Wasl is crucial to maintain microglial core activities during glioblastoma initiation stages

Abstract Microglia actively promotes the growth of high‐grade gliomas. Within the glioma microenvironment an amoeboid microglial morphology has been observed, however the underlying causes and the related impact on microglia functions and their tumor promoting activities is unclear. Using the advantages of the larval zebrafish model, we identified the underlying mechanism and show that microglial morphology and functions are already impaired during glioma initiation stages. The presence of pre‐neoplastic HRasV12 expressing cells induces an amoeboid morphology of microglia, increases microglial numbers and decreases their motility and phagocytic activity. RNA sequencing analysis revealed lower expression levels of the actin nucleation promoting factor wasla in microglia. Importantly, a microglia specific rescue of wasla expression restores microglial morphology and functions. This results in increased phagocytosis of pre‐neoplastic cells and slows down tumor progression. In conclusion, we identified a mechanism that de‐activates core microglial functions within the emerging glioma microenvironment. Restoration of this mechanism might provide a way to impair glioma growth.


| INTRODUCTION
High-grade gliomas represent a complex and devastating disease and are posing an unmet clinical need. These tumors resist multimodal therapies and survival times are only 14 months on average (Gregory et al., 2020;Kadiyala et al., 2019;Lucki et al., 2019;Wen & Kesari, 2008). In recent years, a lot of focus has been on the complex microenvironment of gliomas. Microglia and infiltrating macrophages are the most prominent cell types within the glioma microenvironment and can account for up to 30%-50% of the total tumor mass (for review see (Hambardzumyan et al., 2015;Quail & Joyce, 2017).
While these processes have been described in gliomas, surprisingly little is known about the apparent change of morphology of microglia within the glioma and the possible impact on their functions. Microglia, as the resident innate immune cells of the brain, display unique morphological features. Under physiological conditions microglia are in a surveillance mode and actively and continuously scan their microenvironment using dynamic large processes providing them a ramified morphology (Nimmerjahn et al., 2005). However, once the homeostasis is altered by injury or brain pathologies, microglia retract their processes to acquire an amoeboid shape. This change in morphology can correlate with an either anti-or proinflammatory state of microglia (Bernier et al., 2019;Bolasco et al., 2018;Chia et al., 2018;Karperien, 2013;Kettenmann et al., 2011;Lawson et al., 1992;. Of note, an amoeboid microglial morphology has been observed in vivo across different glioma models at different stages of glioma growth as well as within human glioma samples (Annovazzi et al., 2018;Chia et al., 2018;Juliano et al., 2018;Kvisten et al., 2019;Resende et al., 2016;Ricard et al., 2016). Furthermore, these microglia show a decreased phagocytic activity and motility within the central area of neoplastic lesions (Hutter et al., 2019;Jaiswal et al., 2009;Juliano et al., 2018;Pyonteck et al., 2013;Wu et al., 2010). The mechanisms underlying this rapid and drastic morphological remodeling are still not known. Clearly, these morphological phenotypes must be highly regulated and involve adaptations to the cellular cytoskeleton (Bernier et al., 2019;Okazaki et al., 2020). Cell morphology, phagocytosis and motility are cellular processes known to be actin-dependent and are crucial for the multitasking roles of microglia (Koizumi et al., 2007;Liu et al., 2020;Lively & Schlichter, 2013;Okazaki et al., 2020;Pollard & Cooper, 2009;Uhlemann et al., 2015). An alteration of the microglial actin cytoskeleton and their morphology will most likely affect several of these core functions. Therefore, it is important to understand the impact of an amoeboid microglial morphology on their functions within the tumor microenvironment and its effects on tumor growth.
Here, we investigated the influence of a pre-neoplastic glioma environment on microglia morphology and related functions. We utilized a recently published zebrafish glioblastoma multiforme (GBM) model which is based on expression of the human oncogene HRasV12 in the proliferating domains of the developing brain and gives rise to tumors similar to the mesenchymal subtype of human GBM (Mayrhofer et al., 2017). Analysing larval stages of this zebrafish model allowed us to directly study, the influence of an early pre-neoplastic environment on the morphology and functions of microglia in vivo. Importantly, we detected an immediate impact of pre-neoplastic HRasV12 + cells on the microglia population resulting in an amoeboid phenotype and increased proliferation of the microglia. Furthermore, their phagocytic activity, motility and speed was significantly reduced compared to control microglia. RNA sequencing of microglia revealed significantly lower expression levels of wasla, the zebrafish orthologue of human WASP like actin nucleation promoting factor (WASL, also known as N-WASP), a key regulator of actin cytoskeleton organization (Dart et al., 2012;Linder et al., 1999;Lorenz et al., 2004;Park & Cox, 2009;Yamaguchi et al., 2005;Yu et al., 2012). Importantly, a microglia specific rescue of wasla expression in HRasV12 + larvae restored microglial morphology as well as their number, speed and motility. Furthermore, the wasla rescue in microglia restored their phagocytic activity which resulted in improvements in both engulfment of pre-neoplastic cells and survival.

