Dysregulation of EndMT and cell adhesion markers in cAVM nidus

Our earlier histopathological and immunostaining studies indicated that major vessels in cAVM nidus possess arterial and venous, as well as capillary, endothelial markers [30]. Endothelial cells undergoing mesenchymal transition demonstrate altered morphological features, with subsequent increases in mesenchymal proteins such as transgelin (SM22-α), calponin, etc. We observed that at the mRNA transcript level, SNAI1, Slug (SNAI2), transgelin (TAGLN), and Calponin 1 (CNN1) were highly expressed in cAVM samples. SNAI2 (Slug) had a more prominent expression (3.16 fold) in cAVM than SNAI1 (2.36 fold) when compared with control specimens (Fig. 1A).

Fig. 1

Dysregulation of EndMT and major cell adhesion markers in cerebral AVM nidus. A Scatter plots with bars represent the mRNA fold changes of SNAI1, SNAI2, Calponin 1 (CNN1), transgelin (TAGLN) and cell adhesion-associated integrin α9 (ITGA9), integrin β1 subunits (ITGB1), VE-cadherin (CDH5), and N-cadherin (CDH2) in 18 human cAVM nidi and 15 control brain tissues. GAPDH was used as the endogenous control for mRNA fold analysis. Scatter dots represent expression in individual specimens and bars represent mean value. B Cerebral AVM nidus consisted of enlarged vessels with medial thickening and intimal hyperplasia. Immunohistochemical staining shows the reduced expression of activated integrin α9/β1 and overexpression of SNAI1/2, Transgelin, Calponin1, and N-cadherin in cAVM vessels compared with control brain vessels. All these EndMT-associated markers were expressed in the intima or neointima of huge AVM nidal vessels (scale bar 100 µm, magnification ×20). C Bar graph showing histoscore analysis of SNAI1/2, Transgelin, Calponin1, N-cadherin, and integrin α9/β1 in three random microscopic fields of immunostained sections. Values are mean ± SD. *p < 0.05 versus control tissue, **p < 0.01, ***p < 0.001, ****p < 0.0001. ns nonsignificant difference

During the EndMT process, cell–cell and cell–ECM adhesions of endothelial cells and basement membrane will be disrupted. These molecular changes facilitate the invasiveness of endothelial cells. Hence, we studied the expression of adhesion factors such as cadherins and integrins in cAVM nidus. VE-cadherin (CDH5) was downregulated by 0.84 fold in patient samples, while N-cadherin (CDH2) mRNA was upregulated (2.76 fold) in cAVMs. Integrin α9 subunit (ITGA9) mRNA was overexpressed by 2.78 fold in cAVMs, but integrin β1 (ITGB1) mRNA was reduced by 0.63 fold.

Furthermore, we conducted an immunohistochemical staining-based EndMT protein expression and localization analysis in AVM and control tissues. We observed that there was an overexpression of SNAI1/2 in the intimal regions of large vessels in AVMs (Fig. 1B). SNAI1/2 was not expressed in control brain vasculature. There was an approximate 14-fold increase of SNAI1/2 in nidal samples compared with controls, based on histoscore analysis (Fig. 1C). Calponin 1 and transgelin expressed very intensely in both intima and media of cAVM nidal vessels. Transgelin expression was present in small blood vessels of control brains, but Calponin 1 was not expressed in control vasculature.

N-cadherin was localized to the neointimal regions of the nidus tissues and was not observed in control specimens. We also did not observe N-cadherin, a mesenchymal marker, in the medial layer of any of our patient samples. The activated form of integrin, α9/β1, was found to be downregulated in cAVM nidus in contrast to control vasculature.

Cerebral AVMs express higher levels of NICD3 and NICD4

We initially studied the expression of all four Notch receptor genes, Notch 1–4, in the cAVM nidus. Notch3 and Notch4 mRNAs were significantly overexpressed in nidus specimens, compared with control specimens, by approximately three and two fold, respectively (Fig. 2A). Notch1 was slightly overexpressed in cAVMs (1.26 fold) in comparison with control mRNA. Notch2 expression levels were statistically insignificant in cAVM nidus.

