Systematic pharmacological screens uncover novel pathways involved in cerebral cavernous malformations

Abstract Cerebral cavernous malformations (CCMs) are vascular lesions in the central nervous system causing strokes and seizures which currently can only be treated through neurosurgery. The disease arises through changes in the regulatory networks of endothelial cells that must be comprehensively understood to develop alternative, non‐invasive pharmacological therapies. Here, we present the results of several unbiased small‐molecule suppression screens in which we applied a total of 5,268 unique substances to CCM mutant worm, zebrafish, mouse, or human endothelial cells. We used a systems biology‐based target prediction tool to integrate the results with the whole‐transcriptome profile of zebrafish CCM2 mutants, revealing signaling pathways relevant to the disease and potential targets for small‐molecule‐based therapies. We found indirubin‐3‐monoxime to alleviate the lesion burden in murine preclinical models of CCM2 and CCM3 and suppress the loss‐of‐CCM phenotypes in human endothelial cells. Our multi‐organism‐based approach reveals new components of the CCM regulatory network and foreshadows novel small‐molecule‐based therapeutic applications for suppressing this devastating disease in patients.

Our combined multi-organismal screens revealed a number of important insights into the regulatory network involved in CCM. This has come with many surprises that we discuss in our manuscript. For instance, we find a striking conservation of the effects of drugs on worms and vertebrates, suggesting that some of these compounds may also have beneficial effects in translational therapeutic applications in the human disease. Among the compounds identified in both C. elegans and zebrafish were anti-hypertensive or anti-angiogenic drugs. Of note, worms have no circulatory system and hence, these drugs may act by affecting more fundamental functions of conserved molecular pathways -including roles that go beyond those that have been well-established in vertebrate physiology.
Similarly, neurotransmitter-related agonists or antagonists had beneficial properties in both worm and zebrafish. This class of drugs may function either autonomously on non-neural tissues or noncell autonomously on neural cells in their interaction with the vasculature. Several compounds that affect MAPK signaling exhibited suppressive effects on CCM mutant phenotypes in worms and zebrafish. This finding suggests some cellular role of MAPK signaling in both apoptosis in the worm and roles in the vertebrate vasculature; until now, apoptosis is not among the functional GO-BP terms for the ccm2 mutant transcriptome.
Metabolites of the retinoic acid synthesis pathway and drugs that affected the metabolic enzymes of this biochemical synthesis pathway showed a potent capacity to suppress the CCM phenotypes in zebrafish and C. elegans. This is an exciting finding given a recent study by the team of Andreas Trumpp who showed that Vitamin A-retinoic acid signaling regulates hematopoietic stem cell dormancy (Cabezas-Wallscheid et al., 2017). Further studies are required to assess the molecular role of this pathway upon the loss of CCM proteins.
In conclusion, we identify both novel compounds and the molecular pathways and biological processes that they affect. These are conserved and relevant to both the basic cellular and more complex cardiovascular defects found in CCM animal models relevant to the pathology in humans. Our systems biological approach significantly improved our understanding of how these molecular pathways are interconnected, which will greatly aid in designing effective targeting therapeutics. The inventory of relevant compounds now raises the possibility of benefits to be gained through combinatorial treatments based on drugs which target different molecular pathways involved in the disease. We hope that you will favorably consider our appeal. In addition to the other reviewer suggestions, we would also like to suggest Elisabetta Dejana as an excellent reviewer of our work. Thank you for your e-mail asking us to reconsider our decision on your manuscript. I have now carefully read your letter and article again and discussed it with my colleagues, including our chief editor.
I appreciate that you highlight once more the impact and novelty of the study. Reading your letter, I realized that indeed, some aspects of the work were not clear to us when we first evaluated your article: 1) For example, you mention that your screening strategy is original and novel compared to previous published screens, that these screens might have been biased and as such should not be considered as precedence. This type of information would be relevant for the community to add in the discussion section of the paper and backed up with scientific arguments. We are a general journal and our readership is broad, explanations and clarifications for a general audience are always welcome.
