent ATAC sequencing revealed an increase of binding sites for the transcription factor NKX2.2 in the early invaders. Functional analyses revealed that a knockout of NKX2.2. led to an increase in both invasion and proliferation. Finally, the authors showed with different sorted early and late invaders as well as SEMA3Chigh and SEMA3Clow expressers that pro-invasive features go along with reduced proliferation potential in accordance to previously published data. However, they decrease with time, thus demonstrating a reversion of the phenotype and high plasticity.

    Major comments:

    In general, the paper contains novel and interesting data, is concisely written and supported by replicates. The key conclusion, the presence of a small proportion of highly invasive cells in a seemingly homologous cell population and their striking requirement for lung metastasis, is very convincing. In vitro, SEMA3C was confirmed as a marker for the early invaders in two independent cell lines. However, a few questions remain open, as detailed below:

    We thank the reviewer for their positive assessment of our work. We also thank them for their careful review of our manuscript. Our responses to their specific comments are provided in the lines below.

    1. The relevance of NKX2.2 in the early invaders is currently unclear to me.

    The ATAC sequencing data revealed a high enrichment of accessible NKX2.2 binding sites in early invaders, and data were tested by comparative RNA sequencing of control cells and cells with NKX2.2 ko (Figure 2). The Figure legend of Figure 2 says: "NKX2.2 is a transcription factor that promotes the invasive subpopulation", but the data don`t support this (ko leads to reduced invasion). Accordingly, the authors also state in the Results part "... the direction of the effect is the opposite of what one might have expected".

    To set the role of NKX2.2 into context, it would be useful to confirm the actual involvement of NFX2.2 in the invasive phenotype and clarify if NFX2.2. might probably even suppress some pro-invasive genes. I would advise to investigate the protein levels and/or protein localization of NFX2.2 and probably perform ChIp experiments on selected pro-invasive genes that play a role in the early invaders.

    The reviewer has raised some excellent points about our studies of NKX2.2 and its role in invasion. Indeed, we were also surprised by the fact that NKX2.2 had the opposite effect as expected (its peaks are enriched for accessibility in the early invaders in FS4, but knockout leads to increased invasion). We elected to include the results because it was a hypothesis we tested, so in the interest of full disclosure of results, we chose to leave the result in.

    The reviewer has also made some nice suggestions about how to further explore the role of NKX2.2 in regulation (e.g. ChIP-seq). Owing to the complexity of validating and performing this assay, we felt these experiments were beyond the scope of the current manuscript; we hope to explore these possibilities more fully in the future.

    Another excellent suggestion the reviewer made was to look at the regulatory capacity of NKX2.2 to directly demonstrate the link between NKX2.2 regulation and expression differences between early- and late-invading cells. In order to establish this connection, we used a gene set from molecular signatures database (MSigDB: https://www.gsea-msigdb.org/gsea/msigdb/human/geneset/NKX2_2_TARGET_GENES.html) consisting of genes with an NKX2.2 binding site within their promoter (TSS -1000 bp to TSS +100 bp) identified by the gene transcription regulation database (GTRD–paper here: https://pubmed.ncbi.nlm.nih.gov/33231677/). We used the Fisher’s exact test to see if the overlap between these genes regulated by NKX2.2 and genes that are differentially expressed between early-invading cells versus their respective parental population in both cell lines had more overlap than one would expect by chance. Indeed, the p-values using this approach were 3.937e-16 and 0.037 for the FS4 and 1205Lu cell lines, respectively. These results, combined with the motif analysis with our ATAC-seq data, demonstrated that the activity of NKX2.2 is relevant in the early-invading state. We thank the reviewer for the suggestion and feel this additional analysis has improved our conclusions about NKX2.2.

    Also, we further checked whether NKX2.2 levels correlated in early versus late invading cells across a panel of cell lines (Fig. 2C). We found that in 4/6 of these lines, NKX2.2 expression was higher in the early invaders. These results further support the case that NKX2.2 is an important positive regulator of invasion in multiple contexts.

    “In order to establish the generality of our results, we measured NKX2.2 expression levels across multiple cell lines by single molecule mRNA FISH. We found that the early invaders had higher levels of NKX2.2 expression in four out of the 6 lines tested (Fig. 2C), demonstrating the generality of our results and strengthening the case that NKX2.2 is a potential regulator of early invasiveness. The role of NKX2.2 as a regulator of early invasiveness was further established through comparative analysis between genes with NKX2.2 promoter region binding sites (-1000 bp to +100 bp relative to the transcription start site (TSS) as annotated by the Gene Transcription Regulation Database (GTRD)) and genes differentially expressed in early-invading and parental cells. Analysis using Fisher's exact test revealed a significant overlap between GTRD annotated genes regulated by NKX2.2 and genes expressed in FS4 (****p=3.937e-16) and 1205Lu (*p=0.037) early-invading cells. These results, in complement with our results from ATAC-sequencing motif analysis, further supported the relevance of NKX2.2 regulation in the early-invading state.”

    1. The sequencing data are currently accessible via a Dropbox link. They should be deposited instead in a data repository.

    We thank the reviewer for noting this problem. We have uploaded all data to the SRA/GEO at the following links:

    https://urldefense.com/v3/https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE224772;!!IBzWLUs!SEr5DTViPf08-IBQnv0ml-CoLX3cbaiNlCz-DJbpIKm7UcVXlL9-OD9reVQJs5pm_gzeqJYC_dM-MV8DonwX4c4$

    https://urldefense.com/v3/https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE224769;!!IBzWLUs!SEr5DTViPf08-IBQnv0ml-CoLX3cbaiNlCz-DJbpIKm7UcVXlL9-OD9reVQJs5pm_gzeqJYC_dM-MV8DtY6ZB3A$

    https://urldefense.com/v3/https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE224771;!!IBzWLUs!SEr5DTViPf08-IBQnv0ml-CoLX3cbaiNlCz-DJbpIKm7UcVXlL9-OD9reVQJs5pm_gzeqJYC_dM-MV8Dq_3ghAU$

    Minor comments:

    1. The cell line used for Supplementary Figure 4 should be named in the figure legend.

    We thank the reviewer for the suggestion. We have included the name of the cell line in the figure legend for Supplementary Figure 4. The text reads as follows:

    “A. FS4 melanoma cells were sorted based on SEMA3C expression. Cells were live-imaged for ~10 days every hour and single cells were tracked manually for cell position, cell division and lineage. Lineages were traced manually from single cells. Cell speed was calculated for each cell using the average distance traveled over time.”

