IGFBP5 is an ROR1 ligand promoting glioblastoma invasion via ROR1/HER2-CREB signaling axis

Diffuse infiltration is the main reason for therapeutic resistance and recurrence in glioblastoma (GBM). However, potential targeted therapies for GBM stem-like cell (GSC) which is responsible for GBM invasion are limited. Herein, we report Insulin-like Growth Factor-Binding Protein 5 (IGFBP5) is a ligand for Receptor tyrosine kinase like Orphan Receptor 1 (ROR1), as a promising target for GSC invasion. Using a GSC-derived brain tumor model, GSCs were characterized into invasive or non-invasive subtypes, and RNA sequencing analysis revealed that IGFBP5 was differentially expressed between these two subtypes. GSC invasion capacity was inhibited by IGFBP5 knockdown and enhanced by IGFBP5 overexpression both in vitro and in vivo, particularly in a patient-derived xenograft model. IGFBP5 binds to ROR1 and facilitates ROR1/HER2 heterodimer formation, followed by inducing CREB-mediated ETV5 and FBXW9 expression, thereby promoting GSC invasion and tumorigenesis. Importantly, using a tumor-specific targeting and penetrating nanocapsule-mediated delivery of CRISPR/Cas9-based IGFBP5 gene editing significantly suppressed GSC invasion and downstream gene expression, and prolonged the survival of orthotopic tumor-bearing mice. Collectively, our data reveal that IGFBP5-ROR1/HER2-CREB signaling axis as a potential GBM therapeutic target.

In the manuscript by Weiwei Lin et. al., RNA sequencing analysis revealed IGFBP5 was differentially expressed between invasive or non-invasive subsets of GBM primary cell cultures. They further show through knockdown and gain of function studies that IGFBP5 drives GBM cell invasion (in vitro and in vivo), potentially via ROR1/HER2 heterodimer formation and subsequent activation of CREB. in invasive patient derived GBM models. Finally, nanoparticle delivery of CRISPR/Cas9 targeting IGFBP5 was able to suppress GBM cell invasion and improve the survival of orthotopic tumor-bearing mice. Specific targeting of IGFBP5 via IGFBP5 has recently been shown as having an important role in GBM cell invasion (Dong et. al, 2020). Therefore, the novelty is primarily based on the proposed mechanism of action and the in vivo experiments. Yet, there are numerous areas that need to be addressed to support these conclusions.
Major Points 1) Mechanism of Action: The proposed ROR1/HER2/CREB mechanism is not fully supported by the data. If they are proposing that IGFBP5 mediated ROR1 activation leads ROR1-mediated HER2 phosphorylation, they need to demonstrate this through the introduction of various gain of function and loss of function mutants (e.g., kinase dead ROR1 and/or HER2). Similarly, they also need to show that downstream CREB is necessary and/or sufficient for ROR1/HER2 mediated invasion (e.g., gain of function CREB under loss of function ROR1/HER2). As of now, the data support changes in activity and function of these players correlate with invasion and not necessarily that these factors are all mechanistically linked. Finally, the targets of CREB from Fig 5 have not been confirmed for their functional relevance.
2) Additional in vivo models to support conclusions: The authors should show efficacy of the NPs in at least one additional sensitive model. Moreover, the authors should show whether targeting IGFBP5 is ineffective in a model predicted to be insensitive to this approach. Other points 1) The use of glioma stem cells for the patient derived neurosphere culture is not appropriate. The simple formation of neuosphere cultures cannot by itself define a putative "glioma stem cell" 2) It is unclear what is the difference in the point from  figure 1c, why was the comparison between DEGs identified by the authors' RNAseq analysis compared with TCGA low-grade glioma dataset but not the high-grade glioma dataset 6) For figure 1e, what is the mRNA expression fold change relative to? 7) For figure 1f, possibly perform the same statistical comparison performed in 1b and 1e by comparing the non-invasive vs the invasive models? 8) For figure 2, did the silencing of IGFBP5 with shRNAs also affect cell growth in vitro? Could that have also contributed to the differences seen in invasiveness? Some of this has been shown prior in: Dong, C., Zhang, J., Fang, S. et al. IGFBP5 increases cell invasion and inhibits cell proliferation by EMT and Akt signaling pathway in Glioblastoma multiforme cells. Cell Div 15, 4 (2020). https://doi.org/10.1186/s13008-020-00061-6 9 ) For figure 4, do naturally invasive models (e.g., 448 and X01) have higher basal ROR1, HER2, Editorial Note: Parts of this Peer Review File have been redacted as indicated to remove third-party material where no permission to publish could be obtained. and CREB activation relative to non-invasive models not stimulated by IGFBP5? 10) For figure 4, what are the quantifications of the RTK and kinase array? 11) For figure 7, does the Cas9/CRISPR reduce protein production? 12) Figure 3j, all 6 mice with IGFBP5 died on the same day? Very unlikely to see all mice reach endpoints on the exact same day. How were endpoints established for this experiment?
Reviewer #2 (Remarks to the Author): This MS by Lin et al. IGFBP5, a secreted protein known for its role of binding to IGFs, can act as a ROR1 ligand and promote glioblastoma stem cell (GSC) invasion. The authors provided good evidence indicating that IGFBP5 expression affects GSC migration/invasion. This finding is not surprising because it has been reported by several groups that IGFBP5 can stimulate the migration/invasion of normal and tumor cells, including glioblastoma cells. What is potentially new is the idea that IGFBP5 binds to ROR1, causes ROR1/HER2 heterodimer formation, and increases CREB-mediated gene expression. If proven correct, this would be a new and interesting finding. There are, however, a number of issues and major concerns. To convincingly support the notion that IGFBP5 causes ROR1/HER2 heterodimer formation and increases gene expression and cell invasion vis CREB, more rigorous, complimentray genetic (dominant negative & constitutive active) and pharmacological experiments are needed.
Major comments: 1) In this MS, critical technical details are often missing, making it difficult to assess the quality of the data. To give a few examples, I cannot find information on how different GSC cells (448, X01, 83, are these mouse #?) were isolated and what were the objective criteria to classify them into invasive vs. non-invasive GSCs? There is no description in the Methods section about the RNA-seq. The raw data of RNA-seq data set were not available. In vivo xenograft data in Fig. 3i. There is no description in the Methods section about the xenograft experiment. It is unclear what were the biological replica. The quantification data is not presented.
2) The conclusion that IGFBP5 acts as ROR1 ligand was based on the data shown in Fig. 4. Unfortunately, all biochemical assays lacked quantified results (e.g., Fig. 4a-4h). Were these experiments performed only one time? Quantitative and reproducible biochemical data are needed. Same question with Fig. 7k.
3) It is also troublesome that contradictory results (e.g., Fig.4d, different effects on ROR1 levels; Fig. 4e different effects ROR1 levels) were ignored. How was the staining specificity validated? 4) Fig. 4k , Fig. 6e, 6g: Only one section image is shown per group. What were the biological replica? Are these statistically different changes? How was the staining specificity validated? Are these GSC cells and how do you know? 5) To convincingly support the notion that IGFBP5 causes ROR1/HER2 heterodimer formation and increases gene expression and cell invasion vis CREB, more rigorous, independent genetic (dominant negative & constitutive active) and pharmacological experiments are needed. It will also be critical to demonstrate the specificity of IGFBP5 binding and the Kd value. 6) Fig. 7m -I could not see much difference between the scrambled control and sgIGFBP5 group. Yet, it was stated that in vivo nano capsule-mediated IGFBP5 disruption dramatically increase mouse survival. Did I miss something? Did the authors actually detect any indel in the IGFBP5 gene in these mice?
3) The effects of IGFBP5 on cell migration/invasion have been reported since the late 1990s. These papers should be cited. 4) For in vitro invasion assays using shRNA-mediated knockdown, the possible off-target effects were not addressed. A simple experiment would be to test if adding back rIGFBP5 to the knocked down cells can rescue the cell invasion/gene expression.

