Molecular signature of brain arteriovenous malformation hemorrhage: a systematic review.

BACKGROUND
The mechanisms of brain arteriovenous malformation (bAVM) development, formation and progress are still poorly understood. By gaining more knowledge about the molecular signature of bAVM in relation to hemorrhage, we might be able to find biomarkers associated with this serious complication, which can function as a goal for further research and can be a potential target for gene therapy.


AIMS
To provide a comprehensive overview of the molecular signature of bAVM-related hemorrhage We conducted a systematic review, following PRISMA guidelines, of articles published in Embase, Medline, Cochrane central, Scopus, and Chinese databases (CNKI, Wanfang).


SUMMARYOF REVIEW
Our search identified 3,944 articles, of which 3,108 remained after removal of duplicates. After title, abstract and full text screening 31 articles were included for analysis. The results show an overview of molecular characteristics. Several genetic polymorphisms are identified which increase the risk of bAVM rupture by increasing the expression of certain inflammatory cytokines (IL-6, IL-17A, IL-1β and TNF-α), NOTCH pathways, MMP-9 and VEGFA.


CONCLUSIONS
Several molecular factors are associated with the risk of bAVM-related hemorrhage These factors are associated with increased inflammation on the cellular level and changes in the endothelium leading to instability of the vessel wall. Further investigation of these biomarkers regarding hemorrhage rates, together with their relationship with noninvasive diagnostic methods, should be a goal of future studies to improve the patient specific risk estimation and future treatment options.


Zurich, 24 th September, 2021
Dear Section editor, dear reviewers, We thank you for critically reviewing our manuscript and giving us the opportunity to improve the manuscript to make it suitable for publication in World Neurosurgery. Below are the point-by-point responses to the comments. With these changes, we hope that the manuscript improved in such a way, that it can be accepted in your journal. For your information, the references of the tables failed in the previous version and were added again to the current version.
Kind regards,

Menno Germans
Reviewer #1: Thank you for your supporting comments regarding the importance of this topic.
 I suggest describing more extensively the data of the review in the section "Results", eventually by establishing a statistical significancy for a least several biomolecular markers. We decided to move the text from paragraph 2 in the discussion to the results section. Now, the results section indeed describes the significant results in more detail.
 A further table summarizing the rate of positivity of the single molecular makers in the overall reviewed studies should better focus the results and define the most important molecular signatures at risk of hemorrhage.
To prevent double presentation of results, we decided to present the significant results in bold in table 2. In this way, both an overview of all results and only the significant results are presented in one table.
Reviewer #3: Thank you for your overall positive feedback regarding the quality of the study. We agree that this confuses the reader. We added the complete list to the abstract.  2. In Table 2 We have overseen this typo in the reviewing of the manuscript. We adjusted your comment accordingly.

Response to Reviewers
Reviewer #4:  Please clarify how the studies are able to distinguish between changes leading to hemorrhage vs changes that are caused by the hemorrhage. We agree that this is a major issue in our analysis. In our previous version, we mentioned that this could have led to reverse causality ("…the inclusion of bAVM with different interval between time of hemorrhage and resection might have induced reverse causality, defined as changes to the molecular signature due to the pretreatment or hemorrhage itself.").
To emphasize this issue more we adapted the text: "…or the inclusion of bAVM with different intervals between time of hemorrhage and resection probably lead to some bias in the results. In what extent this reverse causality, defined as changes to the molecular signature due to the pretreatment or hemorrhage itself, caused bias cannot be investigated with the current results."  Please provide the search string and when the databases were last accessed. These data are indeed very important to mention, for transparency of the results and possible newly published studies.
In the first sentence of the 'Information sources' paragraph we mentioned the data of last search ("…from start onwards to Jan 31th 2020"). Additionally, we adapted the searches according to the database with a clinical librarian, which is mentioned at the end of the above-mentioned paragraph ("A modified search strategy was designed according to the specific requirements of each database (see Supplementary In this study we decided to include the molecular signature as a wide definition, as mentioned in the first paragraph of 'Methods': "…is to identify molecular signatures, defined as genetic profiles, protein expression status or cellular signaling pathways…".
 …e.g. "Activated macrophage" is in my opinion not specific enough. Afterall, a molecular signature is to be described in this review, so in this example, what are the actual markers used to identify activated macrophages. Please use more specific items in this table.
We agree on your comment, as the review is about molecular signatures. According to the articles we added the relevant data.
 Moreover, I would recommend an additional column describing the method used in the original papers to characterize the samples, like bulk/ssRNAseq, FACS, ELISA etc.
To clarify the methods which were used in the studies, we added one extra column in table 1 describing the techniques.
 Additionally, either in this table or elsewhere, the authors should specify the timepoint of sample collection in relation to hemorrhage since this can significantly influence the expression of many markers.
We agree that the timepoint of sample collection in relation to hemorrhage is an important issue because it could lead to heterogeneity in the analysis. We reviewed this issue already and came to the conclusion that only a very small number of articles published these data. We therefore decided not to included these data. In our limitation section we addressed your comment, as mentioned in the first answer to your comments. Thank you for this comment, as the review should be up-to-date when published. We performed our final search in January 2021 and included all the relevant articles. We also included some relevant articles in our discussion, which were not selected for the review. In the discussion, we refer to the article you mentioned above: 'All the above-mentioned factors are related with acute inflammation or maturation of the endothelial cell. This finding is supported by previous studies which have shown that an inflammatory reaction in the vascular wall plays an important role in the rupture of a bAVM. 16 As such, the increased risk of bAVM hemorrhage seems to be a result of increased inflammation on cellular level and changes in the endothelium leading to instability of the vessel wall.' The mechanisms of bAVM development, formation and progress are still poorly understood.

