The VE-cadherin/AmotL2 mechanosensory pathway suppresses aortic inflammation and the formation of abdominal aortic aneurysms

Arterial endothelial cells (ECs) have the ability to respond to mechanical forces exerted by fluid shear stress. This response is of importance, as it is protective against vascular diseases such as atherosclerosis and aortic aneurysms. Mechanical forces are transmitted at the sites of adhesion to the basal membrane as well as cell-cell junctions where protein complexes connect to the cellular cytoskeleton to relay force into the cell. Here we present a novel protein complex that connects junctional VE-cadherin and radial actin filaments to the LINC complex in the nuclear membrane. We show that the scaffold protein AmotL2 is essential for the formation of radial actin filaments and the flow-induced alignment of aortic and arterial ECs. The deletion of endothelial AmotL2 alters nuclear shape as well as subcellular positioning. Molecular analysis shows that VE-cadherin is mechanically associated with the nuclear membrane via binding to AmotL2 and Actin. Furthermore, the deletion of AmotL2 in ECs provokes a pro-inflammatory response and abdominal aortic aneurysms (AAA) in the aorta of mice on a normal diet. Remarkably, transcriptome analysis of AAA samples from human patients revealed a negative correlation between AmotL2 expression and aneurysm diameters, as well as a positive correlation between AmotL2 and YAP expression. These findings provide a conceptual framework regarding how mechanotransduction in the junctions is coupled with vascular disease.


Introduction
The blood vessel wall is lined with a thin layer of vascular endothelial cells (ECs), which form a barrier between the blood and tissues. These cells differ in biochemical characteristics depending on their localization in arteries or veins, as well as on the organ in which they reside. The endothelium is continuously exposed to the shear stress exerted by the blood flow. Understanding how ECs respond to shear stress is of importance as it has implications for the development of vascular diseases. Indeed, since the 1870´s it has been postulated that mechanical stress exerted on the blood vessel wall may be a trigger of atherosclerosis. Also, low wall shear stress has been associated with abdominal aortic aneurysm (AAA) rupture (1). AAA is characterized by localized medial and adventitial inflammation and dilatation of the abdominal aorta and is prevalent in men over 65 with significant morbidity and mortality (2). By contrast, areas of laminar flow appeared relatively protected against the development of the inflammatory disease (3).
To explore how the mechanotransductive pathways mediate protection or activation of the vascular disease is of clear importance (4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18). To date, several mechanosensory pathways have been identified that relay external mechanical forces to the endothelial lining (19). In vitro, flow-induced endothelial alignment is dependent on the activation of GTPases and consequent actin reorganization. In vivo, it has been shown that endothelial junctional protein complexes including PECAM-1, VE-cadherin, and VEGFR2 play an important role in the adaptive response to shear stress (18). However, it is still not clear how mechanical forces exerted by the blood flow are transmitted from the junctions via the cytoskeleton into the cell.

Studies of the (Angiomotin) Amot protein family may provide important insights into this
aspect. This is a family of scaffold proteins that link membrane receptors to the actin cytoskeleton, polarity proteins, and are implicated in modulating the Hippo pathway (20)(21)(22)(23)(24)(25)(26).
We have recently shown that one of its members, AmotL2 (p100 isoform), is associated with the VE-cadherin complex in ECs and E-cadherin in epithelial cells (25,27). Silencing of AmotL2 in zebrafish, mouse or cells in vitro results in a loss of radial actin filaments that run perpendicular to the outer cell membrane. These actin filaments mechanically connect cells via binding to junctional cadherins, and thereby transmit force. Conditional silencing of AmotL2 in the endothelial lineage of mice inhibits expansion of the aorta during the onset of circulation resulting in death in utero at embryonic day 10 (25). In this report, we have analyzed the role of AmotL2 in controlling junctional and cytoskeletal components during the alignment of arterial ECs exposed to laminar flow. Here we present a novel mechanical transduction pathway active in arteries that is protective against vascular inflammation as well as the formation of AAAs.

