miR-146 connects stem cell identity with metabolism and pharmacological resistance in breast cancer

Tordonato et al. reveal miRNA-146 as a specific marker for breast stem cells and for cancer stem cells. miR-146 maintains the stem cell identity and coordinates a transcriptional and metabolic program, distinct from bulk cells, connected to the refractoriness to antifolate drugs.

In this paper Tordonato and colleagues identify miR145 as critical determinant of CSC/stem cells self-renewal in breast cancer and postulated its connection with folate metabolism thus suggesting a vulnerability of breast cancer cells that could be exploited therapeutically. The paper is of potential interest, however a few concerns remain to be addressed, find them below: MiR-146a/b expression identifies cells with SC/CSC identity: -"In BC, the proportion of cells with tumour-initiating ability (herein operationally equaled to CSCs) correlates with the molecular/biological characteristics of the tumour and its aggressiveness": do miR146a/b levels correlate with prognosis in TCGA or METABRIC data? -Isolation and analysis of CSC is normally based on a combination of assays, here the authors used PKH26 incorporation. It would be nice to see in supplemental stainings for CD44 and CD24 to estimate the purity of CS isolation. -(Minor comment) "The silencing of miR-146 in these PDXs reduced the frequency of TICs by 5-fold ( Figure 3C-F), in the absence of significant effects on proliferation", correct but it would still be nice to see a FISH for miR146 in PDX and/or a QPCR in FACS sorted CSC subpopulations.
-"miR-146 KD accelerated the conversion of the CD44high/CD24low subpopulation towards the original cell Population". The nature of the assay performed in figure 4N is not clear, it should be better explained in the text.
-"To gain mechanistic insights into miR-146 functions in the mammary SC compartment, we performed a comparative analysis of the transcriptomes": these experiments have been performed in duplicate thus making difficult a statistical evaluation of the results. -Can you retrieve the miR146 seed in the upregulated set upon miR146 inhibition and if so in which bins?
MiR-146a/b regulates quiescence and one-carbon pool metabolism -Can you identify direct targets proving this connection? -Do you detect an inverse correlation between those targets and the miRNA in the PDX tissues upon miR-146 KD? miR-146 depletion synergizes with MTX treatment -Can the effects be reverted by addition of folic acid to the cells?
Reviewer #2 (Comments to the Authors (Required)): In their study Tordonato et. al show that miR-146 is highly expressed in normal mammary stem cells (SCs) and in cancer stem cell (CSC) populations in breast tumors. The authors show that cell populations in normal mammary cell lines and in a breast cancer cell line, SUM159, that show high expression of miR-146 have a higher sphere forming efficiency (SFE) than cell populations from the same cell lines that show low expression of miR-146, arguing that miR-146 high cells have a higher capacity for self-renewal. Knock-down of miR-146 in these cells led to reduced SFE and fewer tumor initiating cells (TICs), arguing that miR-146 plays a functional role in mediating the ability of these cells to self-renew. Through comparative transcriptional analysis of miR-146-high and miR-146 low (endogenously low or low due to knock-down) cells, the authors identify pathways mediated by miR-146, including metabolic processes. Finally, the authors reason that miR-146 might mediate drug response and find that knock-down of miR-146 renders SUM159 cells sensitive to methotrexate but not to other common chemotherapeutics.
This study provides convincing evidence that miR-146 expression mediates SC/CSC phenotypes associated with self-renewal in normal mammary epithelial cells and in breast cancer cell lines. The observation that knock-down of miR-146 renders cells more sensitive to methotrexate treatment is interesting and, though it requires more extensive follow up, suggest therapies that target miR-146 could be effective in the clinic. Overall, the manuscript is well written and the experiments are straight-forward and well-controlled. However, the strength of the study would increase significantly if the authors were able to confirm the observations regarding methotrexate sensitivity in more than one tumor line. Some specific points: Figure 3 and related Figure S2: 1. In the TIC experiments using PDX lines, the number of mice used per line is inconsistent between lines --e.g. for the same dose, 4 for one line, 5 for another, and 8 for the last. The authors also do not use the same 3 doses for all 3 lines. While they use 25k, 10k, and 5k for 2 of the lines (430p and 339p), they use 50k, 25, and 5k for the third line (197p). These discrepancies should be corrected or explained.
2. The percent KI67 quantification is shown for PDX_430p and PDX_197p but not for PDX_339p. 4. The authors do not clearly describe the sensor construct in the text --e.g. NGFR is used as a reporter for the presence of the construct and for normalization. 5. In N the baseline levels of CD44hi cells for the population should be shown on the graph for reference.   figure 3, the authors use PDX lines in which miR-146 has been knocked-down. Does this knock-down also render these PDX lines more sensitive to methotrexate? 9. Does expression of miR-146 correlate with patient response to methotrexate?
10. The author's statement in reference figure S4D, is confusing --"the non-SC population did not show any change in MTX sensitivity upon miR-146 manipulation (KD or overexpression)." This suggests a context specific role for miR-146, which should be discussed briefly in the discussion.
1st Revision -Authors' Response to Reviewers: January 26, 2021 Reviewer #1: (we numbered the reviewer's comments for clarity) 1. In BC, the proportion of cells with tumour-initiating ability (herein operationally equaled to CSCs) correlates with the molecular/biological characteristics of the tumour and its aggressiveness: do miR146a/b levels correlate with prognosis in TCGA or METABRIC data?
R. Agree. We tested this possibility in the METABRIC dataset, which contains better clinical information with respect to the TGCA dataset. The level of expression of miR-146a correlated with disease outcome (HR 1.22, p = 0.04), while no differences could be observed for miR-146b. These data are shown underneath and have been included in Supplemental Figure S1K. Figure S1K: Kaplan Meier analysis of BC patients from Metabric dataset (N= 1217) according to miR-146a-b levels (HR and 95% CI were calculated in univariate analysis). Briefly, we selected from the Metabric dataset, 1217 patients for whom clinical informations were available on cBioportal. miR-146a-b levels were defined high or low, over the median value of expression in all the patients analyzed.

