Fig 1.
Translational repression of VEGF in normoxia and let-7 mediated VEGF desuppression in hypoxia in ECs.
(A) VEGF mRNA were targeted by miRISC and inaccessible for translation in normoxia. (B) HIF-1 proteins that were stabilized in hypoxia induced let-7 biogenesis, which led to the downregulation of AGO1 mRNA, protein and miRISC formation. VEGF mRNA were desuppressed and ready for translation because of reduced miRISC activities.
Fig 2.
Proposed model scheme of the miR control of HIF-VEGF pathway.
HIF-1α is stabilized in hypoxia and the HIF-1 dimer complex transcriptionally induces let-7 production. Mature let-7 represses AGO1 and leads to a global desuppression of VEGF. Model components in the colored backgrounds correspond to the four modules: blue/O2 sensing, pink/VEGF repression by miR-15a, orange/HIF-dependent transcription, green/let-7 biogenesis and targeting. Species whose names end with an N subscript are located inside the nucleus; reactions that point to red signs indicate degradation. The symbols v# refer to the 57 chemical reactions listed in S1 Table.
Fig 3.
Comparisons between model simulations and experimental results.
(A) HIF-1α protein is quickly induced and stabilized in 2% cellular O2 during a 48-hour simulation (line). (B) Hypoxia induces miR-mediated translational repression of AGO1 over time and reduces intracellular AGO1 protein abundancy (line). (A-B) Predicted time course expressions of HIF-1α and AGO1 are normalized and compared with the quantified Western blot data (symbol) in ECs with standard deviation values provided by Chen et al [21]. (C) 1% O2 induces VEGF protein production over 8 hours (line) in SHEP cells and (D) 200 μM of CoCl2 treatment in normoxia can mimic hypoxia in cells and induce VEGF protein synthesis in HepG2 cells during a 48-hour simulation (line). (C-D) We numerically quantified the two sets of raw VEGF Western blot data from literature in SHEP and HepG2 cells by densitometry (symbol), and the results with calculated standard deviation are compared with model simulations [50, 51].
Fig 4.
HIF-1α level at time zero is the steady state normoxic (21% O2) level. (A) Total HIF-1α expression profiles are highly sensitive to oxygen availability. (B) When PHD2 initial concentration is in excess, an oxygen-dependent, switch-like behavior in the amount of hydroxylated HIF-1α is observed. As simulation span increases, a steep reduction in HIF-1α hydroxylation occurs between 2% to 4% O2. (C) TTP is responsible for the delayed drop (initial overshoot) in the induction of HIF-1α in hypoxia. By increasing the dose of a simulated siRNA that silences TTP expression (assuming siRNA binds TTP mRNA potently with a Kd of 1 nM), the duration of the initial overshoot is lengthened. (D) Varying the rate of HIF-1α import from cytoplasm into nucleus (kforward) affects the overall HIF-1α profile in hypoxia. (E) Larger kforward values of HIF-1α nuclear import result in higher levels of HIF-1 transcription factor complex formed. (F) Smaller kforward values lead to lower total HIF-1α levels in normoxia and in hypoxia, while the majority of induced HIF-1α is located only in the cytoplasm and unable to form transcription complex with HIF-1β. (D-F) Magnitude of kforward is set to 10%, 50%, 200% and 500% of its original value respectively in the comparisons. For each kforward value, steady state levels of all species, after the model is simulated in normoxia for a long time span, are collected and considered a new set of initial conditions.
Fig 5.
Let-7 and AGO1 mutually control each other in hypoxia.
(A) Varying the cellular AGO1 abundance by antagonizing or overexpressing its mRNA changes the let-7 availability. (B) Let-7s that are in association with AGO1 are less prone to degradation, so a decrease in the binding strength of let-7 toward AGO1 causes more let-7 to be degraded. Consequently, (C) let-7-mediated activity including AGO1 repression is downregulated, allowing additional AGO1 protein synthesis. (B-C) Association rate of AGO1 and let-7 (kforward) is adjusted to 10%, 20%, and 50% of its original value respectively in the comparisons. (D) AGO1 overexpression leads to an early upstroke in its time course profile but its steady state level changes insignificantly. (E) After 4 hours, almost all the additional AGO1 mRNAs (e.g. 0.01 μM and 0.04 μM) being introduced in the beginning are fully shuttled into the p-body to become inaccessible for translation. (F) In hypoxia, VEGF mRNA released from miRISC, in combination with HIF induction, boosts the pool of free form VEGF mRNA. A simulated AGO1 overexpression rescues the drop in miRISC level and drives free form VEGF mRNA back into miR-mediated repression. (A, D-F) The model assumes that in AGO1 silencing, siRNA binds AGO1 mRNA potently with a Kd of 1 nM; AGO1 overexpression is simplified as an increase of certain amount in the initial concentration of AGO1 mRNA.