Amotl2 is essential for arterial alignment.
We analyzed the expression pattern of AmotL2 in mouse descending aorta (DA) and the inferior vena cava (IVC) as shown in Fig. 1a,b. ECs of the DA were typically elongated and aligned with the direction of blood circulation and contained radial actin filaments that were connected to the cellular junctions (Fig. 1c). AmotL2 localized to EC junctions as previously reported (25). In contrast, ECs of the IVC exhibited a more rounded cellular shape with no or few detectable radial actin filaments as well as a significantly lower expression level of AmotL2. Box plots in Fig. 1d-f represents the statistically significant difference between the DA and IVC with regards to AmotL2 expression, cellular shape and the presence of radial actin filaments. In addition, we mapped AmotL2 expression in mouse retinal vasculature at different ages (Supplementary Fig. 1). AmotL2 was expressed at similar levels in arteries and veins of post-natal day 6 mice, whereas in mature mice (three months) Amotl2 was primarily expressed in retinal arteries. Interestingly, AmotL2 was barely detectable in blood vessels of mice older than 15 months.
We have previously shown that AmotL2 is required for the formation of radial actin filaments both in epithelial and endothelial cells (25,27). The preferential expression of AmotL2 in EC of the aorta raised the possibility that AmotL2 controlled arterial EC shape via the formation of radial actin fibers. To address this question, we used a genetic deletion approach to silence amotl2 gene expression specifically in the endothelial lineage as previously reported (28). In this model system, amotl2 flox/flox mice were crossed with Cdh5(PAC) CreERT2 transgenics as well as ROSA26-EYFP reporter mice (25), from now on referred to as amotl2 ec/ec . This crossing enables efficient inducible conditional recombinase expression and subsequent amotl2 knockout in ECs (amotl2 ec-/ec-) after tamoxifen injections as well as quantification of recombination efficiency by YFP expression (Supplementary  Fig. 2a,b). Adult mice (7-9 months old) were euthanized one month after tamoxifen injections, and aortae were dissected and analyzed by whole-mount immunostaining.
Inactivation of AmotL2 in the DA resulted in the loss of radial actin filaments and altered cell shape (Fig. 2a, quantification in 2b,c). No detectable changes in the actin cytoskeleton and cellular shape were observed in the endothelium of the IVC of amotl2 ec-/ecmice. This observed phenotype change appeared to be arterial-specific as similar effects were observed in arterial, but not venous ECs of other organs such as the urinary bladder (Fig. 2d, The nucleus is the largest organelle of the EC. As such it is exposed to the hemodynamic drag by the blood flow. In response to shear stress, EC as well as EC nuclei elongate, and nuclei orient themselves relative to the direction of flow. Nuclear positioning as well as alignment has previously been shown to be dependent on the association to microfilaments as well as the tubulin network (29). In amotl2 ec+/ec+ mice, nuclei of ECs of the DA were elongated and orientated in parallel with cell alignment in the direction of blood flow.
However, in AmotL2 deficient ECs, the nuclei were more rounded with irregular shapes and positioned close to the cell edge downstream of the flow direction (Fig. 2g, quantification in 2h,i). These changes in nuclear shape and positioning were again not observed in the IVC (Fig. 2j, quantification in 2k,l). Taken together, these data showed that AmotL2 is required for EC elongation as well as positioning of the EC nucleus.