Reviewer #3 (Remarks to the Author):
This interesting study shows a clear effect of IGFBP-5 in promoting GBM stem cell invasion and provides a plausible mechanism involving ROR1, HER2, and CREB activation, together with the CREB-dependent expression of ETV5 and FBXW9. Strengths: In general, the experiments are clearly presented and well interpreted, and the development of an IGFBP-5-targeting nanocapsule that significantly impairs GBM tumor growth and prolongs survival in mice with orthotopic GBM tumors is very impressive. Weaknesses: Although IGFBP-5 is immunoprecipitated using an anti-ROR1 antibody its direct binding to ROR1, as proposed, is not demonstrated, and the existence of a functional HER2-ROR1-IGFBP-5 complex is unclear since it is not precipitated using an anti-HER2 antibody (see specific comment #4). Therefore the precise mechanism of IGFBP-5 action is in some doubt, and a role for IGF1R phosphorylation is not explicitly excluded by the single negative phosphoarray experiment (see specific comment #7). Nevertheless this work adds significantly to current knowedge of the role of IGFBP-5 in GBM.
Specific comments 1. Line 116: "a small number of studies". Since high IGFBP-5 expression in GBM has been reported previously (as far back as 2006, doi: 10.1177/153303460600500303), some of these prior studies should be cited here. 2. Fig. 4c. The pROR1 and pCREB responses seem weaker in 131 cells than 83 cells. These data would benefit from quantitation of phospho/total ratios from several independent assays, and statistical evauation. 3. Fig. 4f, g: Some IP bands are stronger than the input. State what fraction of the total lysate was analyzed in the input lanes. 4. Lines 185-186. If a ROR1-IGFBP-5 complex interacts with HER2, IP with a HER2 antibody should precipitate both ROR1 and IGFBP-5, but this is not seen in Fig. 4f. One interpretation is that IGFBP-5 can bind to ROR1, but this binary complex does not interact with HER2. This is different from the model shown in Fig. 4m. Further interpretation of the coIP data is required, as well as some coIPs using anti-IGFBP-5 as the precipitating Ab. 5. Fig. 4h suggests that both ROR1 and HER2 have a role in CREB phosphorylation, but it provides no information about the role of IGFBP-5 in this process. Neither do Figs. 4i and j provide information about the role of IGFBP-5. So line 190, ".. implying that IGFBP-5 promotes …", overinterprets the data. . " … HER2 knockdown did not alter ROR1 phosphorylation", etc. It is difficult to be confident about this statement in the absence of a quantitative analysis of Fig. 4h. Phospho/total data should be analyzed for several repeat experiments. 7. Lines 332-333. Rather than introducing a new result in the Discussion, the authors should indicate the position of pIGF1R in Fig. 4a to illustrate that IGF1R phosphorylation is not promoted by IGFBP-5. Given that IGF1R phosphorylation may have occurred transiently and not been detected on this array, it would be informative to check the acute pIGF1R response to IGFBP-5 (e.g. a time-course starting at 5 or 10 min) by immunoblot and to ensure that the addition of an IGF1R tyrosine kinase inhibitor such as NVP-AEW541 did not block any downstream effects of IGFBP-5.
Minor point Line 112: "… adverse roles": I believe the authors mean "diverse" roles.
Reviewer #4 (Remarks to the Author): Lin et al. report a comprehensive study into the role of Insulin-like Growth Factor-Binding Protein 5 (IGFBP5) on glioblastoma infiltration in orthotopic mouse models, in particular, as it relates to GSCs. In a GSC-derived brain tumor model, they conducted RNA seq analysis and found that IGFBP5 was differentially expressed between invasive or non-invasive subtypes. The authors then engaged in detailed mechanistic studies to establish the important of IGFBP5 for GSC invasion. Finally, they report data suggesting that the nanoparticle-based delivery of CRISPR/Cas9-based IGFBP5 gene editing results in impressive therapeutic benefits, such as significant increase in survival in a GBM mouse model.
Overall, this is a large research study with an impressive attention to details, in particular on the mechanistic aspects. The initial hypothesis related to IGFBP5 is motivated by a comparative study between invasive and non-invasive GBM subtypes. They follow up with both silencing and know down experiments that clearly establish the link between IFGBP5 and GSC invasion. While this study has clearly merits -in particular as it relates to GBM pathobiology -there are also significant shortcomings that dampen the potential impact. The following aspects should be addressed prior to publication: 1) IFGBP5 is a member of the IGFBP family, and, contrary to what is suggested in the introduction and the discussion sections, a significant bulk of literature clearly links these proteins to GBM infiltration and poor clinical prognosis. The same is also true for IFGBP5, see for example (references not cited by authors): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7885861/ or https://celldiv.biomedcentral.com/articles/10.1186/s13008-020-00061-6.
2) There are fundamental limitations of using orthotopic GBM systems to model GBC invasion. For this particular study, genetically engineered spontaneous glioblastoma mouse models would be a better choice. The authors should at least discuss the limitations of the model they selected and how this could affect the scientific premise of their study.
3) IFGBP5 is overexpressed in a wide range of tissues, and it is unclear how CRISPR/Cas9-based IGFBP5 gene editing would affect these tissues systematically, see, e.g., https://www.proteinatlas.org/ENSG00000115461-IGFBP5/tissue. This seems an important aspect that should be addressed by the authors.
4) The nanoparticle delivery system provides clear benefits in survival, but a more complete characterization would be warranted. What is the biodistribution of the nanoparticles? Where are they accumulating? If they go into liver and spleen, is there any toxicity or off-target activity? In particular, it would be interesting to know if they accumulate in any of the tissues that overexpress IFGBP5.
5) The CRISPR/Cas9-based IGFBP5 gene editing experiments have only been done in the X1 subtype, but at least one non-invasive subtype should be included for comparison. 6) What was the dose used for the CRISPR/Cas9-based IGFBP5 gene editing experiments? A dose escalation study should be included. What is the fraction of the total dose delivered to the brain tumor? How much was delivered to the brain? What is the GBM/brain ratio?
7) The authors claim for their CRISPR-Cas9 system "a much preferrable outcome on tumor growth as well as mouse expectancy, comparing to the well-known siRNA-based therapies targeting the conventional GBM targets (i.e., PLK134 or STAT3)". However, when consulting their cited references, it appears as the opposite is the case, i.e., their survival results, also significant, don't match up with the one achieved by the other therapies.

