Fig 1.
Induction of TSP-1 via multiple mechanisms by hypoxia and TGFβ signaling.
An increase in the abundance of translatable TSP-1 mRNA in hypoxia results from the regulation by different pathways. Arrow symbol denotes activation, ⊣ symbol denotes repression. MicroRNAs that target TSP-1 (e.g. miR-18a) are less abundant in hypoxic conditions, together with the activation of different TSP-1 promoters, lead to an increase in intracellular TSP-1 production.
Fig 2.
Reaction diagram of TSP-1 regulation by hypoxia and TGFβ signaling.
(A) HIF stabilization in hypoxia, induction of let-7 and regulation of TSP-1 mRNA by miR-18a. Transcription of TSP-1 gene is modulated by different factors. (B) TGFβ signaling and calcium-mediated activation of NFATc1. Species whose names end with an N subscript are located inside the nucleus; reactions that point to red signs indicate degradation. The symbols v# in the two subparts (A and B) refer to the chemical reactions listed in S1 Table.
Fig 3.
Model optimization against experimental data in ECs.
(A-D) Experimental measurement and model simulation of phosphorylated SMAD1 and SMAD2 protein in response to 1 ng/ml TGFβ treatment in BAECs: (A) normalized phosphorylated SMAD1 protein level without CHX treatment, (B) normalized phosphorylated SMAD2 protein level without CHX treatment, (C) normalized phosphorylated SMAD1 protein level with CHX treatment, (D) normalized phosphorylated SMAD2 protein level with CHX treatment. The time-course experimental data from Valdimarsdottir et al. and simulation results are both normalized with respect to the corresponding peak value [71]. (E) Experimental data from Goumans et al. and model-generated dose response curve of total phosphorylated SMAD2 protein measured at 60 minutes after TGFβ treatment. X-axis is in log scale. Values are normalized against the maximum of each dataset [72]. (F) HIF-1α protein stabilization and (G) AGO1 protein downregulation are observed in HUVECs in hypoxia (2% O2) by Chen et al [23]. (H-I) Dicer protein is downregulated by hypoxia (1% O2) in HUVECs; hypoxia results in accumulation of HIF-2α protein in human dermal microvascular ECs from Ho et al [59]. (J) P53 protein expression is induced in hypoxic conditions (1% O2) in HUVECs; experimental data from Lee et al [73]. (K) TSP-1 protein expression is increased in hypoxia in HUVECs (2.7% O2); data from Phelan et al [16]. (L) TSP-1 protein expression is induced in human pulmonary aortic endothelial cells in response to 24 hours of hypoxia; data from Labrousse-Arias et al [14]. (A-L) Results (data-point values and simulations) are normalized against the maximum (or the normoxic baseline values) in each dataset. Quantification of experimental data from the literature is done in ImageJ following standard protocols.
Fig 4.
Qualitative model validation against experimental data in different cell types.
(A-C) Experimental data (symbols) of total phosphorylated SMAD1 and SMAD2 protein in BAECs and mouse embryonic ECs when stimulated by different amount of TGFβ [71, 72]. Model simulations are presented by solid curves. (D-F) HIF1/2 protein expression data (symbols) measured in HUVECs and HCT116 colon carcinoma cells at different oxygen tensions [59, 75, 76]. Model simulations presented by solid curves show rapid induction of HIFs in hypoxia. (G) Dicer protein level (symbols) is decreased in hypoxia (1% O2) and measured in human dermal microvascular ECs [59]. Model simulation is presented by the solid curve. (H-I) Data on MXI-1 mRNA in H460 lung cancer cells and PSAP protein in HCT116 colon carcinoma cells indicate upregulation of both molecules in hypoxia [77, 78]. Model simulations confirm the trend. (J-K) Data on Myc protein in hypoxic conditions (symbols) in HCT116 cells, H460 cells, U2OS osteosarcoma cells and Hela cells [75, 77]. The downregulation of Myc is also captured by model simulations (solid curves). (L-M) Levels of miR-18a quantified in HCT116 cells, MGC-803 and HGC-27 gastric carcinoma cells [21, 79]. Model simulation of miR-18a downregulation in hypoxia agrees with experimental findings. The expression of miR-19a, another miR that potentially targets TSP-1, is also downregulated in hypoxia [19]. (N) Let-7a, together with other members of the let-7 miR family, belongs to the HRM group, and its induction is captured by model simulation (solid curve) and supported by experimental data (symbols) in HUVECs [23]. (O) Experimental data (symbols) in NRK52E rat kidney cells indicate that TSP-1 protein expression is stimulated by TGFβ signals [27]. Model simulation is shown by the solid curve. (A-O) Experimental data-point values and simulation curves are all normalized with respect to the peaks (or normoxic values in the bar-graphs). Literature data are quantified by ImageJ following standard protocols.