Fig 6.
Let-7 represses Dicer and AGO1 that both limit miR-15a expression in hypoxia.
(A) Overexpression of AGO1 mRNA by 0.01, 0.04 and 0.08 μM in hypoxia leads to short-term rises in the expression levels of total miR-15a (B) by promoting the association of free form miR-15a with AGO1 to make more miR-15a RISCs. (C) Dicer processing is a limiting step in the production of miR-15a in hypoxia. Introducing both Dicer and AGO1 mRNAs at the beginning of simulation results in elevated miR-15a abundance compared to adding AGO1 mRNA alone. (D) When let-7 no longer inhibits Dicer translation, an overexpression in Dicer mRNA generates a remarkable change in the expression profiles of Dicer with respect to the control situation. With let-7 mediated Dicer silencing, the response of Dicer mRNA overexpression is significantly attenuated. (E) Relative expression of non-translatable VEGF mRNA associated with miR-15a RISC and (F) translatable VEGF mRNA in response to different treatment strategies. Hypoxia causes an initial increase in the binding between VEGF mRNA and miR-15a RISC because of the rapid HIF-1-activated VEGF transcription, but the impact of AGO1 silencing becomes dominated later that, in the long run, miR-15a-bound VEGF is reduced compared to the normoxic level. In hypoxia, enforced let-7 overexpression or AGO1 silencing modestly increases the amount of translatable VEGF, while let-7 antagonists or AGO1 overexpression can remarkably blunt VEGF production.
Fig 7.
Simulations of potential pathway-related therapeutic strategies in tumor and PAD.
Different doses of (A) let-7 antagonists, (B) AGO1 mRNA overexpression, and (C) miR-15a mimics are applied and total VEGF production curves at different O2 tensions are obtained. (D) Either antagonizing let-7 or overexpressing AGO1 strongly suppresses VEGF synthesis compared to the control curve, particularly in severely hypoxic conditions (0–2% O2) that simulate in vivo tumor oxygenation. More specifically, let-7 antagonist is more effective because it maintains a stable and higher intracellular AGO1 expression than direct AGO1 overexpression. Since both AGO1 and Dicer are significantly reduced in hypoxia and are limiting factors in miR maturation and stabilization, the response of miR-15a overexpression is relatively insignificant. The results should be evaluated qualitatively since many proteins could have much lower concentration in vivo compared to in vitro cultures which might invalidate some of the model assumptions. (E-F) Testing the impact of different doses of let-7 mimics and miR-15a antagonists on VEGF release in PAD conditions, which is modeled as a weak let-7 induction by HIF-1 in hypoxia (see Methods for details). Antagonizing miR-15a, compared to overexpressing let-7, results in stronger promotion on VEGF production. (G) The combination of miR-15a antagonist with let-7 mimic positively modulates angiogenesis; it elevates VEGF production by more than two fold with respect to PAD controls in hypoxia.
Fig 8.
Sensitivity analysis of key species in the pathway.
Sensitivity of (A) cytoplasmic HIF1-α, (B) free form AGO1, (C) free form let-7, and (D) VEGF to variations in different sets of kinetic parameters (direct production and degradation rates excluded). (A-D) Kf(X/Y) stands for the forward reaction constant of species X binding species Y; Vm(X) stands for the speed of reaction X; Kf(X) stands for the forward rate constant of species X dissociation. Detailed description of each parameter is available in the supplemental information. (E) At different O2 levels, TTP mRNA overexpression is tested as an anti-angiogenic therapy in silico compared to miR-based therapeutic strategies. (F) Affinity of O2 binding with PHD2 or FIH and HIF-1α translocation rate contribute to the trend of HIF-1α stabilization in hypoxia. (G) Relative downregulation of AGO1 in hypoxia is influenced by its binding with let-7. (F-G) Parameters are set to 500% of their original values in the comparisons. For each new value, steady state levels of all species, after the model is simulated in normoxia for a long time span, are collected and considered a new set of initial conditions. HIF-1α and AGO1 levels in hypoxia are normalized with respect to their concentrations at time zero (normoxic steady states).