Reviewer #1 (Remarks to the Author):
In the manuscript by Weiwei Lin et. al., RNA sequencing analysis revealed IGFBP5 was differentially expressed between invasive or non-invasive subsets of GBM primary cell cultures. They further show through knockdown and gain of function studies that IGFBP5 drives GBM cell invasion (in vitro and in vivo), potentially via ROR1/HER2 heterodimer formation and subsequent activation of CREB. in invasive patient derived GBM models. Finally, nanoparticle delivery of CRISPR/Cas9 targeting IGFBP5 was able to suppress GBM cell invasion and improve the survival of orthotopic tumor-bearing mice.
Specific targeting of IGFBP5 via IGFBP5 has recently been shown as having an important role in GBM cell invasion (Dong et. al, 2020). Therefore, the novelty is primarily based on the proposed mechanism of action and the in vivo experiments. Yet, there are numerous areas that need to be addressed to support these conclusions.
Major Points 1) Mechanism of Action: The proposed ROR1/HER2/CREB mechanism is not fully supported by the data. If they are proposing that IGFBP5 mediated ROR1 activation leads ROR1-mediated HER2 phosphorylation, they need to demonstrate this through the introduction of various gain of function and loss of function mutants (e.g., kinase dead ROR1 and/or HER2).
Response: W e truly appreciate the reviewer's comment. As suggested by reviewer, we have strengthened ROR1/HER2/CREB mechanism by investigating the role of each molecule in GSCs invasion by i) knocking down HER2 using shRNA or pharmacological inhibitor, Irbinitinib (Response i) To determine the functional role of HER2 and ROR1 in GSC invasion, we silenced HER2 and ROR1 in invasive GSCs 448 and X01 with shRNA. As shown in response Fig. 1 and 2, both CREB phosphorylation and the invasive capacity of GSCs 448 and X01 were downregulated by knocking down HER2 (Response Fig. 1a-d) and ROR1 (Response Fig. 2a-d).

Response
Similarly, they also need to show that downstream CREB is necessary and/or sufficient for ROR1/HER2 mediated invasion (e.g., gain of function CREB under loss of function ROR1/HER2). As of now, the data support changes in activity and function of these players correlate with invasion and not necessarily that these factors are all mechanistically linked.

Response:
We thank the reviewer for this comment. As the reviewer suggested, we first overexpressed CREB in non-invasive GSC and investigated the role of CREB in regulating GSC invasion. Next, we also investigated their function by overexpressing CREB under loss-of-function (LOF) of IGFBP5, ROR1 or HER2. Immunoblot analysis verified a successful ectopic CREB overexpression in noninvasive GSCs (83 and 131) (Response Fig. 3a-b). CREB overexpression significantly increased downstream genes ETV5 and FBXW9 expression, and the invasive capacity of GSCs (Response Fig.   3c-h).
To investigate the downstream of CREB in IGFBP5-ROR1/HER2 signaling pathway, we ectopically overexpressed CREB in IGFBP5 knockdown GSCs, and CREB sufficiently restored GSC invasion (Response Fig. 3i, j). Furthermore, overexpression of CERB also rescued HER2 or ROR1 knockdown-mediated repression of GSC invasion (Response Fig. 4a-h). These results indicated that CREB is a key molecule in IGFBP5-mediated ROR1/HER2 signaling axis. Fig. 6d-n). Overexpression of CREB enhances GSCs invasion. a-b, IB analysis of pCREB, and CREB in 83 (a) and in 131 (b) GSCs infected with vector control, or CREB OE lentivirus. GAPDH was used as a loading control. c-d, RT-qPCR analysis of CREB targets ETV5 (c) and FBXW9 (d) expression in 83 GSCs infected with vector control, CREB OE lentivirus. Data are presented as mean ± SEM (n = 3). ** P < 0.01, t-test. e-f, RT-qPCR analysis of CREB targets ETV5 (e) and FBXW9 (f) expression in 131 GSCs infected with vector control, CREB OE lentivirus. Data are presented as mean ± SEM (n = 3). * P < 0.05, ** P < 0.01, t-test. g-h, Invasion assays of 83 (g) and 131 (h) GSCs infected with vector control, or CREB OE lentivirus. Images (left) taken after 48 h of invasion are representative of three independent experiments (scale bar, 100 µm), and the graph (right) shows the mean number of invasive cells ± SEM (n = 3). * P < 0.05, t-test. i, IB analysis of pROR1, pHER2, pCREB, ROR1, HER2, and CREB in X01 GSCs infected with shCtrl, shIGFBP5-1 lentivirus, and then infected with vector control, CREB OE lentivirus. GAPDH was used as a loading control. j, Invasion assays of X01 GSCs infected with shCtrl, shIGFBP5-1 lentivirus, and then infected with vector control, CREB OE lentivirus. Images (left) taken after 24 h of invasion are representative of three independent experiments (scale bar, 100 µm), and the graph (right) shows the mean number of invasive cells ± SEM (n = 3). ** P < 0.01, t-test. Fig. 7). Overexpression of CERB rescues HER2 or ROR1 knockdown-mediated repression of GSC invasion. a-b, IB analysis of pROR1, pHER2, pCREB, ROR1, HER2, and CREB in 448 (a) and X01 (b) GSCs infected with shCtrl, shHER2-1 lentivirus, and then infected with vector control, CREB OE lentivirus. GAPDH was used as a loading control. c-d, Invasion assays of 448 (c) and X01 (d) GSCs infected with shCtrl, shHER2-1 lentivirus, and then infected with vector control, CREB OE lentivirus. Images (left) taken after 24 h of invasion are representative of three independent experiments (scale bar, 100 µm), and the graph (right) shows the mean number of the invasive cells ± SEM (n = 3). * P < 0.05, ** P < 0.01, t-test. e-f, IB analysis of pROR1, pHER2, pCREB, ROR1, HER2, and CREB in 448 (e) and X01 (f) GSCs infected with shCtrl, shROR1-1 lentivirus, and then infected with vector control, CREB OE lentivirus. GAPDH was used as a loading control. g-h, Invasion assays of 448 GSCs (g) and X01 (h) GSCs infected with shCtrl, shROR1 lentivirus, and then infected with vector control, CREB OE lentivirus. Images (left) taken after 24 h of invasion are representative of three independent experiments (scale bar, 100 µm), and the graph (right) shows the mean number of invasive cells ± SEM (n = 3). * P < 0.05, ** P < 0.01, t-test.  2) Additional in vivo models to support conclusions: The authors should show efficacy of the NPs in at least one additional sensitive model. Moreover, the authors should show whether targeting IGFBP5 is ineffective in a model predicted to be insensitive to this approach.