Abbreviations
Some researchers think that the formation and progress of bAVMs may be linked not only to hemodynamic factors, but also genetic factors and somatic mutations. 4,5 Recently, some studies showed that certain angioarchitectural factors, such as venous congestion and progressing ectasia, are associated with an increased risk for hemorrhage. 6,7 Other studies emphasized the influence of genes and proteins on clinical symptoms and angioarchitectural changes. 8,9 The molecular signature of the bAVM, including the genetic profile, protein expression status and cellular signaling pathways, probably dictates the clinical course through their effect on the angioarchitecture or destabilization of the nidus of the bAVM. 10,11 By gaining more knowledge about the molecular signature of bAVM in relation to hemorrhage, we might be able to find biomarkers associated with this serious complication. These biomarkers can function as a goal for further research and can be a potential target for gene therapy.
In order to identify potential biomarkers for bAVM hemorrhage, we systematically reviewed the literature to create an overview of the known molecular factors which are associated with bAVM hemorrhage.  between the two reviewers was resolved by discussion to achieve a consensus.

Assessment of risk of bias
The two main reviewers independently assessed the methodological quality of the included studies using the Newcastle Ottawa Scale (NOS) 12 , modified for case-control studies. Since there was no need for follow-up, we did not assess the risk about response. The risk of bias was assessed as low, moderate or high when the total score was ≥7, 5-6 or <5, respectively. Any disagreements concerning the bias assessment was resolved by consensus.

Presentation of results
An overview of the article selection process is presented in a flow diagram according to the PRISMA guidelines. 13 The results are presented in tables where the table on molecular factors is stratified according to type of sample (bAVM tissue or blood samples).

Data synthesis and analysis
Because we anticipated major heterogeneity among our search, such as the same biological effect of different genes or proteins or different types of hemorrhage, we provided a narrative synthesis of the data. Odds ratios (OR) with their 95% confidence intervals were calculated to assess a potential association between molecular changes and the risk of intracranial hemorrhage, when applicable. We also evaluated whether data were suitable for a metaanalysis by searching for identical molecular signatures which were reported in at least two studies.

Data Availability Statement
All data regarding this study are published in this article. The source data are saved in a secured environment in Covidence (https://www.covidence.org/reviews/38842). The first two authors Germans 6 of this article have access to the data and these will be shared by request.