Fig 5.
Dynamics of SMADs and TSP-1 protein synthesis.
(A) SMAD7, whose expression is induced by TGFβ activation, binds the RSMAD-receptor complex and prevents the phosphorylation of the RSMAD in the complex. (B) SMAD4 abundance is negatively regulated by SMAD7 since accumulated SMAD7 promotes SMAD4 degradation. (C) RSMAD-receptor complex sequestered by SMAD7 upon TGFβ activation is released when SMAD7 expression decreases, giving rise to the tail expression of activated RSMAD-SMAD4. (D) Inhibition of SMAD7 protein synthesis removes the tail expression but significantly prolongs the RSMAD-SMAD4 activation signals. (E) TSP-1 protein synthesis stimulated by different doses of TGFβ and (F) the corresponding dose response curve produced by the model. The peaks of total intracellular TSP-1 levels are normalized with respect to the maximum peak observed in the simulation of 20 ng/ml TGFβ stimulation. X-axis is in log scale.
Fig 6.
Low oxygen induces TSP-1 expression by different mechanisms.
(A) TSP-1 protein expression is increased along with decreased oxygen availability. (B) Hypoxia increases the transcriptional activity of TSP-1 gene through the induction of its promoters. (C-D) miR-18a, a TSP-1 targeting miR, is downregulated in hypoxia due to the downregulation of its promoter Myc, and (E) miR processing molecules AGO1 and Dicer. (F) Hypoxia de-suppresses the TSP-1 mRNA that was originally under miR-mediated repression, which contributes to the increase of total translatable TSP-1 mRNA. The ratio of repressed TSP-1 mRNA over free mRNA decreases more than tenfold in the condition of 2% oxygen. (G) Model prediction of a nonlinear behavior in TSP-1 synthesis with respect to oxygen tension. This relationship may be partially contributed by the switch-like nature of cellular oxygen sensing [22, 82, 84]. (B-G) Results are normalized with respect to the normoxic baseline values.
Fig 7.
Different therapeutic interventions to restore TSP-1 level in Myc-induced tumors.
(A) Hyperactivity of Myc is simulated by increasing its rate of production (baseline rate multiplied by 5), which results in a significant downregulation of TSP-1 protein level. Experimental evidence indicates that Myc expression could be 2–10 times higher in tumor samples compared with control samples [99–102]. (B) The effect of TSP-1 repression by Myc overexpression is less obvious in hypoxia. (C) In the cases of Myc hyperactivity, knockdown of Myc synthesis (protein synthesis rate reduced to 10% of the hyperactive value) leads to a more notable increase in TSP-1 protein abundance in normoxia compared to hypoxia. (D) Hyperactive Myc engages both the direct transcriptional and posttranscriptional pathways (miRs) to repress TSP-1 protein expression. (E) Simulating the application of a Myc inhibitor at different doses under the condition of Myc hyperactivity. (F) Simulating the application of a miR-18a antagonist at different doses under the condition of Myc hyperactivity. Simulation results suggest that miR antagonists reach a maximum efficacy at around 40 nM. (G) Direct transcriptional stimulation by different doses of TGFβ results in a significant early increase in TSP-1 protein expression under Myc hyperactivity compared to the other two strategies. (A-G) Results are normalized with respect to the normoxic steady-state values calculated with baseline Myc synthesis rate. (E-F) The simulations assume that Myc inhibitors potently bind and sequester cytoplasmic Myc with a Kd of 1 nM, and that miR-18a antagonists bind and sequester miR-18a RISC with a Kd of 1 nM [103].