Response Fig. 4 (New Extended Data
Response: As suggested by the reviewer, we performed antitumor evaluation in a non-invasive 83 GSC mice model. The course of treatment was identical to that with X01 GSC model. The tumor bioluminescence intensity in mice treated with Ang-SS-Cas9/sgIGFBP5 increased dramatically, comparable to that with PBS and non-specific sequence control Ang-SS-Cas9/sgScramble (Response Fig. 6a). In addition, the body weight in all the three groups show similar reduction during the treatment period (Response Fig. 6b). Importantly, the Ang-SS-Cas9/sgIGFBP5 nanocapsules did not prolong the survival time of 83 GSCs-bearing mice (Response Fig. 6c). All these results demonstrated that Ang-SS-Cas9/sgIGFBP5 was ineffective in a non-invasive mice model, supporting that the IGFBP5 plays a key role in the invasion of GSCs.
Accordingly, we have added relevant sentences in our revised manuscript as follows:  Page 15 lines 19--23 "To further confirm the specificity of IGFBP5 regulatory role for invasive GSCs, we performed Ang-SS-Cas9/sgIGFBP5 in a non-invasive 83-Luc (stable luciferase-expressing 83 GSCs) GSCbearing mice model. The results showed the identical outcomes of tumor growth and mouse survival among Ang-SS-Cas9/sgIGFBP5, Ang-SS-Cas9/sgScramble and PBS (Extended Data Fig. 9a-c)." Since reviewer 4 (comment 7) asked us whether our CRISPR-Cas9 system is better than siRNA- 2) 448 GSC was dissociated by Dr. Do-Hyun Nam in 2013 3 , which was classified as proneural subtype.
3) 131 GSC was obtained from Dr. Do-Hyun Nam 4 , which was firstly published as mesenchymal type of glioblastoma by upregulation of the mesenchymal related genes. GSCs which were generously donated by our collaborative labs. For the purpose in revealing that IGFPB5 is a universal regulator for invasive GBM, we applied the same assays to our newly isolated 772 GSC. We evaluated the stemness of 772 GSCs by confirming the notable high expression level of stemness markers CD133 and Nestin in comparison to the normal astrocyte cell line HA1800 (Response 3) Related to above, are there models predicted to be insensitive to targeting IGFBP5. For example, shRNAs against IGFBP5 in the other 2 models, 131 and 83 further reduce invasiveness and improve survival?
Response: We thank the reviewer for raising this point. Following your comment, we firstly silenced IGFBP5 in non-invasive GSCs (83 and 131 GSCs) by shRNA and evaluated the invasive ability in vitro.
The results showed that shRNA suppressed IGFBP5 mRNA expression in 83 and 131 GSCs (Response Fig. 10a, b), there was no significant differences on the invasiveness between shCtrl and shIGFBP5 in 83 or 131 GSCs (Response Fig. 10c, d), because of the low basal (Response Fig. 11) and secreted IGFBP5 level (Original Fig. 1f).   Fig. 1c (Fig. 1c left-up side), so that 292 DEGs were selected by both methods (4-fold change in RPKM and FDR < 0.1). Indicating that 292 Genes can be used for the selection of top-ranked DEGs (we revised the 299 Genes to be 292 Genes in revised Fig. 1c). The analysis results suggest that IGFBP5 is ranked on top 27 th in these 292 DEGs, which as the top 1 in the 36 DEGs candidates in Fig. 1c. based on corresponding clinical information for 309 genes obtained from the TCGA GBM microarray dataset, and these DEGs were significantly associated with poor survival (hazard ratio (HR) > 1, P < 0.01, Cox proportional hazards analysis) (Response Fig. 13b). Among them, IGFBP5 showed the 2nd highest differential expression between invasive and non-invasive GSCs, while just S100B (a wellknown astrocyte marker) was ahead of IGFBP5 (Response Fig. 13b).
Taken together, we found that IGFBP5 plays a potential oncogenic role in both invasive high-grade and low-grade glioma (see details in revised Fig. 1 in our manuscript). Since that invasion is considered the most common and pivotal event across the entire tumor development from low-grade to high-grade glioma 6 . We believe that finding out the key driver genes or biomarkers for invasion at the early time point of low-grade glioma is important to glioma therapy and translational research.
We revised Fig. 1h and described this part in the revised manuscript.  Response: We thank the reviewer for this comment. As suggested, we revised Fig. 1f using the same statistical comparison performed in Fig. 1b and 1e (Response Fig. 14).
DMEM/F12 without any growth factors, that provide insufficient supplements for the GSC growth, was used in invasive assay to avoid the cell growth contributed invasive differences.
As for your concern that Dong et al. 7 demonstrated that knockdown IGFBP5 inhibited GBM cell invasion. Actually, their results that IGFBP5 knockdown slightly increased cell growth suggested the effect of IGFBP5 on cell proliferation does not contribute to cell invasion.
Response: Following your comment, we performed immunoblot to detect ROR1, HER2, and CREB expression in the 4 GSCs (448, X01, 131 and 83) with a normal astrocyte cell line HA1800 as a control (Response Fig. 15). The result of immunoblot analysis showed the expression of CREB and ROR1 were enriched in invasive GSCs (448, X01) than non-invasive GSCs (131, 83), but the expressions of total HER2 did not show same pattern with the subtype of GSCs.
10) For figure 4, what are the quantifications of the RTK and kinase array?

Response:
We quantified the spot intensity and normalized it through the reference spots using ImageJ software. After normalized, the expression (mean pixel density) differences of phosphorylated HER2 and ROR1 were shown as rIGFBP5/Vehicle ratio (Response Fig. 16a).
Response Fig. 16. Quantification of RTK/Kinase Array. Quantitative analysis of the mean pixel density ratio of vehicle control (Veh, blue) to rIGF (100ng/ml recombinant IGFBP5 protein treated 6 h, red) for RTK array from manuscript Fig.4a (a), for Kinase array from Fig.4b (b) in 83 GSCs by using ImageJ software. * P < 0.05, ** P < 0.01, t-test. Refer Spots were used as loading controls. figure 7, does the Cas9/CRISPR reduce protein production?

Response:
We thank the reviewer for this comment. To explain the reasons why all mice reach endpoints on the exact same day, we would like to show the body weight and survival data sheets of 83 overexpressing IGFBP5 (Response Fig. 18). We established to sacrifice the mouse when it has decreased around 20% of body weight loss from 0 day of injection. Moreover, if the mice show neurological symptoms such as hunched back and body trembling, we decided to sacrifice the mouse.
As you can see in the table, we could observe all 4 mice had more than 20% decrease in body weight compared to 0 day except for two mice (number 1 mouse in cage 1, 097392 and number 5 in cage 2, 097393) (Response Fig. 18c). Two mice (number 1 mouse in cage 1, 097392 and number 5 in cage 2, 097393) showed severe neurological symptom. This is why we did sacrifice all 6 mice on the same day.
Response Fig. 18 non-invasive GSCs? There is no description in the Methods section about the RNA-seq. The raw data of RNA-seq data set were not available.