    1. In Figures 4H-M and Supplementary Figure 4D-I, the authors describe data performed in "sister" and "cousin" cells. It would be useful to provide a definition for both in the main text or figure legend.

    This is a very good point. We have provided the following definitions in the main text, and have changed the wording from “sister” to “sibling” to avoid gendered terminology:

    “(sibling cells are defined as those that share a common parent cell, and cousin cells are defined as those that share a common grandparent.)”

    1. Discussion: "This lack of permanence may reflect the fact that the invasive cells are not subjected to stress-in our case, cells merely pass through a transwell, which may be the reason for the "burning in" of the phenotype in the case of resistance."

    This sentence is misleading - please clarify.

    We apologize for the confusion caused by this sentence. We have now changed it to the following:

    “It is interesting that the early-invading cells eventually revert to the population average even after going through the transwell. Such a result contrasts with our previous work (Shaffer et al., 2017b), in which a rare subpopulation became permanently therapy resistant and did not revert even after several weeks off-treatment. One possibility is that the stress of undergoing therapy treatment induces a transcriptional rewiring, and this rewiring is not induced by the migration through transwells. Further studies will be required to test these hypotheses.”

    Furthermore, there are some errors in the reference to the Figures throughout the paper. These which should be corrected:

    We thank the reviewer for their detailed reading and finding these issues. We have now fixed them all in our revised manuscript.

    1. Results, section "NKX2.2 is a transcription factor that promotes the invasive subpopulation".

    Here the authors write: "...we performed RNA sequencing on the NKX2.2 knockout cells and compared the effects on gene expression to the gene expression differences between early vs. non- invaders across the two cell lines." This sentence should contain the reference to Supplementary Figure 3B-D (which is otherwise not referred to).

    We thank the reviewer for their detailed reading and noticing this issue. We have now referenced Supplementary Figure 3B-D in the text cited above.

    1. Results: "Overexpression of SEMA3C in FS4 cells revealed no changes in invasiveness, suggesting that SEMA3C is a marker with no functional relevance to invasiveness per se; Fig. 1D, Fig. 2A-B)"

    The correct reference should be: Suppl. Fig. 1D, Fig. 2A-B. Also, in the current manuscript version the authors jump from Figures 1 to Figure 2 A,B, before coming back to Figure 1. To avoid this, I would advise to shift the current Figure 2A, B to Figure 1 or the supplementary information.

    We thank the reviewer for pointing out this error in the reference to these figures. Figure 2A-B is now referenced as “Supp. Fig. 1 E-F”. The figure legend has also been updated.

    1. Results: "We then sampled lungs from mice at various times post-injection to look for metastatic cells (Fig.1F, Suppl. Fig. 2B,C)."

    As Supplementary Figure 2B, C does not show metastasis, but rather primary tumor growth, I would advise the following wording: "We then sampled lungs from mice at various times post-injection to look for metastatic cells (Fig.1F) and overall tumor growth (Suppl. Fig. 2B,C)."

    We thank the reviewer for their advice to reword the sentence cited above. We have now edited the text to read as suggested by the reviewer. In addition, Supp. Fig. 2B,C is not referenced as Supp. Fig. 2C,D.

    "We then sampled lungs from mice at various times post-injection to look for metastatic cells (Fig.1F) and overall tumor growth (Supp. Fig. 2C,D)."

    1. Results: "Interestingly, NKX2.2 knockout cells showed markedly increased invasion and proliferation (Fig. 2A,B), suggesting a change in regulation of both processes. "

    The correct reference is Fig. 2C, D.

    The reviewer is right that we only have results in one cell line, and fully agree that the results in FS4 are only correlative. We have now weakened the language in the abstract and the results to emphasize that this result held in 1205Lu cells only.

    • Given the robust literature regarding phenotypic switching in melanoma, the NKX2.2 knockout increasing both invasiveness and proliferation (figures 2C, 2D) suggests it may not be involved in phenotype switching. Perhaps NKX2.2 is a negative regulator of cell activity/metabolism. We thank the reviewer for highlighting the possible connections with metabolism. To explore this possibility , we performed metabolic assays on NKX2.2 knockout and AAVS control cells and observed no significant changes in Extracellular acidification rate (B). We did observe some differences in oxygen consumption rate in the cells (A), but the differences do not seem to be large enough or systematic enough to be meaningful given the variation within the controls. We have now included these results in Supp. Fig. 3E-F.

    Note, the data previously referenced as Figure 2C,D is now in Figure 2A,B.

    “NKX2.2 is a transcriptional repressor and activator essential for the differentiation of pancreatic endocrine cells (Habener et al., 2005). In mice, deletion of NKX2.2 prevents the specification of pancreatic islet cells resulting in the replacement of insulin-expressing β cells and glucagon-expressing α cells with ghrelin-expressing cells; This lack of specification resulted in mortality of newborn mice due to hyperglycemia (Sussel et al. 1998; Prado et al. 2004). Given the link of NKX2.2 with glucose metabolism, we wondered whether NKX2.2 had an effect on metabolic activity prompting us to test the NKX2.2 knockout lines for metabolic differences in the oxygen consumption rate (OCR; an indicator of oxidative phosphorylation) and the extracellular acidification rate (ECAR; an indicator of glycolysis) of the cells. Seahorse assay analysis revealed no systematic differences in metabolic activity (Supp. Fig. 3E,F).”

    We thank the reviewer for the correction. The reference has now been corrected in the main text.

    __Reviewer #1 (Significance (Required)): __

    Nature and significance of the advance/ literature context:

    In their manuscript, the authors provide interesting biological data about the presence of intrinsically and reversibly pro-invasive / pro-metastatic melanoma cells in a seemingly homogenous subpopulation. With SEMA3C, they also provide a marker for early invading cells, which might be useful in future studies to identify therapeutic vulnerabilities for this subgroup. This study sheds further light on the functional effects of phenotypic plasticity, which was previously described particularly in the context of therapy resistance, as mentioned by the authors.

