Lipoprotein-biomimetic nanostructure enables efficient targeting delivery of siRNA to Ras-activated glioblastoma cells via macropinocytosis

Hyperactivated Ras regulates many oncogenic pathways in several malignant human cancers including glioblastoma and it is an attractive target for cancer therapies. Ras activation in cancer cells drives protein internalization via macropinocytosis as a key nutrient-gaining process. By utilizing this unique endocytosis pathway, here we create a biologically inspired nanostructure that can induce cancer cells to ‘drink drugs' for targeting activating transcription factor-5 (ATF5), an overexpressed anti-apoptotic transcription factor in glioblastoma. Apolipoprotein E3-reconstituted high-density lipoprotein is used to encapsulate the siRNA-loaded calcium phosphate core and facilitate it to penetrate the blood–brain barrier, thus targeting the glioblastoma cells in a macropinocytosis-dependent manner. The nanostructure carrying ATF5 siRNA exerts remarkable RNA-interfering efficiency, increases glioblastoma cell apoptosis and inhibits tumour cell growth both in vitro and in xenograft tumour models. This strategy of targeting the macropinocytosis caused by Ras activation provides a nanoparticle-based approach for precision therapy in glioblastoma and other Ras-activated cancers.

rHDL particles and effective ATF5 suppression by siRNA in vitro are well designed and executed. The evidence that macropinocytosis plays a key role in the process is also convincing. The endogenous Ras and Rac pathways are known to play an essential role in the control of macropinocytosis in both normal and transformed cells, so it is not surprising that knockdown of Ras expression results in reduced uptake of CaP-rHDL nanoparticles. However, the claim that that such nanoparticles represent a specific therapeutic approach for selective targeting of tumors driven by activated Ras is confusing, and the studies fall short of demonstrating this. The problem is, in part, related to the authors' use of the terms "Ras-activated" glioblastoma and "Rastransformed cell line". Activating mutations in Ras are rare in human GBM, and the majority of the studies in this report were performed with the C6 rat glioma cell line, which does not harbor Ras mutations and thus is not a "Ras-activated cell line". It is true that expression of endogenous Ras protein is relatively high in these cells compared with normal astrocytes but this is not the same as a "Ras-activated" tumor where Ras mutation (e.g., V12) results in constitutive activation of Ras pathways and is a driver of tumorigenesis and tumor progression.
Specific Comments: 1) Measurements of total endogenous Ras protein expression levels (Fig. 5a) do not necessarily indicate that Ras is "hyperactivated". Determination of the percentage of activated Ras (e.g., by G-LISA Ras activation assays) would be required to draw this conclusion.
2) The conclusion that CaP-rHDL nanoparticles "penetrated deep and distributed extensive(ly) in the C6 glioblastoma spheroids" (pg 16) seems inconsistent with the confocal images in Supplementary Fig. 2, which show most of the DiI-labeled CaP-rHDL concentrated around the periphery of the spheroids. Regarding the same point, in Fig. 5b, the images of patient-derived spheroids appear to be whole mounts, so it is not clear whether the fluorescent cells are mostly on the surface or the interior.
3) The small sizes of some of the illustrations, particularly in the complex figures, make it impossible to see if the conclusions are supported. For example, in Fig. 8d, the authors conclude that treatment with ATF5 siRNA "led to hyperchromatism of the cell nuclei". That panel should be moved to the supplementary data and expanded so that nuclear morphology can be easily assessed. 4) In Fig. 8e, we are shown one spheroid for each condition evidence that anti-ATF5-CaP-rHDL nanoparticles inhibit growth/viability of C6 spheroids. This conclusion needs to be supported by quantitative data such as average spheroid diameter or volume, with an appropriate n and statistical analysis. 5) The argument that the increased apoptosis ( Fig. 9) and the improved survival ( Fig. 10) of mice treated with anti-ATF5 nanoparticles is specifically related to suppression of ATF5 would be strengthened by IF or WB results demonstrating that ATF5 is suppressed in the tumor tissue. 6) The paper contains numerous grammatical and typographical errors throughout.
Below is a detailed description of our revision according to the review comments.