Response:
We apologize for the missing several critical technical details, especially for the statement "Statistics and reproducibility" in the Method part. We have included this part in our revised manuscript.
We confirm that three technical replicates were performed for all experiments for reproducibility. Part II: Once more, we apologize for the RNA-seq method missing, the detail of RNA-seq have been added in our revised manuscript. To provide you a simple explanation for the raw data set of RNA seq, we actually uploaded the raw data set of RNA sequence followed the editor's suggestion when this study was under consideration. However, we have no idea why you couldn't get our raw data before. Currently, NCBI has released our RNA-Seq raw result. Accordingly, you could get our RNA-Seq raw data from Following purification, the mRNA was fragmented into small pieces using divalent cations under elevated temperatures. The cleaved RNA fragments were copied into first-strand cDNA using SuperScript II reverse transcriptase (Invitrogen) and random primers. This was followed by secondstrand cDNA synthesis using DNA Polymerase I, RNase H and dUTP. These cDNA fragments were then subjected to an end-repair process, the addition of a single 'A' base, and adapter ligation. The products were then purified and enriched with PCR to create the final cDNA library. Indexed libraries were then paired-end sequenced with an Illumina HiSeq 4000 (Illumina, Inc., San Diego, CA, USA) at Macrogen Incorporated.

RNA-seq data analysis
Raw fastq files were aligned to the human reference genome (hg19) using the STAR program (https://github.com/alexdobin/STAR). The aligned sam files were converted to bam files and sorted by coordinate using the Samtools program (http://www.htslib.org). Duplicate reads were removed using Picard (https://github.com/broadinstitute/picard). The gene expression values were calculated based on the read per kilobase of exon per million (RPKM) value. Genes that had greater than 30% missing values were discarded. The expression levels of the filtered genes were globally normalized with the Quantile normalization method using the R limma package. The enrichment score (ES) of four GBM subtypes 73 with each molecular signatures in 772 GSCs was analyzed using single sample gene set enrichment analysis (ssGSEA)." In vivo xenograft data in Fig. 3i. There is no description in the Methods section about the xenograft experiment. It is unclear what were the biological replica. The quantification data is not presented.

Response:
We really appreciate the reviewer for pointing out the missed experimental details of xenograft data from Fig. 3i. We have revised the statement of Histology and immunohistochemistry (IHC) staining method in the revised manuscript as follows:  Page 28 line 5--Page 29 line 2 "Histology, Immunohistochemistry (IHC) and immunofluorescence staining. For histological observations, the brains were removed, fixed with 4% paraformaldehyde for 24 h at 4 ℃, sectioned at a thickness of 4 μm using an essential microtome (Leica RM2125 RTS). For histological observations, haematoxylin and eosin (H&E) stains were carried out with 1% hematoxylin (DaKo) and 0.25% eosin (Merck). Stained sections were dried and mounted in an organic mounting medium. Prior to immunohistochemical (IHC) staining for pHER2 Y1248 (1:100, R&D Systems), pROR1 Tyr786 (1:100, Thermo Fisher) and pCREB Ser133 (1:100, Cell Signaling Technology), the sections were subjected to an antigen retrieval process using citrate buffer (pH 6.0), and endogenous peroxidase was blocked by incubating with 3% hydrogen peroxide. The tissue sections were then incubated overnight at 4 ℃ in a humidified chamber with the primary antibody, diluted with antibody diluent buffer (IHC World). The tissue sections for 3,3′-diaminobenzidine (DAB) staining were developed using DAB (Vector Laboratories) as the chromogen. For immunofluorescence staining against GFP, the tissue sections were subjected to an antigen retrieval process using citrate buffer (pH 6.0). Then incubated overnight at 4 ℃ with anti-GFP antibody (1:500, Abcam) in antibody diluent buffer (IHC World). Secondary staining was performed at RT for 2 h with fluorochrome-conjugated antibody (Alexa 568, Thermofisher Scientific) and 4',6diamidino-2-phenylindole (DAPI; Sigma, 1:5,000). Fluorescence images were acquired with Zeiss LSM 780 confocal laser-scanning microscope (Carl Zeiss, Thornwood, NY, USA)." Additionally, we calculated the number of GFP positive cells which infiltrated into the corpus callosum in the brains of mice. The number of infiltrated cells that had spread outside the tumor mass was extremely rare in Vector-overexpressed 83 GSC bearing mice. However, the expression was over 100 counts representing GFP-positive cells (Mean = 120) in ectopically IGFBP5 overexpressed 83 GSC bearing mice (Response Fig. 19). This quantitative data supports our conclusion that IGFBP5 promotes GSC invasion.
Response Fig. 19. Quantification of GFP-positive (GFP + ) cell numbers spread into the corpus callosum. GFP + cells indicate infiltrating into the corpus callosum outside of tumor mass in orthotopic xenografts of 83 GSCs infected with vector control or IGFBP5 lentivirus. Data are presented as mean ± SEM (n = 3). **** P < 0.0001, t-test.
2) The conclusion that IGFBP5 acts as ROR1 ligand was based on the data shown in Fig. 4. Unfortunately, all biochemical assays lacked quantified results (e.g., Fig. 4a-4h). Were these experiments performed only one time? Quantitative and reproducible biochemical data are needed. Same question with Fig.   7k.

Response:
We would like to explain that three technical replicates were actually performed for all experiments for reproducibility and have included "Statistics and reproducibility" section in the revised manuscript (details as mentioned above in Response to major Comment 1). In the response to major Comment 5, we have demonstrated the specificity of IGFBP5 binding with ROR1 by determining the Kd value (Please see details as below in Response to major Comment 5).
As suggested, the quantification of biochemical assays, immunoblot, and IHC staining has been provided in response Fig. 16 and 20. Kindly remind that we have replaced the siRNA of HER2 and ROR1 by shHER2 or shROR1 in X01 GSCs in response Fig. 21 for our revised Fig.4 k, l, which showed higher efficacy than the siRNA transfected cell, since reviewer 1 suggested us to conduct shROR1 and shHER2 for the further experiments.

Response Fig. 16 (Similar question from Reviewer 1). Quantification of RTK/Kinase Array.
Quantitative analysis of the mean pixel density ratio of vehicle control (Veh, blue) to rIGF (100ng/ml recombinant IGFBP5 protein treated 6 h, red) for RTK array from manuscript Fig.4a (a), for Kinase array from Fig.4b (b) in 83 GSCs by using ImageJ software. * P < 0.05, ** P < 0.01, t-test. Refer Spots were used as loading controls.
3) It is also troublesome that contradictory results (e.g., Fig.4d, different effects on ROR1 levels; Fig. 4e different effects ROR1 levels) were ignored. How was the staining specificity validated?

4)
Are these statistically different changes? How was the staining specificity validated? Are these GSC cells and how do you know?

