Cell to cell communication is an early event in glioblastoma

Glioblastoma (GB) is the most aggressive and lethal tumour of the central nervous system (CNS). GB cells proliferate rapidly and display a network of ultra-long tumour microtubes (TMs) that mediate cell to cell communication. GB TMs infiltrate into the brain, enwrap neurons and facilitate the depletion of Wingless (Wg)/WNT from the neighbouring neurons. GB cells establish a positive feedback loop including Wg signalling upregulation that activates the JNK pathway and matrix metalloproteases (MMPs), in turn, these signals promote TMs infiltration, GB progression and neuronal synapse loss and degeneration. Thus, cellular and molecular signals other than primary mutations emerge as central players of GB. Here we describe the temporal organization of the events that occur in GB. We define the progressive activation of JNK pathway signalling mediated by Grindelwald (Grnd) receptor, is caused by the ligand Eiger (Egr)/TNFα produced by the healthy tissue. We propose that cellular interactions of GB with the rest of the brain is an early event that precedes GB proliferation and expansion. We conclude that non-autonomous signals facilitate GB progression and contribute to the complexity and versatility of these incurable tumours.


Introduction
Glioblastoma multiforme (GB) is the most frequent and aggressive primary malignant brain tumour with a 3 per 100.000 incidence per year (Gallego, 2015). GB patients´ median survival is 12-15 months, with less than 5% of survival after 5 years (Gallego, 2015;Louis et al., 2016;McGuire, 2016;Rogers et al., 2018). The causes of GB are under debate (McGuire, 2016), 5% of the patients develop GB after a low grade astrocytoma (Alifieris and Trafalis, 2015) and the most frequent mutations include gain of function of the epidermal growth factor receptor (EGFR) (97%) and the phosphatidylinositol-3 kinase (PI3K)/phosphatase and tensin homologue (PTEN) pathways (88%) (Hayden, 2010). The diagnosis, and therefore the treatment of GB, requires a mutation analysis considering the high frequency of clones within the same primary GB (Wick et al., 2018). Temozolomide (TMZ) is the only treatment for GB, however, recent discoveries restrict the use of TMZ in GB patients depending on the methylation status of methylguanine DNA methyltransferase (MGMT) (Wick et al., 2018).
Moreover, among other mutations, Isocitrate dehydrogenase (IDH) define the nature and features of GB (Waitkus et al., 2018) together with molecular alterations including 1p/10q deletions and tumour suppressor protein 53 (TP53) and alpha thalassemia/mental retardation (ATRX) mutations (Miller et al., 2017;Waitkus et al., 2018). The genetic and molecular heterogeneity complicate the diagnosis and treatment of these fatal brain tumours.
EGFR mutant forms show constitutive kinase activity that chronically stimulates Ras signalling to drive cell proliferation and migration (Furnari et al., 2007;Maher et al., 2001).
Other common genetic lesions include loss of the lipid phosphatase PTEN, which antagonizes the phosphatidylinositol-3 kinase (PI3K) signalling pathway, and mutations that activate PI3KCA, which encodes the p110a catalytic subunit of PI3K (Furnari et al., 2007;Maher et al., 2001). GBs often show constitutively active Akt, a major PI3K effector. Multiple mutations that coactivate EGFR-Ras and PI3K/Akt pathways are required to induce a glioma (Holland, 2000). In Drosophila, a combination of EGFR and PI3K mutations effectively causes a glioma-like condition that shows features of human gliomas including glia expansion, brain invasion, neuron dysfunction, synapse loss and neurodegeneration This model involves the co-overexpression of constitutively active forms of Epidermal Growth Factor Receptor (dEGFR λ ) and an activated membranelocalized version of the PI3K catalytic subunit p110α/PI3K92E (dPI3K92E CAAX ) under Gal4 UAS control, specifically driven in the glial cells by means of repo-Gal4 (Brand and Perrimon, 1993;Casas-Tinto et al., 2017).
The recent discovery of a network of ultra-long tumour microtubes (TMs) in GB (Osswald et al., 2015), also known as cytonemes in Drosophila (Casas-Tinto and Portela, 2019), improves our understanding of GB progression and therapy resistance (Osswald et al., 2016). In GB, glial cells display a network of membrane projections which mediate cell to cell communication. TMs are actin-based filopodia that infiltrate into the brain and reach long distances within the brain (Osswald et al., 2015). TMs are required in GB cells to mediate Wingless (Wg)/WNT signalling imbalance among neurons and GB cells.
Wg/WNT signalling is increased in GB cells to promote proliferation, at the expense of neuronal Wg signalling that results in neurodegeneration and lethality (Arnes and Casas Tinto, 2017;Casas-Tinto and Portela, 2019;. The central role of TMs in GB biology has emerged as a fundamental mechanism for GB rapid and lethal progression; thus, it is an attractive field of study towards potential GB treatments.