2) You also explain in the letter why the different pathways identified through the screen are particularly relevant in CCM. Our understanding now is that while the drug administered to mice was not identified through the screen, it targets some of the relevant pathways that were found relevant through the screen. Is that correct? If it is, I would strongly encourage you to rephrase this part of the work to make the reasons why you chose IR3mo more evident for the readers, or the study appears split with the screen on one side and a drug testing in mice on the other, with at best a weak link between the two. Moreover, IR3mo has no effect in C. elegans. It is unclear to us why that is. Could you please clarify in the text as well?
This said, after intense internal discussions about your paper, I would be happy to reconsider my decision and send your manuscript out for peer-review, provided you address these two points above to help with the understanding of the paper rationale.
I look forward to hearing from you soon. Should you decide to modify your paper accordingly and resubmit, please let us know.
1st Revision -authors' response 6 th April 2018 Thank you very much for reconsidering our manuscript. We have read your excellent comments and the two major points are well-made. We understand that they are critical for a better understanding of this study and we will be happy to improve upon these points. I would also like to briefly reply to your questions below: Best regards Salim Seyfried POINTS HIGHLIGHTED BY EDITOR.
1) For example, you mention that your screening strategy is original and novel compared to previous published screens, that these screens might have been biased and as such should not be considered as precedence. This type of information would be relevant for the community to add in the discussion section of the paper and backed up with scientific arguments. We are a general journal and our readership is broad, explanations and clarifications for a general audience are always welcome.
We will expand this discussion.
2) You also explain in the letter why the different pathways identified through the screen are particularly relevant in CCM. Our understanding now is that while the drug administered to mice was not identified through the screen, it targets some of the relevant pathways that were found relevant through the screen. Is that correct?
IR3mo has been identified during the screen (e.g. see tables S1,3). We realize that the reader may be surprised that it was not mentioned in the first part of the manuscript. We will strengthen the tie between the first and second part of the manuscript and better reveal the logic of analyzing IR3mo in preclinical trials (which is, among others and as discussed in the first paragraph of the results chapter on IR3mo, that it targets some of the relevant pathways that were found relevant through the screen).
If it is, I would strongly encourage you to rephrase this part of the work to make the reasons why you chose IR3mo more evident for the readers, or the study appears split with the screen on one side and a drug testing in mice on the other, with at best a weak link between the two. Moreover, IR3mo has no effect in C. elegans. It is unclear to us why that is. Could you please clarify in the text as well?
We can only speculate why IR3mo is not working in worm. For instance, the worm cuticle may not be permeable for the drug or the concentration used in the trials may not have been sufficient. Another possibility is that IR3mo targeted some of the pathways that are specific for zebrafish (SRC, MST1R, VEGFR2). However, we decided that IR3mo was still an excellent candidate drug based on the logic as outlined above. Thank you for the submission of your manuscript to EMBO Molecular Medicine. We have now heard back from the three referees whom we asked to evaluate your manuscript.
As you will see from the reports below, while referees 2 and 3 are overall positive and support, in principle, publication of the article in EMBO Molecular Medicine (pending appropriate revisions), referee 1 questions the rationale for investigating IR3mo (as we did and previously discussed with you). Therefore, a more thorough discussion on the choice of IR3mo, as well as addressing the reviewers' concerns in full will be necessary for further considering the manuscript in our journal, with the exception of additional validation in other in vivo CCM models asked by referee 2 (referee 2, comments 2 and 4). EMBO Molecular Medicine encourages a single round of revision only and therefore, acceptance or rejection of the manuscript will depend on the completeness of your responses included in the next, final version of the manuscript. Referee #1 (Remarks for Author): The manuscript attempts to find new small molecule compounds that can therapeutically alleviate the condition, cerebral cavernous malformations (CCM). They use two established CCM model systems in zebrafish, worm, and in huvec knockdown cells, to probe a panel of compounds for those that have effects in all three systems. Surprisingly, although they found 5 compounds that seemed to have effects in each system, they chose a kinase inhibitor that did not work in the worm for further studies. Using this inhibitor, in wt and siKRIT1 HUVECs they report reduced ERK5 phosphorylation, and in two mouse models of CCM disease they report rescue of lesion burdern. However, they do not understand the mechanism of ERK5 phosphorylation reduction, and I think that understanding the mechanism of IR3mo inhibition of ERK5 phosphorylation is required for this study to be clinically informative. Some specific comments are noted below.