| Mounting, immunohistochemistry, image acquisition and live imaging
Whole-mount immunostaining of samples was performed as previously described (Astell & Sieger, 2017

| Phagocytosis assay
Phagocytosis assay was performed using an Eppendorf FemtoJet microinjector to inject custom made zymosan (Sigma) coupled with Pacific Blue fluorochrome (molecular probes) into the larval zebrafish brain. Anesthetized 3 and 5 dpf HRasV12 À , HRasV12 + and HRasV12 + ; wasla larvae were injected into either telencephalon or tectum with approximately 2 nl of injection solution composed of 2.5.10 5 zymosan/ μl, 0.1% phenol red in PBS. Injected larvae were maintained at 28.5 C in E3 containing 200 μM PTU for 6 h, fixed in 4% PFA/1% DMSO in PBS at RT C for 2 h then microglia immunostaining was performed.
Phagocytosis of pre-neoplastic HRasV12 + cells by microglia is described in the microglia/macrophage isolation section.

| Image analysing
Analysis of all images was performed in 3D using Imaris (Bitplane, Zurich, Switzerland). To assess microglia (4C4 + ) morphology and volume measurements, we used the surface-rendering tool in Imaris 8.2.1, which allowed segmentation of individual cells in 3D as well as bigger volumes such as brain and pre-neoplastic mass volume. To assess microglia morphological changes, we calculated the ratio of the cellular surface and cellular volume of individual cells as previously described Gyoneva et al., 2014). Microglia with a ratio smaller than 0.8 were classified as amoeboid. The "Spots" function tool was used to quantify the number of amoeboid microglia related to the total number of microglia within the full brain, and the averaged value expressed as measure of the percentage of amoeboid microglia. To determine the percentage of infiltrated macrophages and microglia at 5 dpf, a 4C4 immunostaining was performed on HRasV12 À and HRasV12 + larvae with GFP + macrophages and microglia (mpeg1:GFP). The number of 4C4 + /GFP + cells (microglia) and 4C4 À /GFP + cells (macrophages) were counted in relation to the total number of myeloid cells (microglia + macrophages) within the full brain and the averaged value expressed as measure of percentage of macrophages and microglia.
To quantify proliferation rates, the number of 4C4 + /EdU + cells (EdU + microglia) were counted in relation to the total number of microglia within the full brain and the averaged value expressed as a measure of percentage of microglia proliferation. To quantify zymosan phagocytosed by microglia of 3 and 5 dpf HRasV12 À , HRasV12 + and HRasV12 + ; wasla larval brains, we used the surface-rendering tool in Imaris 8.2.1 to manually create a surface corresponding to the fluorescent signal of all zymosan particles. An additional surface for microglia allowed us to distinguish zymosan that had been phagocytosed by microglia. This allowed us to read out the sum of fluorescence intensity (AU) of phagocytosed zymosan within microglia surfaces and the total AU of all zymosan surfaces. The percentage of microglia phagocytosis was calculated from the averaged value of the number of phagocytosed zymosan related to the total number of zymosan within either the telencephalon or the tectum (Engulfed zymosan fluorescent intensity:total zymosan fluorescent intensity) x 100).
To assess microglia motility of 3 and 5 dpf HRasV12 À , HRasV12 + and HRasV12 + ; wasla larval brains, we used the "Tracking" function tool and manually tracked mpeg:eGFP + cells along time. Imaris software calculated track length and track speed mean for each cell and microglia tracks were displayed as time color-coded lines for the different conditions. To measure the number of phagosomes per microglia from 5 dpf HRasV12 + and HRasV12 + ; wasla larvae, we used mpeg1:eGFP + cells to visualize phagosomes in black within GFP + cytoplasm. The combination of 3D and slide views on Imaris software allowed us to manually count the number of phagosomes in microglia within the tectum of 10 larvae per condition.
FACS allowed cell separation from debris in function of their size (FSC-A) and granularity (SSC-A). Single cells were then separated from doublets or cell agglomerates (FSC Singlet; SSC Singlet). From the single-cell population, a gate was drawn to separate live cells (DAPIÀ) from dead cells (DAPI+). Unstained and cells incubated with secondary antibody Alx647 only were used as controls to draw gates corresponding to microglia (4C4 + /Alx647 + ) populations. Finally, microglia (4C4 + /Alx647 + ; Figure S2) and microglia/macrophage (mCherry + ; Figure S4A) were segregated from the live cell population gates. FACS data were analysed using FlowJo Software (Treestar, Ashland, OR).
Phagocytosis of pre-neoplastic HRasV12 + cells by microglia was measured using the mean of GFP fluorescent intensity from HRasV12 + cells detected within isolated microglia from 5 dpf HRasV12 + and HRasV12 + ; wasla larvae.