    We thank the reviewer for their kind assessment of the impact of our work.

    __Audience: __

    The study is interesting for scientists from the melanoma field as well as the cancer metastasis field in general.

    Own expertise:

    Melanoma, phenotypic switch, metabolism, signal transduction, stress response

    __Reviewer #2 (Evidence, reproducibility and clarity (Required)): __

    Metastatic potential in clonal melanoma cells is driven by a rare, early-invading subpopulation

    Kaur et al.

    In this manuscript the authors highlight a small subpopulation of "early-invading" melanoma cells and functionally characterize the nuances of these early cells compared to their slowly invading counterparts. A cell surface marker, SEMA3C and the transcription factor NKX2.2 were associated with differences in the invasive rates. Importantly, the group demonstrates that existence of the invasive subpopulation is not reliant on genetic changes, and thus exhibits plasticity. While the underlying concept surrounding this paper (phenotypic plasticity) is not novel, highlighting a surface marker and transcription factor that may, at least in part, be associated with phenotype plasticity is interesting. However, the current study seems underdeveloped. Specific points of concern are listed:

    Major

    • Only two cell lines are used throughout this study. We thank the reviewer for pointing out the need for more cell lines. We have now added two new cell lines to our study, WM793 and WM1799, both of which recapitulate the fundamental phenomenology in question. Although we did not show it in our initial submission, we had originally queried a panel of melanoma cell lines in order to determine their suitability for our study (from which we settled on 1205Lu and FS4). This panel has multiple melanoma cell lines obtained from a variety of melanoma tumor samples from Radial Growth Phase (RGP), Vertical Growth Phase (VGP), and metastatic tissues. We now have included these data in our revised manuscript, since they further support our point.

    “We tested a panel of different melanoma cell lines from Radial Growth Phase (RGP), Vertical Growth Phase (VGP), and metastatic tumor types for the existence of fast invading subpopulations. We used four patient-derived melanoma cell lines, FS4, 1205Lu, WM1799, WM793, all of which have BRAF mutations (V600K for FS4, V600E for 1205Lu, WM1799, and WM793) and are known to be highly invasive in vitro and in vivo (27). Out of the 11 melanoma cell lines tested, the FS4 (not shown) and 1205lu cell lines displayed the highest levels of fast invading subpopulations (Supp. Fig. 1A).”

    First, we showed that they all have an invasive subpopulation, with 1205Lu and FS4 (not shown) having the most invasive cells. Second, validating a central claim of the manuscript, we showed that many of these cell lines, including WM1799 and WM793, showed much higher levels of both SEMA3C (4/6) and NKX2.2 (4/6) expression in the early invading population as compared to the late invading population.

    Together, these data make a strong case that our findings generalize across multiple cell lines, including RGP and VGP models. We have incorporated new text that reads as follows:

    “In order to establish the generality of our results, we measured expression of the surface marker SEMA3C across the early and late invading subpopulations of a panel of melanoma cell lines. We found that SEMA3C levels were higher in the early invading subpopulation in 4 of the 6 lines tested (Supp. Fig. 1H). Thus, these results held across a variety of cell lines and, thus, were not a unique feature of a particular patient sample.”

    • The in vivo metastasis assay in figure 1 is difficult to interpret and presents a number of concerns.
    1. Only ~50% of early invading cells were labeled with GFP, this confounds many aspects of the experiment. The authors comment that in the primary tumor, as expected "...a roughly equal mix of human melanoma cells that were GFP positive and negative." If there was an expectation of equal proliferative rates in the primary tumor of early and late invading cells, given that only 1/2 of the early cells were GFP+, wouldn't we expect only 25% of the human cells to be GFP+?

    The reviewer has raised a very important quantitative question about our experiments, which we have now addressed with a more thorough set of analyses. Initially, we quantified GFP positivity post -transduction by looking at fluorescent protein levels, for which the threshold was fairly arbitrary, and potentially could have miscounted many GFP positive cells as GFP negative due to low but non-zero levels of expression. We hence recalculated our positivity rate based on single molecule RNA FISH for GFP and mCherry, given that the technique is sensitive down to even veryl ow levels of expression.

    As can be seen in Supp. Fig. 2B, the vast majority of transduced cells did indeed get the transgene and had some level of expression of GFP/mCherry. At a threshold of 5/10 molecules (GFP/mCherry, respectively), we obtained 88% and 96.15% positivity rates for GFP and mCherry, respectively. At these rates of positivity, we would expect much closer to 50% of the cells being GFP positive in the tumors, as observed. We thank the reviewer for noticing this discrepancy, and feel that our new analysis clears up the confusion and strengthens our results. These results are described in the main text as follows:

    “We labeled the cells with sufficient virus so that 88% of the early invaders were labeled with GFP and 96.15% of the late invaders were labeled with mCherry (Supp. Fig. 2B). We then sampled lungs from mice at various times post-injection to look for metastatic cells (Fig.1F) and overall tumor growth (Supp. Fig. 2C,D).”

    1. The authors note technical difficulties in detecting mCherry in sections. It seems as though this forced them to use a RNA FISH probe to identify human vs. mouse and by extension/negative selection the human FISH positive, GPF negative cell represented a mCherry stained late-invading cell. This is not ideal and seems over complicated. If the population of interest was engineered to express mCherry, why not directly probe for mCherry?

    The reviewer has raised an important point about our experimental design. Indeed, we attempted multiple times and in multiple ways to detect mCherry protein directly. We tried multiple times with multiple antibodies, but the signal was simply not detectable. Hence, we arrived at the experimental design we outlined. We felt that a fully transparent disclosure of the issues was preferable, even if it did make the design sound overly complex. We will note that our primary result—that the vast majority of the metastatic cells are GFP positive and hence derived from fast invaders—is robust to any detection issues for mCherry.

    1. Given the poor initial labeling/transduction of the early invaders, how can the authors be confident that all human cells without GFP signal are late invaders?

    The reviewer raises a great point that is addressed by our GFP and mCherry RNA FISH analysis above, showing that the transduction efficiency was actually quite a bit higher than initially thought due to low but non-zero GFP signal being counted as GFP negative. With the much higher transduction efficiencies we have now validated, we believe that the vast majority of human cells with no GFP signal should be late invaders.