Response:
We would like to explain that three technical replicates were actually performed for all experiments for reproducibility and have included "Statistics and reproducibility" section in the revised manuscript (details as mentioned above in Response to major Comment 1).
Additionally, the representative images of IHC staining for original Fig. 4k, Fig. 4l, and Fig. 6g (Response Fig. 22-24) were provided and quantified. And for the specificity of the primary antibody, we employed IgG and secondary antibodies in each brain slides of each experimental groups to confirm the negative controls and auto-fluorescence.
As the reviewer suggested, we have detected and quantified the expression of stemness marker Nestin in the brain of X01 GSCs-bearing mice (Response Fig. 25 Fig. 1a-d) and ROR1 (Response Fig. 2a-d).
ii) The pharmacological HER2 inhibitor, Irbinitinib reduced the downstream of CREB phosphorylation (Response Fig. 1e, f) and the GSCs invasion (Response Fig. 1g, h) as well.
iv) We overexpressed CREB in non-invasive GSC and determined the role of CREB in regulating GSC invasion. Immunoblot analysis verified a successful ectopic CREB overexpression in noninvasive GSCs (83 and 131) (Response Fig. 3a-b). CREB overexpression significantly increased downstream genes ETV5 and FBXW9 expression, and GSC invasive capacity (Response Fig. 3c-h). To investigate the downstream of CREB in IGFBP5-ROR1/HER2 signaling pathway, we ectopic overexpressed CREB in IGFBP5 knockdown GSCs, and CREB sufficiently rescued GSC invasion (Response Fig. 3i, j). Furthermore, overexpression of CERB also rescued HER2 or ROR1 knockdown mediated repression of GSC invasion (Response Fig.   4a-h). These results indicated CREB is the key molecule in IGFBP5-mediated ROR1/HER2 signaling axis.
Taken together, these results suggest an essential role of IGFBP5/ROR1/HER2-CREB signaling axis in GSC invasion.
Response Fig. 1 (Revised Fig. 4k and New Extended Data Fig. 4d, f- For the suggestion on specificity of IGFBP5 binding and the Kd value, we have conducted microscale thermophoresis assay (MST) to identify the specificity binding between IGFBP5 and HER2/ROR1 dose responses, and also calculated the Kd value through the Kd fitting mode using MO Affinity Analysis software. According to the analysis of IGFBP5-ROR1 binding affinity curve, we concluded that IGFBP5 and ROR1 could specifically bind with Kd value of 157.5 nM (Response Fig.   26). However, we found that IGFBP5 and HER2 could not bind with no affinity, since the signal to noise ratio less than 5 (Response Fig. 27). Taken together, our results suggest that IGFBP5 serves as a novel ligand for ROR1.
Response Fig. 26 (Revised Fig. 4j) Response Fig. 27 (New Extended Data Fig. 4c). Measurement of IGFBP5-HER2 binding affinity in vitro. The in vitro binding affinity between IGFBP5 and HER2 was tested by MST assay. The concentration of IGFBP5 proteins is kept constant at 50 nM, while the HER2 concentration varies from 1.00 µM to 0.03 nM at response time of 10 sec (a) and 2.5 sec (b). Fig. 7m -I could not see much difference between the scrambled control and sgIGFBP5 group. Yet, it was stated that in vivo nano capsule-mediated IGFBP5 disruption dramatically increase mouse survival. Did I miss something? Did the authors actually detect any indel in the IGFBP5 gene in these mice?

Response:
We have provided the detailed information regarding this as below: In the Fig. 7m, the median survival time of mice in the PBS treatment group is 23 days, sgScramble group is 25 days, while sgIGFBP5 group is prolonged up to 64 days. Accordingly, survival time of mice treated by sgIGFBP5 increased around 200% comparing to these of mice treated with PBS or sgScramble.
For your other concern about indel in the IGFBP5 gene editing in the mice, our previous work using the same nanocapsules except for the target gene PLK1 instead of IGFBP5, showed these nanocapsules induced 38.1% indel frequencies in GBM-bearing mice after successive intravenous injections, which is the highest gene editing efficiency in the brain via non-invasive treatment to the best of our knowledge 8 . To further evaluate whether the efficient gene editing of these nanocapsules resulted from the blood-brain barrier (BBB) penetration and tumor accumulation, we also conducted tumor penetration experiments of Ang-SS-Cas9/sgRNA and free Cas9/sgRNA. Confocal images showed that our Ang-SS-Cas9/sgRNA nanocapsules have much stronger brain targeting ability than free Cas9/sgRNA (Response Fig. 28). Moreover, immunoblot analysis of IGFBP5 in tumor tissue from mice treated with Ang-SS-Cas9/sgIGFBP5 showed much lower expression level than Ang-SS-Cas9/sgScramble or PBS (Original Fig. 7k). Taken together, these data suggest that the Ang-SS-Cas9/sgRNA nanocapsules can efficiently deliver Cas9/sgIGFBP5 to the brain tumor and suppress IGFBP5 gene expression in the tumor.
Response: Fig.7k showed the IB analysis of entire tumor tissue lysates.
3) The effects of IGFBP5 on cell migration/invasion have been reported since the late 1990s. These papers should be cited.

Response:
We appreciate reviewer's helpful comments. We have cited the papers as you suggested.
 Page 6 lines 11--13 "IGFBP5 is a secreted protein of the IGFBP family that mainly regulates the specific binding of insulin-like growth factors (IGFs) to IGF receptors 21-24 , which showed a high correlation with the migration of breast cancer 25,26 ." 4) For in vitro invasion assays using shRNA-mediated knockdown, the possible off-target effects were not addressed. A simple experiment would be to test if adding back rIGFBP5 to the knocked down cells can rescue the cell invasion/gene expression.

Response:
We appreciate reviewer's helpful comments. As suggested, we treated recombinant IGFBP5 back into the IGFBP5 knockdown 448 and X01 GSCs. The result of invasion assay and immunoblot showed rIGFBP5 treatment can rescue the phosphorylation of HER2, ROR1 and CERB (Response Fig.   29a, b) and GSCs invasive ability (Response Fig. 29c, d).

Reviewer #3 (Remarks to the Author):
This interesting study shows a clear effect of IGFBP-5 in promoting GBM stem cell invasion and provides a plausible mechanism involving ROR1, HER2, and CREB activation, together with the CREB-dependent expression of ETV5 and FBXW9. Strengths: In general, the experiments are clearly presented and well interpreted, and the development of an IGFBP-5-targeting nanocapsule that significantly impairs GBM tumor growth and prolongs survival in mice with orthotopic GBM tumors is very impressive. Weaknesses: Although IGFBP-5 is immunoprecipitated using an anti-ROR1 antibody its direct binding to ROR1, as proposed, is not demonstrated, and the existence of a functional HER2-ROR1-IGFBP-5 complex is unclear since it is not precipitated using an anti-HER2 antibody (see specific comment #4). Therefore the precise mechanism of IGFBP-5 action is in some doubt, and a role for IGF1R phosphorylation is not explicitly excluded by the single negative phosphoarray experiment (see specific comment #7). Nevertheless this work adds significantly to current knowledge of the role of IGFBP-5 in GBM.
Specific comments 1. Line 116: "a small number of studies". Since high IGFBP-5 expression in GBM has been reported previously (as far back as 2006, doi: 10.1177/153303460600500303), some of these prior studies should be cited here.

