Sox17 mediates adult arterial endothelial cell response to hemodynamics

Abstract

Sox17 is a critical regulator of arterial identity during early embryonic vascular development. However, its role in adult endothelial cells (ECs) are not fully understood. Sox17 is highly expressed in arterial ECs but not in venous ECs throughout embryonic development to adulthood suggesting that it may play a functional role in adult arteries. Here, we investigated Sox17 mediated phenotypical changes in adult ECs. To precisely control the temporal expression level of Sox17, we designed a tetracycline-inducible lentiviral gene expression system to express Sox17 selectively in cultured venous ECs. We confirmed that Sox17-induced ECs exhibit a gene profile favoring arterial and tip cell identity. Furthermore, in comparison to control ECs, Sox17-activated ECs under shear leads to greater expression of arterial markers and suppression of venous identity. These data suggest that Sox17 enables greater hemodynamic adaptability of ECs in response to fluid shear stress. Here, we also demonstrate key morphogenic behaviors of Sox17-mediated ECs. In both vasculogenic and angiogenic 3D fibrin gel studies, Sox17-mediated ECs prefer to form cohesive vessels with one another while interfering the vessel formation of the control ECs. Sox17-mediated ECs elicit hyper-sprouting behavior in the presence of pericytes but not fibroblasts, suggesting Sox17 mediated sprouting frequency is dependent on supporting cell type. Using a microfluidic chip, we also show that Sox17-mediated ECs maintain thinner diameter vessels that do not widen under interstitial flow like the control ECs. Taken together, these data showed that Sox17 mediated EC gene expression and phenotypical changes are highly modulated in the context of biomechanical stimuli, suggesting Sox17 plays a role in regulating the arterial ECs adaptability under arterial hemodynamics as well as tip cells behavior during angiogenesis and vasculogenesis. The results from this study may be valuable in improving vein graft adaptation to arterial hemodynamics and bioengineering microvasculature for tissue engineering applications.

INTRODUCTION

Sox17 belongs to the Sox (Sry-related high mobility group box) family consisting of 20 members that share a conserved high mobility group (HMG) box domain [1, 2]. This domain is responsible for DNA binding and bending, facilitating protein interactions, and importing/exporting into the nucleus [1, 3].

The SoxF subgroup of this family consisting of Sox7, Sox17, and Sox18 play an important role in cardiovascular development and embryonic hematopoiesis [1, 2, 4–8]. Sox17 itself plays a pleiotropic role in multiple developmental processes: visceral and definitive endoderm differentiation [9–11], oligodendrocyte progenitor cell differentiation and maturation [12], endocardium/cardiomyocyte differentiation and formation [13–16], and definitive hematopoiesis [8, 17, 18]. In the vascular context, Sox17 is a crucial regulator of arterial EC specification, angiogenesis, and vessel maturation [19–24].

The exact developmental role and placement in the mechanistic pathway of Sox17 seems to be context dependent, resulting in sometimes conflicting outcomes in previous studies. For example, Corada et al. demonstrated that Sox17 deletion in a C57Bl/6J mouse embryo resulted in a fusion of the dorsal aorta and cardinal vein inferring that Sox17 is crucial for arterial specification [19]. In contrast, another Sox17 knockout study of a mouse embryo (mixed background of C57BL/6, 129sv and CD1 strains) only led to mild malformations of the anterior dorsal aorta as another SoxF member, Sox18, compensated for the deficiency [25]. Similar observations of partial effects have been observed by other groups, and it is thought that these particular discrepancies are due to differences in genetic background and phenotype variation [26–28].

In another example, Sox17 either promotes or diminishes vessel stability depending on the biological system [20, 23, 29, 30]. Lee et al. demonstrated that EC colonies derived from Sox17-null mouse embryonic stem cells (mESCs) had linear junctional distribution of VE-cadherin and β-catenin, few lamellipodia and microspikes at the periphery, and reduced migratory behavior [20]. They observed Sox17-deficient hyaloid vessels displayed improved postnatal vascular stability, and Sox17 overexpression led to diminished structural integrity in the yolk sac vessels. Yang et al. observed similar destabilizing behavior of Sox17 in B16F10 melanoma tumor vessels promoting angiogenesis and hyper-branching instead [23].

Conversely, the brain vasculature shows a different trend in Sox17’s role in structural integrity[29–32]. Sox17 deletion in the postnatal mouse endothelium induces intracranial aneurysm (IA) under chronic hypertensive stress [29]. Moreover, Sox17-deficient intracerebral arteries displayed luminal dilation, wall thinning, and diminished vascular smooth muscle layers. Correspondingly, the authors found upon analysis of IAs in human patients that the intracerebral arteries also exhibited reduced Sox17 expression, thin arterial walls, and destabilized endothelial junctions. Likewise, Corada et al. demonstrated that EC-specific Sox17 inactivation resulted in a perturbed blood brain barrier (BBB) and increased permeability in brain vasculature for mice in the embryonic and postnatal stage [30]. Moreover, though permeability was not altered in the adult brain vessels, Sox17 deletion led to more severe lesions when cerebral ischemia was artificially induced. Here, Sox17 activation is crucial to maintaining BBB integrity.

Interestingly, conflicting reports on Sox17 arose even when using similar genetic mouse models and focusing on the same biological system [19, 20]. Corada et al. observed that Sox17, besides being localized in arteries, suppresses tip cell specification and angiogenic sprouting [19]. Deletion of Sox17 led to hypersprouting. Meanwhile, Lee et al. demonstrated that Sox17 activation leads to favoring of tip cell phenotype and vascular hypersprouting [20]. Despite the use of similar mouse models, phenotype differences could arise from unexplored mechanisms, such as epigenetic factors.

These evidences suggest that the role of Sox17 is quite context dependent, such as whether it is expressed in embryo or adult, or whether it is expressed in specific cell types or everywhere in animal models. Thus, the results are all different and complex to interpret. The critical role of Sox17 in embryonic development is well appreciated, but whether it plays a continuous role in adult vessel is not yet fully understood. Due to the dynamic nature of Sox17, the use of advanced in vitro systems can be advantageous as certain parameters can be controlled more easily, e.g., shear stress and cell type inclusion. In this study, we created several in vitro arterial vascular models by transducing human umbilical venous endothelial cells (HUVECs) with a tetracycline-inducible lentivirus vector expressing Sox17 and incorporating them in shear devices, 3D hydrogel cultures, and microfluidic devices. With this more controlled microenvironment, we were able to not only recapitulate vascular features that were observed in previous reports but also identify new behaviors of Sox17-mediated vascular ECs.

MATERIALS AND METHODS

Cell culture maintenance

Primary human umbilical venous endothelial cells (HUVECs, Lonza) and primary human aortic endothelial cells (HAECs, Lonza) were cultured at 37°C and 5% CO₂ in EGM-2 medium (PromoCell) on 0.1% gelatin-coated flasks. To support vessel formation, we also used primary human lung fibroblasts (NHLF, Lonza) or primary human brain vascular pericytes (HBVP, ScienCell) in co-culture. NHLFs were maintained in FGM-2 medium (PromoCell) on 0.1% gelatin-coated flasks. HBVPs were maintained in Pericyte Medium (ScienCell) on poly-d-lysine-coated flasks. All cell types were passaged with 0.25% trypsin-EDTA (Gibco).

Lentivirus transduction

All lentiviruses are third generation; they were designed and purchased from VectorBuilder. For HUVECs, the Sox17 lentivirus (VB151201-10016) is a Tet-On inducible system where the gene of interest can be expressed rapidly in high levels with the presence of doxycycline (tetracycline analog). To distinguish them easily through imaging, we also created cell lines that was co-transduced with a lentivirus expressing either tdTomato (VB181014-1005thm), ZsGreen1 (VB181014–1006ufx), and EGFP (VB150915–10026).

For lentivirus transduction, HUVECs were passaged at ~70% confluence onto 6-well plates on Day −1. Lentivirus and 5ug/mL polybrene were added on Day 0. Multiplicity of infection (MOI) of 2 or 5 was used for this study. The virus was removed and fresh medium added on Day 1. On Day 3, the cells were selected for antibiotic resistance. To ensure purity, cells undergo antibiotic selection again before the beginning of an experiment.

3D fibrin dot culture: vasculogenesis

To mimic vasculogenesis, HUVECs were embedded in a fibrin gel with support cells, either NHLFs or HBVPs. If NHLFs were used, a seeding density ratio of 1×10⁶ HUVECs/mL:1×10⁶ NHLFs/mL or 2.5×10⁶ HUVECs/mL:1.25×10⁶ NHLFs/mL was used. For co-culture of Sox17-HUVECs and control-HUVECs, a ratio of 1:1 was used with 1.5×10⁶ cell density for each HUVEC type. NHLF cell density was also 1.5×10⁶ cells/mL (2:1 HUVEC:NHLF ratio).