Reviewer: 1
Comments: This manuscript reports the development of a biomimetic lipoprotein nanoparticle (CaP-rHDL) for the efficient delivery of siRNA as the target treatment of Ras-activated brain cancer. In this work, the authors carefully characterized the Ras-activation dependency of the macropinocytosis-mediated cellular uptake of the CaP-rHDL in both glioblastoma cell line and patient-derived glioblastoma. They further demonstrated the anti-glioma activity of ATF5 siRNA-loaded CaP-rHDL both in vitro and in vivo, resulting in remarkable RNA-interfering efficiency and increased apoptosis at a low siRNA dose (0.36mg/kg). Ras pathway plays essential role in malignant transformation and is an attractive target for cancer therapies. This strategy of targeting the macropinocytosis caused by Ras activation is highly novel and could provide powerful nanoparticle-based treatment strategy for Ras-activated cancers. In my opinion, this study is potentially very exciting and is a rather comprehensive work that could be considered for publication on Nature Communications after addressing following concerns.
1. The stability of control particle, CaP-LNC for in vivo study could contribute to the significant low uptake of CaP-LNC in tumor compared with CaP-rHDL, rather than the less targeting effect assumption made by authors (Fig 6 discussion). Therefore, serum stability and pharmacokinetic study should be conducted to guide in vivo data analysis. In addition, an alternate but better control could be the formulation described in Ref 19 since it contains both cholesterol and PEG-lipid to stabilize the particle and has shown good in vivo behavior and RNAi effect. Further, due to the stability concern of CaP-LNC, a much higher and selective accumulation of CaP-rHDL in the glioma regions in Fig.6 is not a conclusive evidence to support the statement of "ApoE-reconstituted lipoprotein could overcome BBB, BBTB and enormously accumulated at the glioma site where nutrient is lack due to tumor growth".
Response: Good points. We agree with the reviewer that the stability of CaP-LNC might also contribute to the significant low uptake of CaP-LNC in tumor compared with CaP-rHDL, and the serum stability and pharmacokinetic study should be conducted to guide in vivo data analysis. Therefore, we firstly  Table 1 (Fig. 6f-h), suggesting that the incorporation of ApoE in the nanocarrier did play an important role in enhancing its glioblastoma-targeting efficiency. According to the reviewer's suggestion, these additional data were incorporated in the revised manuscript, please see Fig. 1d, Fig. 6f

Comments:
Brain tumor delivery is a significant concern. The studies involve a potential improvement in the specificity of delivery based on a system specifically geared towards ras activated cells. The initial phases of the development are interesting, but the biologic aspects of the studies are relatively weak and would benefit from substantial additional effort. In addition, the data are poorly presented with substantial data potentially better presented as data not shown or in the supplemental data.  Response: Great points. We agree with the reviewer that more models will make the study more reliable. C6 cell, as a classical rat glioma cell line with close relation with RAS pathway for cell proliferation 6 (Fig. 4 and supplementary Fig. 2-3).
To determine whether the cancer cell-targeting delivery strategy can be translated into a human system, we further investigated the cellular uptake of CaP-rHDL and its macropinocytosis dependence in a patient-derived  Fig. 6f-h, Fig. 8e-f, Fig. 9b-c and Fig. 9e), which nicely support our hypothesis that CaP-rHDL not only efficiently delivers siRNA to the target tissue, but also facilitates the gene knockdown activity to exert its antitumor efficacy. The related data have been added in the revised manuscript, please see Fig. 4-5, Fig.   6f-h, Fig. 8e-f, Fig. 9b-c, Fig. 9e and supplementary Fig. 2-4. According to the reviewer's suggestion, using G-LISA Ras activation assay, we

