Timing of Smarcb1 and Nf2 inactivation determines schwannoma versus rhabdoid tumor development

Germline mutations of the SMARCB1 gene predispose to two distinct tumor syndromes: rhabdoid tumor predisposition syndrome, with malignant pediatric tumors mostly developing in brain and kidney, and familial schwannomatosis, with adulthood benign tumors involving cranial and peripheral nerves. The mechanisms by which SMARCB1 germline mutations predispose to rhabdoid tumors versus schwannomas are still unknown. Here, to understand the origin of these two types of SMARCB1-associated tumors, we generated different tissue- and developmental stage-specific conditional knockout mice carrying Smarcb1 and/or Nf2 deletion. Smarcb1 loss in early neural crest was necessary to initiate tumorigenesis in the cranial nerves and meninges with typical histological features and molecular profiles of human rhabdoid tumors. By inducing Smarcb1 loss at later developmental stage in the Schwann cell lineage, in addition to biallelic Nf2 gene inactivation, we generated the first mouse model developing schwannomas with the same underlying gene mutations found in schwannomatosis patients.

The manuscript by Vitte and colleagues addresses the issue of modeling schwannomas and rhabdoid tumors in mice, two tumors with very different aggressiveness that are both associated with mutations of SMARCB1. They show that P0-driven inactivation of Smarcb1 leads to the high penetrance occurrence of rhabdoid-like tumors in the cranial nerves and possibly in the meninges. In contrast, tumors are not observed when the inactivation of Smarcb1 is induced later in neural crest cell development. They have also generated double GFAP-driven Smarcb1 and Nf2 inactivation. These mice develop Schwannomas that are similar to the ones observed in single Nf2 mutant mice. This work constitutes undoubtedly an interesting new model to investigate the oncogenic role of Smarcb1 inactivation but, rather unfortunately, these double mutant mice develop tumors that are similar to the single NF2 mutant mice and therefore fail to model the 4-hits/3-steps hypothesis. While this manuscript relies on an impressive amount of work, some major concerns need to be addressed. 1-Though the Smarcb1-/-tumors they observed have phenotypic similarities with RTs it must also be noted that the tumor localization (mostly cranial nerves) are relatively unusual for human RTs even though sporadic case reports of such localizations (thoroughly quoted by the authors) have been published. This modeling of a relatively narrow spectrum of the human malignancy should be more precisely indicated and discussed. 2-The extensive IHC analyses are interesting but they still only investigate a small subset of markers. A more comprehensive expression profiling should be undertaken. This may allow performing a more thorough comparison with the numerous expression profiles that have been published for human and mouse Smarb1-deficient tumors and that are available in different databases. In particular recent data shows that different subgroups of RTs exist in humans and in mice. Such a profiling may enable a more precise identification of the corresponding human and mouse subgroup(s). 3-The title and abstract are somehow misleading. In fact the authors fail to develop a model of Smarcb1-induced Schwannomas as the double Smarcb1/NF2 mutants do not behave differently than the single NF2 mutant hence providing no clue to elucidate the respective roles of these two genes in schwannoma development. 4-The authors should be more cautious concerning the impact of their work on the identification of the RT cell-of-origin. They indeed show that neural crest cells can be transformed by Smarcb1 inactivation hence suggesting that a minority of RTs (see above comment) may have such a cellof-origin but this should not lead to a broad conclusion that this neural crest cell progenitor is, in general, the RT cell-of-origin. 5-The references need to be reviewed. Some important landmark papers on RTs are not quoted while the authors do reference a high number of relatively marginal reports. 6-The suggestion that the observed tumors correspond to the recently described SHH group is only supported by the expression of Gli1 which constitutes a relatively weak proof. Again, comprehensive transcriptome data and/or a clear demonstration of a nuclear localization of Gli1 would be necessary to further document this hypothesis. 7- Fig.8 is misleading as single Nf2-/-mice which behave as double Smarcb1/Nf2 mutant mice are not indicated. The authors cannot draw conclusions on the 4-hits/3-steps model. 8-Fig7 shows that whorls of Baf47 negative cells are only observed in the mGFAP-cre; Nf2-/-; Smarcb1-/-and not in the mGFAP-cre; Smarcb1-/-background. May this indicate that in this context the loss of Nf2 is necessary for the survival of Smarcb1 cells?
Reviewer #2 (Remarks to the Author): Expert in AT/RT The manuscript by Vitte et al. seeks to elucidate the timing of genetic events leading to schwannoma and a rare cancer called rhabdoid tumor.
The presented data is novel and explains long observed findings and the link between two completely different diseases. The manuscript is well structured and written and especially the pictorial part helps to understand the experimentation and gives in the end an excellent overview of potential events. The application of the 4-hit/3-step model nicely fits the authors line of evidence.
There is only very limited criticism: 1) the functional consequences of inactivation of the genes SMARCB1 and NF2 is mainly described in morphological terms. The results of some expression data is presented in Figure 5. However the consequences of gene inactivation of SMARCB1 and NF2 on the SWI/SNF complex and its remaining members is not addressed at all. It would be very interesting to see whether the complex loses all of its functions and composition or whether other mechanisms are responsible for the morphological change. It might be worthwhile to try and elucidate gene set enrichment in the different constructs. The limited data and discussion on molecular subtyping (shh) is a step in the right direction.
2) the authors should be encouraged to discuss a few findings in human patients in relation to their data. How for example would they explain the phenomenon of synchronous tumors? While the authors found Cre and SMARCB1 deletion in kidneys one of the major target organs of rhabdoid tumors, they did not see any tumors in the kidney. Furthermore there are some human families in which family members are affected by rhabdoid tumors while others have schwannomas. How can this data be reconciled?
3) As the presented mouse models are only the second model of rhabdoid tumors which consistently demonstrate intracranial RT without any other sign of malignancy outside the CNS the manuscript would benefit from a closer comparison to the recent manuscript by Han et al (Nat Comm 2016). 4) On a minor note it would be nice to see a proliferation marker when examining the mouse tumors (e.g. KI67 or mouse equivalent). All in all this is a beautiful manuscript with novel and important findings with interest not only for the specialist, but also for any researcher interested in the mechanisms of disease. Especially the influence of spatial and temporal factors is very interesting and has a potential to help understand many disease states.
I thus suggest to accept the manuscript following minor revisions.