| RNA extraction and cDNA amplification
All experiments were performed in three replicates with a total num-
The expression data and clinical annotation for human glioma samples were downloaded from the Joyce Lab Brain TIME database (Klemm et al., 2020). We retrieved raw count data for MDMs and microglia. The raw counts were normalized and transformed (rlog) using DESeq2 (Love et al., 2014). Eventually, we compared WASL

| Survival assay
HRasV12 À , HRasV12 + and HRasV12 + ; wasla larvae were screened at 2 dpf for positive transgene expression then were housed in a purpose-built zebrafish facility, in the Queen's Medical Research Institute, maintained by the University of Edinburgh Biological Resources.
Larvae were kept by 20 per nursing tanks (three replicates) at 28 C on a 14 h light/10 h dark photoperiod, daily fed by facility's staff members from 5 to 31 dpf. Surviving larvae were counted every day for 1 mpf.

| Statistical analysis
All experiments were performed in replicates. Within the figure, legends "N" indicates the number of replicates while "n" indicates the total number of larval fish analysed. All measured data were analysed using StatPlus (AnalystSoft, Inc.). Unpaired two-tailed Student's t-tests were performed to compare two experimental groups, and one-way ANOVA with Bonferroni's post-hoc tests for comparisons between multiple experimental groups. Statistical values of p < .05 were considered to be significant. All graphs were plotted in Prism 8 (GraphPad Software) and values presented as population means ± SD.

| Pre-neoplastic cells affect the microglia population in the larval zebrafish brain
Although an amoeboid morphology is a consistent feature of microglia within gliomas and has been described across models and species Here, we investigated the effect of pre-neoplastic cells on the microglia population using a zebrafish GBM model (Mayrhofer et al., 2017). This model is based on overexpression of human HRasV12 in the proliferating regions of the developing central nervous system (CNS) and results in aggressive tumors resembling human mesenchymal GBM (Mayrhofer et al., 2017). To analyse preneoplastic stages, we outcrossed the driver fish line Et(zic4:GAL4TA4,-UAS:mCherry) hmz5 (Distel et al., 2009) to the line Tg(UAS:eGFP-HRasV12) io006 (Santoriello et al., 2010; hereafter Zic/HRasV12 model; Figure 1a). The Zic4 enhancer drives Gal4 expression in the proliferating regions of the developing CNS and upon binding of Gal4 to its target UAS sequence, activates expression of mCherry in control larvae (hereafter HRasV12 À ) and additional eGFP-HRasV12 expression in tumor developing fish (hereafter HRasV12 + larvae). In this model, HRasV12 expression is enriched in the telencephalon from 1 day post fertilization (dpf) onwards, followed by the tectal proliferation zone (TPZ) at 2 dpf and in the cerebellum from 3 dpf. The expression is low in the optic tectum ( Figure 1a). This set up results in significantly increased proliferation and measurements based on the mCherry signal that is present in HRasV12 À and HRasV12 + brains revealed an increased brain size in HRasV12 + larvae from 3 dpf onwards (Mayrhofer et al., 2017). Our own brain volume measurements confirmed this, and we observed a significant increase of the larval brain volume at 5 dpf in HRasV12 + brains (5.4 ± 0.8 10 6 μm 3 ) compared to HRasV12 À brains (2.7 ± 0.3 10 6 μm 3 ; p < .0001; Figure 1b).
To visualize microglia, we performed immunohistochemistry using the microglia specific zebrafish 4C4 antibody in HRasV12 À (control) and HRasV12 + larvae (Becker & Becker, 2001;Chia et al., 2018;Mazzolini et al., 2019;Tsarouchas et al., 2018). In line with normal zebrafish development, we did not detect microglia progenitor cells in the brain at 1 and 2 dpf (data not shown), whereas they appear within the tectum by 3 dpf, then spread to the entire brain during development in both conditions (Figure 1c). Under normal physiological conditions, mature microglia are highly ramified and are constantly scanning their microenvironment (Nimmerjahn et al., 2005). Once microglia detect changes, such as microenvironment modifications in brain pathologies, they retract their processes and adopt an amoeboid morphology (Karperien, 2013). During zebrafish development microglia show an amoeboid phenotype during brain colonization stages at 3 dpf and transit toward a ramified phenotype over the next two days of development (Mazzolini et al., 2019;Svahn et al., 2013;Figure 1c). In order to test if the exposure to pre-neoplastic cells has an impact on this process, we analysed microglia morphology in HRasV12 + larvae at 3 dpf and upon exposure to HRasV12 + cells for 2 days (5 dpf). To assess microglial morphology, we used our previously described method to calculate the ratio of the microglia cell surface to microglia cell volume, which can be used as a read out for their morphological changes ; Figure 1d). Within 3 dpf HRasV12 À and HRasV12 + brains, microglia were mostly amoeboid in both conditions with a percentage of 80% ± 4 and 84% ± 5, respectively ( Figure S1A). However, in 5 dpf larvae we detected 67 ± 16% of amoeboid microglia in presence of HRasV12 + cells compared to 17 ± 12% in the HRasV12 À condition (p < .0001; Figure 1d).
