Excessive fluid shear stress-mediated Klf4 leads to arteriovenous pathogenesis

Background: Vascular networks form, remodel and mature under the influence of both fluid shear stress (FSS) and soluble factors. For example, FSS synergizes with Bone Morphogenic Protein 9 (BMP9) and BMP10 to promote and maintain vascular stability. Mutation of the BMP receptors ALK1, Endoglin or the downstream effector SMAD4 leads to Hereditary Hemorrhagic Telangiectasia (HHT), characterized by fragile and leaky arterial-venous malformations (AVMs). But how endothelial cells (EC) integrate FSS and BMP signals in normal vascular development and homeostasis, and how mutations give rise to malformations is not well understood. Results: Here we show that loss of Smad4 in murine ECs increases cells’ sensitivity to flow and the resulting AVMs are characterized by excessive elongation and polarity against the flow. Smad4 deletion also blocks the anti-proliferative response to high FSS, leading to loss of arterial identity. Our data show that loss of cell cycle arrest leads to loss of arterial identity, which is essential in AVM formation upon Smad4 depletion in ECs. Excessive flow-induced activation of KLF4-PI3K/AKT due to Cyclin dependent Kinase (CDK) activation mediates the aberrant morphological responses to flow triggering AVM formation. Conclusions: Our results show that loss of polarization against the flow is not required for AVM formation upon EC Smad4 depletion. Instead, increased EC proliferation-mediated loss of arterial identity due to flow-induced PI3K/Akt/Cdks hyperactivation and Klf4 over-expression are the main events associated with AVM formation. findings that the canonical BMP9/10-Smad4 signaling plays a crucial role in shear stress regulation of vascular homeostasis. We therefore set out to test the concept that AVMs arise from loss of shear stress-mediated EC quiescence using Smad4 EC loss of function (LOF) mice as a model of AKT1/2 siRNA VHS40082 and VHS41339), VEGFR2 CDH5 (Dharmacon/Horizon, using Lipofectamine RNAiMax (Invitrogen) in 2% OPTI-MEM. Transfection efficiency was assessed by western blotting and qPCR. Experiments were performed 48-60 hours post transfection and results were compared with siRNA CTRL (ON-TARGETplus Non-Targeting Pool D-001810-10-05). Inhibition of PI3K was achieved by using Pictilisib (S1065, Selleckchem) in a concentration of 75nM and inhibition of CDK4/6 by using Palbociclib in concentration of 2 µM. Before experiments cells were starved for 8-10 hours in 2% FCS.


Introduction
Vascular networks form, remodel and mature under the influence of multiple biomechanical and biochemical signals, but how these are integrated to promote vascular development and maintain adult homeostasis is poorly understood. Fluid shear stress (FSS) from blood flow is a critical variable that determines vascular endothelial cell (EC) number and shape in vascular development and maintenance 1 . ECs also polarize and migrate according to the flow direction; in different systems this may be with or against the flow 2 , but in the developing retina is against the flow, which is proposed to be important in guiding vessel formation 3,4 . One aspect of EC flow responses is the existence of a cell-autonomous shear stress set-point specific to each vessel type. FSS near the set-point promotes EC elongation and alignment parallel to the flow, and stabilizes the vessel whereas flow that is persistently above or below this level triggers vessel remodeling to restore FSS to the appropriate magnitude and contribute to vascular disease 5 .
We previously found that high shear stress within the physiological range synergizes with secreted Bone Morphogenic Protein (BMP) 9 and BMP10 to activate canonical Smad 1/5, which promotes EC quiescence and vascular homeostasis 6 . This pathway contributes to the inhibition of EC proliferation by FSS and to expression of factors that mediate pericyte recruitment, thus, stabilizing the vessels. By contrast, FSS activates the related Smad2/3 pathway only at low FSS magnitude to induce inward arterial remodeling 7 . In murine models of HHT, AVMs are characterized by a plethora of dysregulated EC events e.g increased in EC size, misdirected migration, increased proliferation and changes in EC fate 6,8,9,[13][14][15] . Yet, if one or a complex of interwined cell events, flow dependent or independent drive AVM formation remains largely elusive.
Mechanistically, one important mediator downstream of this flow-BMP9 crosstalk is PI3K/AKT, which is hyperactivated in HHT lesions in mouse models and in human patients 9,14, 16 .
Pharmacological inhibition of PI3K or depletion of EC Akt1 rescued AVM formation in HHT murine models 9,14 . Interestingly, PI3K/AKT activated in EC by FSS mediates shear stress-induced EC responses downstream of the mechanosensory junctional receptor complex 17 .
These findings suggest that the canonical BMP9/10-Smad4 signaling plays a crucial role in shear stress regulation of vascular homeostasis. We therefore set out to test the concept that AVMs arise from loss of shear stress-mediated EC quiescence using Smad4 EC loss of function (LOF) mice as a model of AVM formation.

