cSTAR analysis identifies endothelial cell cycle as a key regulator of flow-dependent artery remodeling

Abstract

Fluid shear stress (FSS) from blood flow sensed by vascular endothelial cells (ECs) determines vessel behavior, but regulatory mechanisms are only partially understood. We used cell state transition assessment and regulation (cSTAR), a powerful computational method, to elucidate EC transcriptomic states under low shear stress (LSS), physiological shear stress (PSS), high shear stress (HSS), and oscillatory shear stress (OSS) that induce vessel inward remodeling, stabilization, outward remodeling, or disease susceptibility, respectively. Combined with a publicly available database on EC transcriptomic responses to drug treatments, this approach inferred a regulatory network controlling EC states and made several notable predictions. Particularly, inhibiting cell cycle-dependent kinase (CDK) 2 was predicted to initiate inward remodeling and promote atherogenesis. In vitro, PSS activated CDK2 and induced late G₁ cell cycle arrest. In mice, EC deletion of CDK2 triggered inward artery remodeling, pulmonary and systemic hypertension, and accelerated atherosclerosis. These results validate use of cSTAR and identify key determinants of normal and pathological artery remodeling.

Fig. 1.. cSTAR analysis of EC states under different flow conditions.

(A to D) RNA-seq data from ECs under OSS, LSS, PSS, HSS, or STAT conditions for 24 hours were analyzed, and DPD scores were calculated. (A) Two-dimensional (2D) plot of phenotypic scores in the DPD_FSS and DPD_remod plane. (B) 2D plot of phenotypic scores in the DPD_FSS and DPD_OSS plane. (C and D) Top GSEA Hallmark (HM) gene sets that contribute to the STV_remod [(C)] and STV_OSS [(D)]. The components of STVs were used as the GSEA input for the GSEA hallmark gene set. Positive-normalized enrichment score (red) reflects an increase in a molecular process while moving along the corresponding STV, and negative-normalized enrichment score (blue) reflects a decrease. Size of circles corresponds to the number of genes in STV corresponding to the specific GSEA term. (E) Scatter plot of the DPD_remod and DPD_OSS scores calculated for in vivo single-cell RNA-seq data (33) from the left carotid artery partial ligation mice model: 2L, 2 days left carotid artery; 2R, 2 days right carotid; 14L,14 days left carotid; 14R, 14 days right carotid. Histograms on the top and right present distributions of DPD values across one axis.

Fig. 2.. cSTAR reconstruction of the core network driving EC cell fate decisions.

(A) Schematic illustrating the data analysis pipeline and the reconstructed core network. RNA-seq data from different flow conditions were used to obtain the STVs, and perturbation data from LINCS database (https://clue.io/data/CMap2020#LINCS2020) were used to infer the wiring of the core network and its connection to the phenotypic modules (DPD_remod and DPD_OSS). The resulting inferred network is presented at the bottom. Arrowheads indicate activation, and blunt ends indicate inhibition. Line widths indicate the absolute values of interaction strengths. Flow chamber was adapted from (81). (B) Global impact scores of the effect of perturbation of each core network module on the rest of core network components (blue) and on the cells’ phenotype (red) calculated as L2 norm of the relevant column in the global response matrix $\mathit{R}=-{\mathit{r}}^{-1}$ (Materials and Methods) (C) Global impact of the core network modules on the direction of vessel diameter change (calculated using the DPD_FSS) and vessel remodeling program regardless of the direction (calculated using the DPD_remod).

Fig. 3.. Characterization of cell cycle state under different FSS.

(A) Diagram of the FUCCI reporter. (B and C) FUCCI HUVECs were subjected to OSS, LSS, PSS, HSS, or static for 24 hours. Cells were fixed, stained with 4′,6-diamidino-2-phenylindole (DAPI), and imaged as described in Materials and Methods. Representative images for these conditions are shown in (B). Scale bar, 100 μm. (C) Quantification of FUCCI-red, FUCCI-green–positive cells, and double-negative cells. n = 6 experiments. (D and E) HUVECs were subjected to flow patterns as in (A), labeled, and stained for EdU as described in Materials and Methods. Percent of EdU-positive cells was quantified. n = 6 experiments. Scale bar, 100 μm. Data are presented as means ± SD. *P < 0.05, **P < 0.01, and ***P < 0.001. ns, not significant, calculated by one-way analysis of variance (ANOVA) with Tukey’s multiple comparison tests.

