We previously reported that during hypoxia in human ECs HIF-1 is eventually replaced by the HIF-2 [15, 16, 3], and our goal here was to explore the underlying mechanism. As shown in Fig. 1A, we followed the HIF-1α and HIF-2α protein levels in human umbilical vein endothelial cells (HUVECs) at 4-, 8- and 24-hour exposure to hypoxia (1% O2). Consistent with our previous findings[15, 25, 26], we observed that HIF-1α rapidly accumulated in HUVECs exposed to hypoxia at 4 hours and was reduced at 8 hours and dramatically reduced at 24 hours. In contrast, HIF-2α reached maximal levels at 8h, and remained elevated even after 24 hours. Hence, although in ECs both HIF-1α and HIF-2α at first rapidly accumulate during hypoxia, the rapid loss of HIF-1 levels leads to the transitional switch to HIF-2 signaling. The hypoxic changes of mRNA and protein levels for both HIF1A/HIF-1α and EPAS1/HIF-2α were well correlated (Fig. 1B-D). Interestingly, the hypoxic accumulation of HIF-1α subunits was also accompanied by the continuous decline in HIF1A mRNA (Fig. 1D), whereas EPAS1 mRNA levels remained fairly constant during studied time course (Fig. 1E). Furthermore, EPAS1 mRNA levels are two-fold higher than HIF1A in normoxia and five-fold higher during hypoxia at 24 hours (Fig. 1F).
Surprisingly, however, in cells cultured in hypoxia HIF-1α and HIF-2α protein half-life’s were comparable after 4- and 8-hours exposure to hypoxia (Figs. 2 and 3): HIF-1α 4h hypoxia t1/2 ≈ 33 minutes and HIF-2α 4h hypoxia t1/2 ≈ 32 minutes; HIF-1α 8h hypoxia t1/2 ≈ 41 minutes and HIF-2α 8h hypoxia t1/2 ≈ 46 minutes. Whereas after 16 hours of exposure to hypoxia HIF-2α was slightly more stable than HIF-1: HIF-1α 16h hypoxia t1/2 ≈ 42 minutes and HIF-2α 16h hypoxia t1/2 ≈ 56 minutes. Notably, the HIF-1α protein half-life measured in HUVECs exposed to hypoxia for 4 hours is consistent with the previous report of Hagen et al. , whereas to the best of our knowledge, no data regarding other times regarding the half-life HIF-2α protein stability during hypoxia are available. These results suggest that the hypoxic transition from HIF-1 to HIF-2 is a result of decreased HIF1A mRNA levels rather than the small differences in HIF-1α and HIF-2α subunits protein stability at the later time points during hypoxia.
Since, the oxygen-dependent activity of PHDs provides a canonical mechanism of HIF-α subunits destabilization in normoxia, and PHDs (PHD-2 and PHD-3) were shown to have different preference against the HIF-1α and HIF-2α subunits as well as oxygen requirements  , we tested whether the observed earlier degradation of HIF-1α might also result from reactivation of these enzymes during prolonged hypoxia. Given that during hypoxia HIF-1 activity leads to activation of anaerobic glycolysis and inhibition of the mitochondrial aerobic metabolism, we tested whether these changes could lead to reactivation of PHDs [19, 28].
To test this, we used Firefly luciferase reporter vectors (the Renilla luminescence was used as a calibrator of transfection efficiency) containing the respective HIF-1α and HIF-2α oxygen-dependent degradation domains (ODDs) and a control vector, lacking the ODD region, as shown in Fig. 4A [5, 29] and Supplemental Fig. 1. The presence of HIF-1α and HIF-2α ODDs resulted in significantly reduced Firefly luminescence to less than one-third of the observed for control vector, when the transfected HUVECs were cultured in normoxia. Furthermore, as expected, when the transfected cells were incubated at 1% oxygen for 4 hours, the presence of HIF-1α and HIF-2α ODD domains had no significant impact on the Firefly luminesce, that was comparable to the control vector, confirming that PHD-dependent degradation of both HIF-α subunits was inhibited. However, in cells that were incubated at 1% O2 for 8 hours and 24 hours, the Firefly luminescence from both HIF-1α and HIF-2α reporters was significantly reduced, when compared to signal measured after 4 hours incubation at 1% O2 (Fig. 4B). Notably, the luminescence of HIF-1α and HIF-2α reporters after 8- and 24-hours exposure to hypoxia was comparable to the one measured during normoxia, suggesting that at prolonged exposure to hypoxia PHD activity has been restored. Furthermore, HIF-1α and HIF-2α ODDs Firefly luminescence recorded for their respective reporter vectors were not significantly different between each other, neither in normoxia nor during the time course exposure to hypoxia (Fig. 4B), suggesting that PHD-dependent degradation has similar impact on both HIF-1α and HIF-2α ODD domains.
To confirm this result another way, we investigated effects of blocking PHD activity with dimethyloxalylglycine (DMOG), a competitive inhibitor of the endogenous 2-oxoglutarate cofactor for prolyl hydroxylase activity. Following HUVECs transfection with luciferase control, HIF-1α ODD. and HIF-2α ODD reporter vectors, cells were incubated for 4 hours at 1% O2 and then cells were incubated for next 4 hours (total 8 hours at 1% O2) and 20 hours (total 24 hours at 1% O2) in the absence or presence of 2.5 mM DMOG. As shown in Fig. 4B, during hypoxia DMOG prevented the reduction of Firefly luminescence from both HIF-1α ODD and HIF-2α ODD reporter vectors at 8 hours and 24 hours of hypoxia. Taken together, these results confirmed that HIF-α subunits degradation during prolonged hypoxia is due to PHD activity.
