PATJ downregulates after IS and lower expression levels associate with functional independence. The risk allele G of rs76221407 confers higher PATJ mRNA expression.
We first determined in leukocytes how PATJ mRNA levels are modulated at 24 hours (hr) after IS. 50 IS patients from the GODS cohort2 were grouped depending on their 3-months mRS scores. Functional independent patients (mRS ≤0-2, (n = 25)) were compared with patients with marked disability (mRS≥4, (n = 25)). PATJ mRNA levels decay after 24 hr post-IS, but this depletion is significantly more pronounced on those patients with good functional outcome (Fig. 1A).
We next checked whether the risk allele G of SNP rs76221407 influence gene expression. For that purpose, we used the retrospective cohort GRECOS19 of 77 healthy controls. Due to the low frequency of the SNP rs76221407, estimated to be 0.028 (GnomAD), only five individuals presented heterozygosis (AG genotype), while the other 72 were homozygous (AA genotype). We found that the carriers of the G allele at rs76221407 had significantly higher expression of PATJ mRNA in blood (Fig. 1B).
PATJ mRNA expression was also determined in brain homogenates from mice (n = 4/time point) subjected to permanent middle cerebral artery occlusion (pMCAo). 24 hr after ischemia PATJ mRNA expression was significantly lower in the ipsilateral ischemic region (IL) once compared to its contralateral control (CL) (Fig. 1C).
These results demonstrate that PATJ is downregulated both in blood and brain after IS and suggest that PATJ depletion in response to IS could be beneficial, since patients with protective mRS scores have lower PATJ expression levels. Moreover, individuals harboring the risk G allele at rs76221407 present higher PATJ expression.
In human brain ECs PATJ depletion is mediated by the stabilization of Hypoxia Inducible Factor-1α (HIF-1α).
To further explore PATJ expression in response to hypoxia, human brain microvascular ECs (hCMEC/D3) were subjected to HIF-1α-dependent hypoxia, incubating cells with the hydroxylase inhibitors cobalt chloride (CoCl2, 250 µM) or dimethyloxalylglycine (DMOG, 300µM) for 8, 24 and 48 hr. HIF-1a-independent chemical hypoxia was tested by treating cells with the cytochrome c oxidase inhibitor sodium azide (NaN3, 1mM) for 16 hr. PATJ expression was studied by western-blot (WB). Five main PATJ species (230, 200, 135, 80 and 60 kDa) were identified in hCMEC/D3 using the PATJ antibody against the 800–1000 domain20. After 8 and 24 hr of HIF-1α stabilization, the only PATJ specie downregulated was the 80kDa, but prolonged (48 hr) CoCl2 incubation also translated in depletion of the PATJ higher molecular mass species of 230, 200 and 135 kDa. Azide exposure did not alter PATJ levels (Fig. 1D).
PATJ knockdown (KD) causes loss of polarity and disruption of TJs in hCMEC/D3 cells.
To deeper study the molecular events underlying PATJ downregulation in response to IS, we generated hCMEC/D3 stably depleted of PATJ using lentiviral particles harboring shRNA against PATJ. Several clones were generated (Suppl. Figure 1A), three of which were selected for further analysis due to a progressive depletion of the PATJ species (Fig. 2A). PATJ KD1 clone only downregulated the 135kDa (-10,2% ± 1,5) and the 80kDa species (-64,4% ± 1), PATJ KD2 had depleted the 230 kDa (-17% ± 2), 135 kDa (-49,6% ± 3) and the 80 kDa (-83,9% ± 1,5) species, and PATJ KD3 was the only clone with depletion of all five PATJ species (230 kDa (-94,5% ± 2), 200 (-66,7% ± 3), 135 (-42,6% ± 4), 80 (-79,7± 0,7) and 60 kDa (-92,6% ± 1,6) (Fig. 2B).
