Real-Time Imaging of HIF-1α Stabilization and Degradation

HIF-1α is overexpressed in many human cancers compared to normal tissues due to the interaction of a multiplicity of factors and pathways that reflect specific genetic alterations and extracellular stimuli. We developed two HIF-1α chimeric reporter systems, HIF-1α/FLuc and HIF-1α(ΔODDD)/FLuc, to investigate the tightly controlled level of HIF-1α protein in normal (NIH3T3 and HEK293) and glioma (U87) cells. These reporter systems provided an opportunity to investigate the degradation of HIF-1α in different cell lines, both in culture and in xenografts. Using immunofluorescence microscopy, we observed different patterns of subcellular localization of HIF-1α/FLuc fusion protein between normal cells and cancer cells; similar differences were observed for HIF-1α in non-transduced, wild-type cells. A dynamic cytoplasmic-nuclear exchange of the fusion protein and HIF-1α was observed in NIH3T3 and HEK293 cells under different conditions (normoxia, CoCl2 treatment and hypoxia). In contrast, U87 cells showed a more persistent nuclear localization pattern that was less affected by different growing conditions. Employing a kinetic model for protein degradation, we were able to distinguish two components of HIF-1α/FLuc protein degradation and quantify the half-life of HIF-1α fusion proteins. The rapid clearance component (t1/2 ∼4–6 min) was abolished by the hypoxia-mimetic CoCl2, MG132 treatment and deletion of ODD domain, and reflects the oxygen/VHL-dependent degradation pathway. The slow clearance component (t1/2 ∼200 min) is consistent with other unidentified non-oxygen/VHL-dependent degradation pathways. Overall, the continuous bioluminescence readout of HIF-1α/FLuc stabilization in vitro and in vivo will facilitate the development and validation of therapeutics that affect the stability and accumulation of HIF-1α.


Introduction
The HIF-1 (hypoxia inducible transcriptional factor 1) controls the expression of genes involved in critical aspects of cancer biology, such as angiogenesis, glucose metabolism, cell survival, invasion and tumor progression [1,2,3,4,5,6,7]. HIF-1 is a heterodimeric protein complex composed of two subunits: a stable and constitutively expressed HIF-1b, and an inducible, O 2 -and growth factor-regulated HIF-1a-subunit [8,9]. HIF-1a protein is constantly modified posttranslationally by prolyl hydroxylases at Pro402 and/or Pro564 within the oxygen-dependent degradation (ODD) domain which promotes binding with pVHL (von Hippel-Lindau protein) and subsequent targeting for rapid proteasomal degradation. The half-life of HIF-1a protein, as determined by standard immunoblotting method, is about 5-8 min under normal oxygenated conditions [10,11,12]. It has been proposed, that interaction between pVHL and HIF-1a occurs in the nucleus, where HIF-1a protein is ubiquitinated and then exported to the cytoplasm for further proteasomal degradation [13]. Under hypoxic conditions, the prolyl hydroxylation reaction is inhibited and pVHL-HIF-1a interaction is abrogated, resulting in HIF-1a accumulation in the nucleus and dimerization with HIF-1b [14,15]. The degradation of HIF-1a is also regulated in an O 2independent manner by the competitive binding to either heat shock protein 90 (HSP90), which stabilizes the protein [16,17], or to the anchoring protein (RACK1), which leads to HIF-1a degradation by an oxygen-independent process [18].
HIF-1a is overexpressed in many human cancers compared to normal tissues [19,20], and this overexpression is due to the interaction of a multiplicity of factors and pathways that reflect specific genetic alterations and extracellular stimuli (e.g., hypoxia) that impact on both synthesis and degradation [21]. The activation of oncogenes (H-ras and v-src) involving signaling cascades (PI3K and MAPK), as well as loss of function mutations in tumor suppressor genes (VHL, PTEN and p53), have also been shown to result in HIF-1a protein accumulation and increased expression of downstream HIF-1 target genes [9,21,22,23,24].
The objective of this study was to investigate and compare the tightly controlled level of HIF-1a protein in normal and tumor cells by imaging the dynamic response of different fusion variants of the HIF-1a gene linked to the Firefly luciferase (FLuc) gene. To accomplish this goal, we engineered several different reporter constructs and generated reporter cell lines to study the dynamic stability of HIF-1a under hypoxia and in the presence of a prolyl hydroxylation inhibitor -CoCl 2 , a protein synthesis inhibitorcycloheximide, and a proteasomal degradation inhibitor -MG132. Using immunofluorescence microscopy, we observed different patterns of subcellular localization of HIF-1a/FLuc fusion protein between normal cells and cancer cells; similar differences were observed for HIF-1a in non-transduced, wild-type cells. By employing a kinetic model for protein degradation, we were able to quantify the half-life of HIF-1a/FLuc noninvasively in cells with different genetic backgrounds using a more sensitive dynamic bioluminescence imaging technique.
To construct a HIF-1a/FLuc fusion gene with a deleted oxygendependent degradation domain (DODDD), corresponding to 1206-1813 bp in the HIF-1a ORF and 401-603 amino acids the HIF-1a protein, the following approach was used. PCR primers, corresponding to the N-terminal and C-terminal ends of the HIF-1a protein and two additional primers corresponding to the internal part of protein without ODD domain were designed: 59-AAAAGAAAAGTCTCGAGATGCAGC, 59-ATAGTTTAG-CGGCCGCAGCAAAGTTAA, 59-CATAGAAGCGGCCGCA-GACTCAAATAC, 59-GGCGGATCCTTACACGGCGATCT. PCR products with XhoI/NotI and NotI/BamHI ends were subcloned into the SFG-HIF-1a/Fluc-IRES2-GFP between XhoI and BamHI sites to generate the HIF-1a(DODDD)/FLuc fusion gene without the ODDD sequence.

