Neural crest-related NXPH1/α-NRXN signaling opposes neuroblastoma malignancy by inhibiting metastasis

Article activity feed

1. 

   Review Commons

   Nov 14, 2023

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

   Learn more at Review Commons

   ---

   Reply to the reviewers

   1. General Statements [optional]

   On behalf of the authors, I would like to thank the three reviewers for providing a valuable feedback on our manuscript. We appreciate that they all considered that the first part of our study conveys novelty to the field of neuroblastoma research, while being coherent with recent studies aimed at identifying the NB cells-of-origin through transcriptomics and single-cell RNA-sequencing approaches. We indeed identified a transcriptional signature that distinguishes LR-NBs from HR-NBs, revealing that these two NB subgroups are better discriminated by the core transcriptional signature shared by the distinct SA cell types, …

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

   Learn more at Review Commons

   ---

   Reply to the reviewers

   1. General Statements [optional]

   On behalf of the authors, I would like to thank the three reviewers for providing a valuable feedback on our manuscript. We appreciate that they all considered that the first part of our study conveys novelty to the field of neuroblastoma research, while being coherent with recent studies aimed at identifying the NB cells-of-origin through transcriptomics and single-cell RNA-sequencing approaches. We indeed identified a transcriptional signature that distinguishes LR-NBs from HR-NBs, revealing that these two NB subgroups are better discriminated by the core transcriptional signature shared by the distinct SA cell types, rather than by the transcriptional specificities of any of these cell types, as recently debated. Of note, our findings unveil that the sympatho-adrenal transcriptional program facilitates NB formation but concomitantly restricts its malignant potential. We also wish to thank the reviewers for acknowledging that, in contrast to previous studies, we pursued further by testing the functional relevance of this signature through a combination of in vitro and in vivo experiments. We thereby identified NXPH1 and its receptor α-NRXN1 as ones of the very first factors showing an anti-metastatic activity in the context of NB. Uncovering NXPH1/α-NRXN signaling as a possible target to treat metastatic HR-NBs gives our study considerable clinical relevance.

   We consider that the critics and recommendations provided by the three reviewers are positive and pertinent. We are thus willing to address nearly all the reviewers’ concerns and suggestions within the scope of a revision, including performing additional in vivo experiments, as explained in details in the following section. We hope that the planned revisions will be sufficient to make our manuscript suitable for publication.

   __ __

   2. Description of the planned revisions

   Reviewer #1 (Evidence, reproducibility and clarity (Required)):

   **Summary:**

   In the reviewed manuscript, the authors aimed to characterize the sympathoadrenal (SA) transcriptional landscape that defines low- and high-risk neuroblastomas (LR-NBs and HR-NBs respectively). In particular, they analyze previously published Affymetrix U219 expression profiles of 18 low- (n=8) and high-risk (n=10) neuroblastomas, and 2 fetal human adrenal glands. The authors define transcriptional signatures of LR- and HR-NBs, and further unbiasedly classified them in 4 clusters defining groups of patients with different prognosis (as tested using the 498 SEQC cohort). Within these transcriptional signatures, the authors delineated a SA signature using a human fetal adrenal gland transcriptional profile recently published (Kildisiute et al. 2021), that can discriminate low-risk neuroblastomas. From these genes, the authors further select NXPH1 and NRNX1-2 as promising targets for extensive experimental in vitro and in vivo validation, including validation in cell cultures, and xenografts, to determine that NXPH1/alpha-NRXN1-2 signaling is sufficient for NB tumor growth, and that the expression of either stimulates the metastatic potential of NB cells.

   **Major comments:**

   1- The cohorts and data used by the authors to conduct the main analysis of the paper are already published, and thus the contribution of the analysis is incremental. In particular, the authors analyzed arrays from a limited cohort size, in comparison with others available sequenced with RNA-seq (e.g. 176 HR- and 322 non-HR NB in 498 SEQC; 224 HR- and 342 non-HR NBs included in the Westermann-genecode19 cohort; and 80 HR- and 20 non-HR in the Jagannathan cohort). Furthermore, the 12,000 most-expressed genes (out of ~20,000 available) were analyzed by the authors, as opposed to more than 40,000 (coding and non-coding) included in a normal RNA-seq study. Only the 498 SEQC dataset provides 12,000 genes significantly up-regulated in either high-risk (n~5,500) or non-high-risk (n~7,000). The differences between datasets could influence the results of the study. For example, in the reviewed manuscript, genes with a high expression in LR-NB (compared to fAG) included DNMT3B, SEMA5A, SOX5, and TET1, all of which have a significantly higher expression in HR-NBs of the 498 SEQC cohort. The quality of the manuscript will be enhanced with consistent results obtained by conducting the same reported analysis in larger cohorts.

   Reply: To define the transcriptional signatures associated to the aetiology of LR- and HR-NBs and to the malignant behavior of HR-NBs, we first needed to compare the transcriptomic signatures of LR-NB and HR-NB samples to the one of a non-tumoral, healthy tissue such as the fetal adrenal gland. The online databases cited by the reviewer #1 do not contain any healthy samples, thus precluding the possibility to use it for the first step of our analysis. This is why we decided to use as a starting point the cohort of samples from the Hospital Sant Joan de Déu (HSJD cohort), which further provided the advantage of comparing samples from patients all diagnosed and treated at the same facility. We then used the SEQC database in different steps of our analytic strategy (Fig.S1F, 2F-H, 3A-B, S3) to test the relevance and coherence of the results obtained with the HSJD cohort in the context of a larger cohort. As mentioned by the reviewer #1, we indeed focused our analysis on the 12,000 most-expressed genes. We did so to follow the recommendations of the software Phantasus for transriptomic analyses of mammalian datasets.

   2- In the reviewed manuscript, PHOX2A and PHOX2B are significantly more expressed in both LR-NB and HR-NB compared to fAG. This is also the case for other adrenergic markers including TH and DBH. Oppositely, the expression of cortex markers (i.e. STAR and CYP11A1) is significantly higher in fAG. Nevertheless PNMT is not significantly up-regulated in fAG in comparison to LR-NB nor HR-NB. Is it possible that the fetal adrenal glands analyzed include a large proportion of cortex that confounds the transcriptional signals? The quality of the manuscript will be enhance if the authors could establish what proportion of the fAG transcriptional signal belongs to cortex, and if they account for its influence in the analysis.

   Reply: The samples of human fetal adrenal gland from which RNA was extracted were obtained from donations (samples staged at 22 weeks post-conception or 2 days after birth) and evaluated by a pathologist who confirmed their correct preservation before sample processing. As fetal adrenal glands are very small tissues, successfully separating the cortical and adrenal regions requires micro-dissection, which was not applied. These samples were instead processed entirely, presenting a ratio of medulla/cortex tissue according to their developmental stage.

   3- A recent published paper (Bedoya-Reina et al. 2021) study the differences of HR-NB and LR-NB from a single-cell perspective. In the published manuscript, the authors conclude that LR-NBs are enriched in cells that resemble chromaffin and sympathoblast cells, while the high-risk neuroblastomas are enriched in undifferentiated cells that resemble cells with progenitor characteristics in post-natal adrenal gland. This is broadly consistent with the conclusion reached by the authors in the manuscript under review. It will enhance the content of the reviewed manuscript if the authors compare their transcriptional signatures with the recently published transcriptional signatures in this paper to answer the following questions:

   -1) to what extent the transcriptional signatures for HR-NBs (4 hierarchical clusters incl.) in the manuscript under review resembles that from the published undifferentiated cluster (nC3) enriched in HR-NBs, and the progenitor cluster (hC1) in post-natal adrenal gland;

   -2) to what extent the transcriptional signatures for LR-NBs (4 hierarchical clusters incl.) in the manuscript under review resembles that from the published NOR (nC7, nC8, and nC9) enriched in LR-NBs, and that from the chromaffin cells (hC4) enriched in post-natal adrenal gland;

   -3) how is the expression of NXPH1 and alpha-NRXN1-2 in the reported LR- and HR-NBs, and adrenal gland.

   Reply: As stated by the reviewer #1, our findings are indeed in agreement with the main conclusions recently reported by Bedoya-Reina et al (Bedoya-Reina et al, 2021). We agree with the reviewer #1’s suggestion that a detailed comparison of our transcriptomic signatures with those of Bedoya-Reina et al would be interesting. We will thus perform this comparison and provide these complementary results in the revised version of the manuscript.

   4- In the discussion, the authors indicate that they do not aim to identify the transcriptional signature associated to NB origin but rather use the component of the SA lineage that distinguish LR- and HR-NBs. This statement implies that neuroblastoma can originate from any cell in the developing SA lineage (i.e. SCP, bridge, chromaffin and sympathoblast), a controversial assumption that requires further proof. In particular, when discussing about the core sympathoadrenal signatures enriched in LR-NBs and HR-NBs, the authors obtained a SA signature of genes shared by at least 3 of the 4 SA cell signatures. Further justification needs to be provided as for why (in particular) one of these SA cell signatures exclude the sympathoblast/neuroblast contribution.

   Reply: We decided to use as a core SA signature the genes shared by at least 3 of the 4 SA cell types to avoid being too restrictive, the SCP signature being particularly distant from the 3 others. As such, the list of genes shared by all 4 cell types consists of 663 genes, of which only 51 are retrieved in our list of 503 LR vs HR DEGs. Conversely, the list of genes shared by Bridge cells, Chromaffin cells and Sympathoblasts (but not shared by SCPs) consists 3,530 genes, of which 199 are retrieved in the list of 503 LRvsHR DEGs. These complementary results, which can be discussed and provided if required, therefore suggest that the core SA signature discriminating LR-NBs from HR-NBs represents mostly the Bridge-Chromaffin-Sympathoblast lineage and excludes the SCP identity. They are therefore in agreement with the recent notion that NB cells-of-origin derive from the sympathoblast-chromaffin lineage.

   5- Some of the most interesting results in the paper are limited to proportions in a subset of top-ranked genes. It will be valuable to set the analysis in an hypothesis driven context, add probabilities, test names, and corrected p-values to the results.

   Reply: Our study is based on the initial hypothesis that LR- and HR-NBs might differ in the way they exploit the transcriptional program underlying their developmental origin. To get a deeper insight into this notion we performed a sequential differential expression analysis of primary samples of LR-NBs, HR-NBs and human fetal adrenal gland using the web-based Phantasus software. This software identifies differentially expressed genes between groups using the Limma R package, as detailed in the Methods section. As such, basic statistics for significance analysis were performed using a modetared T-test (as specified in the Limma R package) and FDR-adjusted P-values were set to PAdditionally, top-ranked genes of SA clusters were selected as part of a heuristic approach aimed at highlighting the clinical implication of the transcriptional clusters retrieved in our analysis. The relationship between their expression levels and patient survival was further analyzed using the SEQC database (Fig. S1E). Next, and in contrast to previous studies, we tested experimentally the validity of our analytical findings correlating the expression of SA-c1 genes with a better patient prognosis. To this aim, we selected the candidate gene NXPH1, one of the top-ranked genes from the SA-c1 subset, on the basis of several complementary arguments (listed in response to the reviewer #2’ comment #4). We thereby analyzed how modulating the expression of NXPH1 or that of its receptor α-NRXN1 affect the growth and metastatic potential of human NB cells. The results obtained argue for the validity of our model, by proposing that the neural crest-derived sympatho-adrenal developmental program, in particular the SA-c1 signature, plays a complex role in NB tumorigenesis: it facilitates tumor growth but blocks metastasis formation, hence opposing NB malignancy.

   **Minor comments.**

   6- In comparison with other cohorts that include low- and intermediate-risk NBs as non-HR NBs, the reviewed data specifically includes low- and high- risk NBs. It is important that the authors include a characterization of intermediate-risk neuroblastomas in their analysis.