However, the molecular mechanisms underlying the expansion of TMs and the signalling pathways mediating TM infiltration are still poorly understood.
The Jun-N-terminal Kinase (JNK) pathway is a hallmark of GB cells that is associated to glial proliferation and stem-like status, and currently it is a pharmacological target for GB (Matsuda et al., 2012). Moreover, the JNK pathway is the main regulator of Matrix metalloproteases (MMPs) expression and cell motility in many organisms and tissues including tumors like GB (Cheng et al., 2012;Ispanovic and Haas, 2006;Lee et al., 2009;Uhlirova and Bohmann, 2006;Zeigler et al., 1999).
MMPs are a family of endopeptidases capable of degrading the extracellular matrix (ECM). Members of the MMP family include the "classical" MMPs, the membrane-bound MMPs (MT-MMPs), the ADAMs (a disintegrin and metalloproteinase; adamlysins) and the ADAMTS (a disintegrin and metalloproteinase with thrombospondin motif). There are more than 20 members in the MMP and ADAMTS family including the collagenases, gelatinases, stromelysins, some elastases and aggrecanases (Malemud, 2006). The vertebrate MMPs have genetic redundancy and compensation, they have overlapping substrates, and pharmacological inhibitors are non-specific. There are two orthologues to human MMPs in Drosophila, MMP1 and MMP2. MMP1 is secreted and MMP2 is membrane-anchored (Page-McCaw et al., 2003). However, recent reports propose that products of both genes are found at the cell surface and released into media, and that GPI-anchored MMPs promote cell adhesion when they become inactive. Moreover, the two MMPs cleave different substrates, suggesting that this is the important distinction within this small MMP family (LaFever et al., 2017). MMPs are upregulated in several tumours, including GBs. Cancer cells produce MMPs to facilitate tumour progression and invasiveness and MMPs upregulation in GB is associated with the diffuse infiltrative growth and have been proposed to play a role in GB cell migration and infiltration (de Lucas et al., 2016;Veeravalli and Rao, 2012) reviewed in (Nakada et al., 2003). MMPs are upregulated in human GB cell-lines and biopsies as compared with low-grade astrocytoma (LGA) and normal brain samples (Hagemann et al., 2012;Hagemann et al., 2010). In particular, among the 23 MMPs present in humans, MMP9, MMP2 and MMP14 are directly implicated in growth and invasion of GB cells (Munaut et al., 2003).
In consequence, MMPs upregulation in GB is an indicator of poor prognosis (Yamamoto et al., 2002) and the study of the mechanisms mediated by MMPs is relevant for the biology of GB, and cancer in general. GB cells project TMs which cross the extracellular matrix (ECM) and infiltrate in the brain to reach territories distant from the primary GB site (Osswald et al., 2015;Osswald et al., 2016). We have previously demonstrated that GB cells activate the JNK pathway and accumulate MMPs (MMP1 and 2). MMPs contribute to TMs expansion in the brain and facilitate Frizzled1-mediated Wg/WNT signalling in GB cells. Moreover, Wg/WNT signalling mediates JNK activation in GB cells to continue with the TMs infiltration process. We hypothesize that the founder mutations in GB (PI3K and EGFR) initiate the process with the expansion of the TMs; afterwards, the system self-perpetuates (TMs-Fz1/Wg-JNK-MMPs-TMs) to facilitate GB progression and infiltration in the brain (Portela et al., 2019).
We have recently showed that JNK signalling pathway is required for GB progression and proliferation. JNK pathway activation mediated by the receptor Grnd is a requirement for both TMs expansion, and Fz1-mediated Wg depletion from neurons. In turn, Wg pathway upregulation in GB induces JNK activity in GB cells that mediate the production of matrix metalloproteinases (MMPs), moreover MMPs silencing in GB cells is sufficient to rescue neurodegeneration and premature death caused by GB (Portela et al., 2019).
However, the molecular mechanisms that determine JNK pathway activation in GB cells remain unknown.
Here, we describe the role of the ligand Egr in the activation of JNK pathway in GB through the specific receptor Grnd, highlighting again the contribution of communication signals between healthy tissue (neurons) and GB cells to the progression of the disease (Jarabo et al., 2020;. Egr is expressed and secreted by non-tumoral tissue, but it is accumulated in tumoral cells and activates the JNK pathway. In consequence, GB cells produce MMPs that facilitate TM progression and GB dissemination. At the cellular level, we analyse 3 aspects required for GB progression: TMs expansion, GB proliferation and synapse loss in the surrounding neurons, and propose a timeline for these events in GB progression.