Quantitative real-time PCR
RNAs from HUVECs or mouse lung ECs (mLECs) were purified using RNeasy-kit (74106, Qiagen  HUVECs were transduced with optimal volume of lentiviral virus at 50% confluence in MV2 medium and 8μg/ml Polybrene (Sigma). After 24 h, the medium containing viral particles was replaced with fresh medium and after additional 24h, the infected cells were selected with 2 μg/ml puromycin for 48 hours.

Exposure of endothelial cells to increased shear stress
HUVECs transfected with siRNAs or OE-HUVECs were plated in a six-well plate and on an orbital shaker (Rotamax120, Heidolph Instruments) at 50, 150 or 250 rpm to generate laminar sheer stress of 1, 5 or 12 DYNES/cm 2 respectively. Results were confirmed in a µ-Slide VI 0.4 (Ibidi, 80601) using a pump system (Ibidi, 10902).

Western blotting
HUVECs were washed with PBS and lysed with Laemmli buffer (1610740, Biorad intensity were quantified using ImageJ.

Proliferation Assay
Proliferation analysis was performed using Click-iT EdU Alexa Fluor 488 Imaging kit (Life Technologies). P6 pups were injected with 200 μg of EdU (5 mg/mL) and sacrificed 4 hours later.
EdU staining was done according the manufacturer's protocol.

RNA-Sequencing
For RNA-Seq of HUVECs the RNA was isolated with the Quick-RNA Miniprep Kit (#R1054, Zymo Research), 60 hours after transfection of Control or Smad4 siRNA. The RNA 6000 Nano Kit (#5067-1511, Agilent) was used to assess the RNA integrity on a Bioanalyzer 2100 (Agilent).
Both sequencing and library preparation were performed on the BGISEQ-500 platform.

RNA-Sequencing data analysis
Quality of RNA seq reads was assessed with the MultiQC tool (v1.13) and trimmed of adapters using Trimmomatic (v0.39). Reads were mapped by STAR (v2.7.10a) with the following settings: -alignIntronMin 20 and -alignIntronMax 500000 to the hg38 reference genome. Tag directories were created with makeTagDirectory and reads were counted by the analyzeRepeats.pl function (rna hg38 -strad both -count exons -noadj) both from HOMER (v4.7.2). Differential expression was quantified and normalized using DESeq2. Rpkm.default from EdgeR was used to determine average reads per millions mapped (RPKM). Heatmaps were created by using heatmapper.ca from the RPKM values and represent the row-based Z-scores.

Statistical analysis
All data are shown as mean ± standard error of the mean (SEM). Samples with equal variances were tested using Mann-Whitney U test or two-tailed Student's t-test between groups. P value <0.05 was considered to be statistically significant. Statistical analyses were performed for all quantitative data using Prism 9.0 (Graph Pad).