Response:
We thank the reviewer for raising this point. We have cited the relative references in our revised manuscript.
Response: As suggested, we quantified the immunoblot results of Fig.4c, and calculated the ratio of phosphorous to total protein (CREB, HER2, ROR1) estimated from immunoblot results (Response Fig.   30). As the bars showed, the treatment of recombinant IGFBP5 protein (rIGF) to GSCs significantly increased phosph/total ratio compared to control groups.
Response Fig. 30. Quantification of IB analysis in Fig. 4c. Bar graph for IB analysis to calculate the gray value of each band, shown as the ratio of phosphorylation to total CREB/HER2/ROR1 expression level from manuscript Fig. 4c (left 83, right 131) using ImageJ software. * P < 0.05, ** P < 0.01, *** P < 0.001, t-test.
3. Fig. 4f precipitate both ROR1 and IGFBP-5, but this is not seen in Fig. 4f. One interpretation is that IGFBP-5 can bind to ROR1, but this binary complex does not interact with HER2. This is different from the model shown in Fig. 4m. Further interpretation of the coIP data is required, as well as some coIPs using anti-IGFBP-5 as the precipitating Ab.

Response:
We greatly appreciate the reviewer for pointing this out. As suggested, the model in Fig. 4m was not precise model and deleted in the revised Fig. 4. We have conducted additional experiments to support our theory, including revised Fig. 4 and a New Extended Fig. 4.
(1) immunoprecipitation with HER2 or ROR1 antibody in X01 and 448 GSCs, it was found that the IP of HER2 antibody could precipitate ROR1, and the IP of ROR1 antibody could precipitate HER2 (Response Fig. 31a-d), indicating that there was interaction between HER2 and ROR1.
(2) exogenous expression of HA-ROR1 and Flag-IGFBP5 in 293T cells, HA and Flag were used as precipitation antibodies for Co-IP. Co-IP using Flag as the precipitating antibody showed interactions between IGFBP5 and ROR1, and ROR1 interacts with both IGFBP5 and endogenous HER2 when using HA as the precipitating antibody (Response Fig. 31e, f).
(3) the purified His-IGFBP5 and ROR1 recombinant proteins were used for in vitro MST assay to detect the affinity between ROR1 and IGFBP5. The results showed a clear binding curve to IGFBP5-ROR1 with a Kd of 157.5 nM (Response Fig. 31g), but no binding affinity between IGFBP5 and HER2 (Response Fig. 27a, b), which suggested that IGFBP5 serves as a novel ligand for ROR1 to trigger downstream signaling transduction.
We described this part in the revised manuscript.
Please see page 9, line 11--page 10, line 3 in our revised manuscript.

Response:
We thank the reviewer for this comment. We determined overexpression of ROR1 wild type (ROR1 WT) but not ROR1 mut (Response Fig. 2), and overexpression of CREB (Response Fig. 3) rescued knockdown IGFBP5 mediated repression of HER2-CREB phosphorylation and GSC invasion.
As suggested, we deleted overinterpreted sentence ".. implying that IGFBP-5 promotes …", and rewrote this paragraph in our revised manuscript.
Please see page 10, lines 4-16 in our revised manuscript.
Response Fig. 2 (Revised Fig. 4l  6. Lines 193-194. " … HER2 knockdown did not alter ROR1 phosphorylation", etc. It is difficult to be confident about this statement in the absence of a quantitative analysis of Fig. 4h. Phospho/total data should be analyzed for several repeat experiments.

Response:
We thank the reviewer's comment. As reviewer 1 suggested, we have replaced the siRNA of HER2 and ROR1 by shHER2 or shROR1 X01 GSCs, which showed higher efficacy than the siRNA transfected cell. And we quantified the immunoblot results of revised Fig.4k, l (Response Fig. 21). The phosphor/total ratios of CREB decreased in both shHER2 and shROR1 (Response Fig. 32) in X01 GSC.
Response Fig. 32. Quantification of IB analysis in Fig. 4h. Bar graph for IB analysis to calculate the gray value of each band, shown as the ratio of phosphorylation to total CREB/HER2/ROR1 expression level from manuscript Fig.4h (left shHER2, right shROR1) using ImageJ software. Ratio of phosphorylation/total protein expression ± SEM (n = 3). ** P < 0.01, n.s P > 0.05, t-test.
7.Lines 332~333. Rather than introducing a new result in the Discussion, the authors should indicate the position of pIGF1R in Fig. 4a to illustrate that IGF1R phosphorylation is not promoted by IGFBP-5.
Given that IGF1R phosphorylation may have occurred transiently and not been detected on this array, it would be informative to check the acute pIGF1R response to IGFBP-5 (e.g. a time-course starting at 5 or 10 min) by immunoblot and to ensure that the addition of an IGF1R tyrosine kinase inhibitor such as NVP-AEW541 did not block any downstream effects of IGFBP-5.

Response:
We thank the reviewer for this constructive comment. We have labeled the position of pIGF1R in revised Fig.4a (Response Fig. 33), which showed very low expression of IGF1R tyrosine phosphorylation in both Vehicle and rIGFBP5 treated GSCs. Moreover, immunoblot analysis showed no expression changed in pIGF1R expression with rIGFBP5 treatment for 6 h (Response Fig. 34a).
Then we tested the short time treatment of rIGFBP5 into 83 and 131 GSCs for 5, 8 and 10 min, respectively, and examined the acute pIGF1R response of IGFBP5. The results showed that the treatment of rIGFBP5 did not affect IGF1R phosphorylation (Response Fig. 34b, c).

Minor point
Line 112: "… adverse roles": I believe the authors mean "diverse" roles.

Response:
We thank the reviewer for pointing this out. We have revised the "adverse" to "diverse" at page 6, line 13 of our revised manuscript.