2.
Isolation and analysis of CSC is normally based on a combination of assays, here the authors used PKH26 incorporation. It would be nice to see in supplemental staining for CD44 and CD24 to estimate the purity of CS isolation.
R. This is an interesting point. Stem cell (SC) isolation is based on approaches that exploit certain properties of SCs, which might vary as a function of the experimental setting. In particular, PKH staining is based on the co-segregation of SC and quiescence phenotypes, and it has been developed to exploit this property in ex vivo contexts such as organoid or mammosphere cultures (see (Pece et al., 2010). Conversely, CD44/CD24 staining has been developed on primary tumors or in cultured cells, systems in which the SC phenotype segregates with the basal phenotype. Unfortunately, there is no proper context in which both can be used successfully (or readily substitute for each other). In mammospheres, where PKH is used, CD44/CD24 staining is not effective since cells cultured as mammosphere become basallike and there is no further stratification by CD44 high /CD24 low staining. PHK cannot be used on fresh tumor samples for practical reasons, and neither on cultured cells, since there is no segregation of stemness with quiescence (these are transformed cells and they all duplicate, even the stem-like subpopulation). We can support this contention with a relevant case, shown in Appendix A, included here only for the reviewer's perusal.
In the figure below, we show single-cell transcriptional profiles of 16195 cells, in which we compared the HMLE bulk population (8450 cells) with the CD44 high SC-like population (7745 cells, isolated from HMLE cells by FACS sorting). Some conclusions can be drawn. A. The single cell profiles reveal that the two cellular populations overlap to a limited extent, highlighting different subgroups (clusters) composed of CD44 high /CD24 low cells in blue as opposed to bulk cells in red. B-C-D. Some of the CD44 high /CD24 low clusters (from CL1 to CL5) are enriched in the expression of genes related to breast SCs, thereby validating the approach (i.e. EMT genes, or SC-signatures from published works). E-F. Using the expression of cell cycle genes, it is possible to estimate the cell cycle profile of each cells in each sample group (CD44 vs bulk) and in each of the clusters. No difference in cell cycle distribution exists between CD44 high stem-like population and HMLE bulk population. Therefore, PKH staining, which is based on quiescence, cannot be coupled with CD44/CD24 staining. In conclusion, the best practice is to apply each of the two purification procedures (PKH or CD44-CD24 staining) to the contexts to which they are best suited. It is worth mentioning that miR-146 was observed consistently upregulated in different contexts and using each of the two approaches: i) in human BCSCs isolated from tumor biopsies using FACS sorting based on CD44 high CD24 -/low lineage -( Figure 1F), from Shimono et al, Cell 2009 (PMID: 2731699); and in PKH + cells isolated from human and mouse mammospheres ( Figures 1B and 1F). Hence, miR-146 upregulation is consistent between the two approaches and can be considered a unifying element over the different properties used to purify SCs. 3. (Minor comment) -'The silencing of miR-146 in these PDXs reduced the frequency of TICs by 5-fold ( Figure 3C-F), in the absence of significant effects on proliferation', correct but it would still be nice to see a FISH for miR146 in PDX and/or a QPCR in FACS sorted CSC subpopulations.