Reviewer #3 (Remarks to the Author): Expert in Schwannomas
In this manuscript, Vitte et al examine the question of why SMARCB1 mutations result in the development of malignant rhabdoid tumors in some patients and benign schwannomas in others; answering this question required that they look both at whether there were windows of susceptibility that differed for these two tumor types and address the question of the cellular origin of rhabdoid tumors (which has been controversial up to now). To test the hypothesis that Smarcb1 inactivation in neural crest cells and the Schwann cell lineage is sufficient for schwannoma pathogenesis, the authors generated P0-CreC;Smarcb1flox/flox mice. They report that 65% of these mice developed aggressive tumors that resembled human rhabdoid tumors. In contrast, P0-CreB;Smarcb1flox/flox mice mostly died in utero or perinatally; those that did survive, died between 1.2-2 months of age with retroorbital or nasal cavity tumors. In contrast, Dhh-Cre;Smarcb1flox/flox mice did not develop tumors. GFAP-Cre;Smarcb1flox/flox mice developed tumors with rhabdoid features between 2 and 17 months of age. In contrast, GFAP-Cre;Smarcb1flox/flox;Nf2flox/flox mice developed schwannomas. Based on these observations, the authors argue that Smarcb1 loss in neural crest cells early in development leads to rhabdoid tumor formation, while Smarcb1 loss at a later developmental stage combined with Nf2 loss results in schwannoma pathogenesis.
The topic addressed in this manuscript is an important one and potentially of strong interest to the readership of Nature Communications. However, in its current form, the manuscript is very difficult to follow and the arguments presented in support of their conclusions are convoluted and not clearly supported by the data as currently presented. Points that the authors need to address to improve the manuscript are as follows: 1) Although the data in the manuscript are quite interesting, there are a number of sentences in the manuscript that are awkwardly phrased that make it difficult to follow the manuscript. For example, there is a sentence in the abstract that states "The single loss of Smarcb1 in neural crest cells is sufficient to initiate tumorigenesis…". The Results section in particular is very dense and difficult to follow; while this is in part, this is due to the complicated genotypes being presented, it also reflects the often decipherable sentences in this section of the paper. The manuscript would be much easier to follow if the authors thoroughly edited it so as to present their work in a more straightforward and less complicated manner. 2) A major theme of this manuscript is that the timing of Smarcb1 loss (and likely the stage at which this loss occurs during neural crest/Schwann cell development) dictates whether the mouse develops a schwannoma or a rhabdoid tumor. This argument is dependent upon knowing the timing with which the different Cre transgenes they use are expressed in their mouse models. However, with the exception of a statement very late in the Results (when discussing the GFAP-Cre driver), the authors provide no information in the Results about the times when these different Cre transgenes are expressed. This is finally presented in Figure 8, which is first referred to in the next to last paragraph of the Discussion. As a Result, I found myself struggling to make sense of their data (and argument) until the very end of the paper. The authors need to clearly indicate when each of these transgenes are active in the embryo when they first present the Cre driver to the reader. 3) In the first paragraph of the Results, the authors mention that some of the P0-CreC;Smarcb1flox/flox mice had bloated intestines with an emaciated wall and a reduced number of villi. Why did this occur? 4) The authors had their tumors examined by competent pathologists whose opinions I respect. However, many of the journal's readers will likely not know these individuals. Given that, the authors really need to do a better job of presenting the diagnostic criteria that were applied and a critical assessment of how their mouse tumors compared to their human counterparts. There clearly were some differences-for instance, the authors simply stated that their mouse rhabdoid tumors were negative for neurofilament and cytokeratin. Human AT/RTs commonly show foci with expression of both of these antigens and the authors made no mention of this. 5) The authors note that the P0-CreB transgene expresses Cre "in a larger cell compartment" than the P0-CreC transgene. Again, what does this mean? Is the timing of expression also different from P0-CreC? This is not mentioned in the text and it is not indicated in Figure 8. 