measured the Ras activity in C6 glioblastoma cells, patient-derived glioblastoma initiating cells, human glioblastoma cell lines (U87and U251 cells), pancreatic cancer cells (MIA PaCa-2 and BxPC-3 cells) and colorectal cell lines (SW-480
and Caco-2 cells) before and after shRNA-mediated Ras knockdown (Fig. 4-5 and supplementary Fig. 2-3), and found a nice linear relationship between cellular Ras activity and the efficiency of cellular uptake of CaP-rHDL. Please check the detailed data in Fig. 4-5, supplementary Fig. 2-3 (Fig. 9c). In addition, the long term effect of ATF5-CaP-rHDL treatment was reflected by the survival in both animal models following four or five injections (Fig. 9d-e). 8. The ability to deliver across the blood-brain barrier was not tested and the targeting of the reagents are exciting but the entire evaluation is not optimal and are not at the level for a high impact journal, yet.

Response: Thanks for the suggestions. Additional experiments have been
performed to test the ability of CaP-rHDL to penetrate across the blood-brain barrier (BBB) with blood vessels visualized by anti-CD31. As shown in the supplementary Fig. 5, CaP-rHDL was found efficiently penetrate through the BBB in both glioblastoma-bearing animal models.
According to the reviewer's suggestion, in this revised manuscript, additional cell models and animal model bearing patient-derived glioblastoma were further used for studying the mechanism and evaluating the efficacy of CaP-rHDL-mediated glioblastoma-targeting siRNA delivery (Fig. 4, 5, 6, 8 and 9).
Hope all our revisions and responses will be satisfactory and this revised version will be acceptable for publication in Nature Communications.