Fig. 4.. CDK2 activity under FSS.

(A) Schematic of CDK2 sensor. NLS, nuclear localization signal; NES, nuclear export signal; S, serines that are CDK consensus phosphorylation sites. (B) Schematic of CDK2 phosphorylation-mediated translocation of DHB-mVenus. (C and D) HUVECs expressing the DHB sensor were subjected to OSS, LSS, PSS, HSS, or STAT conditions for 6 hours. Cells were fixed and imaged as described in Methods, representative images shown in (C). Scale bar, 25 μm. Arrowheads are CDK2 high activity cells. Cell nuclei were identified using fluorescent H2B images to obtain a mask. Ratio of cytoplasmic/nuclear signal was quantified (D). n = 66 cells for each group from three independent experiments. Data showing all points from minimum to maximum. **P < 0.01 and ***P < 0.001, calculated by one-way ANOVA with Tukey’s multiple comparison.

Fig. 5.. CDK2 depletion induces early G1 arrest.

FUCCI HUVECs were transfected with control (siCtrl) or CDK2 (siCDK2) siRNA and analyzed on day 4. (A) CDK2 knockdown confirmed by Western blotting. (B) Cells were fixed and mounted with DAPI. Representative images are showed. Scale bar, 100 μm. (C) Quantification of cell cycle state. n = 5 experiments. Data are presented as means ± SEM. *P < 0.05, **P < 0.01, and ***P < 0.001, calculated by two-tailed unpaired t tests. (D and E) 2D plots of DPD_FSS and DPD_remod [(D)] and DPD_FSS and DPD_OSS [(E)] for CDK2 KN in STAT and PSS conditions. DAPI, 4′,6-diamidino-2-phenylindole.

Fig. 6.. Deletion of endothelial CDK2.

(A) Schematic for production of CDK2 iECKO mice. Cdh5-CreERT2; CDK2^flox/flox (CDK2 iECKO) and CDK2^flox/flox (ctrl) mice at 6 weeks of age were injected with tamoxifen. (B) Cdk2 deletion efficiency in isolated lung ECs confirmed by quantitative PCR. n = 4 mice per group. (C and D) Representative trace and quantification of right ventricle systolic pressure (RVSP) from control and CDK2 iECKO mice. (E) Quantification of left ventricle systolic pressure (LVSP) from control and CDK2 iECKO mice. n = 6 mice per group. (F) Representative SMA immunostaining in Ctrl and CDK2 iECKO entire lung sections. Scale bars, 1 mm. (G) Quantification of SMA⁺ muscularized arteries (diameter < 50 μm and > 50 μm) in Ctrl and CDK2 iECKO lungs. n = 6 mice per group. (H) Representative Masson’s trichrome staining of Ctrl and CDK2 iECKO lung sections. V, vessels. Arrows indicate narrowed and occlusive vessels. Data are presented as means ± SEM. *P < 0.05, **P < 0.01, and ***P < 0.001, calculated by two-tailed unpaired t tests [(B), (D), and (E)] and two-way ANOVA with Tukey’s multiple comparison [(G)]. i.p., intraperitoneal.

Fig. 7.. Atherosclerosis in CDK2 iECKO mice.

(A) Experiment timeline: CDK2 iECKO and Ctrl mice at ~6 weeks were treated with tamoxifen for five consecutive days. At 9 weeks, mice were injected with PCSK9 virus and 1 week later fed with HCHFD for an additional 12 weeks. (B and C) Representative images and quantification of ORO staining for male and female Ctrl and CDK2 iECKO whole aorta. (D) ORO and H&E staining of aortic root sections from Ctrl and CDK2 iECKO mice. Scale bar, 500 μm. (E) Quantification of ORO lesion area for both male and female Ctrl and CDK2 iECKO mice. n = 6 mice per group per gender. (F) Representative images and quantification of anti-myeloperoxidase (MPO; neutrophil marker) on aortic root sections from Ctrl and CDK2 iECKO mice. Scale bar, 250 μm. (G) Representative images and quantification of CD68 (macrophage marker) on aortic root sections from Ctrl and CDK2 iECKO mice. Scale bar, 250 μm. n = 6 mice per group. Data are presented as means ± SEM. *P < 0.05 and **P < 0.01, calculated by two-tailed unpaired t tests.