We previously reported an amoeboid microglial morphology in the presence of pre-neoplastic myrAKT1 neural cells in larval zebrafish brains . To test if different oncogenes have the same impact on microglial morphology when overexpressed in the same cell type, we decided to overexpress myrAKT1 as well as BRAF-V600E under control of the zic4 promoter. To achieve this, we used the Et Microglia numbers have been shown to be increased in late stage tumors but also during tumor initiation (Badie & Schartner, 2001;Bowman et al., 2016;Chen et al., 2012;Chia et al., 2018;Coniglio & Segall, 2013;Graeber et al., 2002;Li & Graeber, 2012). To test whether the expression of HRasV12 in the developing CNS induces an increase in microglial numbers, we quantified 4C4 + cells (microglia) within 3 and 5 dpf brains of control and HRasV12 + larvae.
The expression of eGFP under the mpeg1 promoter allows to visualize all macrophages including microglia (Ellett et al., 2011) and additional staining using the 4C4 antibody allows to distinguish microglia from macrophages. These experiments revealed that in control brains 97% of the myeloid cell population were microglia and this did not change significantly in HRasV12 + brains ( Figure S1E). Hence, pre-neoplastic HRasV12 + cells do not lead to the infiltration of macrophages.
To understand how the increase of the microglia population is achieved, we assessed microglial proliferation activity by performing an EdU assay, and stained control and HRasV12 + larval microglia (4C4 + ). The number of EdU + /4C4 + cells was significantly higher in HRasV12 + brains (21 ± 19 %) compared to control brains (2 ± 2%; p = .0003; Figure 2b). Thus, the exposure of microglia to preneoplastic HRasV12 + cells from 3 to 5 dpf triggers their proliferation and results in a significant increase of their number.
Taken together, our results show that microglia number and morphology are not affected by the immediate presence of pre-neoplastic cells, however within 2 days, microglia numbers increased significantly, and showed a higher percentage of amoeboid cells compared to controls.

| Pre-neoplastic HRasV12 + cells affect microglial functions
In light of our previous observations on amoeboid microglial morphology, we decided to investigate effects of HRasV12 + cells on fundamental microglial functions such as phagocytosis and motility.
These results indicate that within 2 days of contact with preneoplastic cells, microglia exhibit a strong reduction of their phagocytic activity.
During early stages of brain development, microglia are very efficient at clearing debris and apoptotic cells due to their high motility (Haynes et al., 2006;Kyrargyri et al., 2020;Sieger et al., 2012). To assess if the observed change in microglial morphology/function in HRasV12 + larvae has an impact on their motility, we performed high resolution confocal live imaging on HRasV12 + larvae from the outcross between Tg (zic4:Gal4UAS:mCherry:mpeg1:eGFP) and Tg(UAS:TagBFP2-HRasV12) F I G U R E 1 HRasV12 expression in the proliferating regions of the developing CNS alters microglia morphology. (a) Schematic representation of the zebrafish germline system used to induce HRasV12 expression based on the outcross of the indicated fish lines. Schematic anteriorposterior dorsal view of the brain representing the main sub-divisions: telencephalon (T), optic tectum (OT) cerebellum (CB) and tectal proliferation zone (TPZ) in gray. Confocal images showing mCherry and eGFP-HRasV12 fluorescent signal in the proliferating regions of the developing brain of HRasV12 + larvae from 1 to 5 dpf. White dotted lines mark the main brain subdivisions. Scale bar represents 100 μm. (b) Brain volume was assessed using Imaris surface tool to build the segmented images (right panels) of the mCherry signal (left panels) of proliferating regions of the developing brain from 5 dpf HRasV12 À (top panels) and HRasV12 + (bottom panels). Scale bar represents 100 μm. Brain volumes of 5 dpf HRasV12 À and HRasV12 + larvae were quantified. HRasV12 À : n = 10; HRasV12 + : n = 10; N = 3. (c) Confocal images of microglia (magenta) distribution throughout the developing brain of HRasV12 À (top panels) and HRasV12 + brains (bottom panels) from 3 to 5 dpf, using 4C4 antibody to specifically label microglia. Scale bar represents 100 μm. Close-ups of microglia at 3 dpf and 5 dpf under physiological condition (HRasV12 À ) allow to determine their morphology: amoeboid and ramified. Scale bar represents 10 μm. (d) Close-ups of microglia from 5 dpf HRasV12 À (top panels) and HRasV12 + (bottom panels) larvae (left panels) and their segmented images in 3D (right panels) using Imaris surface tool, to assess microglia morphology. Scale bar represents 10 μm. The number of amoeboid microglia was quantified within the microglial population of 5 dpf control and HRasV12 + larvae. Results are shown as a percentage of total microglia. HRasV12 À : n = 15; HRasV12 + : n = 15; N = 3. Error bars represent mean ± SD. Images were captured using a Zeiss LSM710 confocal microscope with a 20X/NA 0.8 objective. All images represent the maximum intensity projections of Z stacks fish lines, and control larvae from the incross between Tg(zic4:Gal4-UAS:mCherry:mpeg1:eGFP). Based on our previous observations on microglia morphology, number and phagocytosis, we decided to perform confocal live imaging for 13 h on larvae between 3 and 4 dpf and between 4 and 5 dpf. We tracked microglial movement in three-dimensional (3D) for the full duration of the time series and calculated the track length (motility) and their speed of movement in the two different conditions. Interestingly, in presence of HRasV12 + cells at 3 dpf, microglia speed (0.007 ± 0.002 μm/s) and motility (312 ± 165 μm) were similar to control microglia speed (0.007 ± 0.004 μm/s) and motility (335 ± 103 μm; Figure 3c). However, we observed an obvious reduction of microglia motility in presence of pre-neoplastic cells at 5 dpf compared to controls ( Figure 3d). Quantification of microglial speed (0.009 ± 0.003 μm/s) and motility (425 ± 144 μm) in HRasV12 + larvae compared to speed (0.012 ± 0.003 μm/s) and motility of microglia in HRasV12 À larvae (567 ± 173 μm) showed a significant difference (P speed = 0.001; P motility = 0.001; Figure 3d).
In summary, our results show that within a short time window, pre-neoplastic cells alter not only microglial morphology but also key functions such as phagocytosis and motility at early stages of GBM formation. Hence, we speculated that lasting alterations of the microglial actin cytoskeleton might be the underlying cause, which result in a permanent change in morphology and impair related functions such as motility and phagocytosis.

| RNA sequencing of isolated microglia reveals down-regulation of wasla gene expression
Our data show that microglia morphology, phagocytosis and motility are altered upon contact with pre-neoplastic cells. These cellular mechanisms are regulated by a compilation of different signaling pathways, but they are all orchestrated by the actin cytoskeleton organization (Freeman & Grinstein, 2014;Pollard & Cooper, 2009;Svitkina, 2018). Hence, we decided to conduct RNA sequencing to investigate differentially expressed (DE) genes involved in actin cytoskeleton organization and regulation between control and HRasV12 + conditions at 3 and 5 dpf. Following our previously published protocols, we isolated microglia from dissociated brains, stained them using the 4C4 antibody and sorted using flow cytometry (Mazzolini et al., , 2019; Figure S2). For each time point, microglia were pooled from 600 HRasV12 À and HRasV12 + larval brains and three replicates were performed per time point for RNA sequencing.
We evaluated the expression correlation between biological replicates using the whole set of genes from the RNA-seq data set.
Each sample consisted of isolated microglia from 600 brains; of note, scatter plots of the normalized transformed read counts showed that biological replicates were highly correlated (r > 0.75).
Interestingly, correlation between control (HRasV12 À ) samples was higher at 3 dpf (r > 0.83) and 5 dpf (r > 0.81) than between HRasV12 + samples (r > 0.75) and (r > 0.79; Figure S3A). The lower correlation obtained at 3 and 5 dpf from HRasV12 + larval brains could be explained by heterogeneity in those samples due to the presence of pre-neoplastic cells ( Figure S3A). PCA confirmed this correlation by showing HRasV12 À replicates are more clustered than replicates from HRasV12 + conditions. Moreover, clusters corresponding to 3 and 5 dpf samples from both conditions were well segregated ( Figure S3B). A global analysis of our data using the KEGG database showed an enrichment for few pathways at 3dpf, whereas we did not detect an enrichment for specific pathways in microglia from HRasV12 + brains compared to controls (Table S1).