Literature search and selection
The database search identified 3,108 articles. After removal of duplicates, 2,746 articles remained for title and abstract screening and finally the full text of 42 articles was analyzed.
Reasons to exclude articles after full-text screening were: no association was examined between molecular signatures and clinical presentation (n=6), the articles were reviews (n=2), the data on bAVM could not be extracted (n=2), and once the same data was published as in another abstract. The search by hand in the reference lists revealed one additional relevant article for the analysis. The search resulted in a total of 31 articles to be included. Figure 1 presents a flowchart of the review process.  The molecular factors with a significant increase in hemorrhage risk were found in the following genetic polymorphisms: TNF-α rs361525_GG, IL-6 rs1800795_GG, NOTCH4 rs443198_TT, NOTCH4 rs915895_AA, VEGFA rs1547651_AT, IL-17A rs2275913_GG, IL-1β rs1143627_CC and IL-1β rs16944_TT. Those associated with a decreased risk of hemorrhage were MMP-9 rs9509_CC, EPHB4 rs314313_CC and EPHB4 rs314308_TT. Other molecular factors which were found to be associated with bAVM-related hemorrhage were protein upregulation, cell activation and activated pathways. We were not able to calculate OR for these parameters, so the impact of the change in these molecular factors still needs further clarification. A summary of the molecular signatures and their association with hemorrhage is shown in table 2.

Meta-analysis
Five single-nucleotide polymorphisms (IL6 rs1800796, IL6 rs1800795, TNFα rs1800629, TNFα rs361525 and APOE e2+/e2-) were studied in two articles or more. Unfortunately, these studies used (partially) the same databases and could therefore not be included for a metaanalysis.

Risk of bias in included studies
The risk of bias assessment per subdomain and the total risk of bias assessment are presented in supplemmentart table. The majority of the studies scored a low risk of bias, whereas one study had a moderate risk and one study a high risk.

Discussion
In this systematic review we present molecular factors which are associated with bAVMassociated hemorrhage. Most of the molecular factors showed an association with an increase in hemorrhage risk. Because of use of -at least partially-identical data in the studies investigating the same molecular factor, we were not able to perform a meta-analysis.
Our search resulted in a selection of 31 articles, published between 2004 and 2018. The molecular factors with a significant increase in hemorrhage risk were found in the following genetic polymorphisms: TNF-α rs361525_GG, IL-6 rs1800795_GG, NOTCH4 rs443198_TT, NOTCH4 rs915895_AA, VEGFA rs1547651_AT, IL-17A rs2275913_GG, IL-1β rs1143627_CC and IL-1β rs16944_TT. Those associated with a decreased risk of hemorrhage were MMP-9 rs9509_CC, EPHB4 rs314313_CC and EPHB4 rs314308_TT. Other molecular factors which were found to be associated with bAVM-related hemorrhage were protein upregulation, cell activation and activated pathways. We were not able to calculate OR for these parameters, so the impact of the change in these molecular factors still needs further clarification. With this study, we were able to create an extensive overview of the molecular factors which currently have been investigated to be associated with bAVM-associated hemorrhage. These results can be used for further hypothesis-driven molecular analyses.
Interleukin 6 (IL-6) acts as a pro-inflammatory cytokine and is an important mediator of the acute phase of the inflammatory response. It indirectly stimulates leukocyte recruitment, endothelial activation, and vascular smooth muscle cell (SMC) proliferation. 14-16 This Germans 9 proliferation of SMC has been shown to be due to increased MMP-9 and VEGFA release, which was induced by IL-6. 17 Both MMP-9 and VEGFA on its own increase the vascular permeability and vasodilation of the bAVM nidus by breakdown of extracellular matrix and vascular remodeling, leading to susceptibility to hemorrhage. 15,18,19 Additionally, researchers have provided evidence that MMP-9 could be a marker for destabilization of the bAVM nidus and risk for rupture by showing variation of MMP-9 levels after bAVM treatment. 20 The importance of VEGFA in modulating angiogenesis and endothelial cell migration, proliferation and survival, was emphasized by a study of Kim et al. 21 who showed variation in plasma VEGFA levels in bAVM patients before treatment, 24 hours after resection, and 30 days after resection.
The NOTCH-pathway is thought to be involved in the formation of bAVM where it is involved in the endothelial cell selection, vascular smooth muscle cell differentiation and proliferation, arterial maturation, as well as to interact with VEGF. [22][23][24] NOTCH signaling is important to the pathogenesis of vascular disease, and lies both upstream and downstream of critical vascular and lymphatic growth signals such as VEGFA, EphB4/ephrinB2 and adrenomedullin/CLR signaling. 25,26 The NOTCH signaling pathway consists of four receptors (NOTCH 1-4) and several studies have identified NOTCH 1, 3 and 4 overexpression in bAVMs in humans. 27,28 NOTCH pathways are also found to be associated with hemorrhagic presentations of bAVM as summarized in our review. 29,30 However, studies about the effect of NOTCH signaling on bAVM hemorrhagic presentation are still limited and the exact mechanism is still unknown. 31 All the above-mentioned factors are related with acute inflammation or maturation of the endothelial cell. This finding is supported by previous studies which have shown that an Germans 10 inflammatory reaction in the vascular wall plays an important role in the rupture of a bAVM. 16 As such, the increased risk of bAVM hemorrhage seems to be a result of increased inflammation on cellular level and changes in the endothelium leading to instability of the vessel wall.
It remains unclear how the changes in the expression of the molecular factors are induced, leading to upregulation of cytokines with their exciting effects. Taking the angiographic changes which are associated with elevated hemorrhage risk into account, 32, 33 the flow through the bAVM could induce shear stress in certain bAVM vessels, increasing the expression of the above-mentioned molecular factors. On the other hand, an increase in expression profiles in the bAVM could also be caused by the environment, i.e. the brain surrounding the bAVM. This theory can be supported by the study of Fierstra et al. 34 , which showed increased cerebrovascular reactivity in the brain surrounding the bAVM in epilepsy patients. Finally, the molecular signature could also be influenced by previous partial treatment or the bAVM hemorrhage itself, which led to the indication for resection and collection of the samples. This reverse causality can only be overcome by a collection of blood samples in patients without bAVM hemorrhage and follow the cohort for a long period of time.
A review summarizing genetic polymorphisms which were associated with bAVM formation and hemorrhage has been performed before and included studies from 2004 to 2015. 35 These authors performed a meta-analysis with the same studies as mentioned in our review. As mentioned before, the data in the meta-analysis are originating from (at least partial) identical databases and due to on the high risk of selection bias by the use of repetitive data, we decided not to perform a meta-analysis on these data. Our review has Germans 11 included original articles identified using multiple search engines over a longer time period (up to Jan 31 th 2020) and included studies written in Chinese, comparably as well comparedwith to the study ofperformed by WSGao et al. 35 In addition, our search criteria included also molecular signatures associated with hemorrhage and was not restricted to solely genetic polymorphisms, as is the case in the other review. Calculating odds ratios, we were able to identify the strength and significance of molecular factors (restricted to singlenucleotide polymorphisms) contributing to the prediction of hemorrhage of bAVM. Although