Reviewer: 3
Summary: In this manuscript the authors provide evidence that ApoE-rHDL nanoparticles containing a calcium phosphate (CaP) core can be taken up effectively by glioma cells in culture (monolayer and spheroid) and in intracranial glioma in mice. Furthermore, they demonstrate that such nanoparticles can be employed to deliver siRNA targeting a pro-survival protein, ATF5, thereby inhibiting intracranial tumor growth and prolonging survival of tumor-bearing mice. The authors claim two other novel aspects of their work: 1) The mechanism of uptake of the CaP-rHDL nanoparticles entails their incorporation into macropinosomes. 2) Macropinocytosis is Ras-dependent, suggesting that the NPs may be particularly useful for targeting "Ras-activated glioblastoma". Response: Great comments. We agree with the reviewer that activating mutations in Ras are rare in human GBM. According to the previous reports, the RAS/RAF pathway activation in glioblastoma is achieved much more frequently by copy number gains of RAS/RAF and/or upstream growth factor (receptor) than by activating RAS/RAF mutations 12,13 . Simultaneously, the level of Ras-GTP in glioma cell lines and glioma specimens is markedly elevated compared to the normal cells and brain specimens, and is comparable to that in oncogenic Ras transformed fibroblasts [14][15][16][17] . As expected, C6 cell, a classical rat glioma cell line, showed a close relation between RAS activity and cell proliferation ability 6 Fig. 4-5). In addition, reducing Ras activity by knocking down Ras expression also significantly inhibited the macropinocytosis-dependent cellular uptake of CaP-rHDL (Fig. 4-5 and supplementary Fig. 2-3). These data collectively suggested that the selective targeting of CaP-rHDL to tumors is driven by activated Ras. Additional data and discussion please see Fig. 4 1) Measurements of total endogenous Ras protein expression levels (Fig. 5a) do not necessarily indicate that Ras is "hyperactivated". Determination of the percentage of activated Ras (e.g., by G-LISA Ras activation assays) would be required to draw this conclusion.
Response: Very good suggestions. Accordingly, G-LISA Ras activation assay was performed to determine Ras activity in different cell lines before and after shRNA-mediated Ras knockdown (Fig. 4-5 and supplementary Fig. 2-3). It was found that the glioblastoma cells exhibited comparable high Ras activity as that of those cancer cell lines with high Ras mutation, and those cells with higher Ras activity such as MIA PaCa-2, SW480 showed much higher cellular association of DiI-labeled CaP-rHDL, compared with their control cells with lower Ras activity, BxPC-3 and Caco-2, respectively. A straightforward linear relation was achieved between the cellular uptake of CaP-rHDL and the intracellular Ras-GTP level (R 2 =0.8687) (Fig. 4c and Fig. 5e) Related data has been incorporated in the revised manuscript, please see Fig. 4-5, supplementary Fig. 2-3, Line 254-276 and Line 325-329.
2) The conclusion that CaP-rHDL nanoparticles "penetrated deep and distributed extensive(ly) in the C6 glioblastoma spheroids" (pg 16) seems inconsistent with the confocal images in Supplementary Fig. 2 Response: Good points. We agree with the reviewer that our previous conclusion that CaP-rHDL nanoparticles "penetrated deep and distributed extensive in the C6 glioblastoma spheroids" seems inconsistent with the confocal images in Supplementary Fig. 2. Indeed, when the tumor spheroids grow up to a certain extent, necrotic cores occur at the center of the spheroids. This could be the major reason why most of the DiI-labeled NP concentrated around the periphery of the spheroids as demonstrated previously. Reviewer 2 also points out that this 3D C6 glioblastoma spheroid is not properly designed. Therefore, we deleted the related data derived from C6 glioblastoma spheroids. In this revised manuscript, the high permeability of CaP-rHDL was reflected by its deep penetration and extensive distribution at the glioblastoma sites in the brain of the tumor-bearing animals ( Fig. 6 and Supplementary Fig. 5). In the case of patient-derived GICs spheroids, where less necrotic cores occur, as shown in Fig. 5b and   supplementary video 1-2, deep penetration of CaP-rHDL within the spheroids was witnessed.
3  Fig. 8e, we are shown one spheroid for each condition evidence that anti-ATF5-CaP-rHDL nanoparticles inhibit growth/viability of C6 spheroids. This conclusion needs to be supported by quantitative data such as average spheroid diameter or volume, with an appropriate and statistical analysis.

4) In
Response: Thanks for the suggestion. As discussed above, the 3D C6 glioblastoma spheroid model is not properly designed, we deleted the related data. Instead, the anti-glioma activity of ATF5-CaP-rHDL was evaluated by anti-proliferation and cell apoptosis assays with quantitative data provided and an appropriate and statistical analysis performed. As shown in Fig. 8 and supplementary Fig. 9, ATF5-CaP-rHDL exhibited much higher tumor-cellular toxicity and superiority in inducing cell apoptosis with the percentage of early apoptosis in cells treated with ATF5-CaP-rHDL (53.2%±10.5%) much higher than that of those treated with ATF5-CaP-LNC (11.1%±1.3%) and NC-CaP-rHDL (7.1%±0.15%). The related results were shown in Fig. 8, supplementary Fig. 9 and Line 459-471.
5) The argument that the increased apoptosis (Fig. 9) and the improved survival ( Fig.   10) of mice treated with anti-ATF5 nanoparticles is specifically related to suppression of ATF5 would be strengthened by IF or WB results demonstrating that ATF5 is suppressed in the tumor tissue.
Response: Thanks for the suggestion. Accordingly, we examined the level of ATF5 at the tumor site by western blot, finding that compared with saline, ATF5-CaP-rHDL suppressed ATF5 expression by 80% (Fig. 9b). Such description was also added in the revised manuscript, please see Fig. 9b and Line 501-503.

6) The paper contains numerous grammatical and typographical errors throughout.
Response: Thanks for the suggestion. The grammatical and typographical errors have been corrected. We hope this revised version could be much more fluent and neat.