Fig. 2

Expression profile of Notch receptors in cerebral AVM nidal vessels. A Scatter plots with bar diagrams of mRNA fold changes of Notch 1–4 receptors in 18 human cAVM nidi and 15 control brain tissues. GAPDH was taken as the endogenous control for quantification. B Representative photomicrographs of immunostaining illustrate that Notch intracellular domain (NICD) proteins NICD3 and NICD4 are overexpressed in cAVM compared with control brain vasculature (scale bar 100 µM, magnification ×20). C Semiquantitative histoscore analysis shows significant overexpression of NICD3 and NICD4 proteins in cAVMs. D Representative western blot of NICD3 protein expressed in three cerebral AVM nidi and three control tissues. GAPDH was considered as the loading control. Lanes 1–3: cAVM nidus, and lanes 4–6: control tissues. Values are mean ± SD. *p < 0.05 versus control tissue, **p < 0.01, ***p < 0.001, ****p < 0.0001. ns not significant

We specifically investigated the expression of activated Notch receptors i.e., Notch intracellular domains (NICD) in cAVM tissue sections and control brain samples. NICD1 was also included in the immunohistochemical studies considering its reported role in endothelial cells, but none of the control or AVM nidal vessels expressed it [31]. NICD3, the Notch mural variant, was found to be highly expressed and localized to the intima and media of nidal vessels, indicating EndMT (Fig. 2B, C). NICD4 receptor levels in cAVM were lower compared with NICD3 and were seen in a diffused pattern, more toward intima and less in the tunica media of tortuous vessels. Elevated expression of NICD3 in AVM nidus was further corroborated by western blot analysis in three of the control tissues and nidi (Fig. 2D, Additional file 2).

Oscillatory flow promotes gamma-secretase-dependent Notch receptor activation

To identify whether oscillatory shear stress-dependent Notch receptor activation occurs in endothelial cells, we conducted qRT-PCR-based mRNA analysis and protein immunofluorescence assay in hCMECs exposed to defined flow conditions using a microfluidic flow chamber. Initially, hCMECs were characterized for von Willebrand factor expression (Additional file 1: Fig. S1). We then analyzed the gene expression profile in hCMECs exposed to uniform parallel shear stress (15 dyn/cm2) for 24 h. We found that the mRNA expression of Notch1–4 under parallel unidirectional flow conditions did not vary significantly from the expression under static flow conditions (Fig. 3A). Oscillatory flow induced overexpression of Notch3 (4.11 fold) and Notch4 (2.06 fold) mRNAs in endothelial cells, compared with static flow. Compared with cells exposed to parallel flow, oscillatory flow resulted in a 3.2 and 1.96 fold increase of Notch3 and Notch4, respectively. Both parallel and oscillatory flow did not induce significant Notch1 and Notch2 expression in hCMECs, compared with static flow conditions (Fig. 3A).

Fig. 3

Notch receptor activation by oscillatory fluid flow in human cerebral microvascular endothelial cells (hCMEC/d3). A mRNA level expression of Notch receptors upon exposure of hCMEC/d3 to disturbed flow (n = 3). Notch3 and Notch4 become prominent as endothelial cells are exposed to oscillatory flow for 24 h, while Notch1 and Notch2 expressions were not significantly regulated by altered fluid flow. mRNA fold values in parallel and oscillatory flow were calculated relative to the static control. All data were normalized with GAPDH expression and are given as relative to static control. B hCMEC/d3 exposed to oscillatory flow at 15 dyn/cm2 for 24 h resulted in the overexpression of NICD3, which was localized to the nucleus. NICD4 was also overexpressed in cells exposed to oscillatory flow, but cytoplasmic localization was more prominent. NICD1 was very faintly expressed in cells exposed to oscillatory flow, but was detected in mean fluorescence intensity (MFI) analysis. DAPI (blue) was used to counterstain nuclei (scale bar 20 µM, magnification ×40). C MFI was plotted as the average fluorescence intensity ± SD of five microscopic fields per flow condition and from three biological replicates. D, E DAPT and RO4929097 efficiently prevented NICD3 expression in endothelial cells in the presence of continuing altered flow. F Representative western blot of NICD3 protein present in proteins isolated from cells exposed to static (lane 3), parallel (lane 2), and oscillatory (lane 1) shear stress conditions. GAPDH was considered as the loading control. PF indicates parallel uniform shear stress and OF represents oscillatory shear stress. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 versus respective static or parallel uniform shear-treated groups. ns not significant