AmotL2 expression is required for arterial response to flow.
Next we investigated whether AmotL2 is required for arterial endothelial compliance to laminar flow in vitro. For this purpose, we used a short hair-pin Lentiviral approach to deplete AmotL2 in Human Aortic ECs (HAoEC, Fig. 3a). No differences between control and AmotL2 depleted cells in cellular and nuclear shape were detectable under static conditions ( Fig. 3b-d). To recapitulate arterial flow conditions, cells were exposed to 14dynes/cm 2 for 48h in a flow chamber as described in materials and methods. Control HAoECs exhibited an elongated phenotype and aligned in the direction of flow; however, the depletion of AmotL2 resulted in failure to elongate and align in the direction of flow ( Fig. 3b, quantification in 3e,f). We could further show that AmotL2 was required for controlling nuclear shape and positioning. Consistent with the cellular shape change, nuclei also exhibited a rounder shape and could not maintain the positioning at the center of the cells when compared to control HAoECs (Fig. 3g,h).
This complex consisting of Sun-domain proteins (SUN-1 and -2) and KASH domain proteins (Nesprin-2) connects to Lamin A/C of the nuclear lamina. Therefore, we next investigated a possible connection between VE-cadherin, AmotL2, actin and the LINC complex.
We used a Co-IP approach to obtain AmotL2 associated immunocomplexes from both murine endothelial cell line (MS1) and primary Bovine Aortic Endothelial (BAE) cells. By spectrometry analysis (raw data in Supplementary Table 1,2), we identified cellular membrane protein VE-cadherin and α-, β-, δ-catenin, as well as the nuclear laminal proteins such as SUN2 and Lamin A, in all triplicate IP samples from MS1 cells (Fig. 4a).
Consistently, the majority of key membrane proteins were recognized in BAE AmotL2 immunoprecipitates ( Supplementary Fig. 3a). Mouse lungs consist of approximately 10-20% ECs. We performed immunoprecipitation using AmotL2 antibodies and could verify that Amotl2 associates with VE-cadherin, B-catenin, actin, and Lamin A also in the murine lung tissue in vivo (Fig. 4b). It was not possible to analyze Nesprin-2 by western blot due to its high molecular mass (<800 kDa). The association of AmotL2 was dependent on myosin contraction and actin filaments as treatment with the myosin II inhibitor blebbistatin or the actin polymerization inhibitor cytochalasin D both decreased the association of AmotL2 to Lamin A/C and SUN2 (Fig. 4c). In addition, immunoprecipitation of VEcadherin confirmed the association to AmotL2 and Lamin A in HAoECs (Fig. 4d).
Interestingly, shRNA depletion of AmotL2 protein in ECs abrogated the association of VEcadherin to both actin and Lamin A (Fig. 4d).
The higher expression levels of AmotL2 detected in the DA as compared to the IVC in vivo raised the question of whether AmotL2 associated complex is shear stress responsive. We placed primary Human Umbilical Venous ECs (HUVECs) in a 15cm culture dish on an orbital shaker to apply the circulatory flow to the cells (schematic in Supplementary Fig.   3b). As previously described (33), the alignment of the cells at the periphery area was observed after 96h ( Supplementary Fig. 3c,d). Analyses of the AmotL2 IP revealed a dramatically increased binding with VE-cadherin, actin and nuclear membrane proteins compared to the static condition (Fig. 4e).
Taken together, MS and IP data suggest a model where VE-cadherin/AmotL2 forms a complex that is connected to the nuclear LINC complex via actin filaments (schematic in