Smad4 signaling maintains the shear stress set-point-mediated EC responses
Impaired responses of EC to FSS, including migration direction and changes in EC size, have been proposed to mediate HHT lesions, mainly in models of HHT1 and HHT2, ie., mutations in ENG and ALK1 8,15,18 . To explore flow-mediated EC events in JP-HHT where SMAD4 is mutated, we depleted primary human umbilical cord ECs (HUVECs) of SMAD4 using small interfering RNA (siRNA) versus CTRL siRNA (confirmed in Figure 1G). Cells were subject to laminar shear  (Figure 1M,N). The relative position of the Golgi and nuclei were then quantified ( Figure 1O). EC polarization against the predicted flow direction was significantly increased in AVMs in Smad4 iΔEC retinas (Figure 1N,O). Thus, multiple EC morphological responses to shear stress are increased after SMAD4 KD in vitro or Smad4ECko in vivo.
It is well established that physiologically high FSS inhibits EC proliferation 19 . As expected 9 , labelling of Smad4 Fl/Fl and Smad4 iΔEC retinas for the mitotic marker KI67 and the total EC marker IB4 revealed increased EC proliferation within AVMs (Figure 1Q,P; quantified in 1R). In vitro, EdU labeling to identify ECs in S phase showed that SMAD4 depletion increased baseline cell cycle progression and completely blocked the inhibition by high shear stress ( Figure 1S).
Thus, Smad4 is required for flow-mediated repression of EC proliferation.
Taken together, these results show that Smad4 resembles Alk1 and Eng in that it is also required for flow-mediated EC proliferation but is opposite in that it suppresses rather than enhances morphological responses to flow.

KLF4 mediates flow-induced hyper-responsiveness of SMAD4 depleted HUVECs
To identify mediators of increased responsiveness of SMAD4ECko to FSS we performed RNA sequencing in CTRL versus SMAD4 siRNAs HUVECs grown in static or subjected to 2 hours 12 DYNES/cm 2 and focused on shear stress responsive genes. Interestingly, among other flow regulated genes, SMAD4 depletion enhanced FSS-induced Krüppel-like transcription factor (KLF) 2 and KLF4 induction (Figure 2A). As the two mechanosensors show dose-dependent induction by FSS 20 , and to validate our transcriptomic results, we perfomed RT-PCR for KLF2 and KLF4 in HUVECs subjected to increasing magnitudes of laminar FSS (1-5-12 DYNES/cm 2 ) and depleted for SMAD4. Interestingly, loss of SMAD4 moderately augmented flow-induced KLF2 and KLF4 with increasing flow magnitude ( Figure 2B).
As KLF4 showed the highest induction to FSS upon SMAD4 depletion in both transcriptomics and RT-PCR data, we therefore considered the role of KLF4 in the altered behaviours of  Figure 3B,B'). Interestingly, this specific region corresponds to the most frequent site of AVM formation 9 . In Smad4 iΔEC retinas, Klf4 expression was highly upregulated in AVMs at the highest intensity relative to the feeding artery and vein (red arrows in Figure 3D and Figure  To test Klf4 function in vivo, we generated two genetic models. First, we examined EC specific Tx-inducible Klf4 LOF neonates (Klf4 i∆EC ) where AVMs were induced by administration of blocking antibodies for BMP9/10 ( Figure 3F,G). Second, we created EC specific Tx-inducible double ko mice, Smad4;Klf4 i∆EC ( Figure 3H,I). Tx was injected at P1-P3 and retinas were analysed at P6. Efficient Smad4 and Klf4 gene deletion was validated by qPCR from P6 mouse lung endothelial cells (mLECs; Figure 3J)