    • The authors may have missed an opportunity to study FS4 clone F6 and 1205 clone E11. What is the SEMA3C and NKX2.2 status of these clones? Are they able to revert expressions? The reviewer has pointed out an interesting opportunity for further exploration. Unfortunately, because they were identified as part of an initial screening study, those particular clones were not kept for subsequent analysis. However, in our revised manuscript, we have now worked up multiple additional cell lines (WM1799 and WM793), both of which had high expression levels of both SEMA3C (Supp. Fig. 1H, shown above) and NKX2.2 (Fig. 2C) in the early invading subpopulation. Currently, we do not have data on reversion experiments for these two cell lines, but we would expect them to behave similarly to the other cell lines we examined in this study.

    • The lack of statistical analysis/comparisons throughout the paper needs to be addressed. We thank the reviewer for pointing out these deficiencies. We have now added statistical comparisons throughout.

    • In figures 1E and 3B, why do the parental (homogenous) cells demonstrate less invasiveness than the selected for the SEMA3C low or "late-invaders" respectively? This is an important point that the reviewer has raised. The finding did occur in every replicate, so we assume it is biologically and not statistical. We have now included the following language in the discussion noting the issue and some possible explanations.

    “It is worth noting that, while the SEMA3C-high (early-invading) subpopulation drove the highly invasive phenotype, the SEMA3C-low (late-invading) subpopulation also displayed a somewhat more invasive phenotype than the parental population. It is unclear what the underlying cause of this difference in invasive behavior is between the SEMA3C-low and parental populations. One possibility is that paracrine signaling between cells in the parental population confers them with less invasive potential than when the cells are isolated into early- and late-invading subpopulations. Another possibility is that technical factors associated with the sorting of SEMA3C-low cells from the parental population alter their invasive properties, thus making them distinct from the parental population.”

    • Conclusions that NKX2.2 knockout increases invasiveness and proliferation are based on 1 cell line. The comparisons done with FS4 early and late invading cells in Figure 1F may be supportive but is correlative in nature. The reviewer is right that we only have results in one cell line, and fully agree that the results in FS4 are only correlative. We have now weakened the language in the abstract and the results to emphasize that this result held in 1205Lu cells only.

    • Given the robust literature regarding phenotypic switching in melanoma, the NKX2.2 knockout increasing both invasiveness and proliferation (figures 2C, 2D) suggests it may not be involved in phenotype switching. Perhaps NKX2.2 is a negative regulator of cell activity/metabolism. We thank the reviewer for highlighting the possible connections with metabolism. To explore this possibility , we performed metabolic assays on NKX2.2 knockout and AAVS control cells and observed no significant changes in Extracellular acidification rate (B). We did observe some differences in oxygen consumption rate in the cells (A), but the differences do not seem to be large enough or systematic enough to be meaningful given the variation within the controls. We have now included these results in Supp. Fig. 3E-F.

    Note, the data previously referenced as Figure 2C,D is now in Figure 2A,B.

    “NKX2.2 is a transcriptional repressor and activator essential for the differentiation of pancreatic endocrine cells (Habener et al., 2005). In mice, deletion of NKX2.2 prevents the specification of pancreatic islet cells resulting in the replacement of insulin-expressing β cells and glucagon-expressing α cells with ghrelin-expressing cells; This lack of specification resulted in mortality of newborn mice due to hyperglycemia (Sussel et al. 1998; Prado et al. 2004). Given the link of NKX2.2 with glucose metabolism, we wondered whether NKX2.2 had an effect on metabolic activity prompting us to test the NKX2.2 knockout lines for metabolic differences in the oxygen consumption rate (OCR; an indicator of oxidative phosphorylation) and the extracellular acidification rate (ECAR; an indicator of glycolysis) of the cells. Seahorse assay analysis revealed no systematic differences in metabolic activity (Supp. Fig. 3E,F).”

    • Given that sorted SEMA3C high levels did not revert to parental FS4 levels, yet the invasive phenotype reverted to parental-like behavior undermines the usefulness of SEMA3C as a marker of invasiveness. The reviewer has brought up an important point. We were able to show that 1205Lu cells had SEMA3C levels revert to those of the parental. The reviewer is right that FS4 did not, which may be because it takes longer for FS4 to revert. It is true that the phenotypic behavior did revert. We have seen similar things in our therapy resistance work (Shaffer et al. 2017, etc.). One possible reason is that the phenotype is governed by multiple factors, and so the phenotype can revert before the expression of SEMA3C. We still think that SEMA3C is a good marker, just perhaps context dependent. We have added text to the discussion to make these important points.

    “We note that SEMA3C levels in FS4-SEMA3C-high cells did not revert to the parental levels within two weeks. This incomplete reversion may be because SEMA3C takes longer to revert than the tested time period. Interestingly, the invasive phenotype did revert in this time period, suggesting that there may be multiple factors associated with the phenotype beyond SEMA3C. It may thus be that SEMA3C is a marker of the early-invading population, but only in certain contexts.”

    Minor

    • How does SEMA3C and/or NKX2.2 expression (here 1.5% of FS4 cells were noted as "SEMA3C high") of metastatic cell lines (FS4 and 1205) compare to RGP and VGP cell lines? The reviewer has asked a great question about radial and vertical growth phase cells. We have tested several other cell lines to determine cell lines that were suitable for transwell assays. We have now included two figures (Supp. Fig. 1H and Fig. 2C) showing the SEMA3C and NKX2.2 status of each of these cell lines (parental cells) and their different subpopulations (early invaders and late invaders)—see also Reviewer #2, Major point 1. We found that the same pattern of SEMA3C-high cells held for both RGP and VGP cell lines.

    • There were a number of instances throughout the manuscript that were not clear, colloquial, or simply unnecessary - i.e. description of transwell assay. The reviewer has raised a good point about our language. We have gone through and tried to improve the clarity and precision. As for descriptions of the various assays, we have found that some readers of our papers are unfamiliar with these assays, so we elected to keep those descriptions in. We hope the reviewer does not object too strenuously.