Deletion of AmotL2 promotes vascular inflammation.
EC alignment and cytoskeletal reorganization in response to laminar blood flow is protective against inflammation (3). The change in the mechano-response of amotl2 ec-/ec EC raised the question whether this was accompanied by a pro-inflammatory phenotype. The aortic arch is exposed to turbulent blood flow and the descending aorta is exposed to laminar flow. For this reason, we analyzed these areas of the aorta separately (Fig. 5a). mRNA isolated from DAs in both amotl2 ec+/ec+ (n=3) and amotl2 ec-/ec-(n=5) mice were analyzed by mRNA-seq. Due to high variability between individual mice, only 65 genes were identified to be differentially expressed (adjusted p value<0.05) between amotl2 ec+/ec+ and amotl2 ec-/ecgroups (Supplementary Table 3). However, those genes were obviously enriched to immuno-related GO terms, such as "Neutrophil activation involved in immune response", "inflammatory response" and "regulation of immune effector process" (Fig. 5b). Gene expression profile analysis on MS1 cells in vitro treated with control or AmotL2 siRNA was also performed by RNA-seq. Interestingly, we did not observe any up-regulation of inflammatory gene signatures in vitro (Supplementary Fig. 4a and Supplementary Table   4), indicating that the in vivo environment was required for this to occur. We therefore performed whole-mount immunostaining to analyze the presence of CD45 + inflammatory cells in the aortic arch. The whole mount immunofluorescence staining showed the presence of spindle-like CD45 + cells in the inner curvature of the aortic arch in wild-type animals and in arterial bifurcations (Fig. 5c,d). Analyses of the amotl2 ec-/ecmice 1-month postinjection of tamoxifen revealed the infiltration of CD45 + round monocytes in the outer curvature of the aortic arch (Fig. 5c,d). The CD45 + monocyte-like cells were detected in the sub-endothelial compartment of the aorta (Fig. 5e,f). In total, 45% (9/20) amotl2 ec-/ecarches contained more than one CD45 + lesion.
Next, we quantified mRNA expression of inflammatory markers in DAs. The mRNA expression level analyzed by TaqMan qRT-PCR showed a significant increase of inflammatory markers, such as Il6, Tnf, and Cd68a in amotl2 ec-/ec-DA, suggesting a provocation of a general immune response. Interestingly, up-regulation of pro-inflammatory markers was more pronounced in male animals ( Fig. 5g-j).
In response to pathogenic stimuli, expression of various leukocyte adhesion molecules can be augmented as the initial sign of vascular inflammation, such as ICAM-1 and VCAM-1.
Indeed, in amotl2 ec-/ec DA, an increase in Vcam-1 expression on transcriptional level was detected by qPCR, but the change of Icam-1 was less obvious (Supplementary Fig. 4b,c), revealing the activation of endothelium of the DAs in the absence of AmotL2. Remarkably, Vcam-1 up-regulation was also more profound in male mice (Supplementary Fig. 4b).
Once leukocytes attach and migrate into underlying intima, the process of lesion formation begins. This requires the participation of chemoattractant cytokines such as Ccl2, also known as monocyte chemoattractant protein-1 (MCP-1), which is especially important for the recruitment of mononuclear immune cells. Compared to the amotl2 ec+/ec+ aortic tissues, amotl2 deletion caused an increase of Ccl2 expression in the descending regions ( Fig. 5j), as well as Cxcl10, which is able to selectively attract NK cells as well as T and B lymphocytes. (Supplementary Fig. 4d). However, other chemokines such as Ccl5 had no significant changes (Supplementary Fig. 4e), suggesting that the induced inflammatory response was mainly affecting innate immune cells.
By contrast, the T and B cell markers Cd4, Cd8, and Cd19, as adaptive immune markers, showed no differences in expression level of these markers on average was detected in the DAs with or without AmotL2 (Supplementary Fig. 4f-h).
In conclusion, the expression profile of inflammatory-related genes indicates that AmotL2 depletion results in an innate immune response rather than an activation of adaptive immunity. Of note is the preferential upregulation of pro-inflammatory markers in amotl2 ec-/ecmales.