Prior to placing the fibrin dots, the 12-well or 24-well plates were water-marked in the center with 50uL of dPBS-CM buffer. Once placed, the drop was immediately aspirated so that only a thin film of dPBS-CM remained. The reasons for this additional step are two-fold. First, the calcium-magnesium will help with the polymerization of the fibrin dot and continued attachment onto the plate throughout the culture study. Second, the watermark will aid in spreading the fibrin gel enough to minimize overlapping vessels along the z-axis plane and to image through the gel more clearly with a non-confocal fluorescence microscope (Nikon Eclipse Ti2).

For the fibrin hydrogel, thrombin and fibrinogen (Sigma-Aldrich) solution were freshly prepared. Thrombin was freshly diluted with dPBS-CM buffer to 100U/mL. Fibrinogen was dissolved in EGM2 medium at 10mg/mL and sterile-filtered with 0.22um PES syringe filter. The HUVECs and support cells, already in their final cell densities in their respective media, were then mixed together at 1:1 ratio (creating a 2x volume) in an Eppendorf tube, centrifuged, and resuspended in the fibrinogen medium at 1x volume as a working master mix. The fibrin gel was made using a 2.5uL thrombin: 100uL fibrinogen ratio with an aim for a ~1min polymerization time. Once the thrombin is added, a 20uL volume of the cell-fibrin solution was quickly deposited onto the well center. A minimum of 4 fibrin dots were made for each culture condition. The plate was incubated at room temperature for 5 minutes before transferred to 37°C for a further 30min incubation. Lastly, EGM2 with or without doxycycline (1000ng/mL) was then added to the wells.

3D fibrin dot culture: angiogenesis

Angiogenesis was tested with HUVEC spheroids embedded in fibrin gels with NHLFs or HBVPs dispersed throughout as support cells. Spheroids were formed using the hanging drop method. Briefly, HUVECs were resuspended in EGM2+ 0.0625% methylcellulose at 0.8×10⁶ cells/mL. If control and Sox17-activated HUVECs were used in combination, a ratio of 1:1 was used (0.4×10⁶ cells/mL each). Drops of 10uL were then placed onto the inner surface of 10cm petri dish lids, which were then inverted and replaced back onto the dishes. The petri dish wells also contained a small volume of dPBS to mitigate evaporation. Dishes were then incubated overnight at 37°C where gravity and the methylcellulose encourage the HUVECs to pool to the floor of the media droplets. The spheroids were then collected by washing the lids with dPBS several times and spinning this suspension down briefly for 1min at 200g. They were then resuspended with EGM2. One 10cm petri dish can yield ~100 spheroids which is enough for 24 fibrin dot samples. NHLFs or HBVPs were dissociated and resuspended in their appropriate medium at 1.5×10⁶ cells/mL. Fibrinogen solution was freshly prepared at 10mg/mL in EGM2 and sterile-filtered with 0.22um PES syringe filter. A master mix of spheroids and support cells were then prepared at 1:1 ratio, spun down for 3min at 200g, and resuspended gently in the fibrinogen solution. To make the fibrin gel, a ratio of 2.5uL thrombin (from 100U/mL stock): 100uL cell-fibrinogen mix was used. With thrombin added, 40uL cell-fibrin solution was quickly deposited into the wells pre-watermarked by 80uL dPBS-CM buffer. Enforcing a larger diameter in the fibrin dot will ensure greater space among the spheroids for sprouting. A minimum of 4 fibrin dot replicates were made for each culture condition. The plate was incubated at room temperature for 5 minutes before transferred to 37°C for a further 30min incubation. Finally, EGM2 with or without doxycycline (1000ng/mL) was then added to the wells.

AIM Biotech microfluidic chip: vasculogenesis

AIM Biotech chips were used to study the effect of interstitial flow on Sox17-HUVEC vascular formation. NHLFs and HBVPs were used as support cells. For fibroblasts, an even ratio of HUVECs: NHLFs (4×10⁶ cells/mL each) was implemented. For pericytes, a 2:1 ratio of HUVECs: HBVPs (4×10⁶ cells/mL and 2×10⁶ cells/mL, respectively) was used instead. When a mixed population of control HUVECs and Sox17-HUVECs was needed, the seeding density is 2×10⁶ cells/mL each with the overall HUVEC seeding density still adding to 4×10⁶ cells/mL.

The fibrin gel containing the cells was injected into the central chamber (10uL volume) according to the manufacturer’s instructions. The fibrin gel was created using 100U/mL thrombin and 10mg/mL fibrinogen at a 1uL thrombin: 100uL fibrinogen ratio to achieve a ~2min polymerization time. Once the gel has polymerized, media was introduced to the side channels. Static conditions were maintained by adding equal volumes of media to both sides. While, interstitial flow (IF) conditions were created by adding a 40uL volume difference between the two side channels. Media was refreshed daily for both static and IF conditions.

AIM Biotech microfluidic chip: angiogenesis

The AIM Biotech channels were also applied here to study angiogenesis under interstitial flow. Again, NHLFs and HBVPs (1.5×10⁶ cells/mL) were used to encourage stable sprout formation. The support cells were injected with the fibrin gel into the central chamber before polymerization. Same procedure in fibrin gel preparation for vasculogenesis in AIM chips were implemented here. A volume of 10uL of HUVECs (12×10⁶ cells/mL total; 6×10⁶ cells/mL per cell type if combining Sox17 and control) were seeded into one of the side channels, while 10uL of media was injected into the other side channel. The chips were then tilted for 2h in 37°C to encourage the HUVECs to pool between the micropost gaps. Media was then added with appropriate volumes for static or IF conditions (similar to the AIM vasculogenesis protocol) and was changed daily. In this study, the direction of interstitial flow was directed against the direction of sprout growth. Culture was maintained for up to two weeks or shorter with images taken at regular intervals when necessary.

Cone and Plate Shear Cone Device

We designed a cone-and-plate device capable of generating laminar shear on a monolayer of cells in a 6-well plate or 10cm well dish. Detachable cones is made of autoclavable stainless steel with a 1° angle. The cone is placed within 500um from the culture dish bottom. The cone is attached to a NEMA USB stepper integrated motor/controller/driver (Arcus Technology, Inc., DMX-J-SA-17). Shear stress can be calculated and set based on the following equation:

$$\tau =\frac{\mu \omega }{\theta }$$ (1)

where τ is shear stress, μ is fluid viscosity (0.00078 Pa•s, media at 37°C), ω is rotational speed, and θ is cone angle (1°). The device was programmed to provide a gradual ramp up to final rotational speed by 7 hours with experiment timepoint 0 starting once final shear setting is reached (10 dyn/cm²).

For this study, cells were seeded at 0.8×10⁶ cells in a 6-well plate the day before covered with customized aluminum plate lids to maintain sterility. Media was changed daily EGM2 with or without 1000ng/mL doxycycline.

RNA isolation and quantitative reverse transcription-PCR

Cells were seeded at 0.2×10⁶ cells/mL in a 24 well dish. RNA was isolated using the RNeasy Mini Plus and Qiashredder kit (Qiagen). Concentration was quantified using the Nanodrop spectrophotometer (Thermo Scientific). Comparative Ct analysis was performed on the QuantStudio 3 (Applied Biosystems) by using the TaqMan^™ Fast Virus 1-Step Master Mix kit and TaqMan^™ Gene Expression Assay primer sets (Applied Biosystems). Results were expressed as fold change in mRNA expression normalized to GAPDH and then to reference samples using the ddCt method. Reported values were stated as mean ± SD of three biological replicates with technical triplicates. Unpaired two-tailed Student’s t-test was implemented for statistical analysis of two groups. One-way or two-way ANOVA with Bonferroni post hoc test was used to determine significance across three or more groups using GraphPad Prism, *P<0.05, **P<0.01, ***P<0.001.

Microarray data

HUVECs were seeded at 0.8×10⁶ cells/mL in a 6 well dish and exposed to 100ng/mL doxycycline in EGM2 medium for 48h. RNA was isolated, and samples were sent out for microarray analysis to a third-party company. Microarray was conducted using Agilent SurePrint G3 Human GE v3 8×60K Microarray. Repeated measure test was used for statistical analysis. Genes expressing more than two-fold were selected for further analysis (P<0.05). Heatmaps were created with Morpheus software (https://software.broadinstitute.org/morpheus).