Results are expressed as a percentage of total microglia. HRasV12 À : n = 17; HRasV12 + : n = 17; N = 3. Error bars represent mean ± SD. Error bars represent mean ± SD. Images were captured using a Zeiss LSM710 confocal microscope with a 20X/NA 0.8 objective. All images represent the maximum intensity projections of Z stacks (FDR < 0.05, fold change > j2j) in microglia from HRasV12 + brains at 3 dpf and 346 DE genes at 5 dpf (Table S2). As we were speculating that actin cytoskeleton organization is impaired in microglia in HRasV12 + brains, we then used KEGG and defined pathways related to actin cytoskeleton in zebrafish (Regulation of actin cytoskeleton, Focal adhesion, Adherens junction, Tight junction and mTOR signaling pathway) and selected genes from our RNA-seq which are referred to them (Table S3). We compared expression of F I G U R E 3 Legend on next page. these genes in microglia between control and HRasV12 + larval brains at 3 and 5 dpf and focussed on genes that are differentially expressed (FDR < 0.05, fold change > j2j) at 5 dpf when pre-neoplastic cells affect microglial functions. We obtained 11 differentially expressed (DE) genes, 8 with lower expression and 3 with higher expression in microglia from HRasV12 + brains compared to control microglia (Figure 4a, Table S3). The majority of these genes belonged to the "Regulation of actin cytoskeleton" pathway, and the identified top ranked DE gene was wasla (Table S3), the zebrafish orthologue of human WASP like actin nucleation promoting factor (WASL, also known as N-WASP). WASL is a key protein of the actin cytoskeleton organization, necessary to maintain cell shape, efficient phagocytosis and motility (Dogterom & Koenderink, 2019;May et al., 2000;Niedergang & Grinstein, 2018;Yamaguchi et al., 2005). Wasla showed 4.2 times lower expression in microglia in the presence of pre-neoplastic cells compared to microglia from control brains (FDR = 0,005). Intriguingly, wasla was the only gene of the WASP family significantly differentially expressed in microglia in the presence of pre-neoplastic cells, whereas was, wasf and wash expressions were the same as in control microglia (Figure 4b).
To test if WASL expression is altered in microglia within human gliomas, we accessed RNA sequencing data recently generated by Klemm et al. (2020)). Here, we focussed on IDH WT gliomas, which typically represent grade IV glioblastomas and are highly infiltrated by macrophages (MDMs) and microglia. We compared expression levels of WASL in MDMs and microglia isolated from these tumors to isolated cells from non-tumor samples. Interestingly, while WASL expression was not altered in MDMs (p = .89), we detected lower expression levels of this gene in microglia of IDH WT gliomas in comparison to microglia from non-tumor samples (Figure 4c). Although not statistically significant, the data shows a clear trend (p = .14).
Our results show that microglia exposed to pre-neoplastic cells express a lower level of wasla. Thus, we hypothesized that reduced levels of wasla are the underlying cause of the change of microglial morphology and the decrease of their motility and phagocytic capacity.
3.4 | Microglia specific overexpression of wasla restores microglial functions in HRasV12 + brains In order to test our hypothesis that lower expression levels of wasla were responsible for the observed changes in microglial morphology and functions, we performed cell specific overexpression for wasla in microglia. To achieve this, we created a plasmid, which encodes for wasla under the control of the mpeg1 promoter specific to microglia and macrophages (Ellett et al., 2011). Injection of the mpeg1:wasla plasmid into one cell stage embryos resulted in a transient, mosaic expression of wasla in microglia. To verify the efficiency of this strategy, we injected the plasmid into one cell stage Tg(mpeg1:mCherry) embryos, isolated mCherry + microglia/macrophages at 5 dpf by FACS ( Figure S4A) and performed qPCR for wasla. We obtained a 1.46 times higher expression of wasla in mCherry + microglia/macrophages from injected embryos compared to non-injected embryos ( Figure S4B).
Hence, we injected the mpeg1:wasla plasmid into HRasV12 À and HRasV12 + embryos and analysed the impact on microglia at 5 dpf.
In summary, these results reveal that wasla is a key gene to maintain microglial functions and its lower expression levels in HRasV12 + brains are responsible for alterations in microglia morphology, phagocytosis and motility.
3.5 | Rescue of wasla expression in microglia slows down pre-neoplastic growth by restoring an efficient microglial phagocytic activity Intrigued by the restoration of microglia morphology and functions by wasla overexpression in the HRasV12 + condition, we investigated its impact on pre-neoplastic growth and survival of the larval zebrafish. The Zic/HRasV12 model has been previously described with a survival rate of 4% in the first month (Mayrhofer et al., 2017).
As we have shown that microglial morphology and actindependent functions such as phagocytosis were restored in these larvae, we hypothesized that phagocytosis of pre-neoplastic cells by microglia contributed to the reduced pre-neoplastic growth and better survival. To address this hypothesis, we first of all tested the direct impact of rescued wasla expression in microglia on phagosome formation. Here, we made use of the Tg(zic4:Gal4UAS:mCherry:mpeg1:eGFP) and performed high resolution confocal imaging at 5 dpf. This allowed us to directly observe microglia engulfing mCherry labeled cells. These cells can then be detected within vesicular structures, which are the phagosomes of the microglia (Movie S1; Figure S5, red arrowheads).
Of note, not all phagosomes were mCherry positive as microglia also engulfed other cell types that were not labeled with mCherry (Movie S1; Figure S5, yellow arrowheads). These phagosomes simply appear as black holes as the GFP signal within the microglia is restricted to the cytoplasm ( Figure S5, yellow arrowheads).