Combined activation of EGFR and PI3K pathways is required for GB progression
The To determine the separate contribution of PI3K and EGFR signaling pathways to GB progression, we first analyzed EGFR and PI3K/PTEN pathways independently. Fz1 localization in the TMs is key to trigger Wg signaling and GB proliferation. We assessed the presence of Fz1 receptor in glial membranes with a specific monoclonal antibody previously validated . Fz1 receptors signal is localized homogeneously across the brain in control samples ( Figure S1B). However, in the GB model (PI3K+EGFR) Fz1 accumulate in TMs ( Figure S1C  These data suggest that the activation of both pathways together is necessary for the expansion of the TMs network and Fz1 localization in the TMs. PI3K and EGFR pathways converge in dMyc expression and dMyc expression is required for GB development (Annibali et al., 2014;Read et al., 2009;Tateishi et al., 2016;Wang et al., 2017). Thus, to determine if dMyc is sufficient to trigger TMs network and Fz1 accumulation, we ectopically overexpressed dMyc (UAS-dMyc) in glial cells under repo-Gal4 regulation and stained the brains with Fz1 ( Figure S1F), Wg ( Figure   S1G-I) and Cyt-Arm antibodies (Figure S1J-L). The confocal images show no morphological evidence of TMs expansion or GB formation (Figure S1F compare to glioma in S1C). Moreover, Fz1, Wg or Cyt-Arm protein distribution resemble control brains ( Figure S1B, F, G, I, J, L). Taking these results together, dMyc overexpression is not sufficient to reproduce the features of the GB even though it is a convergent point of EGFR and PI3K pathways. These results suggest that both PI3K and EGFR constitutive upregulation in glial cells is necessary for the expansion of glial TMs, Fz1 localization in TMs and activation of the Wg pathway in glial transformed cells.
Next, to determine the epistatic relations behind MMPs upregulation, we used a specific monoclonal antibody to visualize MMP1. The confocal images show that MMP1 is homogeneously distributed across the brain in control samples ( Figure 1A) and accumulates in TMs in GB samples ( Figure 1B). Besides, MMP1 shows a homogeneous distribution through the brain upon constitutive activation of PI3K (dp110 CAAX ) and EGFR (TOR-DER CA ) in glial cells ( Figure 1C-D, quantified in F). Besides, dMyc overexpression in glial cells did not cause significant changes in MMPs localization ( Figure 1E-F). Taking these results together, independent overexpression of EGFR, PI3K or dMyc is not sufficient to reproduce the features of GB. These results suggest that the combined activity of PI3K and EGFR pathways are necessary to activate a downstream pathway responsible for the expansion of TM glial projections and MMP1 accumulation in GB cells, however dMyc overexpression is not sufficient to cause these phenotypes.