1. Their rationale to follow up Indirubin-3-monoxime, which rescued in only two of the models, and not one of the 5 compounds that rescued in all 3 is unclear. For example, one of these (ENMD-2076, which should be listed in the main text), is an kinase inhibitor (ENMD-2076) that also inhibits Aurora, KDR, FLT4, FLT3 and SRC. There seems to be much overlap with targets of indirubin-3monoxime. Also, the protein interaction network (p12 referring to Table S8) discusses the importance of proteins involved in angiogenesis signaling, including FLT1, KDR and FLT4 which they didn't follow up experimentally (are the effects they see due to IR3mo inhibition of KDR or SRC tyrosine kinases, for example, and not MEK5 serine/threonine kinase as shown in Fig 4h, does IR3mo inhibit ROCK?). This leads one to wonder what the best candidate compound from their screens was. Particularly as this study is trying to move towards a clinically-useful compound.
2. Figure 3CDE. The Western blots should be shown in the figure, and a scatter plot is the preferred method of data representation. The individual data points should be shown for each replicate. SD is preferred over SEM; Figure  3. The schematic in figure 4h is misleading: they do not know the mechanism of IR3mo modulation of ERK5 inhibition.
4. P9. Lines 3-4. They should include the full list of compounds shown in Table S2.

They should include line numbers in their manuscript.
Referee #2 (Remarks for Author): Comments to the authors on the article titled, "Systematic Pharmacological suppression uncovers novel molecular pathways involved in cerebral cavernous malformations" This paper nicely reports the application of small-molecule suppression screens in different CCMmutant models which revealed signaling pathways relevant to CCM diseases. The authors also identified indirubin-3-monoxime IR3mo compound as a prime candidate that alleviated the CCMlesion burden in murine models and HUVECs. Altogether, the technicality of this study is robust and therefore, has the possibility to open up new platforms for future studies affecting several signaling pathways and biological processes relevant to CCM pathology.
However, there are a few concerns which need to be addressed: Comment 1. Figure 3. Page 22-23: The authors applied STRING network clustering to visualize protein interaction networks related to CCM disease. This revealed several protein clusters both in the zebrafish and C. elegans model. However, implementing one network clustering is insufficient to determine the efficacy of the results obtained. In order to avoid biased results and achieve consistency, it would be relevant to apply multiple protein-protein interaction web based programs available online (such as: BioGRID, IntAct etc).
Comment 2. Figure 4a. The author used ccm2 mutant as an example to show the effects of IR3mo on zebrafish embryos. It will worth to know whether IR3mo will have any changes on ccm1 and ccm3 mutants on zebrafish heart phenotype. Comment 3. Figure 4b. Similarly it will be relevant to also see the phenotype of IR3mo on CCM2 or PDCD10 siRNA silenced HUVECs. Comment 4. Figure 4f. The authors demonstrated the effect of treatment of IR3mo on iCCM2 and iCCM3 mice which alleviated the lesion burden. One might also try to look if there is any effect on lesion burden of IR3mo on iKRIT1 mice model.
Referee #3 (Comments on Novelty/Model System for Author): The use of multiple animal models for the initial screen, and then a cell culture model for validation of targets makes this a unique study and a very robust study design. The final mouse model, used for only one compound/drug, is important as well as the mouse model is the most relevant for future clinical trials. All in all a very nice and robust study with high scientific and potential medical impact, per my comments below.
Referee #3 (Remarks for Author): This is a real tour de force in terms of effort and organization, and the study design is quite robust. The use of two different animal models for the drug screen makes this study unique, and the confirmatory cell culture model before moving to the more difficult but clinically relevant mouse model is another nice addition.
I am not an expert regarding the data mining and bioinformatic analyses. Since CCM pathobiology is already extensively studied, it was reassuring that quite a few drugs, drug targets, or pathways/biological processes were identified through the analysis that have already been suggested by other published studies. However, there seem to be so many different and apparently distinct pathways and biological processes that have resulted from the analyses that I would wonder how many are truly relevant to CCM pathobiology? When the analysis points in so many different directions, I have less confidence the relevance of each. However, one advantage to including all the bioinformatic analyses in this paper is that now there are new hypotheses to test concerning CCM pathobiology. So if I consider all this data analyses as hypothesis generating, I am more assured of its importance.