6) It is not clear what the authors mean by the sentence that spans lines 200-203. If part of this sentence is trying to argue that the tumors could have originated from the leptomeninges, despite the absence of PGDS, it is unclear how the authors arrived at this possibility. 7) Why is the data included on the DHH-Cre;Smarcb1flox/flox mice (they didn't develop tumors)? Are they trying to argue that these mice help them narrow down the window when Smarcb1 loss results in rhabdoid tumor formation? That possibility seems unlikely, given that GFAP-Cre;Smarcb1flox/flox mice developed tumors with rhabdoid features between 2 and 17 months of age (this Cre driver turns on at E13.5 compared to Dhh-Cre, which turns on at E12). The significance of these negative data are unclear as currently presented. 8) Panel e in Figure 2 is not convincing and needs to be presented with a better image. In addition, the tumor images presented in Figure 2 are at very low power. Following the usual pathology dictum that "blue is bad", it seems likely that these are aggressive tumors, but it will not be possible to be sure that this is the case until higher power images are included.
Introduction (line 72) was changed to: "and sporadic case reports of RTs emanating from cranial nerves" The conclusion paragraph in the discussion (lines 466-472) was changed to: "In conclusion, we have shown that biallelic inactivation of Smarcb1 in the neural crest leads to the development of PNS and meningeal RTs, thus identifying the cell of origin for these RT subsets in an early neural crest population." 2-The extensive IHC analyses are interesting but they still only investigate a small subset of markers. A more comprehensive expression profiling should be undertaken. This may allow performing a more thorough comparison with the numerous expression profiles that have been published for human and mouse Smarb1-deficient tumors and that are available in different databases. In particular recent data shows that different subgroups of RTs exist in humans and in mice. Such a profiling may enable a more precise identification of the corresponding human and mouse subgroup(s). Moreover, the comparison with two existing mouse models (Han et al., Ng et al.) showed that mouse RTs clustered in three molecular subgroups irrespective of the model of origin. All these data were added to the Results and Discussion sections.
3-The title and abstract are somehow misleading. In fact the authors fail to develop a model of Smarcb1-induced Schwannomas as the double Smarcb1/NF2 mutants do not behave differently than the single NF2 mutant hence providing no clue to elucidate the respective roles of these two genes in schwannoma development.
We think that the crucial point here is that in mGFAP-Cre mice the double Smarcb1/Nf2 mutants (schwannoma +) behave differently than the single Smarcb1 mutant (schwannoma -).
The observation of frequent somatic, tumor-specific NF2 mutations, and the loss of the second NF2 allele in schwannomas from patients with germline SMARCB1 mutations (Boyd et al. 2008;Sestini et al. 2008;Hadfield et al. 2008Hadfield et al. , 2010 strongly suggests that the classical two-hit model of tumorigenesis does not pertain in the tumors of schwannomatosis patients, at least in the sense that it would require biallelic SMARCB1 inactivation to be sufficient for tumor initiation or growth (Plotkin et al. 2013, Kehrer-Sawatzki et al. 2016). Our hypothesis is that in schwannomatosis patients the role of SMARCB1 loss in schwannomagenesis is to promote and facilitate biallelic NF2 loss (through the 4-hit/3-step mechanism), and that NF2 loss is necessary and sufficient for schwannoma formation. Since the genetic 4-hit/3-step mechanism of double NF2/SMARCB1 inactivation cannot be reproduced in mice because of the two genes being located in different chromosomes, we used conditional mutagenesis to model the different steps.
In line with this hypothesis, the fact that mGFAP-Cre;Smarcb1 flox/flox mutant mice do not develop schwannoma and that mGFAP-Cre;Smarcb1 flox/flox ;Nf2 flox/flox double mutant mice develop benign schwannomas reinforces and models, for the first time, the necessary role of Nf2 loss for schwannoma formation in mice and schwannomatosis patients. Details on the temporal window of susceptibility to RT have been added to the Results (lines 331-333) and Discussion (lines 403-404) sections. The Abstract was modified to reflect the necessity of the early Smarcb1 loss in inducing RTs. The reason for including "Smarcb1 and Nf2 inactivation" in the title is to stress the fact that individually both genes are crucial in determining the development of RT or schwannoma.