   Reply: The samples forming the HSDJ cohort were all obtained from NB patients diagnosed and treated at the Hospital Sant Joan de Déu. Several clinical and biological parameters were used for classification, among which the age of the patient. If we apply the cut-off point of 1 year (used at the time the samples were obtained), our cohort does not include any intermediate-risk NB. If we apply the cut-off of 18 months which is now more usual, our cohort would contain only one case (#HSJD-NB14 - aged 16 months at the time of diagnosis) that could be classified as intermediate-risk.

   7- Further details and figures on what precise criteria was used to remove the sample #LR-08 is required. How including this sample changes the reported results?

   Reply: As explained in the Methods section, before comparing the transcriptomic landscapes of LR-NBs and HR-NBs, we first assessed the sample dispersion by performing a principal component analysis. This analysis identified one outlier (#LR-08) that was clearly distant from all the other NB samples (both LR- and HR-NBs). We thus removed this sample to limit the dispersion and variability that would have impacted the subsequent analyses, as recommended by the Phantasus guidelines. We will provide an illustration of the PCA including this outlier. We believe that performing de novo the whole bioinformatical analysis including this outlier would not bring any novel significant conclusion.

   8- GO-term distribution was assessed using the 50 most-enriched GO-terms. How would the results change if all the significant GO terms were analyzed?

   Reply: We will re-analyse the GO-term distribution by including all the significant terms.

   9- Was the SEQC 498 (GSE62564) dataset obtained with microarrays (as indicated in the methods) or with RNA-seq (i.e. Illumina HiSeq 2000)?

   Reply: Similar results were obtained using either the SEQC database obtained with microarrays or the one obtained with RNAseq. The data presented in our manuscript correspond to the ones obtained with the RNA-seq SEQC database.

   10- In methods, the first quartile (Q1) in SA-c1 has a higher limit in 487 samples and the fourth quartile the lowest limit in 4, how many samples (out of 498 NBs) were excluded and why?

   Reply: As explained in the Methods section and in Fig. 2F, the complete SEQC cohort was included in this survival analysis. To subdivide the 498 samples of the SEQC cohort into 4 expression quartiles, we evaluated whether the expression level of each of the 242 SA-c1 genes (corresponding to 573 ref-seq IDs) in a given patient sample was above or below the mean expression of that gene in the complete cohort. Samples were then distributed into quartiles based on the number of genes presenting an expression level above the mean. As detailed in the Methods section, the resulting sample distribution was as follows: 487≤Q1≤360 (124 patients); 359≤Q2≤250 (125 patients); 249≤Q3≤135 (126 patients) and 134≤Q4≤4 (123 patients).

   11- In the 503 DEGs between LR-HR NBs, NTRK2 and MYCN are not included, even if the HR samples included MYCN amplified tumors. Can the authors comment on this?

   Reply: MYCN did not pass the cut-off when comparing its expression levels in LR-NB and HR-NB samples (showing an adjusted P=0.10226 for a cut-off of adjusted PNTRK2 was present in our initial 12,000 gene dataset, but it was not differentially expressed in any of the comparisons made (LR vs fAG: adj-P=0.4916; HR vs fAG: adj-P=0.63757; HRvsLR: adj-P=0.90431)

   12- The authors mention that the top 30 genes found in cluster c1 (and also in c2) are correlated with favorable patient prognosis. Is it the case that *all* the genes in c1 (and also c2, c3 and c4) are significantly associated with a favorable or else unfavorable prognosis?

   Reply: As presented in the datasheet “KM analyses” of Table S3:

   • 32 out of the 33 genes (97%) of the cluster c2 correlate to an unfavorable prognosis, the 33th gene showing no particular correlation.

   • 29 out of the 32 genes (91%) of the cluster c3 correlate to a favorable prognosis, 1 gene correlates to an unfavorable prognosis and the last 2 do not show any particular correlation.

   For the clusters c1 and c4, we focused on the top 30 genes because these clusters contain numerous genes (338 and 100, respectively). The results obtained showed:

   • All the top 30 genes (100% of the genes tested) of the cluster c1 correlate to a favorable prognosis

   • 23 out of the top 30 genes (77% of the genes tested) of the cluster c4 correlate to an unfavorable prognosis, whereas the other 7 genes do not show any particular correlation

   We believe that these results are convincing enough. However, if considered mandatory we will assess the prognosis of all the genes found in clusters c1 and c4.

   13- The high expression of a (significant?) number of genes in cluster c4 is observed in patients with worst outcome (i.e. lower event-free survival), including ATR, HIF1A, ING2, POLR2L, SRPRB (498 SEQC, analyzed with R2).

   Reply: Indeed, 23 out of the top 30 genes (77% of the genes tested) of the cluster c4 correlate to an unfavorable prognosis, which appears to us as a significant number of genes. We will test whether the expression of the remaining genes forming the cluster c4 also correlate to an unfavorable prognosis.

   14- Regarding the 242 genes in the core SA signature, although its a smaller number, the expression of several genes in the core SA signature with a higher expression in HR compared to LR belonging to clusters 2, 3, and 4 is observed in worst outcome patients in the 498 SEQC cohort (CHD7, DNMT1, HMGA1, HSD17B12, LBR, LSM7, MCM4, NKAP, POLA1, and others). Is this small fraction significant?

   Reply: The genes presenting a higher expression in HR-NBs than in LR-NBs are found either in cluster c2 or cluster c4. The core SA signature retrieved 262 genes of the 503 LR *vs *HR DEGs, of which 14 belong to cluster c2 (including CHD7, DNMT1, HMGA1, LBR, LSM7, MCM4, and POLA1) and 3 to cluster c4 (UQCRFS1, NKAP and HSD17B12). As shown in the “KM analyses” datasheet of Table S6, these 17 genes all correlated with an unfavorable prognosis.

   15- In Kildisiute et al. 2021, NRXN1 is expressed in SCPs, while NXPH1 is expressed in bridge, chromaffin and sympathoblastic cells. How are the microenviroment of these cells regulating the expression of these genes in a developmental context (particularly as sympathoblastic cells are know to have larger proliferative capabilities than SCPs)? how is this cell heterogeneity replicated by a NB cell line? are mesenchymal and adrenergic cells expressing differentially NRXN1 and NXPH1?

   Reply: Unfortunately, the literature about NXPH1 remains very limited (less than 40 articles referenced in Pubmed) and nothing is known about the regulation of its expression during development. The data from Kildisiute et al (Kildisiute et al, 2021) indeed identified NXPH1 in the signatures of bridge cells, chromaffin cells and sympathoblasts, while its receptors NRXN1 and NRXN2 were found in the transcriptomic signatures of all 4 SA cell types. Interestingly, the data provided by Kildisiute et al further established that the expression of NXPH1, NRXN1 and NRXN2 is specifically enriched during the pseudo-time transition from bridge cells to sympathoblasts. This suggests that NXPH1/α-NRXN signaling might be particularly important at that stage and could participate in regulating this transition. But this remains purely speculative and it will need further investigation.

   We initially used a panel of 10 human NB cell lines harboring distinct characteristics in terms of genetic profile and morphological properties. We did not find any specific correlation between NXPH1 or α-NRXN1/2 expression and the different types of NB cell lines. We will provide an illustration of this observation in the revised version of the manuscript.

   NB cells can convert or be reprogrammed from an adrenergic state, which is less chemoresistant in vitro, to a mesenchymal state (van Groningen et al, 2019). As asked by the reviewer #1, we investigated the expression of NXPH1 and α-NRXN1 in relation with the mesenchymal vs adrenergic status of NB cells (using the dataset GSE90803 from (van Groningen et al, 2019). We found that both genes are expressed at higher levels in cells of the adrenergic phenotype, suggesting that NXPH1/α-NRXN signaling might be particularly relevant for the maintenance of this phenotype. If needed, we will provide an illustration of this observation in the revised version of the manuscript.

   16- Figure 1B and C, 2B,D: might the information provided be enhanced? otherwise these inserts might be excluded.

   Reply: We thought that the panels presented as Fig.1B, C and 2B, D would be helpful to the readers. We could remove them if the editors and reviewers consider it mandatory.

   17- Figure 3D: Kildisiute et al. 2021 data and GTEX available at human protein atlas indicate expression of NRXN1 and NXPH1 in developing and adult adrenal gland. Might the results illustrated suggest a confounding effect in the sampled fetal adrenal glands, perhaps from cortex?

   Reply: The samples of human fetal adrenal gland from which RNA was extracted were obtained from donations (samples staged at 22 weeks post-conception or 2 days after birth) and evaluated by a pathologist who confirmed their correct preservation before sample processing. As fetal adrenal glands are very small tissues, successfully separating the cortical and adrenal regions requires micro-dissection, which was not applied. These samples were instead processed entirely, presenting a ratio of medulla/cortex tissue according to their developmental stage.

   18- The authors conduct extensive experiments in NRXN1, and make conclusions about its role in for instance metastasis, nevertheless the LR-NB/HR-NB SA signal only includes NRXN2. Can the authors comment on the differences between NRXN1s and NRXN2?

   Reply: NRXN1 and NRXN2 were both found to be differentially expressed between LR vs fAG and HR vs fAG, and were thus retrieved among the list of 3.096 common DEGs (Table S3). NRXN2 was further found in the list of 503 LR vs HR DEGs (adjusted P=0.037), showing higher expression levels in LR-NBs than in HR-NBs (see Fig.3D). NRNX1 presented an expression profile comparable to that of NRXN2 (Fig.3D) but did not pass the cut-off (adjusted P=0.38) due to an increased variability in HR-NB samples and was thus absent from the list of LR vs HR DEGs. In vitro, the expression levels of NRXN1 and NRXN2 showed comparable patterns. When we initiated the functional experiments there was no fluorescence-conjugated antibody available to detect and sort α-NRXN2, but there was for α-NRXN1. This is the practical reason that led us to focus on α-NRXN1 in the second part of our study.

   Reviewer #1 (Significance (Required)):

   The significance of the study relies in investigating the role of selected targets in neuroblastomas within a risk group. In particular, HR-NBs have poor outcomes and are generally metastatic at the time of diagnosis.

   The results of the manuscript are somehow consistent with a recently published manuscript analyzing LR- and HR-NBs from a single-cell perspective. The manuscript will be enhanced by conducting the suggested comparison between the reviewed and the reported results. The authors further need to comment why HR-NBs markers, particularly MYCN is not recovered in the LR-NB/HR-NB and the LR-NB/HR-NB SA signals. Also they need to comment on possible confounding effects in the fetal adrenal gland.

   The paper is directed to a broader audience of cancer and developmental biologists, and computational biologist. Yet further statistical support needs to be provided.

   Reviewer #2 (Evidence, reproducibility and clarity (Required)):

   **Summary:**

   Provide a short summary of the findings and key conclusions (including methodology and model system(s) where appropriate). Please place your comments about significance in section 2.

   In this paper, Fanlo-Escudero et al., determines gene signatures differentiating low-risk (LR) neuroblastoma (NB) from high-risk neuroblastoma (HR-NB), as well as LR- and HR-NB as an entity from human fetal adrenal gland. They identify a transcriptional signature corresponding to a core sympathoadrenal lineage that can discriminate between LR-NB and HR-NB. This signature is composed of genes associated with favorable patient outcome. The authors further choose one gene, NXPH1, for functional analysis and investigates the effects this gene has on NB progression using in vitro assays, chick CAM assay and mouse in vivo models. The authors conclude that this transcriptional signature can distinguish LR-NB from HR-NB and that NXPH1 is involved in NB cell growth.