A final point is that IR3mo appears to be a very strong candidate for additional mechanistic studies, but also additional studies in animals that might lead to clinical trials. So the proof of the value of this study is that a new, exciting compound has been identified that might ultimately provide some relief to CCM patients. Congratulations to the investigative team.

Referee #1 (Remarks for Author):
The manuscript attempts to find new small molecule compounds that can therapeutically alleviate the condition, cerebral cavernous malformations (CCM). They use two established CCM model systems in zebrafish, worm, and in huvec knockdown cells, to probe a panel of compounds for those that have effects in all three systems. Surprisingly, although they found 5 compounds that seemed to have effects in each system, they chose a kinase inhibitor that did not work in the worm for further studies. Using this inhibitor, in wt and siKRIT1 HUVECs they report reduced ERK5 phosphorylation, and in two mouse models of CCM disease they report rescue of lesion burden. However, they do not understand the mechanism of ERK5 phosphorylation reduction, and I think that understanding the mechanism of IR3mo inhibition of ERK5 phosphorylation is required for this study to be clinically informative. Some specific comments are noted below.
1. Their rationale to follow up Indirubin-3-monoxime, which rescued in only two of the models, and not one of the 5 compounds that rescued in all 3 is unclear. For example, one of these (ENMD-2076, which should be listed in the main text), is an kinase inhibitor (ENMD-2076) that also inhibits Aurora, KDR, FLT4, FLT3 and SRC. There seems to be much overlap with targets of indirubin-3monoxime. Also, the protein interaction network (p12 referring to Table S8) discusses the importance of proteins involved in angiogenesis signaling, including FLT1, KDR and FLT4 which they didn't follow up experimentally (are the effects they see due to IR3mo inhibition of KDR or SRC tyrosine kinases, for example, and not MEK5 serine/threonine kinase as shown in Fig 4h, does IR3mo inhibit ROCK?). This leads one to wonder what the best candidate compound from their screens was. Particularly as this study is trying to move towards a clinically-useful compound.
Upon completion of the primary screen in worms, zebrafish, and HUVECs, our consortium was facing the difficult decision to select not more than one or two compounds for preclinical trials in the murine ccm models. Given this limitation, we defined a number of criteria that served as a guideline for this decision. As indicated in our manuscript, one key aspect was the possibility of applying the drug safely for long-term usage without causing major side effects in patients. Furthermore, we ranked the positive effects of a drug in the zebrafish ccm model higher compared to the C.elegans ccm model since the zebrafish is a vertebrate with a cardiovascular system. As discussed in the text (lines 296-306), IR3mo is an FDA-approved drug with low toxicity derived from traditional Chinese medicine that has been widely used to treat leukemia and other chronic diseases (Eisenbrand et al, 2004; Williams et al, 2011) and interferes with several molecular pathways relevant to CCM. In comparison, treatment with other strong anti-angiogenic or antiproliferative drugs (such as the multi-kinase inhibitor ENMD-2076) is associated with more severe side effects. Also, as shown in Table EV10 (former suppl. Table S10), IR3mo has at least 10 known protein targets as predicted by DePick and other known targets including AHR, CDK5, CDK5R1, which are described DrugBank interaction partners according to PubChem. Potentially, IR3mo may affect a unique combination of targets that mediate the rescue effect.
As additional information, throughout the revisions phase, the Toronto teams of Brent Derry and Peter Roy applied different IR3mo treatment plans on C. elegans ccm models which did not yield a rescue effect. We conclude that IR3mo is highly effective in the context of vertebrate ccm models and human CCM-depleted endothelial cells but not in C.elegans.
2. Figure 3CDE. The Western blots should be shown in the figure, and a scatter plot is the preferred method of data representation. The individual data points should be shown for each replicate. SD is preferred over SEM; Figure 5 of PMID 25204545 (Motulsky, 2014) illustrates why. Likewise Fig 4g should be shown as a scatter plot using SD not SEM.