4-The authors should be more cautious concerning the impact of their work on the identification of the RT cell-of-origin. They indeed show that neural crest cells can be transformed by Smarcb1
inactivation hence suggesting that a minority of RTs (see above comment) may have such a cell-oforigin but this should not lead to a broad conclusion that this neural crest cell progenitor is, in general, the RT cell-of-origin.
Introduction and Discussion sections were modified according to this comment.

5-The references need to be reviewed. Some important landmark papers on RTs are not quoted while the authors do reference a high number of relatively marginal reports.
As suggested by the reviewer, we have revised the choice of references to include landmark references on RTs.

6-The suggestion that the observed tumors correspond to the recently described SHH group is only supported by the expression of Gli1 which constitutes a relatively weak proof. Again, comprehensive transcriptome data and/or a clear demonstration of a nuclear localization of Gli1 would be necessary to further document this hypothesis.
We have included a comprehensive molecular analysis of RTs obtained in the P0-CreC;Smarcb1 flox/flox mouse model by whole genome RNA-sequencing (12 representative tumors).
See answer to point #2. Fig.8 is misleading as single Nf2-/-mice which behave as double Smarcb1/Nf2 mutant mice are not indicated. The authors cannot draw conclusions on the 4-hits/3-steps model. mGFAP-Cre;Nf2 flox/flox mice were added to now Fig. 7.

7-
Indeed, the original text did not stress adequately that our model does not reproduce the 3 genetic steps of double SMARCB1/NF2 inactivation sequentially in a single mouse (Smarcb1 and Nf2 genes are not syntenic in the mouse). Instead, we generated mice for each of the 4 possible gene hit combinations to dissect the impact of each combination on tumorigenesis.

8-Fig7 shows that whorls of Baf47 negative cells are only observed in the mGFAP-cre; Nf2-/-; Smarcb1-/-and not in the mGFAP-cre; Smarcb1-/-background. May this indicate that in this context the loss of Nf2 is necessary for the survival of Smarcb1 cells?
Comparison of SMARCB1 staining in DRG of control and mGFAP-Cre;Smarcb1 flox/flox mice (now Fig. 6) shows unstained nuclei in the mGFAP-Cre;Smarcb1 flox/flox mice. As SMARCB1-negative cells in mGFAP-Cre;Smarcb1 flox/flox DRG don't proliferate, they are present but in smaller number than in mGFAP-Cre;Smarcb1 flox/flox ;Nf2 flox/flox DRG, thus explaining why they are not easily visible. For more clarity, we added insets with higher magnification of representative SMARCB1-negative cells in mGFAP-Cre;Smarcb1 flox/flox DRG and SMARCB1-positive cells in control DRG.

Reviewer #2 (Remarks to the Author): Expert in AT/RT
... The presented data is novel and explains long observed findings and the link between two completely different diseases.