Endothelial RNA isolation from mouse aorta and vena cava

RNA from mouse aortic and vena cava endothelium was isolated from C57BL/6 mice using a previous published protocol. Briefly, mice were euthanized by intraperitoneal injection of sodium pentobarbital and then the vasculature was perfused with saline solution for 2 minutes via the left ventricle after severing the right atrium. The aorta and vena cava were isolated and cleared of periadventitial adipose tissue. The isolated vessel was quickly flushed with 300 ml of TRIzol reagent (ThermoFisher Scientific) using a 25G needle. Eluates from three animals were collected into a microcentrifuge tube. Chloroform (0.2X volume) was added to the eluate, vortexed and centrifuged to separate the aqueous phase. Isopropanol (1.25X volume) was subsequently added to the collected aqueous phase and transferred to RNeasy Plus (Qiagen) column, and total RNA were isolated following the company protocol. RNA quality was validated by spectroscopy using a NanoDrop 2000 (ThermoFisher Scientific). To validate that quality of RNA for its specificity to endothelium without smooth muscle contamination, RT-PCR was conducted to check the expression of PECAM, VE-Cadherin and α-SMA, and compared that to the RNA isolated from whole vessels.

Immunocytochemistry

For monolayer of cells, samples were fixed in 4% paraformaldehyde in dPBS-CMF for 10min at room temperature or 4°C overnight. They were then quenched in 750mM Tris-HCl/dPBS-CMF (pH 8) solution for 5min and then washed three times in dPBS-CMF solution. Samples were permeabilized in 0.1% Triton X-100/dPBS-CMF for 5min at room temperature. Next, cells were blocked with 3% donkey serum in 0.1% Triton/dPBS-CMF for 30min at room temperature. Primary antibodies were diluted in the blocking solution, and cells were incubated overnight at 4°C. Samples were then washed three times in 0.1% Tween/dPBS-CMF for 5min each. Before use, secondary antibodies were spun down at 12,000g for 10min at 4°C to minimize the presence of aggregates in the staining. Then, they were diluted appropriately in the blocking solution, and cells were incubated with the secondary antibodies for 1hr at room temperature or overnight at 4°C. If nuclei staining was required, samples were treated with Hoescht 33342 in dPBS-CMF solution for 15min. Finally, samples were washed three times with 0.1%Tween/dPBS-CMF solution for 5min each and maintained in 0.1% Tween/dPBS-CMF solution.

For the AIM Biotech microfluidic chips, samples were fixed in 4% paraformaldehyde in dPBS-CMF overnight in 4°C. They were then washed three times, 2hr each, in dPBS-CMF at room temperature. Samples were then permeabilized in 0.1% Triton X-100/dPBS-CMF for 1hr and then blocked with 3% donkey serum in 0.1% Triton/dPBS-CMF overnight at 4°C. Primary antibodies diluted in the blocking solution were next added to the samples for overnight staining at 4°C. They were then washed three times, 2hr each, in 0.1% Tween/dPBS-CMF. Secondary antibodies and Hoescht 3342 diluted in blocking solution were added to the samples for overnight incubation in 4°C. Finally, samples were washed three times, 2hr each, in 0.1% Tween/dPBS-CMF and maintained in the same solution until imaging.

Primary antibodies used for this study include the following: anti-Sox17 (1:200, BD Pharmingen^™, 561590), CoupTFII (1:200, R&D Systems, PP-H7147–00), NG-2 (1:100, eBioscience 14-6504-80).

Secondary antibodies used are listed as follows: Alexa Fluor 647 donkey anti-mouse (1:1000, Invitrogen^™), Alexa Fluor 488 donkey anti-mouse (1:1000, Invitrogen^™), Alexa Fluor 594 donkey anti-mouse (1:1000, Invitrogen^™), Alexa Fluor 647 donkey anti-rabbit (1:1000, Invitrogen^™), Alexa Fluor 488 donkey anti-rabbit (1:1000, Invitrogen^™), Alexa Fluor 594 donkey anti-rabbit (1:1000, Invitrogen^™).

Image Analysis

Cell alignment quantification

A phase contrast image is thresholded to highlight the dark centers of the cells. The subsequent binary image is further processed by filling holes. The Extended Particle Analyzer from the BioVoxxel Toolbox plugin for FIJI was then used to quantify the Feret angle of each ROIs which infer the extent of cell anisotropy [33].

Nuclei fluorescence quantification

Images from 2–4 field areas were taken from each of the 3 biological replicates per sample condition. FIJI software was used for image analysis. For quantifying nuclear protein staining, such as Sox17 and CoupTFII, we used the Hoescht stain as a mask to delineate the nuclei. Using the mask overlay, we localize and quantify the fluorescence intensity of each nucleus. With the median of the readings (mean gray intensity values) and the values of representative nuclei for no/low-to-middle/high protein expression, an arbitrary intensity range was chosen that was then used to categorize each nucleus into the no, low-to-mid, or high expression group. This intensity range criteria was applied globally across all sample conditions.

Vasculogenesis Fibrin Dot Image Analysis

For each culture condition, a minimum of 4 fibrin dots were made. Images were then processed to reduce noise by subtracting background, despeckling, and/or enhancing contrast. They were then thresholded to create binarized images. The Skeletonize (2D/3D) and Analyze Skeleton (2D/3D) plugins in the FIJI software were then applied to quantify the vessel features.

Angiogenesis Fibrin Dot Image Analysis

For spheroid angiogenesis involving NHLFs as support cells, a minimum of 4 fibrin dot samples were analyzed for each culture condition. To determine the percent coverage of sprouts by a particular cell type, the merged image served as an overlay mask to the single channel images from which an area fraction can be calculated (Supplementary Fig. 6). Before this quantification is performed, the image serving as the overlay mask was first processed to remove the spheroid center as well as to subtract background noise. The merged image was then thresholded, converted to binary mask, and saved as an ROI. Corresponding single channel images were also processed and converted to binary images before applying the overlay mask on them for analysis. Values are presented as median percent area coverage ± standard deviation (SD). Besides the percent area coverage, the number of sprouts projecting out from the spheroid center was also counted for each culture condition.

For the spheroid angiogenesis study using HBVPs, an alternative approach was used to quantify the angiogenic sprouts. Four fibrin dots per culture condition were imaged and analyzed. For each spheroid, the percent area coverage of both control and Sox17-activated HUVECs was achieved by binarizing their corresponding channels and calculating the area fraction of the subsequent region of interests (ROIs) within a set sample field.

Vasculogenesis AIM Biotech Chip Vessel Analysis

A minimum of 3 chips were imaged for each culture condition. Using the Zeiss LSM800 confocal microscope, z-stack images of the vascular network within the central channel were acquired. Images were then processed for background subtraction before compressed into maximum intensity projections. Binary images were then created, which were then used to quantify vessel characteristics using the FIJI Skeletonize (2D/3D) and Analyze Skeleton (2D/3D) plugins.

To study HBVP recruitment to the vessels, the vessel to pericyte coverage was quantified. The samples were fixed and stained for NG-2, a pericyte marker, prior to image acquisition. Z-stack image sets were processed and then compressed to maximum intensity projections. The individual channels for pericytes and vessels were then converted into binary images. The vessel binary images then served as a ROI mask for the corresponding pericyte binary images to calculate percent area coverage by pericytes.

Angiogenesis AIM Biotech Chip Sprouting Analysis

In the AIM Biotech chips, HUVECs seeded in the side channel migrate into the exposed fibrin hydrogel between the microposts and form stable angiogenic sprouts due to the support cells (NHLFs or HBVPs) encapsulated inside the fibrin hydrogel. At desired timepoints, z-stack images are acquired, processed for background subtraction, and converted into maximum intensity projections. From these processed images, two features were quantified: the number of sprouts forming between the micropost gaps and the Euclidean distance of a sprout’s farthest endpoint from the hydrogel edge.

Statistical Analysis

All experiments performed in biological triplicates at minimum. Using the GraphPad Prism software, data is reported as mean ± standard deviation (SD) unless otherwise specified. Significance of differences was determined by unpaired Student’s t-test, Mann-Whitney U test, one-way ANOVA, or two-way ANOVA with Bonferroni post hoc test depending on the experimental parameters.

RESULTS

Sox17 drives primary venous endothelial cells toward an arterial gene expression profile.

Sox17 is highly expressed in arterial ECs but not venous ECs in vivo. In vitro cultured ECs (either arterial or venous ECs) lose arterial identity and express very little Sox17. To re-enable Sox17 expression in cultured ECs, we transduced primary human umbilical venous endothelial cells (HUVECs) with a Tet-On lentivirus expressing the Sox17 gene that can be induced temporally with the addition of doxycycline. Sox17 lentivirus was transduced into HUVECs at a multiplicity of infection (MOI) of 2–5 and underwent antibiotic selection to ensure 100% transfection. These Sox17 lentivirus infected HUVECs were used for subsequent experiments within 3 passages post-transduction. qRT-PCR of HUVECs induced with doxycycline for 48h confirmed the increase of Sox17 gene expression in a dose-dependent manner (Fig. 1a). Gene expression levels of Sox17 start to plateau around 5-fold after 100ng/mL. Sox17 protein expression also increases with increasing doxycycline dosage (Supplementary Fig. 1).