    • The authors only analyze/mention lung metastases. Were metastases observed at other sites? The reviewer has posed a very good question about whether metastasis occurred at other locations. We stained additional tissues (liver and kidney) that were collected from the same mice and stained as per our lung invasion assays. As shown in our new Supplemental Fig. 2E, we found a similar pattern with the vast majority of metastatic cells being GFP positive; i.e., early-invaders, just as was the case for lung. We thank the reviewer for this helpful suggestion.

    “In the lung, however, we saw predominantly GFP-positive cells, showing that the vast majority of cells that migrated from the primary tumor site were initially early invading cells (Fig. 1I,J). The number of GFP cells in the lung was variable, but generally increased with time. The liver and kidney also showed an enrichment of GFP-positive cells (early invaders), suggesting that the metastatic potential of these cells is not limited to any one particular metastatic location (Supp. Fig. 2E). Thus, we established that the highly invasive subpopulation was able to drive metastasis in vivo.”

    • What is PE indicating in Figure 1D? Apologies, PE refers to the channel we used for the sorting on the FACS machine and stands for “Phycoerythrin”. To avoid any confusion, we have omitted the “PE” text on the y-axis of Fig. 1D.

    • The number of invaded cells seems to vary quite a bit between experiments - Parental 1205 cells in Fig 2C = ~200, yet 1205 clone F6 and the non-clonal 1205 cell line demonstrate ~10,000. Similar differences observed with Fs4 cells - Parental Fig 1E vs. Empty control Figure 2A. The reviewer has a good eye—indeed, there is a wide variability in the amount of invading cells. We have now remarked on this variability in the results section:

    “We note that the number of invading cells varied significantly between experiments. This variability is due to the fact that we employed transwell dishes with different growth areas, ranging from 0.33 cm2 to 4.67 cm2, leading us to collect different cell numbers for individual experiments. The cell density per cm2, however, was kept constant between experiments.”

    Note that Figure 2C and Figure 2A are now referenced as Figure 2A and Supplemental Figure 1F, respectively .

    __Reviewer #2 (Significance (Required)): __

    This work contributes to the growing fields of phenotypic plasticity and intratumoral heterogeneity. The authors claim to have identified a surface marker SEMA3C and a transcription factor NKX2.2 that may play a role in driving invasive proclivity. Importantly, the group demonstrates that changes in these proteins are not genetic, and therefore represent "intrinsic differences" that are a property of the tumor. Furthermore, the authors indicate how the present observations of early invading cells parallels drug resistance phenomena as their previous works highlights intrinsically resistant subpopulations (Shaffer et al., Nature 2017, Torre et al., Nature Genetics 2021 and others.). Taken together, the current and previous work underscores the importance of cell to cell non-genetic variability in disease progression and response to therapy.

    We thank the reviewer for their kind comments on the significance of our manuscript.

    __Reviewer #3 (Evidence, reproducibility and clarity (Required)): __

    In this study, Kaur et al. intended to use similar strategy that the same group had developed (https://www.nature.com/articles/nature22794) to identify the subpopulation in melanoma responsible for metastasis. In brief, the melanoma cell population was subjected to the selection of a specific phenotype (transwell migration dubbed as "invasiveness" behavior). By comparing the early and late invaders, a cell maker was identified to allow distinguishing the high-invasive subpopulation. A series of experiments were devised to validate the metastatic function of the high-invasive cells and delineate the signaling that drove this phenotype. The authors concluded that this rare subpopulation was originated from transcriptional fluctuation, and invasiveness is a trade-off of cell growth. Therefore, as the cells growing, overtime the phenotype was reverted to low invasiveness.

    Consistency is the most important factor for evaluating observation over temporal and spatial range. Therefore, several controls need to be clarified before further investigation in mechanisms:

    1. If the rare invader cells are arising from gene expression fluctuation, the SEMA3C-low population of parental line should generate SEMA3C-high invader subpopulation over time. This should be addressed.

    The reviewer has made an excellent point. Indeed, it is the case that the SEMA3C-low population starts to regenerate the high invader subpopulation over time. We have re-graphed Figure 3D to demonstrate this fact more clearly (See Supplemental Fig. 5A,B), showing that the SEMA3C low population regenerates many more SEMA-3C high cells after 14 days.

    1. Both early and late invader cells exhibited higher invasiveness than the parental line (Fig. 3B). Therefore, the in vivo metastatic potential of the three lines should be compared to validate the role of the invader cells in the metastatic function.

    We thank the reviewer for their comment about testing all three populations in the in vivo context. It is an excellent suggestion, but in order to fully control the experiment, we would need to add all three populations in three separate colors. Given the difficulties we had with getting even the two colors to work together, we think it is beyond the scope of our current efforts to attempt this complex experiment. We have added the following caveat to the text:

    “For unknown reasons, the parental population consistently showed lower invasiveness than the early- and late-invading subpopulations. Given that we did not test the parental population for invasiveness in vivo, future studies may address the sources and mechanisms by which the parental population differs and how those differences manifest in vivo.”

    1. To evaluate the possible intervention of cellular function by fluorescent proteins (https://doi.org/10.1016/j.ccell.2022.01.015), admix of GFP- and mCherry-labeled populations of early invader cells should be used as a control in Fig. 1F. Noticeably, the labeling ratio of the two populations was not even in Fig. 1F.

    The reviewer has brought up an important point about the potential differences brought about by the fluorescent proteins themselves. At this point, it is difficult to redo these complex in vivo experiments, but we can appeal to the fact that the admixture is maintained throughout time as the primary tumor site still has a roughly equal ratio of GFP and mCherry cells in it (Fig. 1I and Supp. Fig. 2E).

    1. When the invader cells were expanded and passed, their invasiveness will revert to the level similar to parental line in 14 days (Fig. 3B). The isolated cells were expanded for further testing and manipulation in Fig. 1C and 1F, respectively. How long did was the period for cell expansion in these experiments?

    We thank the reviewer for bringing up an important question about the details of cell expansion. For the RNA-seq, the cells were directly processed upon going through the transwell, so there was no expansion period. We have made sure to outline this more carefully in our methods section (see below).