Cells transduction
All cell lines were stably transduced by incubating 50% confluent cell cultures with a virus-containing medium for 12 hours in presence of polybrene (8 mg/ml; Sigma, St.Louis, MO, USA). Cells were sorted using the fluorescence-activated cell sorter (FACS; BD Bioscience, CA, USA) with identical gates to obtain populations of cells with the same GFP fluorescence level (Fig. S1).

In vitro bioluminescence assays
Stably transduced and sorted cells were seeded in 6-well plates.

Protein degradation assay and half-life calculation
Cells were seeded into 96-well plates. After 3-4 hours of incubation in standard conditions, the protein synthesis inhibitor cycloheximide (100 mg/ml) (Sigma-Aldrich, St. Louis, MO, USA) was added for the indicated periods of time to the medium. In the experiments with CoCl 2 treatment, cells were pretreated with CoCl 2 for 3 h which was either maintained in the medium during cycloheximide exposure or was removed before exposure to cycloheximide. Following incubation, the cells were assayed for bioluminescence using Bright-Glo Luciferase solution. FLuciferase activity of the test cells was normalized to that of untreated control cells and expressed as % of control. The times profiles for FLuciferase activity were generated and fitted to either a single or double exponential equation to estimate the degradation rate of FLuicferase activity (reflecting the amount of HIF-1a/FLuc fusion protein): % of control = A 0 ?e^(2k a ?t) or A 0 ?e^(2k a ?t)+-B 0 ?e^(2k b ?t), respectively. The half-life was calculated by: t 1/2 = ln2/k. Data fitting was performed using KaleidaGraph (Synergy Software version 3.6.4).

Immunofluorescence microscopy
Adherent cells were fixed with 4% paraformaldehyde in PBS for 12 min, washed three times with PBS, permeabilized with 0.2% Triton X-100 in Superblock (Pierce, Rockford, Il) for 30 min and washed again with PBS (at room temperature). After blocking nonspecific binding with ImageIT (Molecular Probes, Carlsbad, CA) for 30 min at room temperature, the cells were incubated with mouse anti-HIF-1a (BD Biosciences, San Jose, CA and RD Systems, Minneapolis, MN, USA) at 1:50 dilution in Superblock overnight at 4uC. Primary antibodies were detected by incubation with a goat anti-mouse secondary antibody conjugated to Alexa-568 (Molecular Probes Carlsbad, CA) diluted in Superblock at 1:100, at room temperature for 1 h. Following washing in PBS, cells were mounted in Vectashield (Burlingame, CA) containing 1.5 mg/ml DAPI. Fluorescence images were acquired at 640 magnification using an Olympus BX40 fluorescence microscope (Olympus America, Inc.) with CCD camera. DAPI, GFP and HIF-1a were imaged using blue, green and red filters respectively.