Since PATJ contributes to TJs formation5, 20, 21, 22, we first determined whether PATJ KD affected TJs in hCMEC/D3 cells. Immunofluorescence assay for the TJ marker zonula occludens-1 (ZO-1) revealed complete disappearance of ZO-1 staining at cell-cell contacts in PATJ KD3 clone (Fig. 2C). Loss of TJs in this clone correlated with downregulation of ZO-1, occludin and claudin-11 (Fig. 2D). Interestingly, the two intermediate PATJ KD clones (KD1 and KD2) presented different immunoreactivity patterns than CTL cells for ZO-1 and occludin, with higher molecular species of occludin and lower species at ZO-1 (Fig. 2C). TJs disruption in PATJ KD3 cells was further confirmed with a permeability assay that exhibited one thousand times higher permeability to the Dextran tracer than CTL (Fig. 2E). Potentially higher sensibility of PATJ KD cells to tracer toxicity or reduced general cell survival due to PATJ depletion were discarded since EC viability, determined by MTT assay, demonstrated no differences between PATJ KD3 and CTLs (Suppl. Figure 1B).
PATJ knockdown leads to downregulation of the Notch, PI3K/Akt and Hippo pathways and translates into Endothelial to Mesenchymal Transition.
We noted that hCMEC/D3 cell morphology changed drastically due to PATJ KD, making the cells more elongated and spindle-like (Fig. 2C). This fact prompted us to hypothesize that PATJ depletion could induce a severe transcriptome reprogramming. Gene expression profiles, determined through the microarray platform, detected 341 differentially expressed genes identified through fold-change (FC) and P-value filtering adjusted using Bonferroni correction (FC ≥ 2 or FC ≤ -2 and Padjusted < 0.05). Cluster analysis is shown in Fig. 3A. 137 genes were upregulated and 204 downregulated in PATJ KD cells compared with CTLs (Suppl. Table 1 and Suppl. Figure 2). Gene ontology (GO) analysis identified 124 categories that reached False Discovery Rate (FDR) significance (Padjusted<0.01) and highlighted as major modulated biological processes the actin cytoskeleton organization, endothelial to mesenchymal transition (EndMT) and the endoplasmic reticulum unfolded protein response (Fig. 3B and Suppl. Table 2). The major signaling pathways modulated in PATJ KD cells were Notch, Transforming Growth Factor-β (TGF-β)/Bone Morphogenic Proteins (BMP) signaling, PI3K/Akt, VEGFA-VEGFR2 signaling and Hippo (Fig. 3C, E and Suppl. Tables 2–3). These pathways are strongly related to each other, converging in the regulation of EndMT and angiogenesis (Fig. 3 and Suppl. Tables 2–3).
PATJ KD cells express mesenchymal markers, switch off Notch, PI3K/Akt and Hippo signaling.
To validate the expression array data and to deeper investigate whether PATJ KD promotes EndMT and angiogenesis, protein markers23, 24 were measured by WB. The endothelial marker VE-Cadherin was upregulated in the intermediates PATJ KD1 and KD2 and severely downregulated in PATJ KD3. The expression of PECAM-1/CD31, another endothelial marker, differed within the different PATJ KD clones, being downregulated in PATJ KD1 and PATJ KD3 and upregulated in PATJ KD2. Conversely, the mesenchymal marker Vimentin progressively upregulated in PATJ KD cells, inversely correlating with PATJ levels. Another mesenchymal marker, Fibroblast-specific protein 1 (FSP1/S100A4) was only expressed by PATJ KD3 (Fig. 4A). Loss of cell polarity has been associated with disruption of barrier epithelia and proteolytic cleavage of glycocalyx components as Mucin 1(MUC1)25. Its cytoplasmic tail (MUC1-C) is a signal transducer involved in EndMT and hypoxia-driven angiogenesis26, 27. MUC1-C steady-state levels significantly increased in the three PATJ KD clones, with higher accumulation in the two intermediate PATJ KD1 and KD2 clones (Fig. 4B). Expression of matrix metalloproteinases (MMPs) is another major attribute that ECs acquire after undergoing EndMT24, 28. MMP-3, MMP-2, MT1-MMP, MMP-7 upregulated in PATJ KD clones, as well as the mesenchymal markers tissue inhibitors of MMPs, TIMP-1, 2 and 3 (Fig. 4C). These results suggest that the three PATJ KD clones analyzed were at different stages of EndMT, being PATJ KD1 and KD2 at the transient state, which is referred to “Partial EndMT” or “endothelial mesenchymal activation” (EndMA)24, 28, and only the PATJ KD3 cells acquired a complete mesenchymal phenotype.