R.
Agree. We tried to set up an ISH approach to measure the expression of 146 directly on primary mammary samples; unfortunately we did not obtain clear experimental data due to technical problems related to high background, poor sensitivity of the probes and the low basal expression of the two miRNAs. However, we could measure miR-146 levels in PDX upon silencing by using RT-QPCR. Results demonstrate the effective silencing of the two miRNAs in PDX tumors. These data are now included as Supplemental Figure S2F of the revised paper, and are also shown underneath.

4.
'miR-146 KD accelerated the conversion of the CD44high/CD24low subpopulation towards the original cell Population'. The nature of the assay performed in figure 4N is not clear, it should be better explained in the text.

R.
Agree. The manuscript has been revised accordingly (page 8, lines 162-166 of the revised manuscript and page 25, legend to Figure 4N).

'
To gain mechanistic insights into miR-146 functions in the mammary SC compartment, we performed a comparative analysis of the transcriptomes': these experiments have been performed in duplicate thus making difficult a statistical evaluation of the results.

R.
We apologize for the lack of clarity in our original manuscript. We actually used four independent experiments, coming from two completely independent approaches, each performed in biological duplicates. We then used all four experiments with a ranking approach to select down-and upregulated genes in response to miR-146 KD, as depicted in Figure 5G. Having used a ranking approach that divides the transcriptome into groups, the statistical evaluation was made on a categorical basis using the chi-test, whose values are shown in the same Figure 5G. The high chi-square shows that it is very unlikely that genes are regulated by chance in four independent analyses.

6.
Can you retrieve the miR146 seed in the upregulated set upon miR146 inhibition and if so in which bins?

R.
We apologize for the lack of clarity in our original version; actually, these analyses were already present in the original paper. We looked for miRNA seed in the gene set regulated by miR146 inhibition.
To this end, we used Targetscan, which estimates the probability of a gene of being a target by looking at the type of 'seed' and the overall sequence context, measured as 'context score'. The lower the score the higher is the probability of being a target. As shown in the panel below, genes predicted with high and medium probability (<-0.30 and <-0.15 context score) were effectively upregulated (A). These genes were considered as the predicted targets for miR-146 (945 genes,146 TG) and are reported in Figure  5B-C, as effectively upregulated in miR-146 low cells (miR-low) and upon miR-146 silencing (KD). In addition, we found significant enrichment of predicted targets in the first bins (bins 2-3-4, upregulated genes, B) as shown in Figure 5F. These genes were used to select for direct miRNA targets as shown in Figure 5G-H. In the panel below, we have represented the same data, in a different manner. We believe that our original descriptions are sufficient; hovewer, we would be ready to add the panel below as a Supplemental Figure, if the reviewer thinks that this would help.
A. Genes in the two datasets analyzed (miR-low and miR-146 KD, expressed in log2fold change vs. Ctrl) were divided in 5 classes according to the context score derived from Targetscan prediction (-1.74<high<-0.3, -0.3<medium<-0.15, -0.15<low<-0.1 and -0.1<very low<-0.01). Plotted are the cumulative distributions of mRNA fold changes, comparing the five different classes: the higher the context score, the stronger is the upregulation of the mRNAs with respect to the non-target category. B. The ranked list of genes in miR-146low or 146high+KD (from the most upregulated to the most repressed) were divided in bins of 1100 elements/bin. The heatmap displays, in each bin, the enrichment score of the prediction classes defined in A. MiR-146 predicted targets (n=945, 'high' or 'medium' context score groups) were enriched in the first 4 bins.

7.
MiR-146a/b regulates quiescence and one-carbon pool metabolism. Can you identify direct targets proving this connection?

R.
We thank the reviewer for the comment, and we apologize for not having clearly explained this point in the original version of the paper. Indeed, panel 5G shows the direct targets of miR-146 found in SClike cells. The targets related to one carbon pool metabolism are those highlighted in red in the figure, i.e DHFR, GART, MTHFD1, and MTHFD2. In addition, we have also verified the effects at the protein level, through a WB analysis performed on the SC-like cells (CD44 high) subpopulation of HMLE cells, in which miR-146 was silenced. These data (also shown below) are now reported in Supplemental Figure  S3E and further confirm the effects of miR-146 on enzymes of the one-carbon pool metabolism.