AmotL2 depletion promotes the formation of abdominal aortic aneurysms in male mice.
Aortic inflammation is a predictor of the development of arterial aneurysm(34).
Inflammatory aortic aneurysms account for up to 10% of all AAA cases (35). After dissection of the aorta, we could detect the formation of abdominal aortic aneurysms (AAA, dilatation>1.5 times normal size) in the proximity of the renal arterial branch (Fig. 6a), but not in ascending or thoracic aortae. Interestingly, 20 % of male amotl2 ec-/ec-(5/25) mice developed an AAA; however, no aneurysm was detected in the females (0/20). Furthermore, no aneurysms were observed in the amotl2 ec+/ec+ mice (36 mice, 20 male and 16 female).
Imaging analysis of a typical AAA revealed damage to the endothelium as well as the vessel wall ( Fig. 6b,c). The insult to the vessel wall was further verified by histological sectioning.
The intima of the vessel wall was infiltrated by CD45+ leukocytes (Fig. 6d) and elastin fibers were degraded (Fig. 6e).

AmotL2/YAP correlation in human AAA samples.
Next, we assessed whether AmotL2 gene expression could be correlated to disease progression in human AAA patients. We analyzed mRNA expression in surgically resected materials from both healthy donors and patients diagnosed with AAA and undergoing aneurysm repair at the Karolinska University Hospital. mRNA samples were taken from both medial and adventitial layers of the intact aorta (13 donors) or AAA tissues (35 patients). Interestingly, AmotL2 expression level in the media tissue (containing endothelial cells, schematic in Fig. 7a) in AAA samples was significantly down regulated compared to the intact aortic media, which was not found in adventitia tissues (Fig. 7b). The expression pattern of the first exon of AmotL2 was specifically analyzed since this exon is specific for the p100-AmotL2 isoform that we have previously shown interacts with VE-cadherin (25).
In line with the studies in mice, there was an inverse correlation between the expression of the first exon of AMOTL2 and the luminal/external diameter of AAA (Fig. 7c).
YAP (Yes Associated Protein 1) has been reported to act as a rheostat of external mechanical force (36). Of interest is that YAP binds to the AmotL2 promoter and thus activates its transcription (37). Consistent with these findings, a significant positive correlation was detected between AMOTL2 and YAP mRNA levels (r = 0.800, P < 0.0001) in media tissues containing endothelium in AAA patients of both genders (Fig. 7d). We also found a weaker positive correlation of AMOTL2/YAP in adventitia tissues that did not contain ECs (r=0.592, p=0.0002). A robust correlation of the well-established YAP target genes CTGF and CYR61 with AMOTL2 was also observed, as shown in Fig. 7e and 7f, respectively. C-FOS has been reported to bind to the promoter and drive expression of the p60 AMOTL2 isoform. This isoform is primarily activated by severe hypoxia in ischemic tissues (25). However, no correlation between AMOTL2 and C-FOS mRNA expression could be detected (Supplementary Fig. 5a). YAP and TAZ are paralogs with overlapping function as transcriptional regulators of the Hippo pathway. However, there was no significant correlation between AMOTL2 and TAZ expression, suggesting that YAP is primarily driving AMOTL2 expression in the human aortic tissues (Supplementary Fig.   5b). In addition, there was no correlation of AmotL2 with gender ( Supplementary Fig. 5cg).

Discussion
The vascular endothelium plays an important role in the biomechanical response to hemodynamic forces. Understanding the pathways involved in this response is of importance to comprehend the pathogenesis of vascular disease. In this report we show for the first time that the cellular junctions of arterial ECs are connected via AmotL2 and microfilaments to the nuclear lamina. Interference with this pathway impairs EC alignment in response to shear stress and abrogates nuclear positioning resulting in inflammation and formation of abdominal aortic aneurysms.
We have used an inducible mouse model to target AmotL2 in the EC lineage. In our previous publication, we could show that Amotl2 silencing in ECs in utero resulted in impaired aortic expansion and death at embryonic day 10. Silencing of AmotL2 in adult mice, however, did not affect overall survival or had any obvious negative effects up to six months after AmotL2 depletion. The defect in adult mice was clearly more subtle as it was restricted to ECs exposed to arterial flow. We demonstrated not only that AmotL2 is required for cellular alignment in areas of shear stress, but also provided novel insights into how VE-cadherin is mechanically coupled to the cytoskeleton and thereby controls cell shape. AmotL2 triggers the formation of radial actin filaments that mediate junctional tension between neighboring cells. These radial actin filaments were detected in arterial, but not or at least at lower levels in venous endothelium. AmotL2 is a scaffold protein and as such brings together protein complexes of different functions such as Par3, MAGI-1b, Merlin, Actin, VE-cadherin. Of interest is that, in HUVECs of the venous origin, AmotL2 is sufficient to induce radial actin filaments when the cells aligned under flow condition.
Our data are consistent with the notion that VE-cadherin is part of a mechanosensory complex with VEGFR2 and PECAM-1 as previously described (38). The present data show that the VE-cadherin/AmotL2 protein complex is responsible for the actual cell shape modulation in arterial ECs. These most recent investigations together with our previous findings have shown that the VE-cadherin AmotL2 complex mediates mechanical forces between ECs, suggesting that AmotL2 may also relay mechanical forces not only from circulatory shear stress, but also transfers mechanical signals between cells. Of particular interest is the observation that AmotL2 is required for nuclear shape and subcellular positioning. Ingber and coworkers showed early on that there is a direct linkage between the cytoskeleton and the cell nucleus opening up the possibility of a novel mechanical signaling pathway from the exterior to the nucleus (39). Actin filaments are directly associated with the nuclear lamina by binding to Nesprin2, SUN-1 and -2, and Lamin A that

