Cryptic developmental events determine medulloblastoma radiosensitivity and cellular heterogeneity without altering transcriptomic profile

It is unclear why medulloblastoma patients receiving similar treatments experience different outcomes. Transcriptomic profiling identified subgroups with different prognoses, but in each subgroup, individuals remain at risk of incurable recurrence. To investigate why similar-appearing tumors produce variable outcomes, we analyzed medulloblastomas triggered in transgenic mice by a common driver mutation expressed at different points in brain development. We genetically engineered mice to express oncogenic SmoM2, starting in multipotent glio-neuronal stem cells, or committed neural progenitors. Both groups developed medulloblastomas with similar transcriptomic profiles. We compared medulloblastoma progression, radiosensitivity, and cellular heterogeneity, determined by single-cell transcriptomic analysis (scRNA-seq). Stem cell-triggered medulloblastomas progressed faster, contained more OLIG2-expressing stem-like cells, and consistently showed radioresistance. In contrast, progenitor-triggered MBs progressed slower, down-regulated stem-like cells and were curable with radiation. Progenitor-triggered medulloblastomas also contained more diverse stromal populations, with more Ccr2+ macrophages and fewer Igf1+ microglia, indicating that developmental events affected the subsequent tumor microenvironment. Reduced mTORC1 activity in M-Smo tumors suggests that differential Igf1 contributed to differences in phenotype. Developmental events in tumorigenesis that were obscure in transcriptomic profiles thus remained cryptic determinants of tumor composition and outcome. Precise understanding of medulloblastoma pathogenesis and prognosis requires supplementing transcriptomic/methylomic studies with analyses that resolve cellular heterogeneity.

In this study by Malawsky et al., the authors build off previously published findings that demonstrate differences in medulloblastoma tumor progression depending on whether mice were engineered to express oncogenic SmoM2 in neural stem cells or committed progenitors. They show that while both types of medulloblastoma mouse models have similar transcriptomes as demonstrated by bulk analyses, the stem cell-driven SmoM2 tumors progressed faster and exhibited radioresistance, which is a novel finding. Progenitor-driven tumors contained fewer stem cells, were radio-sensitive and progressed slower. The authors also show differences between the 2 tumor types in terms of stromal populations; however, the biological and potential clinical relevance of this finding was not explored.
Overall, the authors provide some additional insight into differential expression profiles in stem cell vs. progenitor-driven tumors, with the radioresistance data being the most compelling. The fact that the stem cell-driven tumors are more resistant to radiation is interesting. In addition, the preliminary data on niche differences are promising. That being said, the major concern with the paper is the lack of novelty or original conclusions. While the study is technically sound and very well-written, a significant portion of the data recapitulate previously published findings which are frequently referenced by the authors (ie. Schüller et al., Cancer Cell, 2008; Zhang et al., Cancer Cell 2019, and the authors' recent publication Ocasio et al. in Nature Communications, 2019 ). This specifically pertains to the survival data, the stem cell signatures as well as the OLIG2+ cell profiles described in Figure 3. Specific comments are as follow: 1. Figure 1c Presumably, the higher stem cell proportion in these tumors would contribute to the RT resistance but additional insights into mechanism would help to further strengthen the manuscript. 3. The analyses of the tumor niche, particularly the myeloid cells, while interesting, is rather descriptive. What is the significance of the more diverse microglia/macrophage and fibroblast cells in the M-Smo tumors? The authors state that counting MHCII-expressing macrophages to distinguish between radiosensitive and radioresistant could be further explored in a clinical setting for prognostic purposes. If these differences can be probed in existing clinical samples by IHC or flow cytometry to provide validation of this hypothesis, the manuscript, in particular the novelty, would be significantly strengthened. For example, it would be very interesting to determine whether this pattern would hold between human samples representing the different SHH MB subtypes recently described by Cavalli  displayed significant differences in aggressiveness and response to radiotherapy. Bulk transcriptomic analysis revealed very few differences, but the characterization of single cells expression profiles revealed in G-Smo tumors a higer frequency of OLIG2-expressing stem-like cells, as well as a lower frequency of macrophages. This is a very interesting paper, providing not only important new information on features that may help medulloblastoma prognostic assessment, but also underscoring the power of single cell analysis in obtaining crucial biological insight. I have no general or technical remarks: the quality of data/data analysis conforms to high standards and the conclusions are well justified by data. Authors may consider to explain a little better in the discussion why the bulk analysis does not show differences in the expression of proliferation markers (I supppose it is for their oscillatory nature) and of macrophagic markers (I would expect a difference, based on the increased representation revealed by the IF analysis). Moreover, they may consider to discuss the possibility that G-Smo and M-Smo tumors may functionally differ in their intrinsic DNA-repair capability, as an increased mutation burden could probably help bridging the increased rediosensitivity of M-Smo tumors with their increased macrophagic infiltrate. However, I think that the manuscript is already acceptable in the current form.
Reviewer #3 (Remarks to the Author): The approach by the authors to ask whether different outcomes in human SHH MBs may be determined by distinct cryptic events during development appears valid. Also, it is fine to compare therapy effects and tumor micorenvironments in mouse models for MB that are driven by different promoters. Still, novel biological insights appear sparse and, above all, this paper clearly lacks validation in the human system, i.e. in human tumor samples or patients. We thank the reviewers for their thorough and thoughtful reading of our manuscript. We have worked to address all of the reviewer comments, as described in the point-by-point response below. We feel that the manuscript is significantly strengthened by the new data added in response to the feedback and suggestions of the reviewers. We hope the reviewers agree.