Conflicts of interest/Competing interests:
The authors declare that there is no conflict of interest.
Availability of data and material (data transparency) All data regarding this study are published in this article. The source data are saved in a secured environment in Covidence (https://www.covidence.org/reviews/38842). The first two authors of this article have access to the data and these will be shared by request.

Introduction
A brain arteriovenous malformation (bAVM) is a developmental vascular disorder of the brain that exists of a fistulous connection between arteries and veins without an intervening capillary bed. 1 Intracerebral hemorrhage (ICH) is the most common presentation of bAVM, and occurs in 37-71%, followed by epilepsy in 30% of cases. 2,3 The annual risk for hemorrhage is approximately 2.3%, with the risk increasing to 4.8% after a first hemorrhage. [1][2][3] The mechanisms of bAVM development, formation and progress are still poorly understood.
Some researchers think that the formation and progress of bAVMs may be linked not only to hemodynamic factors, but also genetic factors and somatic mutations. 4,5 Recently, some studies showed that certain angioarchitectural factors, such as venous congestion and progressing ectasia, are associated with an increased risk for hemorrhage. 6,7 Other studies emphasized the influence of genes and proteins on clinical symptoms and angioarchitectural changes. 8,9 The molecular signature of the bAVM, including the genetic profile, protein expression status and cellular signaling pathways, probably dictates the clinical course through their effect on the angioarchitecture or destabilization of the nidus of the bAVM. 10,11 By gaining more knowledge about the molecular signature of bAVM in relation to hemorrhage, we might be able to find biomarkers associated with this serious complication. These biomarkers can function as a goal for further research and can be a potential target for gene therapy.
In order to identify potential biomarkers for bAVM hemorrhage, we systematically reviewed the literature to create an overview of the known molecular factors which are associated with bAVM hemorrhage.

Objective
The objective of this review is to identify molecular signatures, defined as genetic profiles, protein expression status or cellular signaling pathways, which are associated with bAVMrelated hemorrhage.