The manuscript is well structured and written and especially the pictorial part helps to understand the experimentation and gives in the end an excellent overview of potential events. The application of the 4-hit/3-step model nicely fits the authors line of evidence.
There is only very limited criticism: 1) the functional consequences of inactivation of the genes SMARCB1 and NF2 is mainly described in morphological terms. The results of some expression data is presented in Figure 5. However the consequences of gene inactivation of SMARCB1 and NF2 on the SWI/SNF complex and its remaining members is not addressed at all. It would be very interesting to see whether the complex loses all of its functions and composition or whether other mechanisms are responsible for the morphological change. It might be worthwhile to try and elucidate gene set enrichment in the different constructs. The limited data and discussion on molecular subtyping (shh) is a step in the right direction.
We performed RNA-sequencing to assess the gene expression profile of the P0-CreC;Smarcb1 flox/flox tumors and compare these tumors with published human AT/RT molecular subgroups (Johann et al. 2016) and other published Smarcb1-deficient mouse models (Han et al. 2016;Ng et al. 2015).
We found that RTs from P0-CreC;Smarcb1 flox/flox mice clustered among the three molecular subgroups recently defined by Johann et al., thus showing that our mouse model accurately reproduces the molecular diversity found in human RTs. The comparison with two existing mouse models (Han et al., Ng et al.) showed that mouse RTs clustered in three molecular subgroups irrespective of the model of origin. All these data were added to the Results and Discussion sections.
Analysis of SWI/SNF complex components demonstrated that tumors from P0-CreC;Smarcb1 flox/flox mice specifically expressed subunits found in the pluripotent embryonic stem cell (esBAF) and in the multipotent neural progenitors (npBAF) complexes. Except for subunit CREST (SS18L1 gene), this results was consistent in human AT/RT samples from Johann et al.

2) the authors should be encouraged to discuss a few findings in human patients in relation to their data. How for example would they explain the phenomenon of synchronous tumors?
Synchronous tumors usually develop in patients with cancer predisposition syndromes and are consistent with a germline (or de novo mutation) and a second genetic hit (chromosome 22 complete or partial loss resulting in the loss of the normal SMARCB1 gene) occurring in different tissues at an early stage of development. This phenomenon has been described in few patients with RTs and carrying SMARCB1 mutations (Biegel et al. 2000, Seeringer et al. 2014, Metts et al. 2017. As stated in the Results, 62.5% of P0-CreC;Smarcb1 flox/flox mice displayed multiple tumors at the time of dissection. We added this comment to the Discussion section (lines 448-450).

While the authors found Cre and SMARCB1 deletion in kidneys one of the major target organs of rhabdoid tumors, they did not see any tumors in the kidney.
In addition to results in this manuscript ( Supplementary Fig. 1)  Furthermore there are some human families in which family members are affected by rhabdoid tumors while others have schwannomas. How can this data be reconciled?
The existence of adult carriers of a SMARCB1 mutation without RT in families presenting either RTs or schwannomatosis suggests that there is a specific developmental time window during which RT progenitor cells are vulnerable to SMARCB1 loss (Swensen et al. 2009, Eaton et al. 2011, Carter et al. 2012. In this model, if loss of the second SMARCB1 allele occurs during early development, the individual will develop a RT. The P0-CreC;Smarcb1 flox/flox mice mimic this situation. If mutation of the second SMARCB1 allele does not occurs during early development, the individual carrying a germline SMARCB1 mutation does not develop a RT. This situation is modeled in our mGFAP-Cre;Smarcb1 del/flox mice that do not develop RT (late Cre expression inducing late Smarcb1 loss). The SMARCB1 mutation carrier can then develop schwannomatosis at a later age when the additional somatic mutation steps occur: loss of SMARCB1 2 nd allele in addition to loss of 1st NF2 allele (2 nd step) and the second NF2 mutation (3 rd step). This combination of SMARCB1 and NF2 loss is modeled in mGFAP-Cre;Smarcb1 flox/flox ;Nf2 flox/flox mice that develop schwannomas. This point was added to the Results (lines 292-294) and Discussion sections (lines 404-406).