Mice and tamoxifen injections
The amotl2 flox/flox mice, carrying a loxP-flanked amotl2 gene, were crossed to Cdh5(PAC) CreERT2 and ROSA26-EYFP double transgenic mice. To induce endothelialspecific amotl2 gene inactivation, tamoxifen was administered by intraperitoneal (IP) injection for 5 continuous days. For adult mice over 6 weeks old, 100µl of tamoxifen (20mg/ml) was administered at each injection site. Analysis of mice samples was performed four to six weeks after injections. All mice in this report were in C57BL/6 background, and both females and males were included. Ethical permits were obtained from North Stockholm Animal Ethical Committee and all experiments were carried out in accordance with the guidelines of the Swedish Board of Agriculture.

Tissue preparation
Mice were euthanized using carbon dioxide. Thoracic cavity was rapidly opened, and the heart was exposed while still beating. Cold PBS was injected through a cannula for perfusion for 1 minute and then changed to 4% paraformaldehyde (PFA) for another 1minute perfusion. Aorta from root to aortic-common iliac bifurcation was dissected followed by careful removal of the connective tissues. After 1h extra fixation in 4% PFA, the entire aorta was opened longitudinally. For the aortic arch part, using spring scissors, the inner curvature was cut along anteriorly, while the outer curvature of the aorta was opened from the aortic root through innominate, carotid, and subclavian arteries until the aortic arch resembled a Y shaped split. The whole flattened-out aorta was pinned onto the wax mold and prepared for immunostaining.
For RNA isolation, the aorta was perfused with cold PBS for 2 minutes before careful dissection. Thoracic and abdominal aorta were taken out and frozen in -80˚C for subsequent RNA extraction.
To isolate the urinary bladder, the mice were handled especially gently before sacrifice, which prevented urine leakage. Full bladders were immersed in fixative for 2 hours (2h) and pinned on the wax mold in an open flower shape for whole-mount staining.
Mouse eyeballs were taken out at postnatal day 6 (P6) and month 6 (week 24). After a 2h fixation in 4% PFA, the retinas were dissected out and prepared for immunofluorescence staining for analysis of vasculature. Medium MV (C-22020). To note, the batch of HAoEC used for this study was from a 55year-old male donor with a Caucasian background.

Lentiviral induced knock-down
For knockdown studies, HAoECs (PromoCell) were infected with customized AmotL2 shRNA Lentiviral particles (Sigma) or scrambled shRNA control virus in complete endothelial cell medium with 5 μg/mL polybrene (Vector Builder). The Lentiviruscontaining medium was removed after overnight incubation and fresh medium was added.
Further analyses of confluent cells (WB, IF, Ibidi flow) were performed at ≥ 72h after infection.

Flow experiments
Ibidi flow system: Flow chamber slides (Ibidi μ-Slides VI 0,4 Ibidi-treated) with a volume of 30 μl per parallel channel were coated with fibronectin (Sigma). HAoECs/HUVECs were grown on the slides for 24h until 30-50% confluency followed by infection with Lentivirus or control treatment. 96h after that cells were subjected to 14 dyn/cm 2 laminar flow using the Ibidi pump system and pump control software or kept at static conditions and cultured for 48h at 37°C 5% CO2. Cells were then harvested and processed for further analysis.
Orbital shaker: The Rotamax 120 (Heidolph) complies well with cell incubator and generates circular motion with the maximum speed of 300rpm. HUVECs on a 15cm culture dish with 16ml medium were placed on the shaker for 96h before the harvesting the cell lysates.