Non-autonomous activation of JNK pathway in GB
In addition to Wg pathway, GB cells activate JNK pathway to maintain the stem-like GB cells, which has become a pharmacological target for the treatment of GB (Matsuda et al., 2012). A recent study using gene expression profiling identified genes that were significantly correlated with the overall survival in patients with GB (Hsu et al., 2019). 104 genes were identified, which are common between patients with GB and those with low grade gliomas and can be used as core genes related to patient survival. Of these, 10 genes (CTSZ, EFEMP2, ITGA5, KDELR2, MDK, MICALL2, MAP 2 K3, PLAUR, SERPINE1, and SOCS3) can potentially classify patients with gliomas into different risk groups. Among these pathways, the TNF-alpha signalling pathway stands out. Four genes from this 10-gene group (MAP 2 K3, PLAUR, SERPINE1, and SOCS3) are involved in TNFα signalling, and they might have potential prognostic value for patients with GB (Hsu et al., 2019).
TNF-like weak inducer of apoptosis (TWEAK, also known as TNFSF12) is one of the ligands of the tumor necrosis factor (TNF) family (Chicheportiche et al., 1997). TNFSF12 preferentially activates non-canonical NF-κB and promotes the invasive properties of glioma cells (Cherry et al., 2015). To determine the contribution of TNFSF12 to GB, we analysed the data from the CGGA dataset from GlioVis (http://gliovis.bioinfo.cnio.es/). In silico analysis showed that TNFSF12 gene is highly upregulated in several samples of CNS tumors, including GB ( Figure 2A). Accordingly, TNFSF12 expression has a negative implication in the overall survival of patients with gliomas ( Figure 2B).
Due to the relevance of JNK pathway and TNFα ligand in GB prognosis we decided to study Egr, that is the solely Drosophila orthologue of the ligand for the mammalian Tumour Necrosis Factor (TNF)-TNF Receptor signalling pathway (Igaki et al., 2002;Moreno et al., 2002) and its receptor Grnd (Andersen et al., 2015), in the Drosophila GB model.

Egr re-localizes from Neuron to Glia in GB
To monitor Egr protein distribution, we used a transgenic Drosophila line in which the endogenous egr gene is GFP tagged (Egr-GFP protein fusion) and monitored GFP signal. The results from confocal microscopy show that 70% of Egr-GFP signal is localized in the neurons that are in the proximity of glial cells and 30% of Egr-GFP signal is localized in the glia ( Figure 2C-D, H). However, in GB samples there is a shift of Egr-GFP signal from neurons to glia (~50%) ( Figure 2E-H). Next, we monitored the expression pattern of the JNK receptor Grnd in GB and control samples, with a specific antibody previously validated (Andersen et al., 2015). The quantification of Grnd signal in glial membranes show an increase of Grnd protein in GB brains ( Figure 2I-K). The abnormal distribution of Egr and Grnd in glioma brains could be due to either an increase in gene and/or protein expression or to redistribution of the proteins. We undertook quantitative Polimerase Chain Reaction (qPCRs) experiments with RNA extracted from control and GB larvae brains, which revealed an that there is a 2-fold increase of egr transcription in GB brains ( Figure 2L) consistent with the increased Egr staining in glioma brains.

Egr expressed in healthy tissue activates JNK in GB
Next we studied which cell type upregulates egr expression in GB brains. To determine whether the source of Egr are the glial cells, we silenced egr expression specifically in glioma cells. We used two different RNAi lines to knockdown egr (egr-RNAi) in GB cells (repo>dp110 CAAX ; EGFR λ ; egr-RNAi). The quantifications show that egr knockdown does not prevent GB cell proliferation (Fig 3A-3D and 3E) nor GB membrane expansion ( Fig   3A-3D and 3F). These results suggest that egr expression in GB cells is not relevant for GB progression in the Drosophila model. Therefore, taking into consideration that there is an increase of Egr protein in glial membranes in GB compared to control brains ( Figure   2), and it is not due to a transcriptional upregulation of egr specifically in the glial cells, we propose that Egr is transcriptionally upregulated specifically in neurons and relocalised from the surrounding neuronal tissue to GB cells in order to activate JNK pathway through Grnd receptor in GB cells.