Do you detect an inverse correlation between those targets and the miRNA in the PDX tissues upon miR-146 KD?
R. We thank the reviewer for the comment. To answer the question, we extracted RNA from the PDX tumors (obtained with or without miR-146 silencing), and looked at the levels of target genes of the onecarbon pool (GART, MTHFD1, MTHFD2 genes) and well-known 146 targets (TRAF6, a positive control) by RT-qPCR. However, we couldn't find any anticorrelation for either the one-carbon pool genes or the positive control. There are several possible explanations for this negative result. Probably in the PDX tumors the effects are masked by the heterogeneity of the tumor population in which only a small number of CSCs is present vs. the bulk non-SC population (progenitors and differentiated cells). In these latter cells, obviously, one cannot expect any effect of miR-146 silencing, since the miRNA is expressed at very low levels. Indeed, miRNA effects on transcripts are clearly visible only in very controlled setting, by limiting the sources of transcriptional variability, and at early time points after miRNA manipulation to avoid indirect transcriptional effects.

miR-146 depletion synergizes with MTX treatment. Can the effects be reverted by addition of folic acid to the cells?
R. This is an interesting point and we agree that it would be interesting to investigate whether folic acid supplementation could reverse some of the effects of miR-146. As explained in the letter to the Editor, we would need more than a year to address this point, for regulatory reasons. Therefore, we believe that such investigation should be pursued in the context of follow-up studies.

Reviewer #2
Figure 3 and related Figure S2: 1. In the TIC experiments using PDX lines, the number of mice used per line is inconsistent between lines --e.g. for the same dose, 4 for one line, 5 for another, and 8 for the last. The authors also do not use the same 3 doses for all 3 lines. While they use 25k, 10k, and 5k for 2 of the lines (430p and 339p), they use 50k, 25, and 5k for the third line (197p). These discrepancies should be corrected or explained.

R.
We thank the reviewer for the comment. There is a technical reason related to these differences, mostly due to the fact that each PDX has different growth properties, and therefore the latency for tumor growth and the number of cells that can be purified at each passage from each single tumor are different. However, the results are correct, as we compared control and KD cells at the same dose in each of the PDXs and we derived the TIC frequency using 3 different doses. To prove that results are valid, we also re-calculated the TIC-frequency taking in consideration only the 25k and 5k doses, which are in common for all the PDXs. As shown in the panel underneath the differences between KD and CTRL are maintained with a statistical significance for all the PDXs.
2. The percent KI67 quantification is shown for PDX_430p and PDX_197p but not for PDX_339p.

R.
We thank the reviewer for the comment. Unfortunately, we were unable to perform Ki67 quantification on the histological material for PDX_339p. This particular PDX has a slow growth kinetics and typically has few epithelial cells. Most of the material was used to check for miR-146 KD by RT-QPCR, leaving a low amount of material included as FFPE for histological analyses. Unfortunately with such material we were not able to quantify the Ki67 levels properly. To acknowledge this, we have amended the manuscript as follows "when we measured proliferation effects by Ki67 staining (which was possible in in two out of three PDXs because of availability of material), we did not score differences in KD vs. SCR. (page 7, lines 139-142 of the revised manuscript). R. Agree. We have amended the manuscript with the appropriate reference (page 7, line 149 of the revised manuscript).

4.
The authors do not clearly describe the sensor construct in the text --e.g. NGFR is used as a reporter for the presence of the construct and for normalization.
R. Agree. We have described more clearly the feature of the sensor in the manuscript (page 7-8, lines 145-149 of the revised manuscript).

5.
In N the baseline levels of CD44hi cells for the population should be shown on the graph for reference. R. Agree. Baseline levels of CD44hi/CD24lo population are 95% for both SCR and 146-KD cells. This information has been added in legend to Figure 4N, and it is also reported underneath.

R.
Agree. The cells used are HMLE, we added this information to Figure 5A of the revised manuscript.

7.
The labeling for this figure is confusing. Instead of "sensor" the authors should use a label that clearly indicates what the sample is --miR-146 low cells.

R.
Agree. We have replaced the term 'sensor' with '146-low' (as compared to 146-high) in Figure 5, panels B, C, E, F, G and H. Figure S4: 8. The observation that miR-146 kd renders xenografted tumor cells sensitive to methotrexate treatment was only performed in a single tumor cell line, SUM159. In figure 3, the authors use PDX lines in which miR-146 has been knocked-down. Does this knock-down also render these PDX lines more sensitive to methotrexate?