Research question
This study investigates which molecular signatures are associated with bAVM-related hemorrhage. To achieve this goal, we performed a systematic review on genetic expression profiles and their mediators, such as protein expression and cellular signaling, in relation to bAVM-related hemorrhage.

Inclusion criteria
All studies investigating the molecular signature of bAVM in relation to hemorrhage were eligible for this review. The search was restricted to only humans without age restriction. Only articles written in English and Chinese were included in the search.

Exclusion criteria
Articles which were reviews, letters, editorials, case reports or case series with less than three cases were excluded. Studies investigating other vascular malformations, such as arteriovenous fistulas or cavernoma, and where single data on bAVM could not be analyzed separately, were also excluded.

Data collection
Information on study characteristics (author, year, ethnicity, type of sample, sample size), genetic expression profiles, protein expression, signaling pathway or target cell type, and their effects on hemorrhage were retrieved. If the study also provided information on the hemorrhage itself or previous bAVM treatment, this was added as a general comment. Any disagreement between the two reviewers was resolved by discussion to achieve a consensus.

Assessment of risk of bias
The two main reviewers independently assessed the methodological quality of the included studies using the Newcastle Ottawa Scale (NOS) 12 , modified for case-control studies. Since there was no need for follow-up, we did not assess the risk about response. The risk of bias was assessed as low, moderate or high when the total score was ≥7, 5-6 or <5, respectively. Any disagreements concerning the bias assessment was resolved by consensus.

Presentation of results
An overview of the article selection process is presented in a flow diagram according to the PRISMA guidelines. 13 The results are presented in tables where the table on molecular factors is stratified according to type of sample (bAVM tissue or blood samples).

Data synthesis and analysis
Because we anticipated major heterogeneity among our search, such as the same biological effect of different genes or proteins or different types of hemorrhage, we provided a narrative synthesis of the data. Odds ratios (OR) with their 95% confidence intervals were calculated to assess a potential association between molecular changes and the risk of intracranial hemorrhage, when applicable. We also evaluated whether data were suitable for a metaanalysis by searching for identical molecular signatures which were reported in at least two studies.

Data Availability Statement
All data regarding this study are published in this article. The source data are saved in a secured environment in Covidence (https://www.covidence.org/reviews/38842). The first two authors of this article have access to the data and these will be shared by request.

Literature search and selection
The database search identified 3,108 articles. After removal of duplicates, 2,746 articles remained for title and abstract screening and finally the full text of 42 articles was analyzed.
Reasons to exclude articles after full-text screening were: no association was examined between molecular signatures and clinical presentation (n=6), the articles were reviews (n=2), the data on bAVM could not be extracted (n=2), and once the same data was published as in another abstract. The search by hand in the reference lists revealed one additional relevant article for the analysis. The search resulted in a total of 31 articles to be included. Figure 1 presents a flowchart of the review process.

Meta-analysis
Five single-nucleotide polymorphisms (IL6 rs1800796, IL6 rs1800795, TNFα rs1800629, TNFα rs361525 and APOE e2+/e2-) were studied in two articles or more. Unfortunately, these studies used (partially) the same databases and could therefore not be included for a metaanalysis.

Risk of bias in included studies
The risk of bias assessment per subdomain and the total risk of bias assessment are presented in supplemmentart table. The majority of the studies scored a low risk of bias, whereas one study had a moderate risk and one study a high risk.