KLF4 mediates the shear stress-induced aberrant EC events within AVMs
To inactivation blunted the increased axial polarity (Figure 4A,B; quantified in Figure 4C). Klf4ko ECs were less elongated than in Fl/Fl, and Klf4 inactivation rescued the excessive elongation of Smad4 iΔEC ECs, confirming our in vitro findings ( Figure 4D). To assay Klf4-mediated EC cell cycle progression, we injected EdU into Fl/Fl, Smad4 iΔEC and Smad4;Klf4 iΔEC P6 pups, 4 hours before collecting tissue and labeling for EdU and Erg ( Figure 4E). As previously observed, EdU+/Erg+ double positive ECs increased markedly in Smad4 iΔEC retinas, exclusively in AVMs.
Klf4 inactivation significantly decreased the number of EdU+/Erg+ in the vascular plexus of Smad4 iΔEC retinas to levels comparable to Fl/Fl retinas ( Figure 4E; quantified in 4F). Elevated

Klf4 thus contributes to increased morphological responses and excessive proliferation in
Smad4ECko AVMs.

Excessive flow-mediated PI3K/AKT activation regulates flow-mediated EC responses
We previously identified increased PI3K/AKT activity upon inactivation of BMP9/10-Alk1-Smad4 in ECs, further augmented by high FSS 9,14 and PI3K downstream of flow mediates EC responses 17 . To further understand if increased responsiveness of SMAD4 deficient cells to FSS is due to PI3K/AKT pathway activation, we subjected CTRL versus SMAD4 siRNAs HUVECs to increasing magnitudes of shear stress (1-5-12 DYNES/cm 2 ). Interstingly, SMAD4 deletion significantly increased AKT phosphorylation at serine 473, a marker of activation, under static condition and increasing flow magnitudes had an additive effect ( Figure 5A; quantified in 5B).
To assess the role of activated AKT in the amplified morphological response to FSS, we inhibited PI3K/AKT signaling for 48 hours using a specific PI3K inhibitor-Pictilisib (confirmed by Western Blot (WB)) ( Figure 5C)

Flow-induced KLF4 acts upstream of mechanosensory complex-PI3K/AKT pathway
To untangle the relationships between KLF4 and PI3K in the context of SMAD4 depletion and flow, we subjected HUVECs to flow in the presence of Pictilisib. RT-PCR results show no effect of Pictilisib on flow-induced KLF4 ( Figure 6A). We also considered the role of the junctional mechanosensory receptor complex that mediates flow responses including PI3K activation 24,25 .
Depletion of each of the components of the mechanosensory receptor complex had no effect on the flow-upregulation of KLF4 expression ( Figure 6B). Thus, flow-induced KLF4 expression does not require PI3K or the mechanosensory junctional receptor complex.
As both, FSS-induced Klf4 expression and AKT activation are partially restrained by Smad4 in a similar manner, we then tested effects of KLF4 on AKT activation, with flow and SMAD4KD.
KLF4 inactivation blunted the increase in AKT activity under flow and rescued AKT hyperactivation in SMAD4 depleted HUVECs (Figure 6C; quantified in 6D). To further test if flow-induced KLF4 is upstream of PI3K we examined the KLF4OE HUVECs. KLF4 upregulation was sufficient to activate AKT, with or without FSS (Figure 6E; quantified in 6F). To test these findings in vivo, we labelled retinas for phosphorylated S6 ribosomal protein (pS6), a downstream target of AKT activation, and with IB4 to label ECs. ECs in Smad4 iΔEC retinas showed high pS6 as expected, which was largely rescued by Klf4 deficiency (Figure 6G,H). Klf4 is thus upstream of PI3K to control flow-mediated EC events after Smad4ECko. normalized the levels of pRB1, E2F1, CDK4 and CDK6 in SMAD4KD cells (Figure 7B).