Reviewer #4 (Remarks to the Author):
Lin et al. report a comprehensive study into the role of Insulin-like Growth Factor-Binding Protein 5 (IGFBP5) on glioblastoma infiltration in orthotopic mouse models, in particular, as it relates to GSCs. In a GSC-derived brain tumor model, they conducted RNA seq analysis and found that IGFBP5 was differentially expressed between invasive or non-invasive subtypes. The authors then engaged in detailed mechanistic studies to establish the important of IGFBP5 for GSC invasion. Finally, they report data suggesting that the nanoparticle-based delivery of CRISPR/Cas9-based IGFBP5 gene editing results in impressive therapeutic benefits, such as significant increase in survival in a GBM mouse model.
Overall, this is a large research study with an impressive attention to details, in particular on the mechanistic aspects. The initial hypothesis related to IGFBP5 is motivated by a comparative study between invasive and non-invasive GBM subtypes. They follow up with both silencing and know down experiments that clearly establish the link between IFGBP5 and GSC invasion. While this study has clearly merits -in particular as it relates to GBM pathobiology -there are also significant shortcomings that dampen the potential impact. The following aspects should be addressed prior to publication: such as nude, NOD-SCID, or NOD-SCID-gamma mice. The immune system in these mice differs innately from that of the host; thus, current PDXs do not represent the host immune system.
We have discussed the limitations of our models in Discussion section of the revised manuscript as follows:  Page 19 lines 22--24 "Although, the current PDXs model only can be established in immunodeficient mice such as nude, NOD-SCID, or NOD-SCID-gamma mice. However, these preclinical data offer a proof-of-concept for the treatment of invasive GBM by targeting IGFBP5." 3) IFGBP5 is overexpressed in a wide range of tissues, and it is unclear how CRISPR/Cas9-based IGFBP5 gene editing would affect these tissues systematically, see, e.g., https://www.proteinatlas.org/ENSG00000115461-IGFBP5/tissue. This seems an important aspect that should be addressed by the authors.

Response:
We appreciate the reviewer's nice comment regarding the safety concerns. It is true that IGFBP5 is also expressed in other normal organs including liver and kidney. However, the Ang-SS-Cas9/sgIGFBP5 nanocapsules caused little off-target in these organs even though they accumulated in and excreted from the liver, kidney and spleen. This is mainly attributed to two reasons: (1) the Angiopep-2 decorated nanocapsules specifically target to the low-density lipoprotein-1 (LRP-1) overexpressed endothelial and tumor cells, while could hardly be taken by other normal cells that express much less LRP-1 receptor; (2) even little nanocapsules are taken by the normal cells, they could not induce IGFBP5 gene editing as the Cas9/sgIGFBP5 cannot be released from the nanocapsules due to the low intracellular reduction microenvironment in normal cells. Because the nanocapsules contain abundant disulfide bond which can be specifically cleaved in excessive glutathione tumor microenvironment, thusly triggering the disassemble of nanocapsules and fast Cas9/sgIGFBP5 release for effective gene knockout in tumor sites. The brain delivery system has been well-established in our lab, see more details in our previous work 8 (Science Advances, 2022). Therefore, the selective cellular uptake and specific intracellular drug release in tumor cells could avoid the off targets in other normal tissues.
4) The nanoparticle delivery system provides clear benefits in survival, but a more complete characterization would be warranted. What is the biodistribution of the nanoparticles? Where are they accumulating? If they go into liver and spleen, is there any toxicity or off-target activity? In particular, it would be interesting to know if they accumulate in any of the tissues that overexpress IFGBP5.
Response: Thank for the reviewer for this constructive comments. (1) The nanocapsule delivery system was similarly with our recently published work 8 . Detailed characterization of the nanocapsules was presented in that paper, the results showed that the targeting nanocapsules achieved approximately 11.8%ID/g (injected dose per gram) in tumor site in both immune-free and immune-competent tumor mice models (Response Fig. 36). Meanwhile, high accumulation was also observed in the main organs including liver, kidney and spleen, indicating they may be excreted from these tissues that are in line with reported results [15][16][17][18] .
(2) Indeed, abundant nanocapsules were accumulated in kidney and spleen, while non-significant toxicity was observed in these tissues from the histological analysis (Response Fig. 37). That is mainly attributed to the specific and selective tumor cell uptake and intracellular drug release. Moreover, we have also performed the blood analysis for the nanocapsules, the results showed that these nanocapsules displayed no obvious side effects as compared with PBS treatment (Response Fig. 38). Collectively, these data suggest that the targeted nanocapsules possess good biocompatibility.
[REDACTED] Editorial Note: see Figure 5 A-H of Yan Z., et al. Sci. Adv. 2022 8: eabm8011 (3) Yes, the nanocapsules also accumulated in normal tissues (kidney and liver) that expressed IGFBP5, while they may induce non-specific gene editing in these tissues due to the relative low cell uptake and Cas9/sgIGFBP5 release (refer to comment 3).
escalation study should be included. What is the fraction of the total dose delivered to the brain tumor?
How much was delivered to the brain? What is the GBM/brain ratio?

Response:
We appreciate the reviewer's constructive comment. The dosage is 1.5 mg Cas9 equiv./kg for the anti-tumor evaluation of the nanocapsules. The mice treated with the nanocapsules showed negligible side effects as observed from both the histological and blood analysis. Thusly, we deduce that these nanocapsules have good biocompatibility in vivo. Considering the high expense of Cas9 protein, we did not carry out the dose escalation study, but the constitutions of the nanocapsules are non-toxic materials, including Cas9/sgRNA complex, polyethylene glycol (PEG), angiopep-2 peptide, acrylate guanidine (AG) and cysteamine bisacrylamide (CBA).
As calculated from the biodistribution analysis, more than 11.8% ID/g Cas9/sgRNA was delivered into the tumor, which was 11.5-fold higher than that into the normal brain (1.03% ID/g). Therefore, the ratio of GBM/brain is 11.5, indicating the specific tumor targeting capability of these nanocapsules.

7)
The authors claim for their CRISPR-Cas9 system "a much preferable outcome on tumor growth as well as mouse expectancy, comparing to the well-known siRNA-based therapies targeting the conventional GBM targets (i.e., PLK134 or STAT3)". However, when consulting their cited references, it appears as the opposite is the case, i.e., their survival results, also significant, don't match up with the one achieved by the other therapies.

Response:
We thank the reviewer for pointing out this. As mentioned that siRNA-based therapies in GBM models 17,18 also showed significant survival results, the reason we mentioned as CRISPR-Cas9 system "a much preferable outcome on tumor growth..." is that we compared the survival curve differences of our Cas9/sgIGFBP5 with siRNA targeted PLK1 or STAT3. Targeting PKL1 by siRNA improved the survival days of GBM mice model from about 28 days to 52 days. Although the conventional GBM therapy of targeting STAT3 SNPs + IR prominent improved of survival curve of GBM models, the mono-targeting STAT3 by siRNA only showed around 5 day-difference in survival length 17 . However, targeting IGFBP5 by CRISPR-Cas9 system in our Fig. 7m promoted survival of around 60-day in GBM mice models.
Additionally, we also performed IGFBP5 siRNA therapeutic efficacy towards CSC-2 GSC orthotopic xenograft model. Although in vivo nanocapsule-mediated siIGFBP5 knockdown significantly increased mouse survival (Response Fig. 6f), the improvement of median survival time was less than 10 days.
Taken together, our results and related references suggest that Cas9/sgIGFBP5 nano-therapy is a more potent method for GBM targeting therapy. Second, as the reviewer suggested, we performed immunoblot analysis to detect the expression of pIGF1R in both invasive GSCs (X01 and 448) and non-invasive GSCs (83 and 131). Results showed higher basic expression levels of pIGF1R in non-invasive GSC 83 compared to GSCs X01, 448, and 131 (New Response Fig. 1), suggesting that 83 GSCs might have an upregulation of autocrine IGF1 or IGF2 signaling.