   **Major comments:**

   • Are the key conclusions convincing? The key conclusions are 1) a core SA lineage signature can discriminate between LR-NB and HR-NB, and 2) NXPH1 represses NB malignancy (in terms of metastatic capacity) and is a therapeutic target. The first conclusion is indeed convincing, and not contradictive to common beliefs. The second conclusion is poorly supported by data. The authors perform a range of experiments using in vitro and in vivo settings, but lack some fundamental experiments and overstate their findings.

   • Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether? Several claims should be softened, re-phrased and/or clearly marked as preliminary or speculative. No data nor claims need to be removed.

   For example, the following statements need to be changed:

   1- Page 10, second paragraph: The authors state "Remarkably, NXPH1 and α-NRXN1/2 levels increased in all the NB cell lines harbouring a sphere-forming capacity (Fig. 3E), thereby revealing a strong positive correlation between the expression of NXPH1 and α-NRXN1/2 and the acquisition of a NCC stem cell identity". The authors only show that NXPH1 is expressed in 8 out of 10 NB cell lines. Sphere-forming capacity is displayed in a relative and not absolute scale which makes it difficult to assess which cell lines that do form spheres and to what extent. The capacity to form spheres (from low to high) does not correlate to the levels of NXPH1 in the different cell lines.

   Reply: In its current form, Fig.3E presents via a heatmap representation how the expression of selected genes changes after growing cell lines in sphere-forming conditions as compared to basal (normal) ones. We understand from various reviewers’ comments that this representation has been misleading. We will change it, showing more explicitly the expression levels both in basal and sphere-forming conditions and will bring further details on how the sphere-forming ability of each cell line was assessed and characterized.

   2- Page 13, paragraph 1. The authors write "...these data revealed that NXPH1/α-NRXN1 signaling is necessary and sufficient for NB tumor growth in vivo". This is an overstatement. Tumors still form, meaning that NXPH1 signaling is not sufficient.

   Reply: Indeed, tumors still form after xenografting sh-NXPH1 or sh-NRNX1 cells but they form with a decreased frequency. Specifically, sh-NXPH1 cells formed tumors in 4 out of 6 xenografted mice, and the 4 tumors all showed a markedly reduced volume. Tumors formed from sh-NRNX1 cells were observed in only 2 out of 5 xenografted mice, with 1 of the 2 tumors showing a markedly reduced volume. We consider that these results support the conclusion that inhibiting NXPH1/α-NRXN1 signaling impairs tumor growth, affecting both tumor initiation and tumor growth. We however understand the reviewer #2’s comment and will thus rephrase this part accordingly. In addition, we will provide complementary data showing the mean volume of the tumors generated in the distinct experimental conditions.

   3- Throughout the text, the authors convert their statements. One example is page 15, first paragraph. They write "...growth of NB cells but markedly restrict their metastatic potential", but they do not show this. Instead, they only address the opposite situation - Knockdown enhances metastasis. This is not equal to their statement. See other experiments in other sections. The authors need to go through the manuscript and make sure that they explain their conclusions to actually fit their experiments.

   Reply: We will follow the reviewer #2’s recommendation and will rephrase the conclusions whenever needed to better fit to the experimental results.

   Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary to evaluate the paper as it is, and do not ask authors to open new lines of experimentation.

   **Issues to respond to:**

   4- The authors should more clearly explain why they choose to further study NXPH1 (second highest on the list), involved in synaptogenesis and neurotransmission, instead of SOX6, highest on the list, that is highly relevant in neural crest development.

   Reply: The combination of several reasons led us to choose NXPH1 for functional studies:

   • NXPH1 is a secreted factor, whose activity might be much more easy to target and modulate in future pharmacological/clinical assays than that of a transcription factor like SOX6
   • NXPH1 and its receptor NRNX2 both came out in the list of SA-c1 genes, and NRXN1 presented a comparable expression profile (higher levels in LR-NBs than in HR-NBs, although it did not pass the cut-off required to appear in the SA-c1 list), emphasizing the putative importance of this signaling pathway in NB tumor biology
   • The functional involvement of NXPH1 or that of its receptors has never been addressed in cancer formation or progression to date

   5- When the authors investigate the expression of NXPH1 and other genes and compare that with sphere-forming capacity (Fig. 3E and page 10) they analyze expression in cells cultured in basal medium while sphere-forming capacity is measured after 5 weeks in restricted medium. How does the expression of the analyzed genes change under these conditions?

   Reply: In its current form, our Fig.3E is actually showing how the expression of the analyzed genes changes after growing cell lines in sphere-forming conditions as compared to basal (normal) ones. As explained above (reviewer #1, point #1), we understand that this representation has been misleading and we will modify it.

   6- Is the subpopulation of NRXN1+ cells low (1.5%) because these samples are from aggressive ("High-risk") cell lines?

   Reply: All the NB cell lines used in this study derive from HR-NB tumors. As mentioned by the reviewer #3, there is no cell line modeling low-risk neuroblastoma to date. We indeed observed that the proportion α-NRXN1+ cells, as detected by FACS, was very low in all three cell lines tested. A similar observation was made using cells dissociated from three different patient-derived xenografts. We believe that a similar observation made in 6 samples of distinct origins highlights the consistency of finding a low proportion of α-NRXN1+ cells in NB samples.

   7- The authors state "...the number of cells quantified per tumor section was decreased by ~50% for the α-NRXN1+-deprived cells relative to their control (Fig. 3J, K), thus revealing that α-NRXN1+ cells are required to support NB tumor growth in vivo". This is not a correct conclusion. This experiment shows that a-NRXN1- cells do not grow and expand to the same extent as control cells. They cannot say that a-NRXN1+ cells support NB growth without comparing growth between a pure a-NRXN1+ and control cells.

   Reply: Unfortunately we could not assess the growth of purified α-NRXN1+ cells, given the low number of α-NRXN1+ cells that could be sorted as compared to the numbers of cells required to perform a CAM assay. We thus opted for comparing the growth of total SK-N-SH cells with that of SK-N-SH cells in which the α-NRXN1+ subpopulation had been experimentally removed. We believe that the results obtained convincingly argue for the importance of the α-NRXN1+ subpopulation in promoting NB proliferation and growth. Nevertheless, we understand the reviewer #2’s comment and will rephrase the conclusion.

   8- The authors use shRNAs to knock down NXPH1. They enrich their cells by two means - puromycin or doxocycline. This results in equal cell populations. The authors however state that they use doxocycline to circumvent the growth arrest they observe with puromycin selection. They need to elaborate on this and show why this would be the case and what difference the two methods do and show.

   Reply: We generated two types of knock-downs: a constitutive one and an inducible one. Puromycin was used to select constitutive knock-downs, whereas doxycycline was used to trigger sh-RNA production in an inducible manner, in stable clones previously established through neomycin selection. We apologize if this was not stated clearly enough in the manuscript and we will correct it.

   9- The major flaw of this paper is that the authors use one cell line in total, and even more that they use the same cell line for both knockdown and activation. Since they do show that different NB cell lines have different expression levels (ranging from high to absent), they should choose one cell line for KD and one for overexpression. The authors could also do a rescue experiment with knockout and gain-of-function (e.g., construct that will not be targeted by the shRNA) in the same cells.

   Reply: Since NXPH1 is a secreted protein, we needed a cell line that expresses both NXPH1 and its receptors to expect noticing effects on NB cell behavior when their expression is reduced. We reasoned that performing a gain-of-function of NXPH1 in a cell line that does not express its receptors would have no interest, and vice versa. We also believe that it is more conclusive to perform gain- and loss-of-function experiments in the same cell line, because of the likely differences in cell behavior and aggressiveness of distinct cell lines. We however agree that our conclusions would be strengthened if similar conclusions were reached using different cell lines. We will thus perform growth and metastasis assays both in vivo and in the CAM using NXPH1 and α-NRXN1 shRNAs in an additional cell line. We will moreover consider performing rescue experiments and will think about the best methodology to do so.

   10- They only use one shRNA after trying several (Fig. S4). The efficiency is substantially variable and not convincing. As stated also elsewhere in this review, they need to check protein level. And to ensure that their results are not off-target they should perform at least some crucial experiments with two shRNAs.

   Reply: We agree that the decreased mRNA levels caused by NXPH1 and α-NRXN1 shRNAs showed variability. Yet, they were sufficient to significantly reduce cell viability, which was impaired by two distinct shRNA constructs, both for NXPH1 and α-NRXN1. To complete these experiments as recommended by the reviewer #2, we will assess how NXPH1 and NRXN1 expression is altered at the protein level by western-blotting. We will moreover address possible off-target effects by RT-qPCR.

   11- Why don't the authors add BrdU post-implantation? This is easily done in the egg considering the accessibility and would better reflect the proliferation in vivo.

   Reply: Adding BrdU pre-implantation allowed us to get a read-out of the global proliferative behavior of NB cells over the whole post-implantation duration. Adding it at the end of the post-implantation would have only allowed us to assess the proliferative behavior of cells at the end of the experiment. We believe that this would have been less informative.

   12- Why do the authors switch between CAM and mouse xenografts? I understand that the mouse model must be employed for "metastasis", but can it be explained why and when they perform the different "tumor growth" experiments?

   Reply: The CAM assay was used for 2 reasons: 1) when cell numbers were limiting (i.e. testing the importance of the α-NRXN1+ subpopulation for tumor growth), and 2) to perform a gain-of-function strategy using a recombinant rNXPH1 protein and testing its effects on tumor growth over a duration of 1 week. Such experiment would not have been possible using mouse xenografts, due to the extended experimental duration (7-8 weeks) of this assay and to the need for repeated rNXPH1 injections. The rNXPH1 gain of function experiment in the CAM and the NXPH1/α-NRXN1 loss-of-function experiment using mouse xenografts were performed concomitantly.

   13- Why do the authors do left ventricle injections for metastatic studies and not AG implantations?

   Reply: We reasoned that cell injections into the left ventricle were ideal to test the organotropism of metastatic NBs, as this methodology facilitates cell dissemination and colonization of organs targeted by NB metastasis such as the liver and bone marrow. Cell implantation into the adrenal gland is especially useful to study cell growth in one of the primary sites of NB growth, which was not our experimental purpose at that stage of the study.

   14- The measurement of bioluminescence is very difficult to interpret. The authors discuss metastatic spread, but the images show only large blobs covering the heart and areas surrounding it, especially for the sh-αNRXN1. To be able to say that the cells have colonized specific organs the authors need to dissect these organs and perform staining. I see it as tumors recur rather than particularly metastasize when they re-appear after 6 weeks.

   Reply: The parameters for bioluminescence detection were set equally for all experimental conditions. The differences in bioluminescence intensities observed in mice injected with control cells compared to those injected with shNXPH1 and shNRXN1 cells explain why it is so intense in the case of shNXPH1 and shNRXN1 cells.

   As suggested by the reviewer #2, we had further dissected the mice and recovered the organs of interest. We will provide additional data confirming that an intense bioluminescent signal was effectively detected in the liver and hind legs (containing bone marrow) of mice injected with shNXPH1 and shNRXN1 cells, but not in those injected with control cells.

   15- How do the authors explain results presented in Fig. S5: There are more HuNu+ cells but tumor size is unchanged?

   Reply: In our experimental CAM setup, 5·105 SK-N-SH cells were embedded in 10μl of Matrigel, which served as a matrix facilitating tridimensional NB cell proliferation. The addition of rNXPH1 increased the number of HuNu+ (NB) cells per section and per mm3 (Fig.4G, H and Fig.S5C), without significantly increasing the tumor area (Fig.S5D) nor the tumor/matrigel volume (data not shown in the current version, but it will be included in the revised version). As illustrated in Fig. 4G, the matrigel was not totally filled with NB cells, even at the time of recovery. We thus deduced from these observations that more NB cells progressively filled the matrigel, without reaching the point where they significantly altered the tumor/matrigel volume. Nevertheless, we can provide additional data revealing that the addition of rNXPH1 caused a slight, yet reproducible increase in the tumor/matrigel weight, in agreement with the increased cell density already shown in Fig.S5C.