Given that PHD activity was present during prolonged hypoxia, in follow-up experiments we tested whether we could prevent of the restoration of cytosolic oxygen homeostasis by exposing HUVECs to 0.3% O2 and whether the PHDs would retain activity. In these experiments, the HUVECs were transfected with luciferase control, HIF-1α ODD, and HIF-2α ODD reporter vectors and were incubated for 4 hours at 1% O2, and transferred to 0.3% O2 for another 4 hours (total 8 hours of hypoxia) or 20 hours (total 24 hours of hypoxia). The Firefly luminescence from both HIF-α ODD reporter vectors after 8 and 24 hours remained at the levels observed when cells were exposed to 1% O2 for 4 hours, and significantly higher than in cells incubated at 1% O2 for the same time (Fig. 4B). These results support the hypothesis that the restoration of PHDs activity during prolonged hypoxia exposure results from the recovery of cellular oxygen levels due to adaptive hypoxia response related to the reduction of mitochondrial oxygen consumption.
To confirm the validity of the luciferase results, we tested blocking the PHD activity during hypoxia with DMOG on the HIF-1α and HIF-2α protein levels. As shown in Fig. 5AB, when HUVECs were exposed to hypoxia for 4 hours, and then DMOG was added, and cells were cultured in hypoxic conditions for another 4 hours (total time of hypoxic exposure was 8 hours), the HIF-1α levels were dramatically (about 2 times) higher than in HUVECs exposed to hypoxia for 8 hours without the DMOG addition. Similar, DMOG-related rescue of α subunit protein was also observed in parallel HIF-2α experiments (Fig. 5AC). The HIF-2α levels doubled after 8 hours of exposure to hypoxia if the DMOG was added after 4 hours. These DMOG-induced HIF-1α and HIF-2α protein levels during hypoxia are consistent with the luciferase reporters results and strongly supports the possibility of the hypoxic reactivation of PHDs during prolonged hypoxia. However, in contrast to luciferase reporter results, no significant HIF-1α and HIF-2α accumulation in the DMOG presence was observed after 24 hours exposure to hypoxia (Fig. 5ABC). Considering that these DMOG treatments did not significantly affected HIF1A and EPAS1 mRNA levels (Fig. 5DE), this data suggest that during prolonged hypoxia, PHDs are active and account for both HIF-1α and HIF-2α degradation. Notably, however, although DMOG has been reported as an effective PHDs activity inhibitor, it has pleiotropic effects on the cellular metabolism and different effects on HIF-1α and HIF-2α have been reported for this hypoxia mimetic . It has also been shown that treatment with DMOG, can result in the accumulation of HIF-1α but not of HIF-2α . Hence, although the expected rescue of HIF-2α protein was not observed in 24 hours DMOG experiments, despite the relatively high EPAS1 mRNA levels, this result needed to be confirmed via another approach.
As an alternative approach, HUVECs were exposed to hypoxia (1% O2) for 4 hours and then O2 was reduced to 0.3%, and cells remained in these conditions for another 4 hours (total time of hypoxic exposure was 8 hours). For the 24-hour time point, the cells were changed to 0.3% oxygen for 20 hours. The results were interesting in that in the 8-hour time point, the HIF-1α levels were increased when incubated in the 0.3% oxygen for 4 hours (Fig. 6AB), and the HIF-2α was also affected under the same conditions (Fig. 6AC). At the 24-hour time point the HIF-2α protein level was significantly elevated in the low oxygen, whereas the HIF-1α protein levels were not (Figure AC). Furthermore, the reduced oxygen levels had no significant impact on HIF1A nor EPAS1 expression (Fig. 6DE). Taken together, HIF-α subunits rescue observed upon further reduction of oxygen availability (0.3%) during hypoxia strongly supports some PHD activity at 1% oxygen caused by the cellular oxygen redistribution potentially mediated by HIF-1 and HIF-2. Importantly, the data support the hypothesis that the inability to effectively rescue HIF-1α during prolonged hypoxia is mediated by both PHD activity and mRNA instability.
During hypoxia, initially both HIF-1α and HIF-2α accumulate due to the reduced PHDs activity. The NO-mediated inhibition of the respiratory chain and metabolic switch to glycolysis, however, results in lower mitochondrial oxygen consumption and leads to restoration of some cellular O2 that is sufficient for PHDs-related degradation of both of these HIF-α subunits [19–22]. Notably, however, the HIF1A mRNA levels are also reduced during hypoxia, limiting the amount of translated HIF-1α, whereas the EPAS1 mRNA is more abundant and much more stable, allowing for the continuous accumulation of HIF-2α during prolonged hypoxia .
Notably, we were able to simulate the effects of changing oxygen concentration and of the reduction of HIF1A transcript in the time-course of hypoxia using a dynamic model, in which the effect of the oxygen level on the hydroxylation of the two HIF-α subunits by PHDs was modeled according to an established dynamic model of HIF-1 signaling . In our model, we assumed that the system contains a single dominant PHD isoform with potentially different activities towards HIF-1α and HIF-2α, in agreement with the experimental evidence for ECs [24, 23]. The effects of transcript abundancies of HIF1A and EPAS were modeled using the Mass-Action Law. This model was first fitted to time-series data of 12 time-points from HUVECs under 1% hypoxia. We then used the fitted model to assess (i.e., predict) the response to a further drop in the oxygen level to 0.3% at 4 h of hypoxia. For details of the modeling see Supplementary Information 1. Our results clearly support the conclusion that the residual PHD activity during prolonged hypoxia together with the HIF1A mRNA instability contribute to the HIF-1α to HIF-2α transition during hypoxia.