3) As the presented mouse models are only the second model of rhabdoid tumors which consistently demonstrate intracranial RT without any other sign of malignancy outside the CNS the manuscript would benefit from a closer comparison to the recent manuscript by Han et al (Nat Comm 2016).
We performed gene expression analysis by RNA-Sequencing of 12 rhabdoid tumors from the P0-CreC;Smarcb1 flox/flox mouse model. Unsupervised clustering analysis demonstrated that the tumors from Han et al. mouse model clustered with two of the three molecular subgroup found in our P0-CreC;Smarcb1 flox/flox mouse model. This is consistent with Han et al. analysis where they show that their tumors clustered in two subgroups. These data are now shown in Fig. 4b.

4) On a minor note it would be nice to see a proliferation marker when examining the mouse tumors (e.g. KI67 or mouse equivalent).
The Ki-67 proliferation marker was used and demonstrated a high proliferative index in the RTs from P0-CreC;Smarcb1 mice ( Fig. 3f) and very low proliferation in the schwannomas from mGFAP-Cre;Smarcb1 flox/flox ;Nf2 flox/flox and mGFAP-Cre; Nf2 flox/flox mice (Fig. 6, last row).

5) very minor problems are on page 20 line 411 it should read "a cell of origin". The references are at times incomplete and not consistently formatted e.g. Ref 66 misses page numbers and edition numbers. Ref 47 should be exchanged fro the new WHO classification publish a few months ago. All references should be checked for correctness.
These items were corrected.
All in all this is a beautiful manuscript with novel and important findings with interest not only for the specialist, but also for any researcher interested in the mechanisms of disease. Especially the influence of spatial and temporal factors is very interesting and has a potential to help understand many disease states. I thus suggest to accept the manuscript following minor revisions.

Reviewer #3 (Remarks to the Author): Expert in Schwannomas
...The topic addressed in this manuscript is an important one and potentially of strong interest to the readership of Nature Communications. However, in its current form, the manuscript is very difficult to follow and the arguments presented in support of their conclusions are convoluted and not clearly supported by the data as currently presented. Points that the authors need to address to improve the manuscript are as follows: 1) Although the data in the manuscript are quite interesting, there are a number of sentences in the manuscript that are awkwardly phrased that make it difficult to follow the manuscript. For example, there is a sentence in the abstract that states "The single loss of Smarcb1 in neural crest cells is sufficient to initiate tumorigenesis…". The Results section in particular is very dense and difficult to follow; while this is in part, this is due to the complicated genotypes being presented, it also reflects the often decipherable sentences in this section of the paper. The manuscript would be much easier to follow if the authors thoroughly edited it so as to present their work in a more straightforward and less complicated manner.
These points were extensively revised with a more straightforward comparison of mouse vs. human phenotypes.
2) A major theme of this manuscript is that the timing of Smarcb1 loss (and likely the stage at which this loss occurs during neural crest/Schwann cell development) dictates whether the mouse develops a schwannoma or a rhabdoid tumor. This argument is dependent upon knowing the timing with which the different Cre transgenes they use are expressed in their mouse models. However, with the exception of a statement very late in the Results (when discussing the GFAP-Cre driver), the authors provide no information in the Results about the times when these different Cre transgenes are expressed. This is finally presented in Figure 8, which is first referred to in the next to last paragraph of the Discussion. As a Result, I found myself struggling to make sense of their data (and argument) until the very end of the paper. The authors need to clearly indicate when each of these transgenes are active in the embryo when they first present the Cre driver to the reader.
Sentences mentioning the embryonic day at which DHH-Cre (line 299) and mGFAP-Cre (line 311) are active are included at the first occurrence in the text of the respective mouse model in the Results sections.
As suggested, information about the timing of expression of P0-Cre has been added in the Result section at the first occurrence in the text of P0-CreC (line 121) and P0-CreB (line 157) mouse models. Reference to now Figure 7 was also added at the beginning of the Results section (line 119).
3) In the first paragraph of the Results, the authors mention that some of the P0-CreC;Smarcb1flox/flox mice had bloated intestines with an emaciated wall and a reduced number of villi. Why did this occur?
Bloated intestines likely resulted from defects of the enteric nervous system. Neurons and glia of the enteric nervous system are derived from vagal and sacral neural crest cells (Coelho-Aguiar et al. 2015). Crone et al. 2003 showed that our P0-Cre transgene is expressed in the enteric nervous system (ganglia). We added this information to the Discussion (lines 429-432).