    “RNA sequencing and analysis:

    RNA collection and library prep: Each treatment/sample was tested in 3 separate biological replicates. Upon passing through the transwell, cells were immediately collected and processed for RNA sequencing. Total RNA isolation was performed using the phenol-chloroform extraction followed by RNA cleanup using RNAeasy Micro (Qiagen 74004) kit. For transwell assays, library preparation was performed using Nebnext single-cell/low input RNA library prep kit (E6420L, NEB). For NKX2.2 CRISPR experiments, library preparation was done using NEBNext Poly(A) mRNA Magnetic Isolation Module (NEB E7490L) integrated with NEBNext Ultra II RNA Library Prep Kit for Illumina (NEB E7770L).

    Mouse tumor implantation and growth:

    All mouse experiments were conducted in collaboration with Dr. Meenhard Herlyn at The Wistar Institute, Philadelphia, PA. NSG mice (NOD.Cg-Prkdcscid* Il2rg*tm1Wjl/SzJ) were bred in-house at The Wistar Institute Animal Facility. All experiments were performed under approval from the Wistar Institute Care and Use Committee (protocol 201174). As in the case of RNA sequencing experiments, cells were not expanded prior to injection into the mouse, but were collected and implanted right after passing through the transwell. 50,000 melanoma cells were suspended in DMEM with 10% FBS and injected subcutaneously in the left flank of the mouse.”

    1. If invasiveness and growth are trade-off, why did the mCherry-labeled cells not dominate the population of primary tumors in Fig. 1J?

    Note that Figure 1J is now referenced as Figure 1I. The reviewer brings up a good point. For potential explanations, first, the difference in growth rate is not large, so we would not necessarily expect mCherry cells to dominate on this timescale. Also, we believe that in vivo, the tradeoff may be mitigated by other factors and cell-cell interactions that are not present in vitro. We have added a note on this point to the results.

    “(Note that these numbers were similar despite the slightly increased growth rate of the late-invading subpopulation; we assume this is due to the relatively small difference and cell-cell interactions that could prevent one population from dominating the other.)”

    1. In Fig. 1G, why RNA FISH was not used to detect mCherry-labeled cells?

    Another excellent point. RNA FISH in tissue sections can often be rather challenging due to various reasons including RNA degradation, and mCherry RNA signal was hard to definitively show in these sections. Hence, we opted for MALAT1, which is very heavily expressed and hence provided a strong and reliable signal.

    “For technical reasons, the mCherry cells were not detectable due to the fluorescence of the mCherry protein not being visible in the mouse sections. Nevertheless, we were able to detect late invaders in the population by using a human-specific MALAT1 RNA FISH probe that binds only to human MALAT1 RNA and not mouse MALAT1 RNA (28).”

    1. In vivo cycling (harvesting the cells from metastatic site and implanting them to the primary site in mouse models) has been employed to select metastatic sublines from a parental line. Could in vivo cycling make the early invader phenotype fixed?

    The reviewer has raised a very interesting point about cycling and selection. Indeed, the 1205Lu cells were derived from repeated cycling of invasive lung cells. That is probably the reason that these cells were useful for our assay, because the percentage of early-invading cells was higher. Nevertheless, the cells still have a significant proportion of late invaders, suggesting that the phenotype has not yet been fixed in the population. Perhaps with further cycling, such a fixation could be achieved. We have now noted this possibility in our discussion.

    “It is also possible that repeated cycles of selection, even of non-genetic phenotypes, could lead to an increased fraction of invasive cells. Indeed, 1205Lu cells were derived by exactly such repeated cycles, which presumably are the reason they have a higher percentage of invasive cells; however, despite these repeated rounds of selection, most cells are still not highly invasive, suggesting that it is difficult for this property to fully fix in the population.”

    __**Referees cross-commenting** __

    Both reviewers' questions are important for adequate controls.

    __Reviewer #3 (Significance (Required)): __

    There are several studies trying to identify subpopulation responsible for the metastasis of melanoma and other types of cancer, and a few mechanisms have been revealed. However, the significance depends on if the results can be validated on clinical data. It is lacking in this study.

    We thank the reviewer for their statement of interest in the problem. We agree that it is helpful to link these results to clinical data. We did perform TCGA analyses of several different genes, including SEMA3C, that emerged from our data, and there were no systematic relationships to phenotype. Of course, the relationship to clinical data is complex and many important factors are not obvious from the TCGA data, so we do not think that necessarily diminishes our results. Rather, we think our results raise a conceptual point that there can be rare cells with non-genetic differences that can drive metastasis. Further work will be required to translate these results to the clinic.

    We have added the following to the main text:

    “We found that the SEMA3C-high cells were far more invasive, intrinsically, than SEMA3C-low cells and the population overall, thus demonstrating that cells vary intrinsically in their invasiveness, and the very invasive subpopulation is marked by the expression of SEMA3C (Fig. 1E). Note, overexpression of SEMA3C in FS4 single cell clones revealed no changes in invasiveness, suggesting that SEMA3C is a marker with no functional relevance to invasiveness per se (Fig. 1D; Supp. Fig. 1E-G). We verified the expression levels of the genes identified in our RNA sequencing study in the The Cancer Genome Atlas (TCGA) data. We combined the list of differentially expressed genes in early invaders with the gene set enrichment analysis (GSEA) “Hallmarks of cancer epithelial-mesenchymal transition” and compared expression in primary vs. metastatic TCGA samples, finding no appreciable difference (Fig. 5A-B). These data suggest that these markers do not have obvious clinical correlates. Moreover, Kaplan Meier analysis comparing the survival time (days to death) between patient cohorts with either high or low SEMA3C expression levels revealed that SEMA3C does not predict survival time post-diagnosis, as both survival curves (p=0.898) follow comparable trends between the two cohorts (Fig. 5C). However, conceptually, our results raise the possibility that a rare, non-genetically defined subpopulation of cells may drive metastasis due to its increased degree of invasiveness, which further data collection efforts in patient samples may help validate.”

  5. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #3

    Evidence, reproducibility and clarity

    In this study, Kaur et al. intended to use similar strategy that the same group had developed (https://www.nature.com/articles/nature22794) to identify the subpopulation in melanoma responsible for metastasis. In brief, the melanoma cell population was subjected to the selection of a specific phenotype (transwell migration dubbed as "invasiveness" behavior). By comparing the early and late invaders, a cell maker was identified to allow distinguishing the high-invasive subpopulation. A series of experiments were devised to validate the metastatic function of the high-invasive cells and delineate the signaling that drove this phenotype. The authors concluded that this rare subpopulation was originated from transcriptional fluctuation, and invasiveness is a trade-off of cell growth. Therefore, as the cells growing, overtime the phenotype was reverted to low invasiveness.