FIGURE 1. Sox17-HUVECs promotes gene expression profile towards arterial identity.

A) Sox17 gene expression after 48h in response to various doses of doxycycline. GAPDH served as endogenous control, and values normalized to 0ng/mL doxycycline-dosed HUVECs. n=3, one-way ANOVA with Bonferroni post-hoc test, column has statistical significance compared to corresponding groups denoted by letter above. B) Microarray heat map comparing relative fold change of dox(+)/dox(−) Sox17-HUVECs to that of in vivo mature mouse aorta/mouse vena cava for arteriovenous identity. Sox17-HUVECs were cultured as a monolayer dosed with 100ng/mL doxycycline in EGM2 medium for 48h and lysed for RNA isolation. C) mRNA expression levels for arterial genes of Sox17-mediated HUVECs cultured with 100ng/mL doxycycline for 48h. GADPH served as endogenous control, and values normalized to control HUVECs. n=3, Student’s t-test, *P<0.05, **P<0.01, ***P<0.001. D) mRNA expression levels for venous genes of Sox17-mediated HUVECs cultured with 100ng/mL doxycycline for 48h. GADPH served as endogenous control, and values normalized to control HUVECs. n=3, Student’s t-test, *P<0.05, **P<0.01, ***P<0.001.

EphrinB2 gene expression, an arterial marker, also increases significantly at the 100ng/mL dose and continues to rise with higher doses of doxycycline. At 2000ng/ml, the fold change peaks around 10-fold.

EphB4, a venous marker, exhibits a downward trend starting at 100ng/mL and shows a statistically significant reduction at 500ng/mL (0.6-fold decrease). The subsequent microarray analysis is more sensitive at detecting gene expression levels, detected decrease in EphB4 expression for Sox17-HUVECs cultured as a monolayer with 100ng/mL doxycycline for 48h (Fig. 1b).

Additionally, analysis of other arteriovenous genes through microarray and qRT-PCR confirms that a simple induction of Sox17 alone is sufficient to promote many early-stage arterial markers while suppressing that of venous (Fig.1b,c,d), such as Nrp1, Notch1, Notch4, Jag1, Cx40, EphrinB2, DLL4 for arterial and COUP-TFII for venous. To verify that Sox17 is also effective to induce arterial markers in another cell system, we also performed the experiments using human aortic endothelial cells (HAECs). Previously, we and others have shown that cultured arterial ECs gradually loss their arterial markers compared to those freshly isolated. Here, we found that HAECs, transduced with Sox17, also exhibited similar gene expression trends: marked increase in Notch4, Jag1, Cx40, EphrinB2, and Dll4 concomitant with decrease in venous markers (Supplementary Fig. 2a,b). For comparison, we also performed gene expression analysis in freshly isolated RNA from ECs of adult mouse aorta and vena cava, as a means to validate arterial vs. venous gene expression pattern in adult ECs. Surprisingly, we found that some of the well-published early arterial markers such as Nrp1, EphrinB2, DLL4, Jag1, and Hey1 were not differentially expressed in adult artery vs. vein. This suggests some of the early arterial markers are not maintained throughout adulthood, and the difference between arterial vs. venous expression disappears. On the other hand, several genes were persistently expressed in adult arterial ECs, notably, Sox17, Notch1, Cx40, EPAS1, Sema3g, which also followed the expression pattern of Sox17-upregulated genes. Together, these data suggest that Sox17 not only plays a critical role in early arterial development as demonstrated by other studies, but also persists throughout adulthood and controls the expression of several adult arterial markers, such as Cx40 and Sema3g. On the other hand, the expression of some other early arterial markers, though under the control of Sox17, may be suppressed by other unknown transcription factors in the adult vessels.

Sox17 enhances arterial genetic profile in response to shear stress.

Previously, it was thought that arterial hemodynamics initiate the arterial program during development when the heart starts pumping. However, recent evidence suggested that the arterial identity is predetermined by endogenous genetic/epigenetic programs even before the onset of blood flow, and hemodynamic flow further enhances or maintains the arterial identity later on after blood flow starts. To evaluate whether arterial flow condition can upregulate arterial markers in adult ECs, we investigated the effect of laminar shear stress on control HUVECs (−Dox) and Sox17-HUVECs (+Dox). To achieve this effect, we designed a cone-and-plate device that can apply constant laminar shear on a monolayer of cells (Fig. 2a). The well was seeded with 0.8×10⁶ cells one day before flow. On Day 0, fresh medium (with or without 1000ng/mL doxycycyline) was added. Shear was gradually increased over a 7-hour period until 10 dyn/cm² was reached, which was then maintained for 48 hours further. Medium was refreshed daily. Within 48 hours, both conditions of Sox17-HUVECs (−Dox, +Dox) exhibit alignment in the direction of shear (Fig. 2b,c).

FIGURE 2. Sox17 enhances arterial genetic profile in response to fluid shear stress.

A) Customized cone and plate shear device capable of applying laminar shear to 6-well plate or 10cm dish. B) Representative brightfield images of Sox17-HUVECs cultured as a monolayer with or without doxycycline (1000ng/mL) under static or constant shear stress (10dyn/cm²) for 48h. Scale bar=500um. C) Frequency distribution of the representative brightfield images depicting the extent of cell alignment under static or shear (10dyn/cm²) conditions with or without doxycycline (1000ng/mL) after 48h. D-F) mRNA expression levels of Sox17-mediated HUVECs cultured for 48h under static or shear (10dyn/cm²) conditions and with or without doxycycline (1000ng/mL). Gapdh served as endogenous control, and values normalized to static, no doxycycline condition. n=3, two-way ANOVA, Bonferroni post-hoc test, *P<0.05, **P<0.01, ***P<0.001.

We found that shear stress alone has very little influence on all arterial markers (Sox17, EphrinB2, Cx40, Cxcr4, Notch1, Sema3g, Nes, Dll4, Nrp1, Notch4) in the control HUVECs (−Dox) (Fig. 2d). Remarkably, when Sox17 is present in HUVECs (+Dox), it not only induces higher expression of arterial markers, but also further enhances their expression under flow condition. Notably, EphrinB2, Cx40, DLL4 and Notch4 are significantly up-regulated by flow. In particular, EphrinB2, Cx40 and Notch4 demonstrate very high expression levels when Sox17 is combined with flow, making them more arterial-like. Interestingly, some other early arterial markers, i.e., Cxcr4 and Nrp1, are diminished when both Sox17 activation and shear are introduced. This is more consistent with adult EC expression pattern as some of these early arterial markers diminish in adult vessel (Fig. 1b). As a negative control, we also checked eNOS, which is not regulated by Sox17. As expected, eNOS gene is unaffected by Sox17 activation but increases under shear for both −Dox and +Dox conditions (Fig. 2e). Venous markers, CoupTFII and Nrp2, reduce expression under Sox17 activation. Furthermore, though shear increases gene expression of both CoupTFII and Nrp2, this rise is significantly less in +Dox Sox17-HUVECs. EphB4 seems to be only modestly reduced by Sox17 activation even in the high dose range of doxycycyline (Fig. 1a,1b,1d,2e). Instead, EphB4 shows to be highly responsive to shear, increasing in gene expression levels in both +Dox and −Dox conditions (Fig. 2f). Overall, compared to static control, the presence of both Sox17 and flow have relatively small influence on the final expression levels of venous markers.

Laminar shear increases Sox17 and decreases CoupTFII protein expression.

Because Sox17 was already induced artificially in the modified cell line, its gene expression in response to laminar shear was not significantly increased (Fig. 2d). However, when cells were stained for Sox17 protein for each condition, we observed an increase in the percentage of cells positive for Sox17 protein expression when shear stress was introduced (Fig. 3a,c,e), suggesting that Sox17 expression in a sub-population of cells are sensitive to flow condition. The percentage of cells with low-to-middle protein expression increased from 18.4% to 33.9%. High intensity Sox17 expression percentages increased from 1.1% to 1.9%. In contrast, −Dox condition for both static and shear conditions exhibited no Sox17 protein expression as expected.

FIGURE 3. Laminar shear increases Sox17 and decreases CoupTFII protein expression.