Within the revised manuscript, we have now improved old Figure 4 which has been divided into new Figure 4 (zebrafish and HUVEC data) and a new Figure 5 (mouse data and model): the Western blots (previous suppl. Figure S6) have been moved into new Figure 4 (Fig 4I). The representation of the quantifications shown in Figure 4C-E has been changed into scatter plots with error bars representing the SD (Fig 4J-M). Mouse data from previous Figure  3. The schematic in figure 4h is misleading: they do not know the mechanism of IR3mo modulation of ERK5 inhibition.
This point has been addressed by modifying the model figure and the text of the figure legend to highlight the finding that IR3mo affects pERK5 protein levels and KLF2 expression levels (which is a strong molecular readout of the CCM pathway). As indicated in that figure, two arrowheads are used to indicate these effects of IR3mo and we do not mean to imply that KLF2 mRNA or ERK5 phosphorylation are directly affected by IR3mo. This is explicitly described within the revised figure legend. Table S2.

P9. Lines 3-They should include the full list of compounds shown in
We have changed the text accordingly and list the other compounds from Table EV2 that were missing in the main text (line 185-186). Similarly, we have also named the 5 th compound that showed some degree of rescue in all three CCM models and that was missing from the main text (Table EV3) (lines 200-201).

They should include line numbers in their manuscript.
We have included line numbers within the revised manuscript.

Referee #2 (Remarks for Author):
Comments to the authors on the article titled, "Systematic Pharmacological suppression uncovers novel molecular pathways involved in cerebral cavernous malformations" This paper nicely reports the application of small-molecule suppression screens in different CCMmutant models which revealed signaling pathways relevant to CCM diseases. The authors also identified indirubin-3-monoxime IR3mo compound as a prime candidate that alleviated the CCMlesion burden in murine models and HUVECs. Altogether, the technicality of this study is robust and therefore, has the possibility to open up new platforms for future studies affecting several signaling pathways and biological processes relevant to CCM pathology.
However, there are a few concerns which need to be addressed: Comment 1. Figure 3. Page 22-23: The authors applied STRING network clustering to visualize protein interaction networks related to CCM disease. This revealed several protein clusters both in the zebrafish and C. elegans model. However, implementing one network clustering is insufficient to determine the efficacy of the results obtained. In order to avoid biased results and achieve consistency, it would be relevant to apply multiple protein-protein interaction web based programs available online (such as: BioGRID, IntAct etc).
The reviewer suggests to apply protein-protein interaction web based programs available online and suggests BioGRID and IntAct as examples of such programs. Unfortunately, BioGRID and IntAct are repositories of protein-protein interaction data but they do not provide online web programs for Network visualization. These (and other) resources storing protein-protein interaction information are instead commonly used by several web tools (programs) dedicated to protein Networks including STRING.
We chose STRING to visualize our Networks for its comprehensiveness -it uses the widest breadth of input sources, including automated text-mining and computational predictions -and more importantly, for its quality control -each interaction is annotated with benchmarked confidence scores-. The scores indicate the estimated likelihood that a given interaction is biologically meaningful, specific, and reproducible, given the supporting evidence (e. g experiments) (PMID: 27924014). We chose to display high confidence Networks by selecting only interactions highly supported (e.g. by more than one evidence) (high confident cut-off :0.7). With the choice of STRING, we also reduce and compensate the individual biases in information that each resource might have. While our Networks are not free of potential general bias in the currently existing protein-protein information data (e.g. certain proteins are better characterized than others and therefore, more interactions for them are known), they provide a more comprehensive and consistent picture of protein CCM Networks than those constructed with only one source of protein-protein interaction information (e.g. only BioGRID or IntAct).
Comment 2. Figure 4a. The author used ccm2 mutant as an example to show the effects of IR3mo on zebrafish embryos. It will worth to know whether IR3mo will have any changes on ccm1 and ccm3 mutants on zebrafish heart phenotype.