Discussion
In this systematic review we present molecular factors which are associated with bAVMassociated hemorrhage. Most of the molecular factors showed an association with an increase in hemorrhage risk. Because of use of -at least partially-identical data in the studies investigating the same molecular factor, we were not able to perform a meta-analysis. were MMP-9 rs9509_CC, EPHB4 rs314313_CC and EPHB4 rs314308_TT. Other molecular factors which were found to be associated with bAVM-related hemorrhage were protein upregulation, cell activation and activated pathways. We were not able to calculate OR for these parameters, so the impact of the change in these molecular factors still needs further clarification. With this study, we were able to create an extensive overview of the molecular factors which currently have been investigated to be associated with bAVM-associated hemorrhage. These results can be used for further hypothesis-driven molecular analyses.
Interleukin 6 (IL-6) acts as a pro-inflammatory cytokine and is an important mediator of the acute phase of the inflammatory response. It indirectly stimulates leukocyte recruitment, endothelial activation, and vascular smooth muscle cell (SMC) proliferation. [14][15][16] This proliferation of SMC has been shown to be due to increased MMP-9 and VEGFA release, which was induced by IL-6. 17 Both MMP-9 and VEGFA on its own increase the vascular permeability and vasodilation of the bAVM nidus by breakdown of extracellular matrix and vascular remodeling, leading to susceptibility to hemorrhage. 15,18,19 Additionally, researchers have provided evidence that MMP-9 could be a marker for destabilization of the bAVM nidus and risk for rupture by showing variation of MMP-9 levels after bAVM treatment. 20 The importance of VEGFA in modulating angiogenesis and endothelial cell migration, proliferation and survival, was emphasized by a study of Kim et al. 21 who showed variation in plasma VEGFA levels in bAVM patients before treatment, 24 hours after resection, and 30 days after Germans 9 resection.
The NOTCH-pathway is thought to be involved in the formation of bAVM where it is involved in the endothelial cell selection, vascular smooth muscle cell differentiation and proliferation, arterial maturation, as well as to interact with VEGF. [22][23][24] NOTCH signaling is important to the pathogenesis of vascular disease, and lies both upstream and downstream of critical vascular and lymphatic growth signals such as VEGFA, EphB4/ephrinB2 and adrenomedullin/CLR signaling. 25,26 The NOTCH signaling pathway consists of four receptors (NOTCH 1-4) and several studies have identified NOTCH 1, 3 and 4 overexpression in bAVMs in humans. 27,28 NOTCH pathways are also found to be associated with hemorrhagic presentations of bAVM as summarized in our review. 29,30 However, studies about the effect of NOTCH signaling on bAVM hemorrhagic presentation are still limited and the exact mechanism is still unknown. 31 All the above-mentioned factors are related with acute inflammation or maturation of the endothelial cell. This finding is supported by previous studies which have shown that an inflammatory reaction in the vascular wall plays an important role in the rupture of a bAVM. 16 As such, the increased risk of bAVM hemorrhage seems to be a result of increased inflammation on cellular level and changes in the endothelium leading to instability of the vessel wall.

Conflicts of interest/Competing interests:
The authors declare that there is no conflict of interest.
Availability of data and material (data transparency) All data regarding this study are published in this article. The source data are saved in a secured environment in Covidence (https://www.covidence.org/reviews/38842). The first two authors of this article have access to the data and these will be shared by request.     Introduction A brain arteriovenous malformation (bAVM) is a developmental vascular disorder of the brain that exists of a fistulous connection between arteries and veins without an intervening capillary bed. 1 Intracerebral hemorrhage (ICH) is the most common presentation of bAVM, and occurs in 37-71%, followed by epilepsy in 30% of cases. 2,3 The annual risk for hemorrhage is approximately 2.3%, with the risk increasing to 4.8% after a first hemorrhage. [1][2][3] The mechanisms of bAVM development, formation and progress are still poorly understood.
Some researchers think that the formation and progress of bAVMs may be linked not only to hemodynamic factors, but also genetic factors and somatic mutations. 4,5 Recently, some studies showed that certain angioarchitectural factors, such as venous congestion and progressing ectasia, are associated with an increased risk for hemorrhage. 6,7 Other studies emphasized the influence of genes and proteins on clinical symptoms and angioarchitectural changes. 8,9 The molecular signature of the bAVM, including the genetic profile, protein expression status and cellular signaling pathways, probably dictates the clinical course through their effect on the angioarchitecture or destabilization of the nidus of the bAVM. 10,11 By gaining more knowledge about the molecular signature of bAVM in relation to hemorrhage, we might be able to find biomarkers associated with this serious complication. These biomarkers can function as a goal for further research and can be a potential target for gene therapy.
In order to identify potential biomarkers for bAVM hemorrhage, we systematically reviewed the literature to create an overview of the known molecular factors which are associated with bAVM hemorrhage.

Objective
The objective of this review is to identify molecular signatures, defined as genetic profiles, protein expression status or cellular signaling pathways, which are associated with bAVMrelated hemorrhage.

Research question
This study investigates which molecular signatures are associated with bAVM-related hemorrhage. To achieve this goal, we performed a systematic review on genetic expression profiles and their mediators, such as protein expression and cellular signaling, in relation to bAVM-related hemorrhage.

Inclusion criteria
All studies investigating the molecular signature of bAVM in relation to hemorrhage were eligible for this review. The search was restricted to only humans without age restriction. Only articles written in English and Chinese were included in the search.