We next investigated the pathways involved in triggering EndMT-angiogenesis upon PATJ depletion. Both, Notch and PI3K/Akt signaling, central coordinators of EndMT-angiogenesis29, 30, 31, 32, were off in PATJ KD3 cells with the mesenchymal phenotype (Fig. 4D, E). By contrary, the transient state of the intermediate PATJ KD clones partially downregulated the PI3K/Akt, as evidenced with upregulation of GSK3b and downregulation of b-catenin (Fig. 4D), and differed in Notch signaling, being partially off in PATJ KD1, with depletion of the transcription factors RUNX3 and HES1, and maintained on in PATJ KD2 (Fig. 4E). These results suggest a modulatory role of PI3K/Akt and Notch pathways along the EndMT process induced by PATJ depletion, being partially active at the transient stages and switched off once the mesenchymal phenotype is achieved.
Hippo signaling, another pathway significantly modulated due to PATJ depletion, has also been described as a driver of EndMT-mediated angiogenesis 33, 34. When the pathway is activated, its downstream transcriptional co-activators YAP (Yes-associated protein) and TAZ (Tafazzin) are phosphorylated and sequestered in the cytosol or degraded by the proteasome. When Hippo is inhibited, YAP/TAZ may enter the nucleus, in where activate the TEA domain family members (TEAD) transcription factors promoting gene expression, including genes involved in EndMT34.
We first checked the steady-state levels of YAP/TAZ, incubating cells with or without the proteasome inhibitor MG132 (Fig. 4F). No differences in YAP levels were found in PATJ KD1, but PATJ KD2 and KD3 cells depleted YAP levels by almost 50% (Fig. 4F, G). Regarding TAZ, a progressive depletion was observed in the three PATJ KD clones, being more evident after MG132 treatment (Fig. 4F, H).
PATJ KD cells recruit to the nucleus the signal transducer MUC1-C and the transcriptional modulators YAP/TAZ, b-catenin and ZEB1.
We next evaluated whether YAP/TAZ could be translocated to the nucleus upon PATJ silencing. Intermediate PATJ KD clones had higher amounts of YAP in the nuclear fraction, contrarily to the low levels observed in PATJ KD3 cells (Fig. 5A, B). These results were also confirmed by YAP immunofluorescence (Fig. 5F). By contrary, TAZ was enriched in the nuclear fraction of all three PATJ KD clones (Fig. 5A, C).
Since MUC1-C has been shown to promote the nuclear translocation of YAP/b-catenin complexes 35, we also checked its subcellular distribution. Interestingly, MUC1-C was mostly recruited to the nucleus of PATJ KD clones (Fig. 5A, D). And as expected, the distribution pattern of b-catenin was like the one observed for YAP, being enriched in the nuclear fraction of the two intermediate PATJ KD clones (Fig. 5A, E). The transcriptional repressor ZEB1, a known EndMT inducer36, which expression is controlled by MUC1-C37, was also enriched in the nuclear fraction of the intermediate PATJ KD clones (Fig. 5A, F). These results suggest that the EndMA phenotype of intermediate PATJ KD clones is achieved by MUC1-C/YAP/TAZ/b-catenin/ZEB-nuclear localization. By contrary, cells that complete the mesenchymal transition, as PATJ KD3, only maintain nuclear MUC1-C/TAZ.
PATJ depletion causes actin cytoskeleton remodeling.