Within the revised version of the manuscript, we have now included an additional figure which shows the rescue effect of IR3mo treatment on the krit1 ty2019c mutant cardiac phenotype (Fig EV5A-D). This experiment supports our findings in the murine preclinical model and show that IR3mo has beneficial effects in different vertebrate ccm models.
Currently, mutants for the two redundant ccm3a/pdcd10a and ccm3b/pdcd10b genes of zebrafish have not yet been described and there have been some controversial reports of ccm3a/b knockdown phenotypes. Hence, we did not attempt to rescue ccm3 knockdown phenotypes in zebrafish using IR3mo.
Comment 3. Figure 4b. Similarly it will be relevant to also see the phenotype of IR3mo on CCM2 or PDCD10 siRNA silenced HUVECs.
Within the revised version of the manuscript, we show the effective rescue by IR3mo of KRIT1-or CCM3-silenced HUVECs, whose phenotype is identical to CCM2-deficient HUVECs (Fig EV5E-J).
Comment 4. Figure 4f. The authors demonstrated the effect of treatment of IR3mo on iCCM2 and iCCM3 mice which alleviated the lesion burden. One might also try to look if there is any effect on lesion burden of IR3mo on iKRIT1 mice model.
We are grateful to the editor for acknowledging that testing IR3mo in a third mouse model would be beyond the scope of this work. Taken together, IR3mo has beneficial effects in different vertebrate ccm models.

Referee #3:
The use of multiple animal models for the initial screen, and then a cell culture model for validation of targets makes this a unique study and a very robust study design. The final mouse model, used for only one compound/drug, is important as well as the mouse model is the most relevant for future clinical trials. All in all a very nice and robust study with high scientific and potential medical impact, per my comments below.

Referee #3 (Remarks for Author):
This is a real tour de force in terms of effort and organization, and the study design is quite robust. The use of two different animal models for the drug screen makes this study unique, and the confirmatory cell culture model before moving to the more difficult but clinically relevant mouse model is another nice addition.
I am not an expert regarding the data mining and bioinformatic analyses. Since CCM pathobiology is already extensively studied, it was reassuring that quite a few drugs, drug targets, or pathways/biological processes were identified through the analysis that have already been suggested by other published studies. However, there seem to be so many different and apparently distinct pathways and biological processes that have resulted from the analyses that I would wonder how many are truly relevant to CCM pathobiology? When the analysis points in so many different directions, I have less confidence the relevance of each. However, one advantage to including all the bioinformatic analyses in this paper is that now there are new hypotheses to test concerning CCM pathobiology. So if I consider all this data analyses as hypothesis generating, I am more assured of its importance.
The reviewer points out correctly that a larger number of pathways are misregulated upon loss-of CCM. Hence, some pathways may be changed secondarily after those primarily affected. In future investigations into the molecular control of CCM formation, we will benefit from these pathway lists that provide candidates. Also, in patients, cavernomas are detected only after first clinical manifestations and by that time secondary pathways may be active. Therefore, it is equally relevant to know which are the initial forces driving CCM formation, as well as which are the secondary pathways that are misregulated at later stages of CCM development.
A final point is that IR3mo appears to be a very strong candidate for additional mechanistic studies, but also additional studies in animals that might lead to clinical trials. So the proof of the value of this study is that a new, exciting compound has been identified that might ultimately provide some relief to CCM patients. Congratulations to the investigative team.
We agree with the reviewer that the discovery of IR3mo as an active compound will be beneficial in two ways: first, to further investigate the pathobiology of the CCM disease, and second, as a promising compound (potentially in combination with other compounds) with low toxicity for patients. Currently, we are in the process of establishing a chronic late-onset mouse model of CCM, which will allow us to test the regression of lesions rather than the prevention of their appearance. It will be very interesting to assess whether IR3mo can also be beneficial in this paradigm, which more closely mimicks the human disease. Thank you for the submission of your revised manuscript to EMBO Molecular Medicine, and my apologies for the unusually long review process. We have now received the enclosed reports from the referees. As you will see the reviewers are now supportive, and I am pleased to inform you that we will be able to formally accept your manuscript after the following final editorial amendments: Referee #1 (Remarks for Author): The manuscript is certainly stronger now, and I think it appropriate for publication.