   Are the suggested experiments realistic for the authors? It would help if you could add an estimated cost and time investment for substantial experiments.

   Some suggested experiments take time (orthotopic implantations, rescue experiments, adding cell lines and # of shRNAs).

   However, as long as the authors discuss and address their methods for in vivo growth (in ovo vs in vivo vs their choice of metastasis model), an orthotopic AG model is not necessary. The authors should however consider it for future studies.

   Experiments required to support their conclusions: A rescue experiment and use of different cell lines for KD and overexpression is somewhat time-consuming, as well as ensuring KD at protein level and include an additional shRNA in crucial experiment. I expect that this would take 2-4 months.

   • Are the data and the methods presented in such a way that they can be reproduced? 16- The authors should elaborate on their methods, but in general they are reproducible.

   Reply: As requested, we will bring further details on the experimental setups.

   Are the experiments adequately replicated and statistical analysis adequate?

   17- In several places the number of replicates is questionable. Especially at the end of page 26, the authors state that they have performed n=1-4 replicates. N=1 replicate is never ok. In several instances they do n=2 replicates. This can be acceptable but the authors could address this.

   Reply: When assessing the sphere-forming capacity of the different NB cell lines, some cell lines only produced spheroids in 1 of the 4 replicates tested. This is a result in itself, which helped us classifying cell lines based on their sphere-forming capacity. We understand that this Methods paragraph was elusive. We apologize for it and will clarify this aspect.

   **Minor comments:**

   • Specific experimental issues that are easily addressable. Mainly text changes can be seen as minor. For experiments and other issues, see other sections.

   • Are prior studies referenced appropriately? Yes. 18- The reference list is not coherently styled.

   Reply: We understand that using numbered references can be annoying. We will adapt the reference format to stick to the guidelines of the specific journal to which the manuscript will be addressed.

   Are the text and figures clear and accurate?

   19- Overall, the paper is written in a complex way and difficult to easily comprehend. For example, the authors need to clarify several issues on the material they use and experiments they perform. I suggest substantial re-writing to better convey their messages. Sentences should be short and clear, data explained in the context of it was derived.

   Reply: We will take into consideration the reviewer #2’s suggestion and modify the manuscript to facilitate its comprehension.

   The following text edits and clarifications are required:

   20- Page 3. The authors write "cell-of-origin" in several places, this should be changed to "cells-of-origin" (i.e., plural). The view that all NBs originate from only one cell is too simplistic, and the authors should definitely edit this considering that they are investigating different subgroups of NB.

   Reply: We will correct this mistake.

   21- Page 5. The authors MUST define where the material from the 18 NB patients as well as fetal AG derive from. There is no reference, and taken from the material&methods section, the transcriptome data from these data has not been generated by the authors themselves?

   Reply: The transcriptomic data used herein have indeed been previously generated by two co-authors of the study, and the corresponding reference is cited (ref #20; (Gomez et al, 2015). All the NB samples included in our transcriptomic analysis were obtained at the time of diagnosis from patients attended at Hospital Sant Joan de Déu (HSJD, Barcelona, Spain). Tumors were evaluated by a pathologist and only the snap-frozen pre-treatment samples showing at least 70% of viable tumor content were included for analysis. Neuroblastoma risk assessment was defined by the International Neuroblastoma Staging System (INSS). Samples of normal fetal adrenal gland (n=2) were used as a non-tumoral reference tissue. Total RNA from frozen samples was extracted by TRIzol® Reagent. High quality RNA (RINe>7.00) was hybridized to Human Genome U219 microarray plates at the Functional Genomic Unit, Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS, Barcelona, Spain) according to Affymetrix standard protocols. Microarray data are deposited in NCBI (#GSE54720). We will include this information in the Methods section of the revised version of the manuscript.

   22- Why do the authors choose to work with the data set from Jansky et al in particular?

   Reply: Actually, we chose to work with the data from Kildisiute et al. (ref #16 (Kildisiute et al, 2021). We selected this study because: 1) they identified the transcriptional signatures of the 4 SA subtypes relevant to our study and additional related signatures of interest (adrenal cortex, mesenchyme…), 2) these data were obtained from human tissues, and 3) these signatures were easily accessible from their supplementary data.

   23- Page 7, paragraph 1. The authors write "Remarkably, 67% of the 503 genes found in this signature, which formed the cluster c1, are both associated with a neural identity and a better patient outcome". I do not agree that this is remarkable and the authors should remove this word.

   Reply: We will remove the term “Remarkably”.

   24- Page 10, second paragraph. The authors should clarify that the expression levels of the displayed genes are derived from qPCR analysis (if I have understood that correctly, I had to find and guess this from the M&M section), as well as explain how they have set the scale for sphere-forming capacity and what this corresponds to. What do the colors actually represent?

   Reply: We apologize for this information lacking in the Results section. We will precise that the expression levels were assessed by RT-qPCR, we will add a representation of their expression levels in basal culture conditions and will improve the explanation of the assessment of the sphere-forming capacity of the NB cell lines.

   25- Page 11 and onwards. The authors write "deprived of their a-NRXN1+ subpopulation". This is highly confusing and difficult to read. The authors should write "a-NRXN1- subpopulation"

   Reply: We will follow the reviewer #2’s recommendation and change the text for "α-NRXN1- subpopulation".

   26- Page 11, end of paragraph 2. The authors write "...arguing that NXPH1/α-NRXN signaling could control NB growth and/or aggressiveness". Number of cells do not directly correlate to aggressiveness, and this needs to be re-phrased to only state what the experiment actually shows - proliferation of a-NRXN1- cells.

   Reply: We will rephrase this sentence according to the reviewer #2’s recommendation.

   27- I am in favor of using the CAM assay as a complementary system. The authors however use this to define "...required to support tumor growth in vivo". The CAM assay using NB (i.e., transformed cells) shows the growth of these cells in response to presence of blood vessels and not a full tumor micro-environment. This should be clarified.

   Reply: We will clarify this aspect.

   28- As stated in the previous comment, the authors write "...required to support tumor growth in vivo". The next paragraph has the following headline "NXPH1/α-NRXN signaling stimulates NB growth". This is to me the same thing. The authors should elaborate on how these differ, or if they use them to show the same thing, clearly state that.

   Reply: We will better explain how these distinct pieces of evidence are complementary and reinforce the conclusion that NXPH1/α-NRXN signaling stimulates NB growth.

   29- The authors conclude that NXPH signaling can be used as a therapeutic target. This would however be extremely difficult considering the opposing effects shown on growth vs metastasis. I agree that it is important to find means to inhibit metastasis, but that does not mean we can allow for enhanced growth of the primary tumor. A better reflection would be to use this as a biomarker, but this can only be predictive/speculative since the authors do not perform for example a tissue microarray to show this at IHC protein level, something that is currently the practice in the clinic.

   Reply: While we agree with reviewer #2 that we cannot allow for enhanced growth of the primary tumor, we believe that having identified a secreted factor whose activity inhibits NB metastatic potential is a novel and important finding. We did not wish to suggest that our experimental setup could be directly applied to inhibit the metastatic dissemination of HR-NB tumors. However, we believe that our findings can set the basis of a therapeutic design in which NXPH1/α-NRXN signaling would be enhanced locally to prevent/block metastases from HR-NB tumors.

   As mentioned in the Discussion section (page 17), one study reported that NXPH1 can be used as a DNA methylation biomarker associated with a good prognosis for NB patients (reference #54, (Decock et al, 2016).

   **Material and Methods:**

   30- The authors have misspelled the cell line SK-N-BE(2)c.

   Reply: Indeed. We will correct this mistake.

   31- Why are some cell lines grown in 20% FBS? This is not standard and could impact the results.

   Reply: The 3 cell lines of our panel which were grown in 20% FBS correspond to the 3 cell lines of the mixed subtype, including SK-N-SH, SK-N-Be(2)c and IMR-32 cells. These cell lines have been established and originally grown in presence of a high FBS content (Tumilowicz et al, 1970; Biedler et al, 1973; Ciccarone et al, 1989). In our hands and as recommended by the colleagues that provided us with these cell lines, growing the cell lines in presence of 20% FBS was indeed crucial to sustain their morphological heterogeneity. While we agree with the reviewer #2 that culture conditions could affect cell behavior in vitro, we wish to emphasize that our main conclusions are derived from in vivo experiments. We are thus convinced that our main findings were not impacted by in vitro culture conditions.

   32- The authors should state what the tumor volume limit in their ethical permit is (page 30).

   Reply: The tumor volume limit was set at 1,500 mm3 as specified by our ethical committee.

   Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

   33- Fig. 3E. As discussed elsewhere, the authors should clarify the scale/meaning for sphere-forming capacity

   Reply: As explained above, we will modify the representation of the results shown in Fig.3E to improve their clarity and facilitate their comprehension.

   34- Fig. 4D. What do the numbers (i - vi) refer to? I cannot find this in the figure, figure legend, text or material&methods.

   Reply: These numbers were used to call different tumors and show their GFP content, as appearing in the lower part of this panel. We will precise this information in the corresponding figure legend.

   35- The authors do not present the tumor volume in Fig. 4. The authors discuss tumor growth in the text and this data should be included.

   Reply: We agree with the reviewer #2’ suggestion and will present tumor volumes in the revised version of the manuscript.

   36- Fig. S4. Knockdown efficiency is variable, and efficiency does not correlate to growth capacity. Especially because of this, the authors need to investigate this at protein level.

   Reply: As requested by the reviewer #2, we will investigate the knockdown efficiency at the protein level.

   Reviewer #2 (Significance (Required)):

   • Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field. This paper is partly within the scope of the ongoing debate of NB origin (see references below). This paper is however not as extensive, provides no new patient material (applies data from Jansky et al), and do not address the actual cell-of-origin (which the authors themselves also clearly states). This paper provides new gene signatures that can be used to define low-risk vs high-risk patients which is highly important to the field, but these signatures are not unexpected and does not add a significant advance to the field. The authors also do not address how these signatures could be applied clinically. The authors do not use any new methodology. With this said, with revision of the paper, it will still add to the current focus on NB biology research.

   • Place the work in the context of the existing literature (provide references, where appropriate). 37- There is a recently initiated, important and extensive debate about the cells-of-origin for NB. The authors indeed bring this up in the paper and also state that they do not intend to add to this debate. They use data from one paper from referenced debate above, and I would argue that because of this fact, and that they extract gene signatures from it, they do, at least partly, touch on the NB cells-of-origin debate, and the authors should put their results into context, from a big picture perspective. As of now, they dodge this complex issue.

   References: Dong et al., Cancer Cell 2020; Hanemaaijer et al., PNAS 2021; Jansky et al., Nat Genet 2021; Kameneva et al., Nat Genet 2021; Kildisiute et al., Sci Adv 2021; Furlan et al., Science 2017).

   Reply: We understand the reviewer #2’s argument. We will thus try to put our results in perspective regarding the NB cells-of-origin.

   State what audience might be interested in and influenced by the reported findings.

   This paper will be interesting for scientists within the neuroblastoma field, in particular those working on defining NB subgroups in correlation to developmental stages.

   • Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate. Reviewer's expertise: Neuroblastoma, neural crest, trunk neural crest, chick embryos, mouse models, in vitro models

   Parts of paper outside expertise of the reviewer: Analysis of the bioinformatics data (i.e., extracting signatures).