    Consistency is the most important factor for evaluating observation over temporal and spatial range. Therefore, several controls need to be clarified before further investigation in mechanisms:

    1. If the rare invader cells are arising from gene expression fluctuation, the SEMA3C-low population of parental line should generate SEMA3C-high invader subpopulation over time. This should be addressed.
    2. Both early and late invader cells exhibited higher invasiveness than the parental line (Fig. 3B). Therefore, the in vivo metastatic potential of the three lines should be compared to validate the role of the invader cells in the metastatic function.
    3. To evaluate the possible intervention of cellular function by fluorescent proteins (https://doi.org/10.1016/j.ccell.2022.01.015), admix of GFP- and mCherry-labeled populations of early invader cells should be used as a control in Fig. 1F. Noticably, the labeling ratio of the two population was not even in Fig. 1F.
    4. When the invader cells were expanded and passed, their invasiveness will revert to the level similar to parental line in 14 days (Fig. 3B). The isolated cells were expanded for further testing and manipulation in Fig. 1C and 1F, respectively. How long did was the period for cell expansion in these experiments?
    5. If invasiveness and growth are trade-off, why did the mCherry-labeled cells not dominate the population of primary tumors in Fig. 1J?
    6. In Fig. 1G, why RNA FISH was not used to detect mCherry-labeled cells?
    7. In vivo cycling (harvesting the cells from metastatic site and implanting them to the primary site in mouse models) has been employed to select metastatic sublines from a parental line. Could in vivo cycling make the early invader phenotype fixed?

    Referees cross-commenting

    Both reviewers' questions are important for adequate controls.

    Significance

    There are several studies trying to identify subpopulation responsible for the metastasis of melanoma and other types of cancer, and a few mechanisms have been revealed. However, the significance depends on if the results can be validated on clinical data. It is lacking in this study.

  6. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #2

    Evidence, reproducibility and clarity

    RC-2022-01474 Metastatic potential in clonal melanoma cells is driven by a rare, early-invading subpopulation Kaur et al.

    In this manuscript the authors highlight a small subpopulation of "early-invading" melanoma cells and functionally characterize the nuances of these early cells compared to their slowly invading counterparts. A cell surface marker, SEMA3C and the transcription factor NKX2.2 were associated with differences in the invasive rates. Importantly, the group demonstrates that existence of the invasive subpopulation is not reliant on genetic changes, and thus exhibits plasticity. While the underlying concept surrounding this paper (phenotypic plasticity) is not novel, highlighting a surface marker and transcription factor that may, at least in part, be associated with phenotype plasticity is interesting. However, the current study seems underdeveloped. Specific points of concern are listed:

    Major

    • Only two cell lines are used throughout this study.
    • The in vivo metastasis assay in figure 1 is difficult to interpret and presents a number of concerns.
      • 1.) Only ~50% of early invading cells were labeled with GFP, this confounds many aspects of the experiment. The authors comment that in the primary tumor, as expected "...a roughly equal mix of human melanoma cells that were GFP positive and negative." If there was an expectation of equal proliferative rates in the primary tumor of early and late invading cells, given that only 1/2 of the early cells were GFP+, wouldn't we expect only 25% of the human cells to be GFP+?
      • 2.) The authors note technical difficulties in detecting mCherry in sections. It seems as though this forced them to use a RNA FISH probe to identify human vs. mouse and by extension/negative selection the human FISH positive, GPF negative cell represented a mCherry stained late-invading cell. This is not ideal and seems over complicated. If the population of interest was engineered to express mCherry, why not directly probe for mCherry?
      • 3.) Given the poor initial labeling/transduction of the early invaders, how can the authors be confident that all human cells without GFP signal are late invaders?
    • The authors may have missed an opportunity to study FS4 clone F6 and 1205 clone E11. What is the SEMA3C and NKX2.2 status of these clones? Are they able to revert expressions?
    • The lack of statistical analysis/comparisons throughout the paper needs to be addressed.
    • In figures 1E and 3B, why do the parental (homogenous) cells demonstrate less invasiveness than the selected for the SEMA3C low or "late-invaders" respectively?
    • Conclusions that NKX2.2 knockout increases invasiveness and proliferation are based on 1 cell line. The comparisons done with FS4 early and late invading cells in Figure 1F may be supportive but is correlative in nature.
    • Given the robust literature regarding phenotypic switching in melanoma, the NKX2.2 knockout increasing both invasiveness and proliferation (figures 2C, 2D) suggests it may not be involved in phenotype switching. Perhaps NKX2.2 is a negative regulator of cell activity/metabolism.
    • Given that sorted SEMA3C high levels did not revert to parental FS4 levels, yet the invasive phenotype reverted to parental-like behavior undermines the usefulness of SEMA3C as a marker of invasiveness.

    Minor

    • How does SEMA3C and/or NKX2.2 expression (here 1.5% of FS4 cells were noted as "SEMA3C high") of metastatic cell lines (FS4 and 1205) compare to RGP and VGP cell lines?
    • There were a number of instances throughout the manuscript that were not clear, colloquial, or simply unnecessary - i.e. description of transwell assay.
    • The authors only analyze/mention lung metastases. Were metastases observed at other sites?
    • What is PE indicating in Figure 1D?
    • The number of invaded cells seems to vary quite a bit between experiments - Parental 1205 cells in Fig 2C = ~200, yet 1205 clone F6 and the non-clonal 1205 cell line demonstrate ~10,000. Similar differences observed with Fs4 cells - Parental Fig 1E vs. Empty control Figure 2A.

    Significance

    This work contributes to the growing fields of phenotypic plasticity and intratumoral heterogeneity. The authors claim to have identified a surface marker SEMA3C and a transcription factor NKX2.2 that may play a role in driving invasive proclivity. Importantly, the group demonstrates that changes in these proteins is not genetic, and therefore represents "intrinsic differences" that is a property of the tumor. Furthermore, the authors indicate how the present observations of early invading cells parallels drug resistance phenomena as their previous works highlights intrinsically resistant subpopulations (Shaffer et al., Nature 2017, Torre et al., Nature Genetics 2021 and others.). Taken together, the current and previous work underscores the importance of cell to cell non-genetic variability in disease progression and response to therapy.