Immunofluorescence (IF) staining
Immunofluorescence staining on MS1 cells, HAoEC, and HUVECs was performed as previously described (25). To stain open aorta pinned on wax, endothelium exposed on the top layer was carefully treated using the same protocol as for cell staining, with the exception that the aorta was permeabilized for 20 minutes with 0.1% Tween 20 in PBS.  Supplementary Table 5.

Co-immunoprecipitation (Co-IP) analysis
Mouse lung tissues were cut into small pieces before transferring into lysis buffer for western blot. Tissue homogenizer was gently applied for better protein extraction.
Cell/tissue lysates were incubated with protein G sepharose beads (GE, 17-0618-01) for 1.5 hours at 4˚C as pre-cleaning. Afterwards 2µg Amotl2, Lamin A/C, VE-cadherin antibody or control IgG were added in the lysates overnight at 4˚C. The next morning, the immunocomplexes were enriched by protein G beads for 2h at 4˚C, followed by five-time washing with lysis buffer. The final protein samples were fractionated by polyacrylamide gel and fractions were western blotted for evaluation of IP protein input level. Online LC-MS was performed using a Dionex UltiMate™ 3000 RSLCnano System coupled to a Q-Exactive-HF mass spectrometer (Thermo Scientific). IP samples were trapped on a C18 guard-desalting column (Acclaim PepMap 100, 75μm x 2cm, nanoViper, C18, 5µm, 100 Å), and separated on a 50cm long C18 column (Easy spray PepMap RSLC, C18, 2μm, 100 Å, 75μm x 50cm). The nano capillary solvent A was 95% water, 5% DMSO, 0.1% formic acid; and solvent B was 5% water, 5% DMSO, 95% acetonitrile, 0.1% formic acid. All TaqMan probes involved were purchased from Thermo Fisher Scientific and detailed information can be found in Supplementary Table 5.

RNA-seq
Total RNA purified from full-length aorta (thoracic and abdominal aorta) of amotl2 ec+/ec+ (n=3) and amotl2 ec-/ecmice (n=3) were sent for RNA-seq analysis (Novogene, Beijing, China). Libraries were prepared from 4-5µg of total RNA. PolyA RNA was purified using the Dynabeads mRNA purification kit (Ambion) and fragmented using Fragmentation Reagent (Ambion). First strand cDNA was synthesized from polyA RNA using the SuperScript III Reverse Transcriptase Kit with random primers (Life Technologies). Second strand cDNA synthesis was performed using Second Strand Synthesis buffer, DNA Pol I, and RNase H (Life Technologies). cDNA libraries were prepared for sequencing using the mRNA TruSeq protocol (Illumina).
The genes with significantly differential expression were input to online Enrichr (Ma'ayan Laboratory, Computational systems biology) for GO term analysis (Biological process 2018) (45,46).

Statistical analysis
All statistical figures and analyses were performed using GraphPad Prism software, with the exception of the gene correlation graphs, which were generated using R (https://www.rproject.org/index.html). Statistical analysis of in vivo results was based on at least three animals per group. Comparisons between two groups with similar variances were analyzed by the standard unpaired two-tailed Student t-test, while comparisons between multiple groups were analyzed by Kruskal-Wallis test. The correlation between two genes were analyzed by Pearson Correlation and pearson correlation coefficient is referred to r. A value of P<0.05 was considered as statistically significant (n.s. not significant, * P <0.05, ** P <0.01 and *** P <0.001). No statistical method was used to predetermine sample size in animal studies and the experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment.      Table 3) and presented in the graph ranking by -log10 adjusted P value. mRNA isolated from DAs of amotl2 ec+/ec+ (n=3) and amotl2 ec-/ec-(n=5) mice were sent for RNA-sequencing analysis. 65 significant genes which were differentially expressed were subjected to GO terms matching. c, Whole-mount staining on aortic arch of amotl2 ec+/ec+ and