   Reviewer #3 (Evidence, reproducibility and clarity (Required)):

   **Summary: short summary of the findings and key conclusions (including methodology and model system(s) where appropriate).**

   Dr. Le Dréau and colleagues provide a transcriptional analysis comparing low versus high-risk neuroblastomas using a cohort of 18 patients. By comparing them to normal transcriptional signatures obtained from human fetal adrenal glands, they conclude that low risk tumors share a specific core sympatho-adrenal developmental program, which is associated with favorable patient prognosis. Among this signature, they specifically assess the role of NXPH1/a-NRXN in vitro and in vivo using different models (cell lines, PDX-derived cell lines, mouse and chick xenografts). They propose that NXPH1/a-NRXN axis promotes neuroblastoma tumor growth via cell proliferation, and inhibits metastases.

   **Major comments:**

   • Are the key conclusions convincing?

   There are major points that need to be addressed before being convinced by the conclusions.

   1- The choice for using the SKNSH cell line, among others, to study the role of NXPH1/a-NRXN axis need to be explained. Indeed, whereas there is no cell line modeling low-risk neuroblastoma, authors should properly illustrate the basal expression of NXPH1/a-NRXN in the cell lines and not the ratio provided in Fig3E. If I understood well, according to Fig3F, only 1.5% of SKNSH cells express NRXN1. Could the authors provide FACS plots? And explain how they can sort high and low expressing cells among 1.5%? And then how do they justify using shRNA approach in a cell line in which only 1.5% are concerned cells? Also, the authors should prove the efficiency of the shRNA used on each specific target for both in vitro and in vivo studies by WB.

   Reply: We chose the SK-N-SH cell line for experimental assays based on the following facts: 1) the expression levels of NXPH1 and α-NRXN1/2, 2) this cell line showed the highest sphere-forming capacity, 3) this cell line harbored the highest percentage of α-NRXN1+ cells detected by FACS among the different cell lines tested, and 4) this cell line is of a mixed type, which is supposed to encompass more cell heterogeneity than other (N, I and S) types, thus reproducing more faithfully the complex heterogeneity of primary NB tumors.

   In a revised version of the manuscript we will provide data on NXPH1 and a-NRNX1/2 expression levels in basal culture conditions, and will improve the representation of Fig.3E to facilitate its comprehension. The method used to sort α-NRXN1+high, α-NRXN1+low and α-NRXN1- cells is explained in the Methods section (page 25). As requested by the reviewer #3, we will provide FACs plots to illustrate the sorting method.

   As requested by the reviewers #2 and #3, we will assess the shRNA efficiency by testing the knockdown at the protein level.

   2- The conclusion of inhibition of metastatic process is not supported by enough data. To achieve such a conclusion, authors should provide more than one in vivo experiment (which need to be completed already with the proof of protein deregulation). Some in vitro characterization of metastatic properties such as invasion and migration assays and/or transcriptional analyses could be done.

   Reply: We agree with the reviewer #3 that that our conclusion on the anti-metastatic ability of NXPH1/α-NRXN signaling would be reinforced by complementary experiments. To this aim, we propose to test how NXPH1/α-NRXN knockdown affects the metastatic potential of the SK-N-SH cell line and of another cell line in the CAM assay. This assay can indeed be used to assess not only NB tumor growth (as we already did), but also NB cell invasion in target organs (such as the liver and the bone marrow), thus mimicking a metastatic dissemination.

   3- Why don't the authors use the transcriptome dataset of 498 patients to realize a more powerful comparative study of low versus high risk tumors (Zhang et al, Genome biology, 2015)? The authors should show the expression plot of NXPH1/a-NRXN in low versus high risk patients, in their cohort of 18 patients but also in the cohort of 498 patients. How do the authors reconciliate the idea that NXPH1/a-NRXN could be associated to stem cell identity but low risk tumors?

   Reply: As explained in response to the reviewer #1’ (comment #1), our strategy entailed comparing the transcriptomic signatures of LR-NB and HR-NB samples to the one of a non-tumoral, healthy tissue such as the fetal adrenal gland. The online SEQC database cited by reviewers #1 and #3 does not contain healthy samples, and therefore could not be used for this initial step of the analysis. This is why we decided to use as a starting point the cohort of samples from the Hospital Sant Joan de Déu, which further provided the advantage of comparing samples from patients all diagnosed and treated at the same facility. We then used the SEQC database at different steps of our analytic strategy (Fig.S1F, 2F-H, 3A-B, S3) to test the relevance and coherence of the results obtained with the HSJD cohort in the context of a larger cohort.

   As requested by the reviewer #3, we will provide a dot-plot representation of the expression levels of NXPH1 and its receptors in both the HSDJ and SEQC cohorts.

   At this point we can only speculate on the association between NXPH1/α-NRXN expression and stem cell identity. This correlation might simply reflect the fact that cells from human NB cell lines return to a transcriptional program closer to their neural crest-derived identity when grown in sphere-forming conditions (as suggested by the increased expression of p75/NTR). Alternatively, this correlation might reflect the ability of NXPH1/α-NRXN signaling to retain cells in an immature state. Such ability could explain how NXPH1/α-NRXN signaling participates in promoting primary tumor growth, and the fact that their expression is increased in LR-NBs as compared to normal fetal adrenal gland. On the other hand, NXPH1/α-NRXN expression is higher in LR-NBs than in HR-NBs, and our findings suggest that this is linked to the anti-metastatic ability of NXPH1/α-NRXN. We could further imagine that by favoring stem cell identity NXPH1/α-NRXN signaling might also provide LR-NB cells with an enhanced ability to “re-enter” a normal developmental path or to be eliminated. In this regard, it is worth reminding that LR-NBs are detected earlier during development than HR-NBs, and sometimes show the puzzling ability to regress spontaneously.

   • Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

   Details of experiments and supplementary experiments have to be provided (see previous question).

   • Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.

   4- Additional details of the experiments have to be provided (cf above): levels of expression of NXPH1/a-NRXN in cell lines, FACS analyses to prove enrichment and/or depletion, WB for validation of shRNA knock-down etc...Moreover, the conclusion that NXPH1/a-NRXN axis inhibits metastatic potential of neuroblastoma is supported by only one in vivo experiment using SKNSH cell line with shRNAs anti-NXPH1/a-NRXN. It should be completed with invasion/migration assays in vitro for example, and/or transcriptional signature of tumors obtained +/- shRNAs anti-NXPH1/a-NRXN. Ideally, these results could be validated in an additional model.

   Reply: As mentioned above, we will perform additional experiments following the reviewer #3’s recommendations, including assessing NXPH1 and α-NRXN1 knockdown efficiency at the protein level, assessing how knocking down NXPH1 and α-NRXN1 expression alters the in vivo growth and metastatic potential of an additional cell line, and studying how knocking down NXPH1 and α-NRXN1 alters the metastatic potential of NB cells using a second and complementary metastasis assay (CAM assay).

   • Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.

   Yes

   • Are the data and the methods presented in such a way that they can be reproduced?

   Yes, Material and methods section is well described. But, several points are missing:

   5- The paragraphs describing how RNAs from 18 patients and fetal adrenal glands were obtained, how good was the quality, and how transcriptomes have been realized and sequenced are missing.

   Reply: The transcriptomic data used herein have indeed been previously generated by two co-authors of the study, and the corresponding reference is cited (ref #20; (Gomez et al, 2015). All the NB samples included in our transcriptomic analysis were obtained at the time of diagnosis from patients attended at Hospital Sant Joan de Déu (HSJD, Barcelona, Spain). Tumors were evaluated by a pathologist and only the snap-frozen pre-treatment samples showing at least 70% of viable tumor content were included for analysis. Neuroblastoma risk assessment was defined by the International Neuroblastoma Staging System (INSS). Samples of normal fetal adrenal gland (n=2) were used as a non-tumoral reference tissue. Total RNA from frozen samples was extracted by TRIzol® Reagent. High quality RNA (RINe>7.00) was hybridized to Human Genome U219 microarray plates at the Functional Genomic Unit, Institut d’Investigacions Biomèdiques August Pi i Sunyer (IDIBAPS, Barcelona, Spain) according to Affymetrix standard protocols. Microarray data are deposited in NCBI (#GSE54720). We will include this information in the Methods section of the revised version of the manuscript.

   6- The sequences of shRNAs have to be provided.

   Reply: We will provide the oligo sequences used to generate shRNAs against NXPH1 and aNRNX1.

   -Are the experiments adequately replicated and statistical analysis adequate?

   Yes

   **Minor comments:**

   • Specific experimental issues that are easily addressable.

   7- The authors should show the basal expression of their genes of interest in the patients, cell lines and PDX-derived cell lines, to justify the choice to work then with SKNSH cell line. In addition to the ratio they show in Fig3.E.

   Reply: As explained above, we will provide more details on the expression levels of NXPH1 and α-NRXN1/2 in the HSJD and SEQC cohorts and in the NB cell lines used in our study.

   8- Authors should also illustrate the FACS analyses they realized in this study, in order to appreciate the quantity of positive cells that are either enriched or depleted.

   Reply: The methodology used to sort α-NRXN1+high, α-NRXN1+low and α-NRXN1- cells is explained in the Methods section (page 25). As requested by the reviewer #3, we will provide FACs plots to illustrate this methodology.

   9- Authors could precise in their schemes that DOX is maintained in vivo.

   Reply: We will follow the reviewer #3’s suggestion.

   10- The number of mice need to be integrated in each experiment.

   Reply: The numbers of mice used to assess growth and metastasis in vivo were included in the corresponding figure legends (Figs. 4D, 4E and 5C) and appear discreetly on the panel 4D (to the right). We will follow the reviewer #3’s recommendation and add these details on the corresponding figure panels.

   • Are prior studies referenced appropriately?

   11- No some elements are not right in the introduction:

   • Mutations of PHOX2B are not associated to poor prognosis.

   • Original publications could be provided instead of reviews.

   • Genetic alterations in NB are not only 16%, the authors forgot to mention TERT and ATRX.

   • Maybe the NXPH1 methylation in ref 54 could be more explicit, if DNA methylation is detected on NXPH1, it would be of poor prognosis because driving low expression ..?

   Reply: We will follow the reviewer #3’s critics and correct the mistakes and information lacking in the introduction and discussion sections.

   • Are the text and figures clear and accurate?

   Text is very clear and well written, as are the figures.

   • Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

   Already mentioned before: FACS and WB are needed.

   Reviewer #3 (Significance (Required)):

   • Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field. This work aimed first at better understanding the fundamental transcriptional differences between two risk groups in Neuroblastoma. The authors defend the conceptual idea that these two groups represent two distinct diseases, which is not new in the area but requires attention, and indeed supported here by distinct transcriptional signatures.

   Using published signatures of fetal sympatho-adrenal system, they define a core transcriptional program that is more strongly expressed in low-risk tumors, but we already know that tumors of this category are often more differentiated, and by definition expressed strong markers of SA differentiation, whereas high-risk tumors have a more undifferentiated phenotype.

   Not surprisingly, several genes as well as a specific gene signature could be associated to better prognosis, a well-known characteristic of low-risk tumors. However, among them, a novel axis (NXPH1/a-NRXN) is proposed to explain the proliferation but absence of metastasis that define the low-risk group.

   • Place the work in the context of the existing literature (provide references, where appropriate).

   Neuroblastoma is a rare and very heterogeneous disease, in terms of biology and clinical presentation. To try to decipher such a heterogeneity, recent works have allied single cell transcriptome analyses on tumors and on human fetal cells during development. These studies are well cited in the discussion, and I agree that they yielded discrepant conclusions concerning the cell(s) of origin of Neuroblastoma. By comparing the normal developing human adrenal gland cells to cells from series of neuroblastomas, most of the studies converge towards that the tumors resembled differentiating adrenal neuroblasts. In one study, MYCN-amplified neuroblastoma cells (high risk group) were most similar to normal neuroblasts from seven- or eight-week post-conception, while lower-risk neuroblastomas included more cells resembling late neuroblasts (Janksy et al, 2021).