A,B) Representative immunofluorescent images of Sox17-mediated HUVECs cultured as a monolayer with or without doxycycline (1000ng/mL) under static or constant shear (10dyn/cm²) for 48h. Scale bar=100um. C-F) Quantitative image analysis counting the percentage of Sox17-HUVECs expressing Sox17 or CoupTFII protein at various intensities. Samples were either under static or shear conditions; doxycycline dosage is 1000ng/mL for 48h. Expression intensity determined by setting global arbitrary pixel limits (Here: none: x ≤ 500, low to mid: 500 < x < 950, high: x ≥ 950). n=3; minimum of 5 field areas taken per sample; two-way ANOVA and Bonferroni post-hoc test; *P<0.05, **P<0.01, ***P<0.001.

We also investigated CoupTFII protein expression as +Dox samples remained significantly suppressed in gene expression under shear condition (Fig. 2f). For −Dox samples, low-to-middle expression of CoupTFII was quantified at 76% and 83.3% for static and shear conditions, respectively (Fig. 3b,d,f). With +Dox samples, low-to-middle expression of CoupTFII reduced to 56.3% and 51.3% for static and shear conditions, respectively. A more interesting trend is observed when looking at high intensity CoupTFII protein expression. In −Dox samples, the percentage of cells expressing high protein levels of CoupTFII increased from 0.82% to 15.9% when shear was introduced. The enhancing effect of shear is also seen in +Dox conditions as the percentages increased slightly from 4.5% to 8.7% though this wasn’t statistically significant. However, Sox17 clearly has an overarching suppressive effect in the number of cells expressing CoupTFII.

Sox17-mediated de novo vessel formation earlier in fibrin gel, and prefer to form cohesive vascular networks with one another rather than with venous ECs.

Next, we studied the Sox17-mediated vasculogenic potential in ECs. When Sox17-HUVECs were embedded in fibrin gel with primary human lung fibroblasts (NHLFs) as support cells, the +Dox condition led to earlier vessel formation as elongation of cells were seen as early as Day 3 (Fig. 4a, Supplementary Fig. 4a). Indeed, image analysis of the Day 3 +Dox vessels indicate that an average of ~9 vessels per field area were measured >100um in length compared to that of −Dox condition, which averaged ~2 vessels (Fig. 4b, Supplementary Fig. 4b). This observation was reproduced in an alternative experiment with different HUVECs:NHLFs seeding ratio (Supplementary Fig. 3).

FIGURE 4. Sox17-mediated HUVECs form de novo vessels earlier in fibrin gel, and prefer to form cohesive vascular networks with one another rather than with control venous HUVECs.

A) Representative images on Day 3, Day 5, and Day 8 of 3D vascular network formation in fibrin gel of Sox17-HUVECs co-transduced with second lentivirus containing tdTomato gene (red). NHLFs co-seeded as support cells to encourage vessel formation at 1:1 ratio (HUVECs:NHLFs). Doxycycline dose, if added, is 1000ng/mL. Scale bar = 500um. B) Image analysis of the Day 3 fibrin dot encapsulating Sox17-HUVECs (red) and NHLFs. The number of branches that were above 100um in length were quantified. Sox17 activation encourages earlier elongation of vascular cells. Values are mean ± SD of 6 biological replicates. Mann-Whitney U test, **P<0.01. C) Representative images of vascular formation on Day 7 of Sox17-HUVECs (ZsGreen, green) and control HUVECs (tdTomato, red) seeded at 1:1 ratio in fibrin gel. Sox17-HUVECs prefer to form vessel networks with one another compared to the no doxycycline condition where there is a more even distribution of both cell types. NHLFs used as support cells. Doxycycline dose, if added, is 1000ng/mL. Scale bar = 500um. D) Image analysis of the Day 7 vessel network for the fibrin dot co-culture containing Sox17-HUVEC (green) and control HUVECs (red). Graphs show average branch length (um) of each cell type quantified in no doxycycline and doxycycline (1000ng/mL) conditions. Branch length tends to be longer for both control and Sox17(+) HUVECs under dox(+) conditions indicating there’s wasn’t as much integration of the two cell types. Values are mean ± SD of 6 biological replicates. Mann-Whitney U test, **P<0.01.

Since Sox17-mediated HUVECs exhibit an arterial-like phenotype, their vessel formation and interaction with venous vascular cells were investigated. When control HUVECs were combined alongside with Sox17-HUVECs at 1:1 in the fibrin gel, Sox17-HUVECs tended to form networks with one another in the +Dox condition compared to that of control condition (−Dox), which show equal and random interaction between red and green cells (Fig. 4c). Analysis of average branch length per cell type indicate that both Sox17-HUVECs and control HUVECs show increased average branch length (79.3um and 59um, respectively) and elevated number of branches (11 and 9, respectively) >300um within a given field area when Sox17 is induced (Fig. 4d). In comparison, −Dox conditions led to lower average branch length (71um and 48.8um for Sox17-HUVECs and control HUVECs, respectively) and reduced branch number (4 and 5 for Sox17-HUVECs and control HUVECs, respectively).

Sox17-activated HUVECs exhibit a gene expression profile favoring a tip cell identity and demonstrate a preference to form sprouts with each other instead of with venous ECs.

Given that Sox17 has been implicated in tip cell phenotype, we next investigated their angiogenic ability in vitro. Sox17-HUVECs induced with doxycycline do upregulate many genes involved in tip cells, e.g. Pdgfb, Esm1, Dll4, Vegfr2, Unc5b, Tie1, Nrp1, Notch4, EphrinB2, Cxcr4, Igf2, Igf1r, and CD34 (Fig. 5a). Meanwhile, a few stalk markers, such as Tie2 and Robo4, are diminished.

FIGURE 5. Sox17-activated HUVECs exhibit a gene expression profile favoring a tip cell identity. Sox17-HUVECs angiogenic capability is similar to control venous HUVECs, but again demonstrates a preference to form sprouts with other Sox17-HUVECs when in co-culture with control HUVECs.

A) Microarray heat map comparing relative fold change of dox(+)/dox(−) Sox17-HUVECs for tip cell/stalk cell identity. Sox17-HUVECs were cultured as a monolayer dosed with 100ng/mL doxycycline in EGM2 medium for 48h and lysed for RNA isolation. B) Schematic detailing the protocol of spheroid construction and encapsulation in fibrin gel with NHLFs dispersed throughout as support cells. C) Representative images of Day 7 Sox17-HUVEC spheroids (red) encapsulated in fibrin gel with NHLFs and cultured in dox(−) or dox(+) conditions. Spheroids under both conditions formed angiogenic sprouts with hollow lumens in some areas (white arrowheads). Doxycycline, if added, was at 1000ng/mL. Scale bar=500um. D) Image analysis quantifying the number of sprouts per Sox17-HUVEC spheroid on Day7 for both media conditions. Though there seems to be a downward trend with increasing doxycycline dose, there was no significant difference, on average, in the number of sprouts forming from the spheroid between the two media conditions. Values are mean ± SD of 6 fibrin dots as biological replicates. One-way ANOVA with Bonferroni post-hoc test, ns=not significant. E) Representative images of spheroids on Day 7 composed of Sox17-HUVECs (green) and control HUVECs (red) in the fibrin gel with NHLFs as support cells. Scale bar=500um. F) Image analysis of the Day 7 Sox17-HUVEC and control HUVEC co-culture spheroids quantifying the percent area coverage by each cell type along the length of the sprouts under both media conditions. Sox17-HUVECs, when activated with doxycycline, form sprouts predominantly with themselves while suppressing control HUVECs sprouts. Values are mean ± SD of 6 fibrin dots as biological replicates. Student’s t-test with Welch’s correction, *P<0.05, **P<0.01, ***P<0.001. G) Image analysis of Day 7 Sox17-HUVEC and control HUVEC co-culture spheroids quantifying the number of sprouts per spheroid for both media conditions. With doxycycline added, the co-culture spheroids produced less sprouts, on average, than their doxycycline(−) counterparts. Values are mean ± SD of 6 fibrin dots as biological replicates. Student’s t-test with Welch’s correction, *P<0.05, **P<0.01, ***P<0.001.

To test its angiogenic ability, Sox17-HUVECs were aggregated into spheroids by the hanging drop method and encapsulated into fibrin hydrogel with fibroblasts to promote stable sprout formation (Fig. 5b). After 7 days of culture, stable sprouts are observed in both −Dox-) and +Dox conditions (Fig. 5c). Indeed, lumen was observed in some areas of the vessels in both conditions (white arrowheads). There is no statistical significance to the average number of sprouts emerging per spheroid in both −Dox and +Dox conditions (Fig. 5d).