Hence, the restoration of phagocytic activity in microglia results in an increased phagocytosis of pre-neoplastic HRasV12 + cells. This might F I G U R E 4 The actin nucleation promoting factor wasla is less expressed in microglia from HRasV12 + larvae. (a) Heatmap of differentially expressed (DE) genes (FDR < 0.05, Fold Change > j2j) from microglia transcriptome comparisons between 5 dpf HRasV12 À and HRasV12 + larval brains (11 genes), belonging to zebrafish actin cytoskeleton KEGG pathways [Regulation of actin cytoskeleton, Focal adhesion, Adherens junction, Tight junction and mTOR signaling pathway]. The actin nucleation promoting factor wasla is the top ranked gene of the list. See also Table S3. (b) Dot plots of normalized transformed read counts of the representative WASP family genes. Black plots represent non-DE genes whereas, red plots correspond to DE genes at 5 dpf between control and HRasV12 + conditions. (b) The means ± SD of three independent experiments are plotted. (c) Dot plots of normalized transformed read counts of WASL gene expression of monocyte-derived macrophages (MDMs) and microglia in human non tumor and glioma IDH WT brain samples (Klemm et al., 2020) explain the smaller pre-neoplastic mass volume in HRasV12 + ; wasla larvae.

| DISCUSSION
In this study, we revealed the impact of pre-neoplastic cells on the microglial population, their morphology and functions during tumor initiating stages. Several elegant studies have shown crosstalk between GAMs and neoplastic cells in the brain creating a microenvironment favorable to tumor growth and maintenance (for review see (Gutmann & Kettenmann, 2019). However, while the amoeboid microglial morphology has been described across models and species (Annovazzi et al., 2018;Bayerl et al., 2016;Chia et al., 2018;Juliano et al., 2018;Kvisten et al., 2019;Resende et al., 2016;Ricard et al., 2016), the underlying causes and the timing for the change in morphology are not understood. To our knowledge, this is the first study to provide mechanistic insights of an alteration of microglia morphology and functions due to lower expression levels of the wasl gene in presence of tumor initiating cells. We utilized a wellestablished larval zebrafish brain tumor model to address the earliest stages of tumor induction due to activation of oncogenes. By overexpressing a constitutively active form of the human HRas gene, we induced cellular alterations in the larval zebrafish brain that lead to the formation of tumors similar to the mesenchymal subtype of human GBM by 1 month postfertilization (Mayrhofer et al., 2017).
Mesenchymal subtype GBMs have been found to correlate with a stronger enrichment of GAMs compared to proneural and classical GBM subtypes (Bhat et al., 2013;Wang et al., 2017). Here, we strategically worked with 3 and 5 dpf larvae to monitor the pre-neoplastic cell impact on microglia. By using immunohistochemistry, transgenic zebrafish lines for microglia, functional assays and in vivo imaging we revealed that increased microglial numbers, morphological changes, WASL is a key protein of the actin cytoskeleton organization and hence expressed in almost every cell type. Nevertheless, differences in expression levels can be observed. In the human and mouse brain, astrocytes and neurons for example show higher expression levels compared to other cells such as microglia and oligodendrocytes (Zhang et al., 2014(Zhang et al., , 2016. Among myeloid cells in the brain, Wasl can be detected in every subset with varying expression levels . Interestingly, there seems to be controversy on the role of WASL during cell division. Cytokinesis is the final step of cell division taking place at the end of mitosis, this mechanism is characterized by the formation of a contractile actomyosin ring necessary for the separation of the newly forming daughter cells. Wang et al. con-cluded that WASL has a role in cytokinesis during porcine oocyte maturation , whereas others consider that mechanism as WASL independent (Bompard et al., 2008;Deschamps et al., 2013;Schwayer et al., 2016). Our data support the hypothesis that WASL is not involved in cell proliferation but is crucial for other microglial functions.