   During the submission of this work, another paper using single cell technologies was published and supported the idea of two distinct tumor entities (Bedoya-reina et al, 2021), with also a stronger signature of sympatho-adrenal cells in low risk tumors.

   • State what audience might be interested in and influenced by the reported findings. As the current clinical classification based on various criteria already allows clinicians to identify well low-risk tumors, I think this work would mainly attract fundamental researchers on the molecular differences between low-risk and high-risk tumors.

   • Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate. I'm specialized on pediatric cancers, and especially Neuroblastoma for the past 3 years. I'm interested in tumor cell identity and cell plasticity, particularly in response to treatment. I think that I have sufficient expertise to evaluate all parts of the manuscript.

   __ __

   3. Description of the revisions that have already been incorporated in the transferred manuscript

   Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript. If no revisions have been carried out yet, please leave this section empty.

   __ __

   4. Description of analyses that authors prefer not to carry out

   #1: The reviewers #1 and #3 suggested using a large dataset (the SEQC database containing 498 samples) to realize a more powerful comparative transcriptomic study. However, to define the transcriptional signatures associated to the aetiology of LR- and HR-NBs and to the malignant behavior of HR-NBs, we first needed to compare the transcriptomic signatures of LR-NB and HR-NB samples to the one of a non-tumoral, healthy tissue such as the fetal adrenal gland. The online SEQC does not contain any healthy samples, thus precluding the possibility to use it for the first step of our analysis. This is why we decided to use as a starting point the cohort of samples from HSJD cohort, which further provided the advantage of comparing samples from patients all diagnosed and treated at the same facility. In different steps of our analytic strategy (Fig.S1F, 2F-H, 3A-B, S3), we already used the SEQC database to test the relevance of the results in the context of a larger cohort. In doing so we obtained coherent results. As recommended by the reviewer #1 (comment #3), we will further compare our data with the data from another study (Bedoya-Reina et al, 2021). We thus believe that the request of the reviewers #1 and #3 does not need to be addressed within the scope of a revision.

   #2: The reviewer #2 suggested that some crucial functional experiments should be repeated with another shRNA construct to ensure that our results are not off-target and because the knock-down efficiency of the shRNAs used was variable and not convincing. We have tested two distinct shRNA constructs for both NXPH1 and α-NRXN1, which all comparably reduced cell viability. Following this reviewer’s suggestion we will assess possible off-target effects of the sh-NXPH1 and sh-aNRXN1 constructs by RT-qPCR. Moreover, we will also test the effects of these sh-NXPH1 and sh-aNRXN1 constructs in another cell line and using a novel and complementary metastasis assay (CAM assay).

   #3: The reviewer #2 suggested to study the metastatic potential of NB cells by performing orthotopic implantations of NB cells into the mouse adrenal gland instead of performing cell injections into the mouse left cardiac ventricle. We reasoned that cell injections into the left cardiac ventricle were ideal to test the organo-tropism of metastatic NBs, as this methodology facilitates cell dissemination and colonization of organs targeted by NB metastasis such as the liver and bone marrow, as shown in Fig.5B. Cell implantation into the adrenal gland is especially useful to study cell growth in one of the primary sites of NB formation. We thus believe that our current experimental approach is more relevant to the question we wish to address within the scope of a revision.

   References:

   Bedoya-Reina OC, Li W, Arceo M, Plescher M, Bullova P, Pui H, Kaucka M, Kharchenko P, Martinsson T, Holmberg J, et al (2021) Single-nuclei transcriptomes from human adrenal gland reveal distinct cellular identities of low and high-risk neuroblastoma tumors. Nat Commun 12: 1–15

   Biedler JL, Helson L & Spengler BA (1973) Morphology and Growth, Tumorigenicity, and Cytogenetics of Human Neuroblastoma Cells in Continuous Culture. Cancer Res 33: 2643–2652

   Ciccarone V, Spengler BA, Meyers MB, Biedler JL & Ross RA (1989) Phenotypic Diversification in Human Neuroblastoma Cells: Expression of Distinct Neural Crest Lineages. Cancer Res 49: 219–225

   Decock A, Ongenaert M, Cannoodt R, Verniers K, Wilde B De, Laureys G, Van Roy N, Berbegall AP, Bienertova-Vasku J, Bown N, et al (2016) Methyl-CpG-binding domain sequencing reveals a prognostic methylation signature in neuroblastoma. Oncotarget 7: 1960–72

   Gomez S, Castellano G, Mayol G, Sunol M, Queiros A, Bibikova M, Nazor KL, Loring JF, Lemos I, Rodriguez E, et al (2015) DNA methylation fingerprint of neuroblastoma reveals new biological and clinical insights. Epigenomics 7: 1137–1153

   van Groningen T, Akogul N, Westerhout EM, Chan A, Hasselt NE, Zwijnenburg DA, Broekmans M, Stroeken P, Haneveld F, Hooijer GKJ, et al (2019) A NOTCH feed-forward loop drives reprogramming from adrenergic to mesenchymal state in neuroblastoma. Nat Commun 10: 1–11

   Kildisiute G, Kholosy WM, Young MD, Roberts K, Elmentaite R, van Hooff SR, Pacyna CN, Khabirova E, Piapi A, Thevanesan C, et al (2021) Tumor to normal single-cell mRNA comparisons reveal a pan-neuroblastoma cancer cell. Sci Adv 7: eabd3311

   Tumilowicz JJ, Nichols WW, Cholon JJ & Greene AE (1970) Definition of a continuous human cell line derived from neuroblastoma. Cancer Res 30: 2110–2118

   Read the original source
2. 

   Review Commons

   Nov 14, 2023

   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

   Summary: short summary of the findings and key conclusions (including methodology and model system(s) where appropriate).

   Dr. Le Dréau and colleagues provide a transcriptional analysis comparing low versus high-risk neuroblastomas using a cohort of 18 patients. By comparing them to normal transcriptional signatures obtained from human fetal adrenal glands, they conclude that low risk tumors share a specific core sympatho-adrenal developmental program, which is associated with favorable patient prognosis. Among this signature, they specifically assess the role of NXPH1/a-NRXN in vitro and in vivo using different models (cell lines, …

   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

   Summary: short summary of the findings and key conclusions (including methodology and model system(s) where appropriate).

   Dr. Le Dréau and colleagues provide a transcriptional analysis comparing low versus high-risk neuroblastomas using a cohort of 18 patients. By comparing them to normal transcriptional signatures obtained from human fetal adrenal glands, they conclude that low risk tumors share a specific core sympatho-adrenal developmental program, which is associated with favorable patient prognosis. Among this signature, they specifically assess the role of NXPH1/a-NRXN in vitro and in vivo using different models (cell lines, PDX-derived cell lines, mouse and chick xenografts). They propose that NXPH1/a-NRXN axis promotes neuroblastoma tumor growth via cell proliferation, and inhibits metastases.

   Major comments:

   • Are the key conclusions convincing? There are major points that need to be addressed before being convinced by the conclusions. • The choice for using the SKNSH cell line, among others, to study the role of NXPH1/a-NRXN axis need to be explained. Indeed, whereas there is no cell line modeling low-risk neuroblastoma, authors should properly illustrate the basal expression of NXPH1/a-NRXN in the cell lines and not the ratio provided in Fig3E. If I understood well, according to Fig3F, only 1.5% of SKNSH cells express NRXN1. Could the authors provide FACS plots? And explain how they can sort high and low expressing cells among 1.5%? And then how do they justify using shRNA approach in a cell line in which only 1.5% are concerned cells? Also, the authors should prove the efficiency of the shRNA used on each specific target for both in vitro and in vivo studies by WB. • The conclusion of inhibition of metastatic process is not supported by enough data. To achieve such a conclusion, authors should provide more than one in vivo experiment (which need to be completed already with the proof of protein deregulation). Some in vitro characterization of metastatic properties such as invasion and migration assays and/or transcriptional analyses could be done. • Why don't the authors use the transcriptome dataset of 498 patients to realize a more powerful comparative study of low versus high risk tumors (Zhang et al, Genome biology, 2015)? The authors should show the expression plot of NXPH1/a-NRXN in low versus high risk patients, in their cohort of 18 patients but also in the cohort of 498 patients. How do the authors reconciliate the idea that NXPH1/a-NRXN could be associated to stem cell identity but low risk tumors?
   • Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether? Details of experiments and supplementary experiments have to be provided (see previous question).
   • Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation. Additional details of the experiments have to be provided (cf above): levels of expression of NXPH1/a-NRXN in cell lines, FACS analyses to prove enrichment and/or depletion, WB for validation of shRNA knock-down etc...

   Moreover, the conclusion that NXPH1/a-NRXN axis inhibits metastatic potential of neuroblastoma is supported by only one in vivo experiment using SKNSH cell line with shRNAs anti-NXPH1/a-NRXN. It should be completed with invasion/migration assays in vitro for example, and/or transcriptional signature of tumors obtained +/- shRNAs anti-NXPH1/a-NRXN.

   Ideally, these results could be validated in an additional model.

   • Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments. Yes
   • Are the data and the methods presented in such a way that they can be reproduced? Yes, Material and methods section is well described. But, several points are missing:
   • The paragraphs describing how RNAs from 18 patients and fetal adrenal glands were obtained, how good was the quality, and how transcriptomes have been realized and sequenced are missing.
   • The sequences of shRNAs have to be provided.
   • Are the experiments adequately replicated and statistical analysis adequate? Yes

   Minor comments:

   • Specific experimental issues that are easily addressable. The authors should show the basal expression of their genes of interest in the patients, cell lines and PDX-derived cell lines, to justify the choice to work then with SKNSH cell line. In addition to the ratio they show in Fig3.E.

   Authors should also illustrate the FACS analyses they realized in this study, in order to appreciate the quantity of positive cells that are either enriched or depleted.

   Authors could precise in their schemes that DOX is maintained in vivo. The number of mice need to be integrated in each experiment.

   • Are prior studies referenced appropriately? No some elements are not right in the introduction:
   • Mutations of PHOX2B are not associated to poor prognosis.
   • Original publications could be provided instead of reviews.
   • Genetic alterations in NB are not only 16%, the authors forgot to mention TERT and ATRX.
   • Maybe the NXPH1 methylation in ref 54 could be more explicit, if DNA methylation is detected on NXPH1, it would be of poor prognosis because driving low expression ..?
   • Are the text and figures clear and accurate? Text is very clear and well written, as are the figures.
   • Do you have suggestions that would help the authors improve the presentation of their data and conclusions? Already mentioned before: FACS and WB are needed.

   Significance

   • Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field.. This work aimed first at better understanding the fundamental transcriptional differences between two risk groups in Neuroblastoma. The authors defend the conceptual idea that these two groups represent two distinct diseases, which is not new in the area but requires attention, and indeed supported here by distinct transcriptional signatures.

   Using published signatures of fetal sympatho-adrenal system, they define a core transcriptional program that is more strongly expressed in low-risk tumors, but we already know that tumors of this category are often more differentiated, and by definition expressed strong markers of SA differentiation, whereas high-risk tumors have a more undifferentiated phenotype.

   Not surprisingly, several genes as well as a specific gene signature could be associated to better prognosis, a well-known characteristic of low-risk tumors. However, among them, a novel axis (NXPH1/a-NRXN) is proposed to explain the proliferation but absence of metastasis that define the low-risk group.