We then investigated the sprouting ability of Sox17-mediated HUVECs and control HUVECs when co-cultured together. Both cell types were seeded in equal parts within the spheroid, and this was confirmed visually on Day 1 (Supplementary Fig. 5). As expected, when doxycycline is absent, Sox17-HUVECs integrate without bias alongside the control HUVECs (Fig. 5e, Supplementary Fig. 5). There is a near equal coverage of Sox17-HUVECs (60%) and control HUVECs (40%) within a given sprout (Fig. 5f). However, when doxycycline (1000ng/mL) is added, Sox17-HUVECs out-compete control HUVECs in sprout formation and in fact, suppress the sprouting capability of the control HUVECs. The median percent area coverage of Sox17-HUVECs is 95%, while that of control HUVECs is 5% in the +Dox condition.

Sox17-HUVECs exhibit hypersprouting when co-cultured with pericytes while suppressing angiogenic behavior of control HUVECs.

Again, Sox17-induced HUVECs spheroids co-cultured with NHLFs doesn’t increase the number of angiogenic sprouts (Fig. 5c). This observation is contrary to that seen in previous studies [20, 22, 23]. Therefore, we looked into the angiogenic behavior of these modified cells when co-cultured with primary human brain vascular pericytes (HBVPs). Spheroids consisting of either control HUVECs (green) or Sox17-HUVECs (red) were embedded together in the fibrin gel with pericytes, and sprouting capability was measured (Fig. 6a). Unlike fibroblasts, the presence of pericytes causes hypersprouting of Sox17-induced HUVECs. The percent area coverage by the vessel outgrowth from each spheroid for Sox17-HUVECs is significantly higher in the +Dox condition (43%) than that of the -Dox condition (15.5%) (Fig. 6b). Similar as before, Sox17 activation restrains sprouting of the surrounding control HUVEC spheroids.

FIGURE 6. Sox17-HUVECs exhibit hypersprouting when co-cultured with HBVPs while suppressing angiogenic behavior of control HUVECs.

A) Representative stitched panoramic image on Day 7 of control HUVEC spheroids (eGFP, green) and Sox17-HUVEC spheroids (tdTomato, red) encapsulated in a fibrin dot co-seeded with HBVPs. Doxycycline dosage is 1000ng/mL. Scale bar = 500um. B) Image analysis quantifying the percent coverage in a given sample field by Day 7 control HUVEC (eGFP, green) and Sox17-HUVEC (tdTomato, red) spheroids. Sox17-HUVEC spheroid sprouts migrate farthest and cover more area than their control counterparts. Values are mean ± SD of 4 fibrin dots as biological replicates. Student’s t-test with Welch’s correction, *P<0.05, **P<0.01, ***P<0.001. C) Representative stitched panoramic image on Day 7 where each spheroid contains both control HUVECs and Sox17-HUVECs and immobilized in fibrin gel co-seeded with HBVPs. Doxycycline dosage, if added, is 1000ng/mL. Scale bar = 500um. D) Image analysis of the mixed spheroids (Sox17: tdTomato and control: eGFP) on Day 7 measuring the percent coverage of a given sample field. The Sox17-HUVEC spheroid angiogenic sprouts cover more area than their control counterparts and impedes sprout formation of the control HUVECs. Values are mean ± SD of 4 fibrin dots as biological replicates. Student’s t-test with Welch’s correction, *P<0.05, **P<0.01, ***P<0.001.

Similar to the NHLF co-culture study, we also looked at the angiogenic potential of spheroids composed of both Sox17-HUVECs and control HUVECs. In the presence of pericytes, Sox17-activated HUVECs again exhibit excessive sprouting, increasing the area coverage from 10.2% to 21% (Fig. 6c). In contrast, control HUVECs show reduced percent area coverage, from 2.6% to 0.8%, when compared to that of −Dox condition (Fig. 6d). Again, it appears that Sox17-activated HUVECs suppress the control HUVECs.

Contrary to control HUVECs which widen under interstitial flow, Sox17-HUVECs form thinner vessels under both static and interstitial flow conditions.

Next, we investigated the effect of interstitial flow on vascular formation using Sox17- HUVECs. To test this, we used AIM Biotech microfluidic chips in which a fibrin hydrogel containing HUVECs and NHLFs can be injected into the central channel and interstitial flow can be applied across this hydrogel. Interstitial flow is created by hydrostatic pressure from the difference of media volumes in the two side channels flanking the hydrogel channel (Fig. 7a). In this study, a 40uL volume difference which generates ~1mm H₂O pressure difference was applied. By Day 10, both doxycycline conditions formed cohesive vascular networks with lumen (Fig. 7b, white arrowheads). However, there are clear differences in vessel characteristics. In control HUVECs, vascular lumen typically widened under interstitial flow; however, it remained very thin in Sox17-HUVECs vessels. Branch analysis indicates longer total branch length, overall thinner vessels with smaller lateral diameter, and increased sum number of branches and skeletons (Fig. 7c).

FIGURE 7. Compared with control HUVECs which widen under interstitial flow, Sox17-HUVECs form thinner vessels and more branches when cultured with NHLFs as support cells. Furthermore, when cultured concurrently with control HUVECs under interstitial flow, vascular network formation of both cell types is disrupted.

A) Schematic of culture setup in the AIM Biotech microfluidic chips. Central channel contains the fibrin gel encapsulating HUVECs and NHLFs. Side channels contain media with one side containing a higher volume (40uL difference) to create interstitial flow across the central channel. B) Representative confocal images on Day 10 of Sox17-HUVECs (red) seeded with NHLFs in AIM Biotech chips under interstitial flow (blue arrows indicate direction) with or without doxycycline (1000ng/mL). Both doxycycline conditions yield uninterrupted vessels with lumen formation (white arrowheads). Scale bar = 500um. C) Image analysis of the Day 10 vascular networks formed by Sox17-HUVECs in AIM biotech chips under interstitial flow. Sox17-HUVECs tends toward longer branch length, more branching, and small diameter vessels under interstitial flow. Student’s t-test with Welch’s correction, *P<0.05, **P<0.01, ***P<0.001. D) Representative images on Day 14 of AIM Biotech chips co-seeded with Sox17-HUVECs (red), control HUVECs (green), and NHLFs with interstitial flow applied (blue arrows indicate direction). Addition of interstitial flow leads to disruption of vessel formation of both Sox17-HUVECs and control HUVECs for the doxycycline condition. Doxycycline concentration is 1000ng/mL if added. Scale bar = 500um.

Since Sox17-HUVECs demonstrated preferential vessel formation with their own cell type in static fibrin dot culture (Fig. 4c), we subsequently examined if this behavior persists under interstitial flow conditions. In the microfluidic chips, Sox17-HUVECs (tdTomato) and control HUVECs (eGFP) at 1:1 ratio were seeded with NHLFs inside the fibrin scaffold and cultured for 14 days. Unlike the static culture where Sox17-induced HUVECs demonstrated a bias to form cohesive branches with their own cell type resulting in longer segments of same-cell vessels, introduction of interstitial flow led to disrupted vascular network in the +Dox condition for both cell types (Fig. 7d).

We also examined if exchanging the support cells from NHLFs to HBVPs will alter the effects of interstitial flow on the Sox17-mediated vascular network. Once again, AIM Biotech microfluidic chips were set up in the same manner with Sox17-HUVECs and pericytes suspended in the fibrin hydrogel at the central chamber with a 40uL media volume difference applied across the two side channels. There was successful network formation by Day 7 in that pericytes were recruited to the vessels and lumen formation was detected for all four conditions (Fig. 8a,b). When the vessel characteristics were analyzed more closely, a couple of features align similarly with that of the vascular network co-cultured with NHLFs (Fig. 7b). Under the −Dox condition, the vessel diameter widened with the addition of IF. When Sox17 is induced (+Dox), it displayed thinner vessels (smaller lateral area and diameter) under static as well as after IF was applied. Sox17 activation also led to longer total branch length compared to the −Dox conditions. Here, we also observed the typical behavior of venous HUVECs when under interstitial flow. Besides the widening of the lumen (lateral vessel area and diameter), there was increased connectivity of the vessels depicted by longer total branch length, increased branching, and lower number of skeletons (Fig. 8c). In contrast, IF introduction to Sox17-HUVECs did not cause drastic morphological differences when compared to its static counterpart.

Figure 8. With HBVPs used as support cells, Sox17-activated HUVECs continue to exhibit thinner diameter lumenized vessels in both static and interstitial flow conditions when compared to the control HUVEC vascular network.