Increased EC proliferation-mediated loss of arterial identity is the main driver of AVMs
To further test whether effects on cell cycle are the main drivers of AVMs, we treated Smad4 Fl/Fl and Smad4 iΔEC pups with Palbociclib, an inhibitor of CDK4/6 activity that efficiently blocks cell cycle progression 27 . We first confirmed efficacy of Palbociclib in lungs isolated from treated pups. In contrast to PBS treated pups, Palbociclib treatment decreased the expression of Cdk4 and Cdk6 and inhibited activation of Cdk2 (p-Cdk2) (Figure 7C).
In retinas labelled for IB4 and KI67, Palbociclib treatment decreased EC proliferation as quantified by the number of KI67+ ECs per vascular area in both Control and Smad4 iΔEC retinas ( Figure 7D; quantified in 7E) and significantly rescued the number of AVMs (Figure 7D; quantified in 7F). We further assessed the impact of Palbociclib on morphological responses in Smad4ECKo retinas. Interestingly, similarly to KLF4-PI3K inhibition, Palbociclib treatment efficiently normalized the length/width ratio of Smad4ECko ( Figure 7G) and also blunted the elevated orientation against the flow direction ( Figure 7H). Taken  To test these results in vivo, we labeled retinas for the arterial marker Sox17 (Figure 7K). In Ctrl retinas, Sox17 was confined to ECs in main arteries and a few arterioles. In AVMs in Smad4-deficient mice, Sox17 expression was completely abrogated. In Klf4ECko retinas, Sox17 expression expanded towards the vein and capillary ECs. Klf4 inactivation in Smad4 iΔEC retinas largely rescued Sox17 expression in arteries. Palbociclib treatment led to even greater expansion of Sox17 expression into capillary and venous ECs ( Figure 7K). Collectively, these results suggest that increased EC proliferation-mediated loss of arterial identity is a central cell event in AVM formation.

Discussion
It has been proposed that blood flow is 'a second hit' in HHT, as murine AVMs develop in regions of high shear stress 6,9 , but the mechanisms by which shear stress contributes to AVM pathogenesis remained largely undefined. ECs display an intrinsic set-point for physiological flow-induced shear stress that determines the signaling and gene expression outputs that control EC phenotype.
VEGFR3 expression levels is one factor that can determine shear stress set-point for different types of vessels 5 . Also, non-canonical WNT signaling was proposed to modulate axial polarity set-point to control vessel regression in low flow regions 29 . We report here that loss of Smad4 in mouse ECs increased sensitivity to FSS with enhanced elongation and polarization in FSS together with diminished FSS-mediated cell cycle blockade and loss of EC arterial fate. Thus, Smad4 signaling is a novel mechanism that "sets the set-point" for high flow-mediated EC quiescence responses, e.g elongation, alignment and orientation. Smad4 is also critical for FSS-mediated growth suppression and arterial EC fate, though it remains to be determined if these events are also linked to disruption in the set point.
We previously reported that loss of BMP9-SMAD4 signaling potentiates PI3K/AKT or Eng, implying distinct mechanisms for AVM formation due to different mutations. But at present, the hypothesis that migration direction is not the key event triggering AVMs appears both simpler and consistent with older studies demonstrating that depending on variables such as animal species, age and specific vascular location, ECs can migrate either against or with the flow 2,35 .
FSS induces cell cycle arrest in late G1 to enable maintenance of arterial identity via a Notch-Cx37-p27 signaling axis 27 . Notch and Smad1/5 co-regulate a number of genes, raising the possibilities that these two pathways function together 36 . We now report that Smad4 is also required for flow-induced cell cycle arrest-mediated arterial identity by restricting flow-induced Klf4-PI3K/Akt-CDK signaling. Genetic loss of Klf4 or pharmacological inhibition of PI3K or CDK4/6 rescues EC proliferation and restores arterial identity leading to EC normalization in Smad4ECKO retinas. Our findings support the concept that in Smad4 deficient EC, AVMs arise from loss of shear stress-mediated repression of EC proliferation and arterial identity due to excessive flow-induced Klf4-PI3K/Akt-Cdk (Figure 8).
These studies raise a number of new questions. Why does Klf2/4 induced by physiological flow stabilize vessels whereas higher levels promote cell proliferation and contribute to pathologies?
Klf4 amplification of Akt activation is likely key but this mechanism is also unknown. Further work will be required to elucidate mechanism by which Smad4 "sets the set-point" for FSS-