   • Place the work in the context of the existing literature (provide references, where appropriate). Neuroblastoma is a rare and very heterogeneous disease, in terms of biology and clinical presentation. To try to decipher such a heterogeneity, recent works have allied single cell transcriptome analyses on tumors and on human fetal cells during development. These studies are well cited in the discussion, and I agree that they yielded discrepant conclusions concerning the cell(s) of origin of Neuroblastoma. By comparing the normal developing human adrenal gland cells to cells from series of neuroblastomas, most of the studies converge towards that the tumors resembled differentiating adrenal neuroblasts. In one study, MYCN-amplified neuroblastoma cells (high risk group) were most similar to normal neuroblasts from seven- or eight-week post-conception, while lower-risk neuroblastomas included more cells resembling late neuroblasts (Janksy et al, 2021).

   During the submission of this work, another paper using single cell technologies was published and supported the idea of two distinct tumor entities (Bedoya-reina et al, 2021), with also a stronger signature of sympatho-adrenal cells in low risk tumors.

   • State what audience might be interested in and influenced by the reported findings. As the current clinical classification based on various criteria already allows clinicians to identify well low-risk tumors, I think this work would mainly attract fundamental researchers on the molecular differences between low-risk and high-risk tumors.
   • Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate. I'm specialized on pediatric cancers, and especially Neuroblastoma for the past 3 years. I'm interested in tumor cell identity and cell plasticity, particularly in response to treatment. I think that I have sufficient expertise to evaluate all parts of the manuscript.

   Read the original source
3. 

   Review Commons

   Nov 14, 2023

   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

   Summary:

   Provide a short summary of the findings and key conclusions (including methodology and model system(s) where appropriate). Please place your comments about significance in section 2.

   In this paper, Fanlo-Escudero et al., determines gene signatures differentiating low-risk (LR) neuroblastoma (NB) from high-risk neuroblastoma (HR-NB), as well as LR- and HR-NB as an entity from human fetal adrenal gland. They identify a transcriptional signature corresponding to a core sympathoadrenal lineage that can discriminate between LR-NB and HR-NB. This signature is composed of genes associated with favorable patient outcome. The …

   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

   Summary:

   Provide a short summary of the findings and key conclusions (including methodology and model system(s) where appropriate). Please place your comments about significance in section 2.

   In this paper, Fanlo-Escudero et al., determines gene signatures differentiating low-risk (LR) neuroblastoma (NB) from high-risk neuroblastoma (HR-NB), as well as LR- and HR-NB as an entity from human fetal adrenal gland. They identify a transcriptional signature corresponding to a core sympathoadrenal lineage that can discriminate between LR-NB and HR-NB. This signature is composed of genes associated with favorable patient outcome. The authors further choose one gene, NXPH1, for functional analysis and investigates the effects this gene has on NB progression using in vitro assays, chick CAM assay and mouse in vivo models. The authors conclude that this transcriptional signature can distinguish LR-NB from HR-NB and that NXPH1 is involved in NB cell growth.

   Major comments:

   • Are the key conclusions convincing?

   The key conclusions are 1) a core SA lineage signature can discriminate between LR-NB and HR-NB, and 2) NXPH1 represses NB malignancy (in terms of metastatic capacity) and is a therapeutic target. The first conclusion is indeed convincing, and not contradictive to common beliefs. The second conclusion is poorly supported by data. The authors perform a range of experiments using in vitro and in vivo settings, but lack some fundamental experiments and overstate their findings.

   • Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

   Several claims should be softened, re-phrased and/or clearly marked as preliminary or speculative. No data nor claims need to be removed.

   For example, the following statements need to be changed:

   • Page 10, second paragraph: The authors state "Remarkably, NXPH1 and α-NRXN1/2 levels increased in all the NB cell lines harbouring a sphere-forming capacity (Fig. 3E), thereby revealing a strong positive correlation between the expression of NXPH1 and α-NRXN1/2 and the acquisition of a NCC stem cell identity". The authors only show that NXPH1 is expressed in 8 out of 10 NB cell lines. Sphere-forming capacity is displayed in a relative and not absolute scale which makes it difficult to assess which cell lines that do form spheres and to what extent. The capacity to form spheres (from low to high) does not correlate to the levels of NXPH1 in the different cell lines.
   • Page 13, paragraph 1. The authors write "...these data revealed that NXPH1/α-NRXN1 signaling is necessary and sufficient for NB tumor growth in vivo". This is an overstatement. Tumors still form, meaning that NXPH1 signaling is not sufficient.
   • Throughout the text, the authors convert their statements. One example is page 15, first paragraph. They write "...growth of NB cells but markedly restrict their metastatic potential", but they do not show this. Instead, they only address the opposite situation - Knockdown enhances metastasis. This is not equal to their statement. See other experiments in other sections. The authors need to go through the manuscript and make sure that they explain their conclusions to actually fit their experiments. • Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary to evaluate the paper as it is, and do not ask authors to open new lines of experimentation.

   Issues to respond to:

   • The authors should more clearly explain why they choose to further study NXPH1 (second highest on the list), involved in synaptogenesis and neurotransmission, instead of SOX6, highest on the list, that is highly relevant in neural crest development.
   • When the authors investigate the expression of NXPH1 and other genes and compare that with sphere-forming capacity (Fig. 3E and page 10) they analyze expression in cells cultured in basal medium while sphere-forming capacity is measured after 5 weeks in restricted medium. How does the expression of the analyzed genes change under these conditions?
   • Is the subpopulation of NRXN1+ cells low (1.5%) because these samples are from aggressive ("High-risk") cell lines?
   • The authors state "...the number of cells quantified per tumor section was decreased by ~50% for the α-NRXN1+-deprived cells relative to their control (Fig. 3J, K), thus revealing that α-NRXN1+ cells are required to support NB tumor growth in vivo". This is not a correct conclusion. This experiment shows that a-NRXN1- cells do not grow and expand to the same extent as control cells. They cannot say that a-NRXN1+ cells support NB growth without comparing growth between a pure a-NRXN1+ and control cells.
   • The authors use shRNAs to knock down NXPH1. They enrich their cells by two means - puromycin or doxocycline. This results in equal cell populations. The authors however state that they use doxocycline to circumvent the growth arrest they observe with puromycin selection. They need to elaborate on this and show why this would be the case and what difference the two methods do and show.
   • The major flaw of this paper is that the authors use one cell line in total, and even more that they use the same cell line for both knockdown and activation. Since they do show that different NB cell lines have different expression levels (ranging from high to absent), they should choose one cell line for KD and one for overexpression. The authors could also do a rescue experiment with knockout and gain-of-function (e.g., construct that will not be targeted by the shRNA) in the same cells.
   • They only use one shRNA after trying several (Fig. S4). The efficiency is substantially variable and not convincing. As stated also elsewhere in this review, they need to check protein level. And to ensure that their results are not off-target they should perform at least some crucial experiments with two shRNAs.
   • Why don't the authors add BrdU post-implantation? This is easily done in the egg considering the accessibility and would better reflect the proliferation in vivo.
   • Why do the authors switch between CAM and mouse xenografts? I understand that the mouse model must be employed for "metastasis", but can it be explained why and when they perform the different "tumor growth" experiments?
   • Why do the authors do left ventricle injections for metastatic studies and not AG implantations?
   • The measurement of bioluminescence is very difficult to interpret. The authors discuss metastatic spread, but the images show only large blobs covering the heart and areas surrounding it, especially for the sh-aNRXN1. To be able to say that the cells have colonized specific organs the authors need to dissect these organs and perform staining. I see it as tumors recur rather than particularly metastasize when they re-appear after 6 weeks.
   • How do the authors explain results presented in Fig. S5: There are more HuNu+ cells but tumor size is unchanged? • Are the suggested experiments realistic for the authors? It would help if you could add an estimated cost and time investment for substantial experiments.

   Some suggested experiments take time (orthotopic implantations, rescue experiments, adding cell lines and # of shRNAs).

   However, as long as the authors discuss and address their methods for in vivo growth (in ovo vs in vivo vs their choice of metastasis model), an orthotopic AG model is not necessary. The authors should however consider it for future studies.

   Experiments required to support their conclusions: A rescue experiment and use of different cell lines for KD and overexpression is somewhat time-consuming, as well as ensuring KD at protein level and include an additional shRNA in crucial experiment. I expect that this would take 2-4 months.

   • Are the data and the methods presented in such a way that they can be reproduced?

   The authors should elaborate on their methods, but in general they are reproducible. • Are the experiments adequately replicated and statistical analysis adequate?

   In several places the number of replicates is questionable. Especially at the end of page 26, the authors state that they have performed n=1-4 replicates. N=1 replicate is never ok. In several instances they do n=2 replicates. This can be acceptable but the authors could address this.

   Minor comments:

   • Specific experimental issues that are easily addressable.

   Mainly text changes can be seen as minor. For experiments and other issues, see other sections. • Are prior studies referenced appropriately? Yes. The reference list is not coherently styled. • Are the text and figures clear and accurate?

   Overall, the paper is written in a complex way and difficult to easily comprehend. For example, the authors need to clarify several issues on the material they use and experiments they perform. I suggest substantial re-writing to better convey their messages. Sentences should be short and clear, data explained in the context of it was derived.

   The following text edits and clarifications are required:

   • Page 3. The authors write "cell-of-origin" in several places, this should be changed to "cells-of-origin" (i.e., plural). The view that all NBs originate from only one cell is too simplistic, and the authors should definitely edit this considering that they are investigating different subgroups of NB.
   • Page 5. The authors MUST define where the material from the 18 NB patients as well as fetal AG derive from. There is no reference, and taken from the material&methods section, the transcriptome data from these data has not been generated by the authors themselves?
   • Why do the authors choose to work with the data set from Jansky et al in particular?
   • Page 7, paragraph 1. The authors write "Remarkably, 67% of the 503 genes found in this signature, which formed the cluster c1, are both associated with a neural identity and a better patient outcome". I do not agree that this is remarkable and the authors should remove this word.
   • Page 10, second paragraph. The authors should clarify that the expression levels of the displayed genes are derived from qPCR analysis (if I have understood that correctly, I had to find and guess this from the M&M section), as well as explain how they have set the scale for sphere-forming capacity and what this corresponds to. What do the colors actually represent?
   • Page 11 and onwards. The authors write "deprived of their a-NRXN1+ subpopulation". This is highly confusing and difficult to read. The authors should write "a-NRXN1- subpopulation"
   • Page 11, end of paragraph 2. The authors write "...arguing that NXPH1/α-NRXN signaling could control NB growth and/or aggressiveness". Number of cells do not directly correlate to aggressiveness, and this needs to be re-phrased to only state what the experiment actually shows - proliferation of a-NRXN1- cells.
   • I am in favor of using the CAM assay as a complementary system. The authors however use this to define "...required to support tumor growth in vivo". The CAM assay using NB (i.e., transformed cells) shows the growth of these cells in response to presence of blood vessels and not a full tumor micro-environment. This should be clarified.
   • As stated in the previous comment, the authors write "...required to support tumor growth in vivo". The next paragraph has the following headline "NXPH1/α-NRXN signaling stimulates NB growth". This is to me the same thing. The authors should elaborate on how these differ, or if they use them to show the same thing, clearly state that.
   • The authors conclude that NXPH signaling can be used as a therapeutic target. This would however be extremely difficult considering the opposing effects shown on growth vs metastasis. I agree that it is important to find means to inhibit metastasis, but that does not mean we can allow for enhanced growth of the primary tumor. A better reflection would be to use this as a biomarker, but this can only be predictive/speculative since the authors do not perform for example a tissue microarray to show this at IHC protein level, something that is currently the practice in the clinic.

   Material and Methods:

   The authors have misspelled the cell line SK-N-BE(2)c. Why are some cell lines grown in 20% FBS? This is not standard and could impact the results. The authors should state what the tumor volume limit in their ethical permit is (page 30).

   • Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

   • Fig. 3E. As discussed elsewhere, the authors should clarify the scale/meaning for sphere-forming capacity
   • Fig. 4D. What do the numbers (i - vi) refer to? I cannot find this in the figure, figure legend, text or material&methods.
   • The authors do not present the tumor volume in Fig. 4. The authors discuss tumor growth in the text and this data should be included.
   • Fig. S4. Knockdown efficiency is variable, and efficiency does not correlate to growth capacity. Especially because of this, the authors need to investigate this at protein level.

   Significance

   • Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field.

   This paper is partly within the scope of the ongoing debate of NB origin (see references below). This paper is however not as extensive, provides no new patient material (applies data from Jansky et al), and do not address the actual cell-of-origin (which the authors themselves also clearly states). This paper provides new gene signatures that can be used to define low-risk vs high-risk patients which is highly important to the field, but these signatures are not unexpected and does not add a significant advance to the field. The authors also do not address how these signatures could be applied clinically. The authors do not use any new methodology. With this said, with revision of the paper, it will still add to the current focus on NB biology research.

   • Place the work in the context of the existing literature (provide references, where appropriate).

   There is a recently initiated, important and extensive debate about the cells-of-origin for NB. The authors indeed bring this up in the paper and also state that they do not intend to add to this debate. They use data from one paper from referenced debate above, and I would argue that because of this fact, and that they extract gene signatures from it, they do, at least partly, touch on the NB cells-of-origin debate, and the authors should put their results into context, from a big picture perspective. As of now, they dodge this complex issue. References: Dong et al., Cancer Cell 2020; Hanemaaijer et al., PNAS 2021; Jansky et al., Nat Genet 2021; Kameneva et al., Nat Genet 2021; Kildisiute et al., Sci Adv 2021; Furlan et al., Science 2017)

   • State what audience might be interested in and influenced by the reported findings.

   This paper will be interesting for scientists within the neuroblastoma field, in particular those working on defining NB subgroups in correlation to developmental stages.

   • Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate.

   Reviewer's expertise: Neuroblastoma, neural crest, trunk neural crest, chick embryos, mouse models, in vitro models

   Parts of paper outside expertise of the reviewer: Analysis of the bioinformatics data (i.e., extracting signatures).

   Read the original source
4. 

   Review Commons

   Nov 14, 2023

   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:

   In the reviewed manuscript, the authors aimed to characterize the sympathoadrenal (SA) transcriptional landscape that defines low- and high-risk neuroblastomas (LR-NBs and HR-NBs respectively). In particular, they analyze previously published Affymetrix U219 expression profiles of 18 low- (n=8) and high-risk (n=10) neuroblastomas, and 2 fetal human adrenal glands. The authors define transcriptional signatures of LR- and HR-NBs, and further unbiasedly classified them in 4 clusters defining groups of patients with different prognosis (as tested using the 498 SEQC cohort). Within these transcriptional signatures, the authors …

   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:

   In the reviewed manuscript, the authors aimed to characterize the sympathoadrenal (SA) transcriptional landscape that defines low- and high-risk neuroblastomas (LR-NBs and HR-NBs respectively). In particular, they analyze previously published Affymetrix U219 expression profiles of 18 low- (n=8) and high-risk (n=10) neuroblastomas, and 2 fetal human adrenal glands. The authors define transcriptional signatures of LR- and HR-NBs, and further unbiasedly classified them in 4 clusters defining groups of patients with different prognosis (as tested using the 498 SEQC cohort). Within these transcriptional signatures, the authors delineated a SA signature using a human fetal adrenal gland transcriptional profile recently published (Kildisiute et al. 2021), that can discriminate low-risk neuroblastomas. From these genes, the authors further select NXPH1 and NRNX1-2 as promising targets for extensive experimental in vitro and in vivo validation, including validation in cell cultures, and xenografts, to determine that NXPH1/alpha-NRXN1-2 signaling is sufficient for NB tumor growth, and that the expression of either stimulates the metastatic potential of NB cells.

   Major comments:

   The cohorts and data used by the authors to conduct the main analysis of the paper are already published, and thus the contribution of the analysis is incremental. In particular, the authors analyzed arrays from a limited cohort size, in comparison with others available sequenced with RNA-seq (e.g. 176 HR- and 322 non-HR NB in 498 SEQC; 224 HR- and 342 non-HR NBs included in the Westermann-genecode19 cohort; and 80 HR- and 20 non-HR in the Jagannathan cohort). Furthermore, the 12,000 most-expressed genes (out of ~20,000 available) were analyzed by the authors, as opposed to more than 40,000 (coding and non-coding) included in a normal RNA-seq study. Only the 498 SEQC dataset provides 12,000 genes significantly up-regulated in either high-risk (n~5,500) or non-high-risk (n~7,000). The differences between datasets could influence the results of the study. For example, in the reviewed manuscript, genes with a high expression in LR-NB (compared to fAG) included DNMT3B, SEMA5A, SOX5, and TET1, all of which have a significantly higher expression in HR-NBs of the 498 SEQC cohort. The quality of the manuscript will be enhanced with consistent results obtained by conducting the same reported analysis in larger cohorts.

   In the reviewed manuscript, PHOX2A and PHOX2B are significantly more expressed in both LR-NB and HR-NB compared to fAG. This is also the case for other adrenergic markers including TH and DBH. Oppositely, the expression of cortex markers (i.e. STAR and CYP11A1) is significantly higher in fAG. Nevertheless PNMT is not significantly up-regulated in fAG in comparison to LR-NB nor HR-NB. Is it possible that the fetal adrenal glands analyzed include a large proportion of cortex that confounds the transcriptional signals? The quality of the manuscript will be enhance if the authors could establish what proportion of the fAG transcriptional signal belongs to cortex, and if they account for its influence in the analysis.

   A recent published paper (Bedoya-Reina et al. 2021) study the differences of HR-NB and LR-NB from a single-cell perspective. In the published manuscript, the authors conclude that LR-NBs are enriched in cells that resemble chromaffin and sympathoblast cells, while the high-risk neuroblastomas are enriched in undifferentiated cells that resemble cells with progenitor characteristics in post-natal adrenal gland. This is broadly consistent with the conclusion reached by the authors in the manuscript under review. It will enhance the content of the reviewed manuscript if the authors compare their transcriptional signatures with the recently published transcriptional signatures in this paper to answer the following questions: 1) to what extent the transcriptional signatures for HR-NBs (4 hierarchical clusters incl.) in the manuscript under review resembles that from the published undifferentiated cluster (nC3) enriched in HR-NBs, and the progenitor cluster (hC1) in post-natal adrenal gland; 2) to what extent the transcriptional signatures for LR-NBs (4 hierarchical clusters incl.) in the manuscript under review resembles that from the published NOR (nC7, nC8, and nC9) enriched in LR-NBs, and that from the chromaffin cells (hC4) enriched in post-natal adrenal gland; and 3) how is the expression of NXPH1 and alpha-NRXN1-2 in the reported LR- and HR-NBs, and adrenal gland.

   In the discussion, the authors indicate that they do not aim to identify the transcriptional signature associated to NB origin but rather use the component of the SA lineage that distinguish LR- and HR-NBs. This statement implies that neuroblastoma can originate from any cell in the developing SA lineage (i.e. SCP, bridge, chromaffin and sympathoblast), a controversial assumption that requires further proof. In particular, when discussing about the core sympathoadrenal signatures enriched in LR-NBs and HR-NBs, the authors obtained a SA signature of genes shared by at least 3 of the 4 SA cell signatures. Further justification needs to be provided as for why (in particular) one of these SA cell signatures exclude the sympathoblast/neuroblast contribution.

   Some of the most interesting results in the paper are limited to proportions in a subset of top-ranked genes. It will be valuable to set the analysis in an hypothesis driven context, add probabilities, test names, and corrected p-values to the results.

   Minor comments.

   1. In comparison with other cohorts that include low- and intermediate-risk NBs as non-HR NBs, the reviewed data specifically includes low- and high- risk NBs. It is important that the authors include a characterization of intermediate-risk neuroblastomas in their analysis.

   2. Further details and figures on what precise criteria was used to remove the sample #LR-08 is required. How including this sample changes the reported results?

   3. GO-term distribution was assessed using the 50 most-enriched GO-terms. How would the results change if all the significant GO terms were analyzed?

   4. Was the SEQC 498 (GSE62564) dataset obtained with microarrays (as indicated in the methods) or with RNA-seq (i.e. Illumina HiSeq 2000)?

   5. In methods, the first quartile (Q1) in SA-c1 has a higher limit in 487 samples and the fourth quartile the lowest limit in 4, how many samples (out of 498 NBs) were excluded and why?

   6. In the 503 DEGs between LR-HR NBs, NTRK2 and MYCN are not included, even if the HR samples included MYCN amplified tumors. Can the authors comment on this?

   7. The authors mention that the top 30 genes found in cluster c1 (and also in c2) are correlated with favorable patient prognosis. Is it the case that all the genes in c1 (and also c2, c3 and c4) are significantly associated with a favorable or else unfavorable prognosis?

   8. The high expression of a (significant?) number of genes in cluster c4 is observed in patients with worst outcome (i.e. lower event-free survival), including ATR, HIF1A, ING2, POLR2L, SRPRB (498 SEQC, analyzed with R2).

   9. Regarding the 242 genes in the core SA signature, although its a smaller number, the expression of several genes in the core SA signature with a higher expression in HR compared to LR belonging to clusters 2, 3, and 4 is observed in worst outcome patients in the 498 SEQC cohort (CHD7, DNMT1, HMGA1, HSD17B12, LBR, LSM7, MCM4, NKAP, POLA1, and others). Is this small fraction significant?

   10. In Kildisiute et al. 2021, NRXN1 is expressed in SCPs, while NXPH1 is expressed in bridge, chromaffin and sympathoblastic cells. How are the microenviroment of these cells regulating the expression of these genes in a developmental context (particularly as sympathoblastic cells are know to have larger proliferative capabilities than SCPs)? how is this cell heterogeneity replicated by a NB cell line? are mesenchymal and adrenergic cells expressing differentially NRXN1 and NXPH1?

   11. Figure 1B and C, 2B,D: might the information provided be enhanced? otherwise these inserts might be excluded.

   12. Figure 3D: Kildisiute et al. 2021 data and GTEX available at human protein atlas indicate expression of NRXN1 and NXPH1 in developing and adult adrenal gland. Might the results illustrated suggest a confounding effect in the sampled fetal adrenal glands, perhaps from cortex?

   13. The authors conduct extensive experiments in NRXN1, and make conclusions about its role in for instance metastasis, nevertheless the LR-NB/HR-NB SA signal only includes NRXN2. Can the authors comment on the differences between NRXN1s and NRXN2?

   Significance

   The significance of the study relies in investigating the role of selected targets in neuroblastomas within a risk group. In particular, HR-NBs have poor outcomes and are generally metastatic at the time of diagnosis.

   The results of the manuscript are somehow consistent with a recently published manuscript analyzing LR- and HR-NBs from a single-cell perspective. The manuscript will be enhanced by conducting the suggested comparison between the reviewed and the reported results. The authors further need to comment why HR-NBs markers, particularly MYCN is not recovered in the LR-NB/HR-NB and the LR-NB/HR-NB SA signals. Also they need to comment on possible confounding effects in the fetal adrenal gland.

   The paper is directed to a broader audience of cancer and developmental biologists, and computational biologist. Yet further statistical support needs to be provided.

   Read the original source
5. 

   Version published to 10.1101/2021.11.26.470092v1 on bioRxiv

   Nov 27, 2021