Exclusion criteria
Articles which were reviews, letters, editorials, case reports or case series with less than three cases were excluded. Studies investigating other vascular malformations, such as arteriovenous fistulas or cavernoma, and where single data on bAVM could not be analyzed separately, were also excluded.

Data collection
Information on study characteristics (author, year, ethnicity, type of sample, sample size), genetic expression profiles, protein expression, signaling pathway or target cell type, and their effects on hemorrhage were retrieved. If the study also provided information on the hemorrhage itself or previous bAVM treatment, this was added as a general comment. Any disagreement Germans between the two reviewers was resolved by discussion to achieve a consensus.

Assessment of risk of bias
The two main reviewers independently assessed the methodological quality of the included studies using the Newcastle Ottawa Scale (NOS) 12 , modified for case-control studies. Since there was no need for follow-up, we did not assess the risk about response. The risk of bias was assessed as low, moderate or high when the total score was ≥7, 5-6 or <5, respectively. Any disagreements concerning the bias assessment was resolved by consensus.

Presentation of results
An overview of the article selection process is presented in a flow diagram according to the PRISMA guidelines. 13 The results are presented in tables where the table on molecular factors is stratified according to type of sample (bAVM tissue or blood samples).

Data synthesis and analysis
Because we anticipated major heterogeneity among our search, such as the same biological effect of different genes or proteins or different types of hemorrhage, we provided a narrative synthesis of the data. Odds ratios (OR) with their 95% confidence intervals were calculated to assess a potential association between molecular changes and the risk of intracranial hemorrhage, when applicable. We also evaluated whether data were suitable for a metaanalysis by searching for identical molecular signatures which were reported in at least two studies.

Data Availability Statement
All data regarding this study are published in this article. The source data are saved in a secured environment in Covidence (https://www.covidence.org/reviews/38842). The first two authors Germans 6 of this article have access to the data and these will be shared by request.

Literature search and selection
The database search identified 3,108 articles. After removal of duplicates, 2,746 articles remained for title and abstract screening and finally the full text of 42 articles was analyzed.
Reasons to exclude articles after full-text screening were: no association was examined between molecular signatures and clinical presentation (n=6), the articles were reviews (n=2), the data on bAVM could not be extracted (n=2), and once the same data was published as in another abstract. The search by hand in the reference lists revealed one additional relevant article for the analysis. The search resulted in a total of 31 articles to be included. Figure 1 presents a flowchart of the review process. The included studies were published between 2004 and 2018. All investigations were performed either on bAVM tissues or blood samples (see table 1). None of the studies provided sufficient information on the severity or type of hemorrhage, neither on previous bAVM treatment.

Meta-analysis
Five single-nucleotide polymorphisms (IL6 rs1800796, IL6 rs1800795, TNFα rs1800629, TNFα rs361525 and APOE e2+/e2-) were studied in two articles or more. Unfortunately, these studies used (partially) the same databases and could therefore not be included for a metaanalysis.

Risk of bias in included studies
The risk of bias assessment per subdomain and the total risk of bias assessment are presented in supplemmentart table. The majority of the studies scored a low risk of bias, whereas one study had a moderate risk and one study a high risk.