In vivo bioluminescence imaging
The animal protocol was approved by the Memorial Sloan-Kettering Institutional Animal Care and Use Committee. The U87/HIF-1a/Fluc, U87/HIF-1a(DODDD)/Fluc and U87/Fluc reporter cells (4610 6 ) were injected subcutaneously onto the right shoulder of 6-weeks old male Ncr nu/nu nude mice. After 1 week of tumor growth, a size of the xenografts reached a diameter of ,6-7 mm. For the FLuciferase substrate delivery into mice, the micro-osmotic pumps (Alzet Model 1007D) (Durect Corporation, CA, USA), loaded with D-Luciferin (50 mg/ml) (Caliper Life Sciences, CA, USA), were surgically implanted in the dorsal back subcutaneously. Mice were allowed to recover for 48 h and than imaged using IVIS imaging system before and after intraperitoneal injection of CoCl 2 (60 mg/kg). Photons emitted from the tumor region were quantified using Living Image software (Caliper Life Sciences, CA).

Statistical analyzes
Each experiment or assay was performed at least three times in triplicate and representative examples are shown. Data are reported as means6SD, applying Dunnett's ANOVA-based multiple comparisons procedure for comparing several treatments (CoCl 2 and hypoxia) with a single control and separately for each reporter. P,0.05 is considered significant.

Bioluminescence imaging of HIF-1a levels in cells
To characterize the reporter systems, we performed several experiments to show the functionality of the HIF1a/FLuc and HIF-1a(DODDD)/FLuc fusion proteins in NIH3T3 reporter cells. Bioluminescence intensity was measured following the addition of an inhibitor of prolyl hydroxylation, a hypoxia mimetic, CoCl 2 (100 mM), to the culture medium ( Fig. 1B, C). An increase of HIF-1a/FLuc expression was detectable within the first several hours, reaching a maximum value within 3-8 h, followed by a return toward basal levels following 24 h of treatment. These data were confirmed by Western blotting for the fusion HIF-1a/FLuc and for endogenous HIF-1a proteins (Fig. 1B). In the case of reporter cells bearing the HIF-1a(DODDD)/FLuc fusion, we did not see any significant changes in bioluminescence signal over 24 h as well as in the fusion protein (Fig. 1C).
We also measured bioluminescence of cells transduced with a mutant-variant of HIF-1a, containing a deleted oxygen-dependent degradation domain (DODDD) in the fusion construct ( Fig. 2A). As expected, bioluminescence activity from NIH3T3 and U87 HIF-1a(DODDD)/Fluc reporter cells under baseline normoxic conditions was much higher than that in cells transduced with the HIF-1a/Fluc fusion. The bioluminescence signal from the NIH3T3/HIF-1a(DODDD)/Fluc reporter cells under normoxia was ,300-fold greater (p,0.0001) than that of HIF-1a/Fluc transduced cells. A similar comparison of U87 reporter cells demonstrated a ,50-fold increase (p = 0.0002) in bioluminescence signal as a result of ODDD deletion. The expression levels of the mutated HIF-1a(DODDD)/Fluc reporter in NIH3T3 and U87 cells were similar to each other and not influenced by treatment with CoCl 2 or hypoxia ( Fig. 2A).
To examine whether the increase in bioluminescence was due to HIF-1a stabilization or to an effect of CoCl 2 on the FLuc reporter, the same experiments were performed using cells constitutively expressing Firefly Luciferase. FLuc bioluminescence intensity was not influenced by exposure to 100 mM CoCl 2 or hypoxia (2.5% O 2 ) (data not shown). All bioluminescence data were confirmed by Western blotting using specific antibodies (Fig. 2B).