A small typo: line 433 -models is still -> models are still Referee #3 (Remarks for Author): The data shown in figures should satisfy the following conditions: the data were obtained and processed according to the field's best practice and are presented to reflect the results of the experiments in an accurate and unbiased manner. figure panels include only data points, measurements or observations that can be compared to each other in a scientifically meaningful way. graphs include clearly labeled error bars for independent experiments and sample sizes. Unless justified, error bars should not be shown for technical replicates. if n< 5, the individual data points from each experiment should be plotted and any statistical test employed should be justified Source Data should be included to report the data underlying graphs. Please follow the guidelines set out in the author ship guidelines on Data Presentation.

Captions
Each figure caption should contain the following information, for each panel where they are relevant: a specification of the experimental system investigated (eg cell line, species name). the assay(s) and method(s) used to carry out the reported observations and measurements an explicit mention of the biological and chemical entity(ies) that are being measured. an explicit mention of the biological and chemical entity(ies) that are altered/varied/perturbed in a controlled manner.
the exact sample size (n) for each experimental group/condition, given as a number, not a range; a description of the sample collection allowing the reader to understand whether the samples represent technical or biological replicates (including how many animals, litters, cultures, etc.). a statement of how many times the experiment shown was independently replicated in the laboratory. definitions of statistical methods and measures: Any descriptions too long for the figure legend should be included in the methods section and/or with the source data.
In the pink boxes below, please ensure that the answers to the following questions are reported in the manuscript itself. Every question should be answered. If the question is not relevant to your research, please write NA (non applicable). We encourage you to include a specific subsection in the methods section for statistics, reagents, animal models and human subjects.

B--Statistics and general methods
Please fill out these boxes ê (Do not worry if you cannot see all your text once you press return) 2) HUVECS: No explicit power analysis was used. Experiments were performed on populations of HUVEC cells four times as generally accepted in the field of biology. 3) WORM: In the C. elegans screens, 20 synchronized L1 larval stage worms were added to each well of 96 well plates containing either a small molecule or DMSO. After incubation, the number of living worms in each well were counted to asses the degree of pharmacological suppression of the kri--1(ok1251); ccm--3 (RNAi) semi--lethal phenotype. Use of 20 L1 worms per well is standard in our screens as 20 healthy L1 larvae will develop to adulthood and produce hundreds of progeny per well within the incubation period. This allows us to easily detect suppression of the CCM semi--lethal phenotype in the presence of a small molecule using our counts. 4) FISH: For the zebrafish screen experiments, we chose the 24--well format with 16--20 embryos per well in order to have an optimal compromise between using as few zebrafish embryos as possible and nevertheless having a sample size large enough. We screened for compounds that reverted the ccm2 mutant phenotype to wild--type, but treated the embryos from 16hpf onwards, at which timepoint the mutants are undistingushable from the wild--type siblings. Therefore, for each well, we assumed that on average 75% of the embryos are wild--type, resulting in a maximum possibility of 0.75^16 = ca. 1% false--positive compounds. In addition, the technician was screening thousands of distinct small molecules, the identies of which were simply coded by plate and well number. Hence, in practice, the technician was blinded to the identities of the molecules they were screening. 4) FISH: For the zebrafish screen experiments, we obtained 24--well plates labelled with a code and containing the compounds from the group of Jens von Kries, but did not get the information about the contents of the wells until the analyses had been conducted. Therefore, the analyses were performed blindly, except for the control wells that were always well C6 and D6 and contained no compound. For the zebrafish IR3mo experiments, blinding was not necessary.
2) HUVECS: N/A 3) WORM: See above. 4) FISH: Blinding was done for the zebrafish screen experiments, but not for the IR3mo experiments.
2) HUVECS: Yes 3) WORM: In the C. elegans screens, worms were added to each well of 96 well plates containing either a small molecule or DMSO. After incubation, the number of living worms in each well were counted. The degree of rescue of the kri--1(ok1251); ccm--3 (RNAi) semi--lethal phenotype induced by each compound was assessed and broadly categorized based on a comparison of the number of living worms in wells exposed to the small molecule versus the number of living worms in DMSO controls. No statistical tests were applied. 4) FISH: yes??????