Discussion
In this systematic review we present molecular factors which are associated with bAVMassociated hemorrhage. Most of the molecular factors showed an association with an increase in hemorrhage risk. Because of use of -at least partially-identical data in the studies investigating the same molecular factor, we were not able to perform a meta-analysis.
Our search resulted in a selection of 31 articles, published between 2004 and 2018. With this study, we were able to create an extensive overview of the molecular factors which currently have been investigated to be associated with bAVM-associated hemorrhage. These results can be used for further hypothesis-driven molecular analyses.
Interleukin 6 (IL-6) acts as a pro-inflammatory cytokine and is an important mediator of the acute phase of the inflammatory response. It indirectly stimulates leukocyte recruitment, endothelial activation, and vascular smooth muscle cell (SMC) proliferation. [14][15][16] This proliferation of SMC has been shown to be due to increased MMP-9 and VEGFA release, which was induced by IL-6. 17 Both MMP-9 and VEGFA on its own increase the vascular permeability and vasodilation of the bAVM nidus by breakdown of extracellular matrix and vascular remodeling, leading to susceptibility to hemorrhage. 15,18,19 Additionally, researchers have provided evidence that MMP-9 could be a marker for destabilization of the bAVM nidus and risk for rupture by showing variation of MMP-9 levels after bAVM treatment. 20 The importance of VEGFA in modulating angiogenesis and endothelial cell migration, proliferation and survival, was emphasized by a study of Kim et al. 21 who showed variation in plasma VEGFA levels in bAVM patients before treatment, 24 hours after resection, and 30 days after resection.
The NOTCH-pathway is thought to be involved in the formation of bAVM where it is involved in the endothelial cell selection, vascular smooth muscle cell differentiation and proliferation, arterial maturation, as well as to interact with VEGF. [22][23][24] NOTCH signaling is important to the pathogenesis of vascular disease, and lies both upstream and downstream of critical vascular and lymphatic growth signals such as VEGFA, EphB4/ephrinB2 and adrenomedullin/CLR signaling. 25,26 The NOTCH signaling pathway consists of four receptors (NOTCH 1-4) and several studies have identified NOTCH 1, 3 and 4 overexpression in bAVMs in humans. 27,28 NOTCH pathways are also found to be associated with hemorrhagic presentations of bAVM as summarized in our review. 29,30 However, studies about the effect of NOTCH signaling on bAVM hemorrhagic presentation are still limited and the exact mechanism is still unknown. 31 All the above-mentioned factors are related with acute inflammation or maturation of the endothelial cell. This finding is supported by previous studies which have shown that an inflammatory reaction in the vascular wall plays an important role in the rupture of a bAVM. 16 As such, the increased risk of bAVM hemorrhage seems to be a result of increased inflammation on cellular level and changes in the endothelium leading to instability of the vessel wall.
It remains unclear how the changes in the expression of the molecular factors are induced, leading to upregulation of cytokines with their exciting effects. Taking the angiographic changes which are associated with elevated hemorrhage risk into account, 32, 33 the flow through the bAVM could induce shear stress in certain bAVM vessels, increasing the expression of the above-mentioned molecular factors. On the other hand, an increase in expression profiles in the bAVM could also be caused by the environment, i.e. the brain surrounding the bAVM. This theory can be supported by the study of Fierstra et al. 34 , which showed increased cerebrovascular reactivity in the brain surrounding the bAVM in epilepsy patients. Finally, the molecular signature could also be influenced by previous partial treatment or the bAVM hemorrhage itself, which led to the indication for resection and collection of the samples. This reverse causality can only be overcome by a collection of blood samples in patients without bAVM hemorrhage and follow the cohort for a long period of time.
A review summarizing genetic polymorphisms which were associated with bAVM formation and hemorrhage has been performed before and included studies from 2004 to 2015. 35 These authors performed a meta-analysis with the same studies as mentioned in our review. As mentioned before, the data in the meta-analysis are originating from (at least partial) identical databases and due to on the high risk of selection bias by the use of repetitive data, we decided not to perform a meta-analysis on these data. Our review has included original articles identified using multiple search engines over a longer time period (up to Jan 31 th 2020) and included studies written in Chinese, comparablywith the study of Gao et al. 35 In addition, our search criteria included also molecular signatures associated with hemorrhage and was not restricted to solely genetic polymorphisms, as is the case in the other review. Calculating odds ratios, we were able to identify the strength and significance of molecular factors (restricted to single-nucleotide polymorphisms) contributing to the prediction of hemorrhage of bAVM.

Conclusion
The molecular signature of bAVM hemorrhage consists of a variety of factors, such as presence of certain single-nucleotide polymorphisms, increased expression of cytokines, destabilization of the extracellular matrix and vascular remodeling. These factors are associated with increased inflammation on the cellular level and changes in the endothelium leading to instability of the vessel wall. Further investigation of these biomarkers regarding hemorrhage rates, together with their relation to noninvasive diagnostic methods, should be a goal of future studies to improve the patient specific risk estimation and future treatment options.

Declarations
Funding: Wenhua Sun received funding of the China Scholarship Council to study and do research in University hospital Zurich, Switzerland.

Conflicts of interest/Competing interests:
The authors declare that there is no conflict of interest.
Availability of data and material (data transparency) All data regarding this study are published in this article. The source data are saved in a secured environment in Covidence (https://www.covidence.org/reviews/38842). The first two authors of this article have access to the data and these will be shared by request. and future trends in treatment.