Progressive JNK activation in glioma
JNK is upregulated in several tumours including GB and it is related to glioma malignancy (Hagemann et al., 2005;Huang et al., 2003;Mu et al., 2018;Zeng et al., 2018).
Moreover, JNK is a target for specific drugs in combination with temozolomide treatments as it was proven to play a central role in GB progression (Feng et al., 2016;Kitanaka et al., 2013;Matsuda et al., 2012;Okada et al., 2014). However, little is known about the molecular mechanisms underlying JNK activation in GB cells and the functional consequences for GB progression.
We had previously confirmed JNK pathway activation in Drosophila GB cells , by using the TRE-RFP reporter that confer transcriptional activation in response to JNK signalling (Chatterjee and Bohmann, 2012;Jemc et al., 2012;Ruan et al., 2016). We decided to study the temporal activation of the JNK pathway to uncover the order the signalling events that occur in GB cells and healthy surrounding tissue. We took advantage of the JNK reporter (puc-LacZ) that monitors the transcriptional activation of the downstream JNK target puckered (Langen et al., 2013;Martin-Blanco et al., 1998) to analyse JNK activity in GB at 48 and 96h after GB induction. To control the temporal induction of the tumour, we used the thermo sensitive repression system Gal80 TS .
Individuals maintained at 17ºC did not activate the expression of the UAS constructs, after larvae were switched to 29ºC, the protein Gal80 TS was not longer able to repress Gal4 system and the genetic induction of GB was activated. In addition, GB cells were marked with RFP (UAS-myr-RFP) to distinguish GB cells and healthy tissue. The previous results indicate that JNK signalling pathway activation is progressive, and therefore GB progression follows a chronogram. To evaluate the cellular events associated to GB (progressive growth of the glial membrane projections and the number of glial cells), we dissected larval brains at 24, 48 and 72h after tumour induction (referred to as day 1D, 2D or 3 days in Figure 5A-C). GB cells membrane was visualized with UAS-myr-RFP and we stained the glia nuclei with a specific anti-Repo antibody to quantify the glial cell number by analysing confocal microscopy images from entire brains ( Figure 5A-C).
The statistical analysis of glial membrane volume quantifications ( Figure 5D We quantified the number of synapses in the NMJ of third-instar larvae 1, 2 and 3 days after GB induction. NMJs were stained with anti-bruchpilot to visualize active zones (synapses) by using confocal microscopy ( Figure 6A-C).
The statistical analysis of synapse number ( Figure 6D) show progressive synapse loss between day 1 and day 2, and between day 2 and day 3 after tumor induction. The overall loss of synapses at the NMJ is highly significant ( Figure 6D).
To determine whether there is an association between the volume of the glial network and the number of synapses at the NMJ, we plotted the number of synapses against the volume of the glial network in a correlation graph ( Figure 6E). The correlation index is -0.966 (3s.f.) indicating there is a negative association between the volume of the TMs and the number of synapses at the NMJ. Therefore, larger glial membrane network leads to a greater synapse loss. Taking all these results together, the first cellular event happening after the tumor induction is the neurodegeneration caused by GB, suggesting that the interaction between GB cells and neurons is an early event. Then we observed the expansion of the TMs network volume and finally, the increase in the number of glial cells. These results suggest that the interaction and activation of signaling pathways between GB cells and neurons occurs very early in the process of GB progression, bringing the study of cell to cell communication as an intriguing feature of GB.

Discussion
Activating mutations for EGFR and PI3K pathways are the most frequent initial signals in GB. However, the attempts to treat GB reducing the activation of these pathways have so far been limited by acquired drug resistance. GB cells show a high mutation rate and usually present more than two sub-clones within the same patient and from the same

Progressive neurodegeneration
There is a significant progressive decrease in the number of synapses at the NMJ. We have observed a negative correlation between the volume of the membrane network and the number of synapses at the NMJ, suggesting that the expansion of the membrane projections is responsible for the neurodegeneration. Previous studies from our laboratory have proved that TMs surround neurons and deplete WNT from the neurons , which leads to synapse loss and neurodegeneration (Rich and Bigner, 2004). We can conclude that as the tumor progresses, TMs grow progressively, infiltrate in the brain and surround neurons leading to their degeneration, process  (Boukhatmi et al., 2020). This cell to cell interaction is similar to our findings in glianeuron communication in GB, where glial cells use cytonemes/TMs to hijack neuronal Egr and activate JNK signaling in the tumor cells to sustain tumor progression.

Progressive activation of JNK pathway via Egr/Grnd
One of the main molecular events in GB is the activation of JNK signalling that regulates MMPs expression, TMs expansion and infiltration. Previous results indicate that Wg pathway responds to JNK and network expansion, these three events were described as a regulatory positive feedback loop in GB progression. However, the signals that specifically activate JNK pathway are under debate. Grnd is the receptor that upon binding to the ligand Egr, activates JNK pathway. Our qPCR results and the localization analysis of Egr-GFP fused protein showed that Egr is produced in the surrounding neurons, and re-localized to the membrane of GB projections in contact with healthy neuronal tissue. The progressive activation of JNK pathway in glial cells correlate with the morphological changes (membrane network expansion and number of glial cells) that GB undergo 48h after tumor induction. Therefore, this example of neuron-GB interaction where non-autonomous signals facilitate tumor progression and GB-mediated neurodegeneration, contribute to the complexity and versatility of these incurable tumours. This neuron-GB communication mechanism highlights the relevance of the interactions and signalling between GB and the surrounding healthy tissue. Therefore, it is of interest to uncover the regulatory mechanisms that mediate Egr expression and secretion in neurons in response to GB induction, and the associated response in GB cells mediated by Grnd/Egr as a potential modulator of GB progression.
antibodies, and FlyBase for its wealth of information. We acknowledge the support of the Confocal Microscopy unit and Molecular Biology unit at the Cajal Institute for their help with this project. MP holds a fellowship from the Juan de la Cierva program IJCI-2014-19272 and SCT holds a contract from the Ramón y Cajal program RYC-2012-11410 from the Spanish MICINN. Research has been funded by grant BFU2015-65685P.
Authors declare no conflicts of interest.