A) Representative images on Day 7 of vascular network formed de novo in the AIM Biotech chip with Sox17-HUVECs (Td-Tomato, red) and HBVPs (NG2 stained, green) under static and interstitial flow conditions. Ratio of seeding density for Sox17-HUVECs:HBVPs is 2:1. Doxycycline dose, if added, is 1000ng/mL. Scale bar = 100um. B) Orthogonal projection of Day 7 vasculature confirming lumen formation with pericytes surrounding the vessel. Scale bar = 100um. C) Image analysis quantifying the Day 7 vessel characteristics of Sox17-HUVECs under the four conditions. Values are mean ± SD of 4 biological replicates. Two-way ANOVA with Bonferroni post-hoc test, *P<0.05, **P<0.01, ***P<0.001. D) Image analysis quantifying the percent area coverage by HBVPs on the Sox17-HUVEC vascular network on Day 7. Sox17-activated HUVECs show no change, if not a slight downward trend, in pericyte percent coverage for Day7 in response to interstitial flow when compared to that of static culture. This results contrast with the no doxycycline condition where pericyte coverage increased with interstitial flow. Values are mean ± SD of 4 biological replicates. Two-way ANOVA with Bonferroni post-hoc test, *P<0.05, **P<0.01, ***P<0.001.

We also examined the pericyte coverage of the vessel network. To determine the level of coverage, the vessel network served as a binary mask to quantify how much the pericytes overlap the ROI. By Day 7, the pericyte coverage was highest for the IF −Dox condition where the vessels had widened substantially in response to flow caused by the hydrostatic pressure difference. In comparison, in this particular experiment, Sox17-activated vascular networks show no enhancement in pericyte coverage under IF condition. When looking solely at static cultures, Sox17-induced vessels do show a minor increase in pericyte coverage when comparing to the -Dox condition, but again, this was not seen to be significant.

Under static conditions, Sox17 activation leads to increased angiogenic sprouting in the AIM Biotech microfluidic chips. However, in the presence of interstitial flow this increase is diminished.

Finally, we explored the effect of interstitial flow on the angiogenic potential of the Sox17-induced HUVECs. To set up this study, the AIM microfluidic chips’ central channel was filled with support cells (either NHLFs or HBVPs) suspended in fibrin hydrogel (Fig. 9a). Sox17-HUVECs and control HUVECs were co-seeded at high density in one of the side channels while the other side channel was filled solely with media. Interstitial flow for this study was set against the direction of angiogenesis, so the higher volume of media (40uL) was introduced into the reservoirs opposite the channel seeded with HUVECs.

Figure 9. Under static conditions, Sox17 activation leads to increased angiogenic sprouting in the AIM Biotech microfluidic chips. However, addition of interstitial flow diminishes the increased sprouting. Sox17(+) sprout lengths under interstitial flow are less than those under static condition.

A) Schematic of culture setup for AIM Biotech microfluidic chips. Central chamber contains the support cells (NHLFs or HBVPs) in fibrin gel. Side channels contain either media or HUVECs (Sox17 and control). Interstitial flow is set against the sprout direction by adding a 40uL media height difference to the wells on the right side. B) Representative images on Day 10 of sprouts forming at the gaps between microposts. The left side channel was seeded with an equal mixture of Sox17-HUVECs (ZsGreen, green) and control HUVECs (tdTomato, red). Central channel is composed of NHLFs (1M/mL) embedded in fibrin gel. Doxycycline, if added, was 1000ng/mL. Scale bar = 100um. C) Quantification of the average number of sprouts (regardless of cell type) found per gap in (B). Highest number of sprouts come from the doxycycline(+) static condition. Values presented as mean ± SD of 3 biological replicates. Two-way ANOVA with Bonferroni post-hoc test, *P<0.05, **P<0.01, ***P<0.001. D) Representative images on Day 14 of sprouts. Left side channel was seeded with an equal mixture of Sox17-HUVECs (tdTomato, red) and control HUVECs (eGFP, green). Central channel consists of HBVPs (2M/mL) embedded in fibrin gel. Doxycycline concentration is 1000ng/mL, if needed. Scale bar = 500um. E,F) Quantification of sprout characteristics for channels in (D) with sprout lengths measured in Euclidean distance (E) and sprouting frequency depicted as average number of sprouts per gap (F). Sprout lengths for doxycycline(+) interstitial flow condition is statistically shorter than doxycycline (+) static condition. Both doxycycline(+) conditions, static and interstitial flow, yield more sprouts than doxycycline(−) static condition. No difference seen if comparing the two interstitial flow culture conditions with each other. Values presented as mean ± SD of 3 biological replicates. Two-way ANOVA with Bonferroni post-hoc test, *P<0.05, **P<0.01, ***P<0.001.

First, we investigated angiogenesis with NHLFs. Both Sox17-mediated HUVECs (green) and control HUVECs (red) were seeded into the side channel and cultured for 10 days (Fig. 9b). In the static conditions, Sox17-HUVECs activated by doxycycline do engage more readily in sprouting, i.e., more Sox17-HUVEC cells place themselves at the vascular front compared to the control HUVECs. In the IF conditions, similar trends are observed. Sprouting frequency for the Sox17-activated HUVECs show modest, elevated levels (Fig. 9c). The +Dox static condition averaged ~2 sprouts per gap compared to the −Dox static cultures, which is ~1 sprout per gap. No statistical significance was seen in the sprouting frequency between the two IF conditions though the +Dox condition trends slightly higher.

We also examined Sox17-induced angiogenesis with HBVPs as support cells. For this study, both Sox17-HUVECs (red) and control HUVECs (green) were injected into the side channel and cultured for 14 days. Similar to the NHLFs, doxycycline exposure results in greater instances of Sox17-HUVECs leading the sprout formation (Fig. 9d). Interestingly, when we measured the Euclidean distance of the sprouts from side channel-central channel boundary to the outermost section of the sprout, the IF +Dox condition yielded shorter sprouts than its static counterpart (Fig. 9e). Furthermore, interstitial flow did not significantly affect sprouting frequency in the +Dox conditions (Fig. 9f). No significance was also observed when comparing the sprouting frequency to that of IF −Dox condition as well. However, all three conditions averaged higher sprouting number per gap (above 2) when compared to the static −Dox condition.

DISCUSSION

Due to the contextual nature of Sox17, there is a potential benefit to simplifying the microenvironment using in vitro platforms to elucidate more clearly the factors influencing Sox17-mediated vascular phenotypical changes. With that in mind, we converted HUVECs, which is venous by nature, into Sox17-activated vascular cells using the Tet-On inducible lentivirus system. We then cultured these cells under shear devices, within 3D hydrogels, and within microfluidic chips. We were successful in clarifying some features of the Sox17-induced vascular system observed in previous reports [19, 20] as well as revealing some new findings.

First, upon Sox17 activation by doxycycline, these modified HUVECs exhibited an arterial gene expression profile. Venous genes became more obviously suppressed with higher concentrations of doxycycline. Sox17 regulates many early arterial markers as well as several adult artery markers. In particular, some of differentially expressed genes, such as Cx40 and Sema3g in adult arterial vs. venous ECs, are regulated by Sox17. The sustained expression of Sox17 in adult arterial ECs suggest that Sox17 is not only critical for early arterial development as demonstrated by other studies, but also plays a role in maintaining the adult arterial EC identity, which warrant further studies.

We found that Sox17 can also induce the arterial gene expression in cultured human aortic ECs. Previously, we and others have shown that in vitro cultured arterial ECs display phenotypical drifting and gradually lose their arterial markers in culture, and exposing these cells to arterial flow has very little effect on re-introducing arterial markers. Thus, it is very difficult to recapitulate arterial phenotype in vitro in the cultured ECs, no matter whether it is arterial or venous origin. Our finding provides a new approach to reinstate the arterial markers in cultured ECs by reprogramming cells with Sox17, thus enable arterial EC phenotype in culture and can be applied in various organ/tissue-on-chip system where arterial ECs are needed.

When we applied arterial level shear stress, the Sox17-HUVECs retained their ability to align themselves in the direction of flow and demonstrated preference for arterial phenotype by upregulating arterial genes while downregulating venous ones compared to the control HUVEC shear culture. Importantly, some of the arterial genes, such as Cx40, EphrinB2 and Notch4, are further dramatically enhanced in response to fluid shear stress when Sox17 is present. In contrast, the control HUVECs show very little response to fluid shear stress in all the arterial markers. This observation corroborates with recent findings that vein grafts, when implanted into arterial circulation, demonstrate incomplete adaptation to arterial hemodynamics with essentially no increase in arterial markers. The incomplete adaptation of vein grafts to arterial circulation may contribute to the graft failure in the long term. Our findings, along with these studies, confirm that adult ECs have limited hemodynamic adaptability, whereas embryonic ECs have plasticity in which arterial identity is highly influenced by hemodynamics. Our finding that Sox17 enhances adult ECs adaptability to hemodynamics may have clinical significance. Re-introducing Sox17 to the vein grafts may promote better vessel adaptation to arterial environment thus improve clinical outcomes of vein bypass surgery.