In this study by Malawsky et al., the authors build off previously published findings that demonstrate differences in medulloblastoma tumor progression depending on whether mice were engineered to express oncogenic SmoM2 in neural stem cells or committed progenitors. They show that while both types of medulloblastoma mouse models have similar transcriptomes as demonstrated by bulk analyses, the stem cell-driven SmoM2 tumors progressed faster and exhibited radioresistance, which is a novel finding.
We appreciate the reviewer recognizing the novelty of our data on differences in response to therapy, which are an important aspect of our study.
Progenitor-driven tumors contained fewer stem cells, were radio-sensitive and progressed slower. The authors also show differences between the 2 tumor types in terms of stromal populations; however, the biological and potential clinical relevance of this finding was not explored.

We have added new analyses to address the biological and potential clinical relevance, as we describe in the point-by-point response below. These changes include new data showing that :  fewer cells undergo apoptosis after radiation in G-Smo tumors,  phosphorylated OLIG2 is increased in G-Smo tumors,  more myeloid cells in G-Smo tumors express Igf1, and  HLA-DR proteins, the human orthologs of mouse H2-Ea, show varied expression in patient-derived samples of different medulloblastoma subgroups and subtypes.
Overall, the authors provide some additional insight into differential expression profiles in stem cell vs. progenitor-driven tumors, with the radioresistance data being the most compelling. The fact that the stem cell-driven tumors are more resistant to radiation is interesting. In addition, the preliminary data on niche differences are promising. That being said, the major concern with the paper is the lack of novelty or original conclusions.  Figure 3. Our findings that developmental timing of oncogenesis affects tumor stem cell regulation, tumor-TME I interactions and therapeutic outcomes are novel and are not contained within the papers cited above. In the revision, we have clarified the novelty of this finding, and added new data regarding mechanisms affected by the developmental timing of initial oncogenic event. These changes have strengthened the our conclusion that the developmental timing of oncogenesis is a key determinant of outcome.
Specific comments are as follow: 1. Figure 1c- In an effort to provide mechanistic molecular information, we analyzed OLIG2 phosphorylation. Prior studies have shown that phosphorylated OLIG2 (pOLIG2) disrupts p53 function [1], and we, and others, have previously shown that p53 function is essential for apoptosis in response to radiation in SHH medulloblastoma [2]. Our pOLIG2 studies required a bulk approach by western blot, as all pOLIG2 antibodies that we tested failed in immunohistochemistry assays. Western blot effectively detected pOLIG2 and comparison between tumor genotypes showed increased pOLIG2 G-Smo tumors versus M-Smo tumors. We have added these data, which highlight a potential mechanism for radio-resistance in G-Smo tumors, to the manuscript and to the revised Fig.3.

Figure 3: What is the significance of the dynamic changes in the OLIG2+ cells in the M-Smo vs. the G-Smo treated tumors? We revised the text to clarify the significance of the dynamic changes by stating "The difference in the developmental timing of oncogenesis in G-Smo and M-Smo tumors thus continued to affect stem cell dynamics weeks after the oncogenic events."
Again, more mechanistic insight would strengthen the manuscript. For example, are there differences in proliferation of this cell population at the P15 timepoint?    We found significantly increased HLA-DR in the SHH subgroup tumors compared to the other subgroups, demonstrated by higher HLA-DR scores. Between SHH subtypes, we noted similar, relatively higher scores in the SHH beta and delta subtypes, and markedly lower scores in the SHH alpha subtype. These new data are presented in the revised Fig. 7. While a statistical analysis of SHH subtypes will require more samples and resources beyond the scope of this project, we point out that "SHH beta and SHH delta are the infant-predominant subtypes [6], and the trend toward higher HLA-DR expression in these subtypes suggests that a developmental process affects myeloid subgroups in human tumors, as in our mouse models".