Fig 8.
Different therapeutic interventions to reduce TSP-1 levels in PAD.
(A) Hypoxia and TGFβ together contribute to a much higher intracellular TSP-1 protein expression in the simulated condition of CLI than the normoxic control condition (with no TGFβ). TSP-1 expression curves in response to different doses of (B) let-7 antagonists, (C) miR-18a mimics, and (D) p53 inhibitors. The effect of let-7 antagonist peaks at around 40 nM. (E) A combination of inhibiting both p53 production and NFAT activities achieve the most significant downregulation of TSP-1 in the simulated condition of CLI. In the simulations, the strength of NFAT inhibitor VIVIT is estimated from literature data to be a 70% decrease in the rate of calcineurin-mediated NFAT dephosphorylation when applied at micromolar doses [111]. (B-E) MiR-18a overexpression is simulated as an increase in the initial condition of precursor miR-18a; let-7 antagonists bind and sequester let-7 RISC with a Kd of 1 nM; small molecule inhibitor of p53 binds and sequesters cytoplasmic p53 with a Kd of 1 nM.
Fig 9.
Global sensitivity analysis of model parameters.
Global sensitivity analysis of parameters that control (A) the area under curve (AUC) of HIF-1 dimer in a span of 48 hours in 2% oxygen, (B) AUC of activated SMAD2-SMAD4 complex in nucleus in a span of 24 hours upon 2.5 ng/ml TGFβ stimulation, (C) AUC of TSP-1 in a span of 24 hours under the condition of normoxia plus Myc hyperactivity and (D) AUC of TSP-1 in a span of 24 hours upon 2.5 ng/ml TGFβ stimulation plus hypoxia (2% oxygen). (A-D) Rate descriptions—kf1: HIF-1α translocation into nucleus; kf2: HIF-1α binds FIH complex; kf4: oxygen binds FIH complex; kf7: HIF-1α binds PHD complex; kf8: oxygen binds PHD complex; kf11: HIF1α-OH-FIH dissociation; kf13: HIF1α-OH-PHD dissociation; kf18: TTP synthesis; kf19: HIF-1α binds HIF-1β; vm1: HIF-1α synthesis; vm2: HIF-2α synthesis; vm3: Myc synthesis; vm5: let-7 synthesis; vm6: MXI-1 synthesis; vm7: miR-18a synthesis; vm9: LIN28B synthesis; vm11: PSAP synthesis; vm13: p53 synthesis; vm18: AGO1 synthesis; vm19: Dicer synthesis; vm20: TSP-1 synthesis; kf66: TGFβR internalization; kf72: TGFβR degradation; kf73: TGFβ binds its receptor; kf74: receptor dimer binds R-SMAD; kf75: R-SMAD phosphorylation; kf76: R-SMAD translocation into nucleus; kf77: phosphorylated R-SMAD (pR-SMAD) binds SMAD4; kf78: pR-SMAD-SMAD4 complex translocation into nucleus; kf79: SMAD4 translocation into nucleus; kf82: pSMAD2-SMAD4 dephosphorylation; kf89: SMAD7 sequesters activated R-SMAD; vm27: TGFβ-mediated calcium influx; vm29: NFAT dephosphorylation; vm32: SMAD7 synthesis; vm34: SMAD4 degradation; vm36: SMAD4 synthesis.