Next, we investigated the NICD 1, 3, and 4 protein-level expression and their cellular localization in hCMECs exposed to various flow conditions. There was an overexpression of NICD3 in endothelial cells exposed to oscillatory shear stress compared with parallel flow and static conditions. Oscillatory shear stress altered the subcellular localization of the Notch3 receptor by promoting the nuclear localization of NICD3 (Additional file 1: Fig. S2A). Cells were also found to be elongated with a spindle-like morphology after 24 h of flow. NICD4 expression was marginally elevated, but the localization was predominantly cytoplasmic. NICD1 expression was not significantly observed in cells exposed to any of the three flow conditions, but MFI analysis indicated a slight overexpression in cells exposed to oscillatory flow (Fig. 3B, C). To rule out non-specific staining, we performed negative control assays with secondary antibodies in the absence of primary antibodies (Additional file 1: Fig. S3).

Furthermore, we studied whether oscillatory shear stress directly induces Notch receptor activation by inducing gamma-secretase activity. MTT assay was conducted to decide the optimum DAPT and RO4929097 concentrations (Additional file. 1. Fig. S4). On the basis of the viability offered, hCMECs were treated with 0.5 µM DAPT and 250 nM RO4929097 during the 24 h oscillatory flow exposure. Both inhibitors negatively affected the shear stress response of the gamma-secretase-induced cleavage of the Notch3 receptor in hCMECs (Fig. 3D, Additional file 1: Fig. S2A). However, the efficacy of RO4929097 was superior to DAPT and statistically significant (p = 0.0007) (Fig. 3E). Elevated NICD3 protein-level expression in cells exposed to oscillatory flow was further substantiated by western blot analysis (Fig. 3F, Additional file 2).

Oscillatory flow-induced EndMT requires Notch receptor activation

hCMECs were used to study the loss of endothelial markers and the gain of mesenchymal characteristics in response to oscillatory shear stress. mRNA expression analysis revealed the upregulation of mesenchymal CNN1 (3.21 fold) and TAGLN (3.37 fold) in cells exposed to oscillatory shear stress when compared with static shear stress. The mRNAs of key EndMT markers SNAI1 and SNAI2 were also overexpressed by 3.59 and 3.83 fold, respectively, in oscillatory flow-exposed cells (Fig. 4A). As expected, Slug expression was marginally higher compared with SNAI1 in cells exposed to oscillatory flow, corroborating the findings in cAVM.

Fig. 4

Differential expression of EndMT and cell adhesion markers by oscillatory flow in microvascular endothelial cells. A mRNA expression of genes coding for SNAI1, SNAI2, Calponin1, Transgelin, integrin α9 subunit, integrin β1 subunit, VE-cadherin, and N-cadherin upon exposure of hCMEC/d3 to oscillatory flow (n = 3). The mRNA expression folds of SNAI1, SNAI2 (Slug), Calponin1, transgelin, integrin α9 subunit, and N-cadherin were significantly higher after exposure to oscillatory flow for 24 h. Integrin β1 subunit and VE-cadherin were found to be downregulated in cells exposed to oscillatory flow. mRNA fold values were calculated relative to static control. All data were normalized with GAPDH expression and are given as relative to static control. B hCMEC/d3 exposed to oscillatory flow for 24 h increased nuclear SNAI1/2 and N-cadherin expression, while integrin α9/β1 was highly downregulated, indicating active EndMT and reduced cell adhesion among cells exposed to oscillatory flow (scale bar 20 µM, magnification ×40). C Mean fluorescence intensity was plotted as the average fluorescence intensity ± SD of five fields per flow condition and from three biological replicates. PF indicates parallel uniform shear stress and OF represents oscillatory shear stress. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 versus respective static or parallel uniform shear-treated groups. ns not significant

Immunofluorescence analysis was performed for SNAI1/2 in hCMECs exposed to static, parallel, and oscillatory shear stress. SNAI1/2 was overexpressed and localized to the nucleus in cells exposed to oscillatory flow compared with other flow regimes (Fig. 4B, (Additional file 1. Fig. S2B). MFI analysis indicated a 3.8 and 3.1 fold increase in SNAI1/2 expression, when compared with static and parallel flow, respectively (Fig. 4C).