R.
We agree that it would be interesting to further confirm these effects using PDX as well, as suggested by the reviewer. However, as explained in the letter to the Editors (and also following their suggestions), we would need more than a year to address this point, for law regulatory reasons. Therefore, we believe that such investigation should be pursued in the context of follow-up studies 9. Does expression of miR-146 correlate with patient response to methotrexate?

R.
We agree with the reviewer that this would be an interesting analysis. However, in breast cancer, MTX in not generally used either as first line of therapy, or as monotherapy in relapsed patients; it is usually associated with other drugs, typically 5-FU (Yang et al., 2020). Thus, we could not identify any suitable patient cohort to perform the experiment. 1st Revision -Editorial Decision Thank you for submitting your revised manuscript entitled "miR-146 connects stem cell identity with metabolism and pharmacological resistance in breast cancer". We would be happy to publish your paper in JCB pending final revisions necessary to meet our formatting guidelines (see details below).
To avoid unnecessary delays in the acceptance and publication of your paper, please read the following information carefully.

A. MANUSCRIPT ORGANIZATION AND FORMATTING:
Full guidelines are available on our Instructions for Authors page, https://jcb.rupress.org/submission-guidelines#revised. **Submission of a paper that does not conform to JCB guidelines will delay the acceptance of your manuscript.** 1) Text limits: Character count for Articles is < 40,000, not including spaces. Count includes title page, abstract, introduction, results, discussion, acknowledgments, and figure legends. Count does not include materials and methods, references, tables, or supplemental legends. ). Also, if you used parametric tests, please indicate if the data distribution was tested for normality (and if so, how). If not, you must state something to the effect that "Data distribution was assumed to be normal but this was not formally tested." 5) Abstract and title: The abstract should be no longer than 160 words and should communicate the significance of the paper for a general audience. The title should be less than 100 characters including spaces. Make the title concise but accessible to a general readership.
6) Materials and methods: Should be comprehensive and not simply reference a previous publication for details on how an experiment was performed. Please provide full descriptions in the text for readers who may not have access to referenced manuscripts. 7) Please be sure to provide the sequences for all of your primers/oligos and RNAi constructs in the materials and methods. You must also indicate in the methods the source, species, and catalog numbers (where appropriate) for all of your antibodies. Please also indicate the acquisition and quantification methods for immunoblotting/western blots. 11) eTOC summary: A ~40-50-word summary that describes the context and significance of the findings for a general readership should be included on the title page. The statement should be written in the present tense and refer to the work in the third person.
12) Conflict of interest statement: JCB requires inclusion of a statement in the acknowledgements regarding competing financial interests. If no competing financial interests exist, please include the following statement: "The authors declare no competing financial interests." If competing interests are declared, please follow your statement of these competing interests with the following statement: "The authors declare no further competing financial interests." 13) ORCID IDs: ORCID IDs are unique identifiers allowing researchers to create a record of their various scholarly contributions in a single place. At resubmission of your final files, please consider providing an ORCID ID for as many contributing authors as possible.
14) A separate author contribution section following the Acknowledgments. All authors should be mentioned and designated by their full names. We encourage use of the CRediT nomenclature.

B. FINAL FILES:
Please upload the following materials to our online submission system. These items are required prior to acceptance. If you have any questions, contact JCB's Managing Editor, Lindsey Hollander (lhollander@rockefeller.edu).
--An editable version of the final text (.DOC or .DOCX) is needed for copyediting (no PDFs).
--Cover images: If you have any striking images related to this story, we would be happy to consider them for inclusion on the journal cover. Submitted images may also be chosen for highlighting on the journal table of contents or JCB homepage carousel. Images should be uploaded as TIFF or EPS files and must be at least 300 dpi resolution.
**It is JCB policy that if requested, original data images must be made available to the editors. Failure to provide original images upon request will result in unavoidable delays in publication. Please ensure that you have access to all original data images prior to final submission.** **The license to publish form must be signed before your manuscript can be sent to production. A link to the electronic license to publish form will be sent to the corresponding author only. Please take a moment to check your funder requirements before choosing the appropriate license.** Thank you for your attention to these final processing requirements. Please revise and format the manuscript and upload materials within 7 days. If complications arising from measures taken to prevent the spread of COVID-19 will prevent you from meeting this deadline (e.g. if you cannot retrieve necessary files from your laboratory, etc.), please let us know and we can work with you to determine a suitable revision period.