4) The authors had their tumors examined by competent pathologists whose opinions I respect.
However, many of the journal's readers will likely not know these individuals. Given that, the authors really need to do a better job of presenting the diagnostic criteria that were applied and a critical assessment of how their mouse tumors compared to their human counterparts. There clearly were some differences-for instance, the authors simply stated that their mouse rhabdoid tumors were negative for neurofilament and cytokeratin. Human AT/RTs commonly show foci with expression of both of these antigens and the authors made no mention of this.
As recommended, we added the information about focal expression of neurofilament and cytokeratin staining in human AT/RTs and the methodology used to assess the immunophenotype in the Methods section (lines 207-212, and lines 499-501, respectively). In addition, to reinforce the histological diagnosis, we performed whole genome expression analysis. Unsupervised clustering analysis demonstrated that RTs from P0-CreB;Smarcb1 flox/flox mice clustered with the three molecular subgroups found in human AT/RTs (Johann et al. 2016). Figure 8.

5) The authors note that the P0-CreB transgene expresses Cre "in a larger cell compartment" than the P0-CreC transgene. Again, what does this mean? Is the timing of expression also different from P0-CreC? This is not mentioned in the text and it is not indicated in
We clarified this information in the Results section (line 158): "Use of the P0-CreB transgenic mice, which express Cre recombinase at E9.5, similarly to P0-creC, but in a larger number of cells". The citation (Giovannini et al. 2000) shows the difference of Cre expression between the different P0-Cre mouse transgenic strains. As suggested, we added the P0-CreB;Smarcb1 flox/flox mouse model to now Figure 7.

6) It is not clear what the authors mean by the sentence that spans lines 200-203. If part of this sentence is trying to argue that the tumors could have originated from the leptomeninges, despite the absence of PGDS, it is unclear how the authors arrived at this possibility.
We have demonstrated that after birth PDGS is expressed at later differentiation stages in the telencephalon neural crest-derived meninges (Kalamarides et al. 2009). The absence of PDGS staining shows that the tumors developed before the meninges differentiated and expressed the PDGS marker, which correlates with our hypothesis that the tumors developed in cranial neural crest cells. The sentence (lines 200-203, now lines 202-205) was rephrased to be more specific and synthetic.

7) Why is the data included on the DHH-Cre;Smarcb1flox/flox mice (they didn't develop tumors)?
Are they trying to argue that these mice help them narrow down the window when Smarcb1 loss results in rhabdoid tumor formation? That possibility seems unlikely, given that GFAP-Cre;Smarcb1flox/flox mice developed tumors with rhabdoid features between 2 and 17 months of age (this Cre driver turns on at E13.5 compared to Dhh-Cre, which turns on at E12). The significance of these negative data are unclear as currently presented. DHH-Cre;Smarcb1 flox/flox and mGFAP-Cre;Smarcb1 flox/flox mice were generated to explore the effect of Smarcb1 loss during later stages of Schwann cell development. The fact that mice of both genotypes did not develop tumors allowed us to narrow down the time window during which loss of Smarcb1 results in RT development. The data cited by the reviewer ("developed tumors with rhabdoid features between 2 and 17 months of age") refer to the Smarcb1 del/+ mice and were presented at the end of the mGFAP-Cre section. To avoid confusion, we moved these data at the beginning of the results section (lines 110-115). Figure 2 is not convincing and needs to be presented with a better image. In addition, the tumor images presented in Figure 2 are at very low power. Following the usual pathology dictum that "blue is bad", it seems likely that these are aggressive tumors, but it will not be possible to be sure that this is the case until higher power images are included.

8) Panel e in
Panel in now Figure 1j was replaced by a picture of a X-gal staining, performed more recently, of a trigeminal nerve section dissected from a P0-CreC;AZCL (lacZ reporter) mouse. Figure 1h and 1j represent tissues from P0-CreC;AZCL mice that are wild-type for Smarcb1. The scarce number of blue cells is indicative of mosaic Cre expression in this transgenic mouse line. In, contrast the large number of blue cells in Figure 1d reflects the clonal origin and expansion of the Smarcb1-deleted cells resulting in a RT from a P0-CreC;Smarcb1 flox/flox ;ACZL mouse.
As recommended, in addition to low magnification images focusing on anatomical origin of the tumors, high magnification insets we added to panels 1f, 1i, 1l, 1m and 1n to better evaluate the aggressive pathological features of the tumors shown in Figure 1. Panel 1l is a higher magnification of panel 1k.