To assess the involvement of the Notch pathway in shear stress-induced EndMT in hCMECs, we used GSI DAPT (0.5 µM) and RO4929097 (250 nM) during oscillatory flow and monitored the expression of EndMT factors. Inhibition of Notch receptor cleavage abrogated the downregulation of SNAI1/2 observed during oscillatory stress by approximately 60% with DAPT and 80% with RO4929097 (Fig. 5A, B, Additional file 1: Fig. S2B). Taken together, these data indicate that oscillatory shear stress-induced EndMT requires Notch signaling. Gamma-secretase inhibitor RO4929097 is very efficient in reducing mesenchymal phenotype in endothelial cells during altered flow conditions.

Fig. 5

Gamma-secretase inhibitors (GSI) modulate oscillatory shear-induced EndMT and cell invasiveness. A, B Inhibition of Notch receptor cleavage by 500 nM DAPT and 250 nM RO4929097 prevented the overexpression of SNAI1/2 and N-cadherin in hCMEC/d3 exposed to oscillatory flow for 24 h. Integrin α9/β1 expression was augmented by GSI even in the continuing presence of oscillatory shear stress. EndMT-associated molecular changes were significantly reduced in the presence of RO4929097 (scale bar 20 µM, magnification ×40). C Fluorescent microscopic images of invaded cerebral microvascular endothelial cells, with prior exposure to control and oscillatory flows, at the lower surface of the transwell Matrigel-coated filter membrane stained with 5 μg/ml of nuclear stain Hoechst 33342 in the presence and absence of each inhibitor (250 nM RO4929097 and 500 nM DAPT) (scale bar 100 µM, magnification ×10). D Invasion assay was done in triplicate, and invaded cells were counted by ImageJ and plotted graphically. As noted in the graph, cells exposed to 24 h oscillatory flow invade faster than normal endothelial cells. The presence of DAPT (p < 0.001) and RO4929097 (p < 0.0001) effectively reduced the percentage of invaded oscillatory shear-exposed endothelial cells when compared with cells exposed to oscillatory flow alone. OF represents oscillatory shear stress. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 versus respective static or parallel uniform shear-treated groups

RO4929097 promotes cell adhesion and reduces invasiveness in cells exposed to oscillatory flow

To corroborate the findings on dysregulation of cell adhesion factors in cAVM tissues, we studied, by qRT-PCR, the mRNA expression of genes coding for endothelial VE-cadherin, mesenchymal N-cadherin, and integrin subunits α9 and β1 in endothelial cells exposed to various flow regimes. N-cadherin (2.85 fold) and integrin α9 (2.06 fold) were upregulated, while VE-cadherin (0.71 fold) and integrin β1 (0.54 fold) were downregulated at mRNA level in cells exposed to oscillatory shear stress (Fig. 4A).

Immunofluorescence assay of these adhesion factors in cells exposed to shear stress conditions indicated that there is an upregulation of N-cadherin in cells exposed to oscillatory shear stress versus cells exposed to static and parallel shear stress. Activated integrin α9/β1 was found significantly reduced in cells exposed to oscillatory flow for 24 h (Fig. 4B, C).

DAPT and RO4929097 significantly reduced the N-cadherin in cells exposed to continuing oscillatory flow. Interestingly, integrin α9/β1 expression was augmented in the presence of DAPT and RO4929097 by 45.5% and 56.7%, respectively, when the MFI of confocal images were analyzed (Fig. 5A, B).

Increased cell invasiveness is an important consequence of EndMT. To study the invasive properties of the hCMECs in response to flow regimes, we used a Matrigel transmembrane invasion assay, where migration of cells toward a source of serum attractant, across a membrane coated with Matrigel, was measured. Prior exposure of cells to oscillatory shear stress for 24 h induced an increase in endothelial invasiveness, which was substantially blocked by both DAPT and RO4929097 (Fig. 5C). RO4929097 effectively prevented cell invasiveness compared with DAPT (Fig. 5D). This finding suggests that pharmacological intervention targeting Notch signaling can effectively reduce blood flow-induced invasion capacity and EndMT of vascular endothelial cells.