YAP/TAZ regulate actin remodeling at filopodia and cell-cell junctions34, and actin cytoskeleton remodeling is necessary for EndMT and angiogenesis38. Moreover, “regulation of actin cytoskeleton organization” was the most significant GO biological process associated with PATJ knockdown (Fig. 3B). Therefore, we investigated the actin cytoskeleton dynamics in the PATJ KD clones. Immunofluorescence analysis using labelled phalloidin evidenced loss of stress fibers, but increased membrane ruffles and lamellipodia in the PATJ KD cells, although the two intermediate PATJ KD clones (KD1 and KD2) tended to maintain the cortical endothelial organization of the actin filaments (Fig. 5G). Loss of filamentous actin (F-actin) and concurrent increase in globular (G-actin) was further demonstrated by quantitative WB analysis in the three PATJ KD clones (Fig. 5H, I). Interestingly, key regulatory proteins of the actin cytoskeleton as the Rho A inhibitor Rho GTPase Activating Protein 6 (ARHGAP6) and Myosin Light Chain (MLC) were severely downregulated only in PATJ KD3 cells (Fig. 5J), suggesting a complex pattern of actin dynamics regulation.
PATJ depletion induces endothelial activation and vascular inflammation through nuclear recruitment of MUC1-C and YAP.
Since nuclear MUC1-C is also known to interact with the pro-inflammatory transcription factor NF-κB (nuclear factor kappa B) p65 39 and cytoplasmic YAP is needed to prevent TRAF6 (tumor necrosis factor receptor-associated factor 6)-mediated NF-κB activation40, we checked endothelial activation dependent on PATJ levels, studying the expression of the adhesion molecules ICAM-1 (intercellular adhesion molecule-1) and CD44. ICAM-1 was highly upregulated in PATJ KD2 cells and downregulated in PATJ KD3, meanwhile CD44 was detectable in PATJ KD2 and highly expressed in PATJ KD3 cells (Fig. 6A). TRAF6 levels were also upregulated in PATJ KD3 cells. By contrary, the anti-inflammatory heat shock protein ab-crystallin41 was upregulated in PATJ KD1 and highly expressed in PATJ KD2 (Fig. 6A), suggesting that in these intermediate PATJ KD clones ab-crystallin may contribute to maintain the transient EndMA phenotype. We next measured cytokines secretion in the culture media. Interestingly, only PATJ KD3 cells produced high levels of interleukin-6 (IL-6) under regular growth conditions (Fig. 6B), meanwhile IL-8 production was detectable in PATJ KD2 and KD3 cells (Fig. 6C). None of the other cytokines measured, IL-1b, IL-15, IL-18 and TNF-α, could be detected (data not shown). Once cells were challenged with lipopolysaccharide (LPS), all clones secreted equal amounts of both IL-6 and IL-8 after 6 hr of incubation (Fig. 6D, E), demonstrating that all cells kept the ability to secrete both IL-6 and IL-8 once Toll-like receptor 4 is stimulated with LPS. Our results showed that only the PATJ KD3 cells with the mesenchymal phenotype constitutively secrete IL-6, probably as an autocrine or paracrine feedback loop to maintain the mesenchymal phenotype, as previously shown in other cellular systems42, 43, 44.
The EndMA transient state of intermediate PATJ KD clones favors cellular migration and tubular network formation.
Vascular regeneration is accompanied by vessel elongation, which is promoted by EC migration. During this process the expression of EndMT-related transcription modulators, such as ZEB1, are upregulated36, 37. Thus, the transient EndMA state is thought to exhibit reversible changes during vascular remodeling, such as transient occurrence and return to ECs28. To demonstrate that PATJ depletion exacerbates cellular migration, a scratch wound healing assay was performed. As expected, intermediate PATJ KD clones with the EndMA phenotype (KD1 and KD2) significantly increased cell migration 4 and 8 hr after the scratch (Fig. 7A, B), meanwhile PATJ KD3 was unable to heal the wound during the 48 hours that the experiment lasted, probably due to the severe depletion of key regulatory proteins of the cytoskeleton, as MLC and ARHGAP6 (Fig. 5D).
Cell tube formation assay was performed to prove that the EndMA state of intermediate PATJ KD clones allows cells to conserve the ability to organize forming tubular networks, once grown on Matrigel. In fact, PATJ KD1 had even improved sprouting ability than CTLs (Fig. 7C-D). By contrary, the mesenchymal PATJ KD3 cells, completely lost their potential for endothelial morphogenesis and were unable to form tubular networks. These results confirmed that the EndMA state of intermediate PATJ KD clones allows cells to migrate and tubulogenesis.