All images represent the maximum intensity projections of Z stacks F I G U R E 6 Microglial wasla expression is crucial to slow down tumor progression. (a) Kaplan-Meier survival plot of HRasV12 À , HRasV12 + and HRasV12 + ; wasla larvae control over 31 days, n = 50/60, 4/60 and 3/60, respectively. p = .0017 (Gehan-Breslow-Wilcoxon test between HRasV12 + and HRasV12 + ; wasla conditions). Error bars represent mean ± SD. (b) Brain and pre-neoplastic mass volume were measured using the mCherry signal (brain, top panels) and eGFP signal of HRas + cells (pre-neoplastic mass; bottom panels) of proliferating regions of the developing brain from 5 dpf HRasV12 À (left panel), HRasV12 + (middle panel) and HRasV12 + ; wasla (right panel) larvae. Scale bar represents 100 μm. Brain and pre-neoplastic mass volume from 5 dpf HRasV12 À , HRasV12 + and HRasV12 + ; wasla larvae are quantified using Imaris surface tool. HRasV12 + : n = 9; HRasV12 + ; wasla: n = 6; N = 3. Error bars represent mean ± SD. Red dotted line indicates the brain volume mean in control condition. (c) Close-up confocal images of microglia (mpeg1:eGFP + cells) from 5 dpf HRasV12 + (left panel) and HRasV12 + ; wasla (right panel) brains. Phagosomes are indicated by red asterisks. Scale bar represents 10 μm. The number of phagosomes per microglia from 5 dpf HRasV12 + and HRasV12 + ; wasla brains were quantified. HRasV12 + : n = 80; HRasV12 + ; wasla: n = 80; N = 3. Error bars represent mean ± SD. (d) Mean of GFP fluorescent intensity (MFI) from phagocytosed pre-neoplastic cells detected by flow cytometry within isolated microglia from 5 dpf HRasV12 + and HRasV12 + ; wasla larvae. The means ± SD of two independent experiments are plotted. Images were captured using a Zeiss LSM880 confocal microscope with a 20X/NA 0.8 objective. All images represent the maximum intensity projections of Z stacks cell line (CHME-5) with C6-glioma cells and reported a significant reduction of microglia phagocytic activity after 24 h exposure (Voisin et al., 2010). In HRasV12 + larvae, microglia showed alterations of their functions 2 days after exposure to pre-neoplastic cells. Microglial Of note, isolated microglia from human IDH WT gliomas expressed lower levels of WASL. Although these results were statistically not significant, they revealed a strong trend (p = .14). We speculate that heterogeneity within the microglia population as well as differences in tumor stages might be the underlying explanation. Interestingly, expression levels of WASL were not altered in MDMs isolated from human IDH WT gliomas. This is in line with previous results on a different impact of the glioma environment on MDMs and microglia and further underpins potential differences in their role within gliomas.
Interestingly, amongst the differentially expressed genes extracted from our transcriptomic data belonging to KEGG pathways linked to "Regulation of actin cytoskeleton" some of the other genes might also contribute to the observed effects. In presence of pre-neoplastic cells microglia expressed lower levels of fibroblast growth factor 2 (fgf2) and fibroblast growth factor receptor 3 (fgfr3), which have been shown to increase microglial migration and phagocytic activity (Noda et al., 2014). Furthermore, fibroblast growth factor 10a (fgf10a) showed lower expression in microglia of HRasV12 + larvae. FGF10 treatment has been shown to inhibit microglial pro-inflammatory cytokine secretion and proliferation via regulation of the TLR4/NF-K B pathway in an animal model after spinal cord injury (Chen et al., 2017). Hence, the reduced expression levels of this gene could contribute to the increased microglial proliferation.
Among the DEG that showed higher expression levels in microglia from HRasV12 + brains, we detected G protein subunit gamma 12 (gng12). Interestingly, gng12 is known to be highly expressed after LPS stimulation of microglia and to offset the inflammatory response by reducing levels of nitric oxide and TNFα (Larson et al., 2010). Moreover, long non-coding gng12 RNAs are highly expressed in glioma tissues and its downregulation inhibits proliferation, migration and epithelial-mesenchymal transition of glioma cells (Xiang et al., 2020). These results suggest that high expression levels of gng12 in microglia exposed to pre-neoplastic cells might contribute to the generation of a pro-tumoral response.
Phagocytic events are associated with cytokine secretion as part of the innate immune response, and in preparation for adaptive immunity (Acharya et al., 2020;Chung et al., 2006;Fu et al., 2014;Heo et al., 2015;Murray et al., 2005). To precisely determine the changes in microglial functions in presence of HRasV12 + cells, it is important to understand the strategy applied by these cells to alter microglial gene expression levels. Extracellular vesicles (EVs) contain proteins, lipids and different RNA species that change the activity of recipient cells (Tkach & Théry, 2016;Verweij et al., 2019). Several studies have shown the implication of EVs secreted by tumor cells on microglia/macrophages (Abels et al., 2019;Hyenne et al., 2019;Vos et al., 2015). Of note, wasl expression has been shown to be regulated by various miRNAs (Bettencourt et al., 2013;Schwickert et al., 2015). Data from the Mione laboratory shows that some of these miRNAs have significantly increased expression levels in HRasV12 + cells compared to control cells (Anelli et al., 2018). This includes the miRNA Let-7g-1, which targets the wasla gene. Thus, it is tempting to speculate that EVs In conclusion, we show for the first time that during tumor initiation stages, pre-neoplastic cells influence microglial functions by altering their gene expression profiles resulting in an alteration of microglial morphology and related functions. We identify wasla as a key component in the regulation of microglial morphology, phagocytosis and migration. Our findings provide a mechanism that empowers pre-neoplastic cells to trap microglia within their vicinity, deactivate their phagocytic functions and promote the generation of an antiinflammatory tumor promoting microenvironment. Marshall-Phelps for proofreading the manuscript.

CONFLICT OF INTEREST
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are included in the supplementary tables and are available from the corresponding author upon reasonable request ORCID Julie Mazzolini https://orcid.org/0000-0001-9347-5635 Dirk Sieger https://orcid.org/0000-0001-6881-5183