Similar to observations by Corada et al., Sox17 activation led to increase in gene expression of those in the Notch pathway, e.g., Notch1, Notch4, Dll4, and Jag1 [19]. Notch inhibition by DAPT, a Notch inhibitor, did lead to an increase in percentage of cells expressing high intensity levels of Sox17 protein within the nucleus (Supplementary Fig. 7). This result aligns with those seen by Lee and colleagues where Notch inhibition by DAPT or α-Dll4 antibody led to increased Sox17 protein expression in HUVECs, postnatal lung ECs, and tumor ECs [20]. This increase of Sox17 may be a compensatory mechanism to maintain certain level of Notch activity when endogenous Notch pathway is inhibited by DAPT. What is new and intriguing in this DAPT study though is that the overall percentage of Sox17+ cells (both mid and high expression) did not change, so it seems that DAPT only affects cells already actively expressing Sox17 protein.

It should be noted that the fold change of Sox17 transgene in the shear study is somewhat different than those observed in protein staining. One possibility is the difference between mRNA and protein expression. As another possibility, Sox17 protein expression in the nuclei was not uniform even though 100% of the cells have been transduced with Sox17 under antibiotic selection and all experience the same concentration of doxycycline. Immunofluorescent staining of Sox17 was found to be highly heterogenous among the cell population. This heterogeneity can be seen in protein staining, but not in RT-PCR results, which only reflect the average of all cells.

This high level of heterogenous Sox17 expression puzzled us as we see some cells do not express Sox17 even under Dox treatment. We initially hypothesized that Sox17 expression in one cell was causing an inhibitory cascade in the adjoining cells through the Notch mechanism. However, addition of DAPT did not drastically change this heterogenicity (Supplementary Fig. 7). A similar phenomenon has been documented before by Yu et al. where they found several breast cancer cell lines as well as mammary epithelial cells transduced with a Tet-On lentivirus drastically lose inducibility when not exposed to doxycycline for a period as short as 7 days [34]. They learned that this silencing is variable and cell-line dependent, but can at least be prevented with continuous doxycycline inducer. In terms of cause, the expression loss is due to chromatin modification leading to an inactive promoter. Indeed, they demonstrated that chromatin silencing of their cell lines can be reversed with an HDAC (histone deacetylase) inhibitor, sodium butyrate. Our own preliminary studies also show that sodium butyrate does improve the overall percentage of cells expressing Sox17 as well as the intensity of protein expression, reversing the suppressive effects that DAPT could not accomplish (Supplementary Fig. 8).

Likely, Sox17 transgene repression in the sub-population of transduced HUVECs was also due to chromatin silencing but why this is the case is still uncertain. It could be due to the limitations of the Tet-On lentivirus platform or due to existing epigenetic mechanisms inherent in a shear responsive venous ECs. However, it is confirmed in this present study that shear stress or histone modifications can reverse this loss of inducibility as the number of cells positive for Sox17 nuclear protein expression increased overall after only two days of arterial shear or three days of sodium butyrate treatment. Nevertheless, with this unexpected limitation of the Tet-On system in mind, care was taken to use only recently transduced cell lines for the rest of the present study.

Through the use of 3D fibrin hydrogel cultures, we’ve uncovered several key and novel morphogenic behaviors of Sox17-mediated HUVECs. In vasculogenic studies where Sox17-HUVECs and control HUVECs were dispersed together, Sox17-HUVECs under doxycycline induction demonstrated a bias in forming vessels with its own cell type while suppressing the control HUVECs to form vessels. This preference was reflected in the presence of longer branch lengths compared to the −Dox condition. In a similar fashion, in the angiogenic studies featuring both Sox17-HUVECs and control HUVECs within the same spheroid, Sox17-HUVECs dominated sprouting and suppressed vessel outgrowth from the venous control HUVECs. This striking behavior remains consistent when using either fibroblasts or pericytes as support cells. The preference of Sox17-HUVECs to form vessels of its own while suppressing the control HUVECs may be related to EphrinB2-EphB4 mediated repulsion during early arterial-venous patterning, as Sox17-HUVECs have higher EphrinB2 while control HUVECs have higher EphB4 expression. In the future, such model can be used to investigate the mechanisms of arterial-venous patterning when selected population or patterns of ECs are upregulated with Sox17.

Sprouting frequency, however, appears to be one characteristic that is dependent on support cell type. When performing spheroid angiogenesis with NHLFs on either Sox17-HUVECs alone or in a combined approach with control HUVECs, differences in sprouting frequency between the −Dox and +Dox conditions were minor, if not statistically significant. However, when HBVPs were used, Sox17-HUVECs yielded vastly different morphogenic trends than the control HUVECs. Regardless whether or not Sox17-HUVECs were co-seeded with control HUVECs within the same spheroid, Sox17 activation led to aggressive vessel outgrowth and branching. This hypersprouting behavior is consistent with previous reports [20, 22, 23]. Collectively, this study infers that the surrounding cells within a vascular microenvironment significantly dictates the sprouting intensity of Sox17-mediated vascular cells. One of the possibilities that explains the different behaviors of NHLFs and HBVPs in supporting vascular formation is that ECs and pericytes interactions involve many intimate receptor-ligands, such as Notch, Ephrin and Neuropilin families. Sox17 upregulates many of these in ECs, such as Nrp1, EphrinB2, Notch4, DLL4, and Jag1, which may interact directly with the corresponding receptors on pericytes. On the other hand, NHLFs are not the endogenous supporting cell types in the body; they may support the vessel formation via secreting growth factors instead of direct ligand-receptor interactions. Thus, Sox17 mediated hypersprouting was not observed when NHLFs were used. In the future, such receptor-ligand interactions between ECs and pericytes mediated by Sox17 induction in the vasculogenesis and angiogenesis process can be further studied using our model.

Lastly, we incorporated interstitial flow into the 3D fibrin hydrogel cultures through the use of the AIM Biotech microfluidic chips. In both the fibrin dot cultures and AIM chips in static conditions, we observed overall thinner diameter vessel formation with Sox17 activation. However, it is in the AIM chip where we see for the first time clear morphological differences between −Dox and +Dox cultures in response to IF. First, IF does not lead to widening of the vessel diameter in Sox17-activated HUVECs, which is what was observed in control HUVECs. This morphological response is observed in both NHLF and HBVP culture conditions. In the control HUVECs, the lumen has a much larger diameter (40um) in response to IF, resembling malformed vasculature. On the other hand, the Sox-HUVECs were able to maintain the lumen diameter at 20um, which is closer to a capillary diameter in vivo under flow condition. These results suggest that Sox17 may enable vessel to maintain its diameter in response to varying hemodynamic conditions in an autonomous fashion. Second, interstitial flow diminishes sprouting in Sox17-HUVECs. During arterial venous differentiation, recent studies suggest that venous cells migrate against the flow toward the arterial side and acquire arterial phenotypes. This behavior is also observed in our control HUVECs angiogenic assay, in which angiogenic sprouts favored going against the interstitial flow direction more so than the static condition. However, when Sox17 is induced, this angiogenic increase in response to flow appears attenuated. This may suggest that cells may slow down in their migration after acquiring an arterial fate.

Taken together, this present study recaptures successfully in various in vitro models morphogenic trends of Sox17 observed in previous reports and also adds new insights into the Sox17-mediated vascular phenotypes. We found that Sox17 enables hemodynamic adaptability to arterial flow by upregulating arterial markers that were otherwise not responsive in venous ECs, and able to maintain vascular lumen diameter under changing conditions from static to interstitial flow. These observations suggest Sox17 may play a role in mechanobiology of arterial ECs adaptation under varying hemodynamic conditions. Furthermore, interstitial flow negatively impacts the sprouting ability of Sox17-mediated cells, which may attribute to the divergent findings for in vivo sprouting. It is intriguing to speculate to what extent physical stimuli play a role in Sox17-mediated vascular remodeling especially given that members of Notch pathways, which are heavily regulated by Sox17 as observed in this study, are known mechanosensors [35, 36]. The use of in vitro microphysiological system devices along with precisely controlled modulation of Sox17 in selected cells offer a unique opportunity to answer such questions and warrant future studies.

DATA AVAILABILITY

The raw data required to reproduce these findings are available to download from https://drive.google.com/open?id=1yRl6808ljyom9NCrNiv1h4p0h_zGLOcb&authuser=guohao.dai%40gmail.com&usp=drive_fs.

The processed data required to reproduce these findings are available to download from https://drive.google.com/open?id=1yRl6808ljyom9NCrNiv1h4p0h_zGLOcb&authuser=guohao.dai%40gmail.com&usp=drive_fs.