  7. Review Commons

    Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

    Learn more at Review Commons

    Referee #1

    Evidence, reproducibility and clarity

    Summary:

    The manuscript entitled „Metastatic potential in clonal melanoma cells is driven by a rare, early-invading subpopulation" by Kaur and colleagues provides a phenotypical analysis of the invasive potential of established melanoma cell lines on single cell level. The aim of the study was to answer the question if even homologous tumor cells bear the intrinsic potential to give rise to cells with high invasive (and therefore potentially metastatic) capacity in absence of selection pressure from the tumor microenvironment. The authors used clones from two different melanoma cell lines (to prevent the accumulation of random (epi)genetic changes during cultivation) and performed invasion assays with Matrigel-coated transwell inlays to differentiate between cells that were able to invade early (up to 8 h, approx. 1% of the total cell population) or late (8-24 h; approx. 3% of the total cell population) after plating. Comparative RNA sequencing of early invaders and non-invaders populations revealed a high expression of SEMA3C in early invaders, which was then established as marker in the used cell lines. Interestingly, in vivo models using NSG mice injected with a mixture of early and late invading melanoma cells revealed that both contributed similarly to the primary tumor, while metastatic cells in the lung consisted almost exclusively of early invaders. Subsequent ATAC sequencing revealed an increase of binding sites for the transcription factor NKX2.2 in the early invaders. Functional analyses revealed that a knockout of NKX2.2. led to an increase in both invasion and proliferation. Finally, the authors showed with different sorted early and late invaders as well as SEMA3Chigh and SEMA3Clow expressers that pro-invasive features go along with reduced proliferation potential in accordance to previously published data. However, they decrease with time, thus demonstrating a reversion of the phenotype and high plasticity.

    Major comments:

    In general, the paper contains novel and interesting data, is concisely written and supported by replicates. The key conclusion, the presence of a small proportion of highly invasive cells in a seemingly homologous cell population and their striking requirement for lung metastasis, is very convincing. In vitro, SEMA3C was confirmed as marker for the early invaders in two independent cell lines. However, a few questions remain open, as detailed below:

    1. The relevance of NKX2.2 in the early invaders is currently unclear to me. The ATAC sequencing data revealed a high enrichment of accessible NKX2.2 binding sites in early invaders, and data were tested by comparative RNA sequencing of control cells and cells with NKX2.2 ko (Figure 2). The Figure legend of Figure 2 says: "NKX2.2 is a transcription factor that promotes the invasive subpopulation", but the data don`t support this (ko leads to reduced invasion). Accordingly, the authors also state in the Results part "... the direction of the effect is the opposite of what one might have expected". To set the role of NKX2.2 into context, it would be useful to confirm the actual involvement of NFX2.2 in the invasive phenotype and clarify if NFX2.2. might probably even suppress some pro-invasive genes. I would advise to investigate the protein levels and/or protein localization of NFX2.2 and probably perform ChIp experiments on selected pro-invasive genes that play a role in the early invaders.
    2. The sequencing data are currently accessible via a Dropbox link. They should be deposited instead in a data repository.

    Minor comments:

    1. The cell line used for Supplementary Figure 4 should be named in the figure legend.
    2. In Figures 4H-M and Supplementary Figure 4D-I, the authors describe data performed in "sister" and "cousin" cells. It would be useful to provide a definition for both in the main text or figure legend.
    3. Discussion: "This lack of permanence may reflect the fact that the invasive cells are not subjected to stress-in our case, cells merely pass through a transwell, which may be the reason for the "burning in" of the phenotype in the case of resistance." This sentence is misleading - please clarify.

    Furthermore, there are some errors in the reference to the Figures throughout the paper. These which should be corrected:

    1. Results, section "NKX2.2 is a transcription factor that promotes the invasive subpopulation". Here the authors write: "...we performed RNA sequencing on the NKX2.2 knockout cells and compared the effects on gene expression to the gene expression differences between early vs. non- invaders across the two cell lines." This sentence should contain the reference to Supplementary Figure 3B-D (which is otherwise not referred to).
    2. Results: "Overexpression of SEMA3C in FS4 cells revealed no changes in invasiveness, suggesting that SEMA3C is a marker with no functional relevance to invasiveness per se; Fig. 1D, Fig. 2A-B)" The correct reference should be: Suppl. Fig. 1D, Fig. 2A-B. Also, in the current manuscript version the authors jump from Figures 1 to Figure 2 A,B, before coming back to Figure 1. To avoid this, I would advise to shift the current Figure 2A, B to Figure 1 or the supplementary information.
    3. Results: "We then sampled lungs from mice at various times post-injection to look for metastatic cells (Fig.1F, Suppl. Fig. 2B,C)." As Supplementary Figure 2B, C does not show metastasis, but rather primary tumor growth, I would advise the following wording: "We then sampled lungs from mice at various times post-injection to look for metastatic cells (Fig.1F) and overall tumor growth (Suppl. Fig. 2B,C)."
    4. Results: "Interestingly, NKX2.2 knockout cells showed markedly increased invasion and proliferation (Fig. 2A,B), suggesting a change in regulation of both processes. " The correct reference is Fig. 2C, D.

    Significance

    Nature and significance of the advance/ literature context:

    In their manuscript, the authors provide interesting biological data about the presence of intrinsically and reversibly pro-invasive / pro-metastatic melanoma cells in a seemingly homogenous subpopulation. With SEMA3C, they also provide a marker for early invading cells, which might be useful in future studies to identify therapeutic vulnerabilities for this subgroup. This study sheds further light on the functional effects of phenotypic plasticity, which was previously described particularly in the context of therapy resistance, as mentioned by the authors.

    Audience:

    The study is interesting for scientists from the melanoma field as well as the cancer metastasis field in general.

    Own expertise:

    Melanoma, phenotypic switch, metabolism, signal transduction, stress response

  8. Version published to 10.1101/2022.04.17.488591v3 on bioRxiv
  9. Version published to 10.1101/2022.04.17.488591v2 on bioRxiv
  10. Version published to 10.1101/2022.04.17.488591v1 on bioRxiv