Drosophila glioblastoma model
In Drosophila, a combination of EGFR and PI3K mutations effectively causes a gliomalike condition that shows features of human gliomas including glia expansion, brain invasion, neuron dysfunction, synapse loss and neurodegeneration (Kegelman et al., 2014;Read, 2011;Read et al., 2009). To generate a glioma in Drosophila melanogaster, the Gal4/UAS system (Brand and Perrimon, 1993) was used as described above (repo-Gal4>UAS-EGFR λ ,UAS-dp110 CAAX . To restrict the expression of this genetic combination and control it in a temporal manner, we used the thermo sensitive repression system Gal80 TS . Individuals maintained at 17ºC did not activate the expression of the UAS constructs, but when the larvae were switched to 29ºC, the protein Gal80 TS changed conformation and was not longer able to bind to Gal4 to prevent its interaction with UAS sequences, and the expression system was activated and therefore the GB was induced. Tub-GAL80 TS ; repo-Gal4>UAS-EGFR λ , UAS-dp110 CAAX animals were raised at 17 C, shifted to 29 C for 1, 2, 3 or 4 days at 8, 6.3, 3.7 or 2 days after egg laying (AEL) respectively, and subjected to dissection immediately after.

Imaging
Fluorescent labelled samples were mounted in Vectashield mounting media with DAPI (Vector Laboratories) and analysed by Confocal microscopy (LEICA TCS SP5). In all experiments the whole brain lobes were acquired individually and whole stacks were analysed. The images shown in the figures are a single plane images to facilitate visualization. In some panels, higher magnifications of single plane images are shown in the figures to facilitate the visualization of the antibody staining's in more detail. Images were processed using Leica LAS AF Lite and Fiji (Image J 1.50e). Images were assembled using Adobe Photoshop CS5.1.

Quantifications
Relative MMP1 and Grnd staining within brains was determined from images taken at the same confocal settings. Average pixel intensity was measured using measurement log tool from Fiji 1.51g and Adobe Photoshop CS5.1. Average pixel intensity was measured in the glial tissue and in the adjacent neuronal tissue (N<10 for each sample) and expressed as a Glia/Neuron ratio. Glial network volume was quantified using Imaris surface tool (Imaris 6.3.1 Bitplane Scientific Solutions software).
The number of Repo + cells, the number of synaptic active sites and the number of puc-lacZ positive cells was quantified by using the spots tool Imaris 6.3.1 software, we selected a minimum size and threshold for the puncta in the control samples of each experiment. Then we applied these conditions to the analysis of each corresponding experimental sample. For the puc-lacZ glia or neuron co-localization studies we quantified the total number of puc-lacZ + cells and then applied a co-localization filter (intensity center of the channel of interest) using the Spots tool from the Imaris 6.3.1 software.
For the co-localization of Egr-GFP in glial cells, GFP channel volume was quantified using Imaris surface tool. We selected a specific threshold for the total volume in the control samples and then we applied these conditions to the analysis of the corresponding experimental sample. Then we applied a co-localization filter (intensity mean of the red channel)

Statistical Analysis
To analyse and plot the data, we used Microsoft Excel 2013 and GraphPad Prism 6. We performed a D'Agostino & Pearson normality test and the data found to have a normal distribution were analysed by a two-tailed t test with Welch-correction. In the case of multiple comparisons, we used a One-way ANOVA with Bonferroni post-test. The data that did not pass the normality test were subjected to a two-tailed Mann-Whitney U test or in the case of multiple comparisons a Kruskal-Wallis test with Dunns post-test. Error bars represent standard error of the mean. * represents p value ≤ 0.05; ** p value ≤ 0.01; *** p value ≤ 0.001. Statistical values of p value >0.05 were not considered significant, (n.s.).