As we discuss in the revised Discussion section, different SHH medulloblastoma studies have identified tumor suppressive and tumor supportive functions of myeloid cells [4, 5]. Our new finding that Igf1+ and Igf1-subsets vary in SHH-driven medulloblastomas with different progression patterns provides a mechanistic framework for understanding how differences in myeloid populations can alter outcomes.
To increase the statistical power using available, published data, we added a new analysis of HLA-DRA mRNA expression in the published dataset from Cavalli et al [6]. These data also show increased expression in SHH subgroup tumors and within the SHH subgroup show that SHH alpha tumors show lower average expression compared the other SHH subtypes. Moreover, analysis of SHH subgroup tumors by age, rather than subtype show significantly higher HLA-DRA in infant-onset SHH medulloblastomas. While the differences in average expression are statistically significant, we note large individual variation in each subtype, suggesting heterogeneity within each subtype. We suggest in the Discussion that additional large studies of clinical samples using cell-resolved methods such as scRNA-seq or IHC may allow the definition of more homogeneous subsets.

Reviewer #2 (Remarks to the Author):
This We appreciate that the reviewer considered the conclusions were well justified.

Authors may consider to explain a little better in the discussion why the bulk analysis does not show differences in the expression of proliferation markers (I suppose it is for their oscillatory nature) and of macrophagic markers (I would expect a difference, based on the increased representation revealed by the IF analysis).
We added a paragraph to the Discussion section that considers factors that may obscure important signals in bulk transcriptomic studies. We attribute the lack of significant differences in proliferation markers in the bulk studies to the high levels of proliferative cells in all groups, which may prevent proliferation marker mRNAs from showing statistically detectable variation. We attribute the lack of different signals from myeloid subsets as related to their relatively small fractions in the tumors, and we attribute the lack of differences in Olig2 signals to the common expression of Olig2 by both stem cells and oligodendrocytes.

Moreover, they may consider to discuss the possibility that G-Smo and M-Smo tumors may functionally differ in their intrinsic DNA-repair capability, as an increased mutation burden could probably help bridging the increased radiosensitivity of M-Smo tumors with their increased macrophagic infiltrate.
We appreciate the suggestion that a difference in DNA repair capability may contribute to the difference in radiosensitivity, which is consistent with our new data showing increased pOLIG2 expression in G-Smo tumors. We discuss the possibility that inhibition of p53dependent apoptosis by pOLIG2 may allow increased time for stem cells to repair DNA after xRT. We cite a prior paper showing that Bax-deleted medulloblastomas which are apoptosis incompetent, repair DNA breaks after xRT.

However, I think that the manuscript is already acceptable in the current form.
Reviewer #3 (Remarks to the Author): The approach by the authors to ask whether different outcomes in human SHH MBs may be determined by distinct cryptic events during development appears valid. Also, it is fine to compare therapy effects and tumor microenvironments in mouse models for MB that are driven by different promoters.

Still, novel biological insights appear sparse
The revised text includes new mechanistic data that connect our observations of radioresistance and increased OLIG2+ stem cells in G-Smo tumors. The revised manuscript better emphasizes our primary biological insight, which is that cellular heterogeneity and stem cell maintenance can be determined by the timing of initial oncogenic event and can, in turn, act as determinants of clinical outcomes. The new Figure  1 includes data showing that a large fraction of G-Smo tumors escape apoptosis after radiation, and the revised Figure 3 shows that pOLIG2, which is known to disable p53, is increased in G-Smo tumors. The revised Figure 6 shows that Igf1, which provides essential paracrine support in medulloblastoma, is increased in G-Smo tumors and correlates with increased mTORC1 activation (demonstrated by increased p4EBP1). These new data support a model in which OLIG2+ stem cells, which in G-Smo tumors are more numerous and better supported by Igf1+ myeloid cells, evade p53-dependent apoptosis and drive recurrence. This model provides a mechanistic framework connecting radioresistance to developmental differences.
and, above all, this paper clearly lacks validation in the human system, i.e. in human tumor samples or patients.
In response to this comment, as well as comments of Reviewer 1, we have added new data derived from a collaboration with additional investigators, now added to the authorship, in which we study HLA-DR protein expression in patient-derived medulloblastoma samples. We also collaborated on a new analysis of HLA-DRA mRNA expression in published bulk transcriptomic data from medulloblastoma patients. These new data validate in the human system specific observations made in our genetic models.
Other issues:

Use hGFAP-cre instead of Gfap-cre, if appropriate
We have changed the text to read hGFAP-Cre, as suggested. Figure 2 and 5 are too tiny. Please make any annotation in any figure readable in a way that a regular print out of the paper is sufficient. We appreciate the suggestion revise the figure annotations and labels. We have addressed this issue in each figure, and the revision has greatly increased the readability. Figure 2f

Please update ref 22, which has been published in the meantime
We have made the update. Figure 3, I don't really see the purpose of panels b-g. What's the message that the readers should get away with?

As for
We have revised the presentation of the data in Figure 3. The revised text clarifies that panels that were labelled b-g are representative images provided to give the readers a sense of the spatial distribution of OLIG2+/SOX10-cells throughout the tumors. In the revision, these panels as a group are labeled 3b and the results are quantified in the adjacent panel 3c. These visual data are needed because the spatial information is removed from the quantitative analysis of panel 3c.
What do the squares indicate (please indicate in the legends)?