The analysis of HIF-1a distribution in cells by immunofluorescence microscopy
Having characterized the behavior of the reporter proteins in the stably transduced cell lines, we sought to investigate whether the different genetic backgrounds of the cells influenced the intracellular localization of HIF-1a and its fusion proteins. To show that endogenous HIF-1a protein possess the same cytoplasmic-nuclear trafficking patterns as the HIF-1a/FLuc reporter, we also performed immunofluorescence staining of non-transduced, wildtype NIH3T3, U87 and HEK293 cells. As seen by immunofluorescence microscopy under normoxia HIF-1a/FLuc localization in NIH3T3 reporter cells (Fig. 3A) and HIF-1a in wild-type NIH3T3 cells (Fig. 4A) was primarily cytoplasmic, compared to the more equal distribution between nucleus and cytoplasm in HEK293 reporter and wild-type cells (Fig. 3C, 4C). In contrast, more intense nuclear staining of HIF-1a/FLuc (Fig. 3B) and wild-type HIF-1a ( Fig. 4B) was observed in U87 reporter and non-transduced cells, respectively. Following exposure to CoCl 2 there was a consistent shift in HIF-1a/FLuc and HIF-1a localization to the nucleus in all three cell lines. NIH3T3 reporter and wild-type cells showed increased accumulation in both compartments, with more intense cytoplasmic accumulation (Fig. 3A, 4A). In contrast, U87 and HEK293 reporter-transduced and wild-type cells showed much greater accumulation of HIF-1a/FLuc and HIF-1a in the nucleus following CoCl 2 treatment (Fig. 3B, 3C, 4B, 4C). The same pattern of the HIF-1a/FLuc and HIF-1a re-distribution was observed after exposure of all three reporter-transduced cell lines to hypoxia (2.5% O 2 ) (Fig. S2, S3, S4). Interestingly, the deletion of the oxygen-dependent degradation domain from HIF-1a resulted in a predominant nuclear accumulation of the mutant fusion protein under normal growth conditions, hypoxia (2.5% O 2 ) or CoCl 2 treatment, for both NIH3T3 and U87 HIF-1a(DODDD)-transduced cell lines (Fig. 5 and Fig. S2, S3).
GFP fluorescence, the identifier of reporter-transduced cells, remained unchanged under normoxia, hypoxia or CoCl 2 treatment and served as an internal control (Fig. 3, 4, 5 and Fig. S2, S3,  S4). Furthermore, the data presented in Figure 4 and Figures  S2, S3, S4 clearly demonstrate that the subcellular distribution of the endogenous HIF-1a in wild-type cells is the same as the HIF-1a/FLuc fusion protein in reporter-transduced cells under normal basal growing conditions, CoCl 2 treatment or hypoxia (2.5%).

Real-time imaging of HIF-1a stabilization in tumor xenografts
To characterize the reporter systems in vivo, a series of experiments were performed to image the stabilization of HIF-1a following CoCl 2 administration and inhibition of prolyl hydroxylases in animals bearing a HIF-1a/FLuc reporter xenograft. Three sets of mice bearing one of three constitutively expressing reporter xenografts were studied: HIF-1a/FLuc, HIF-1a(DODDD)/FLuc and FLuciferase (control). After a single intraperitoneal injection of CoCl 2 (60 mg/kg), a time-dependent increase in tumor bioluminescence was observed in U87/HIF-1a/ FLuc xenografts. The maximum bioluminescence intensity, ,6fold over pretreatment values, was observed 40-90 min after CoCl 2 administration. The signal decreased gradually to basal levels over 5 h (Fig. 6A, B). In contrast, only a slight increase in bioluminescence after CoCl 2 injection was detected in mice bearing the U87/HIF-1a(DODDD)/FLuc or U87/Fluc control xenografts.

Assessment of HIF-1a/FLuc and HIF-1a(DODDD)/FLuc degradation in different cell lines
To assess pVHL-ODDD-dependent degradation of HIF-1a/ FLuc in NIH3T3 and U87 reporter cells, the time-course and the degradation rate of the fusion proteins (HIF-1a/FLuc and HIF-1a(DODDD)/FLuc), as well as FLuc alone was determined by sequential bioluminescence measurements following the addition of a protein synthesis inhibitor (cycloheximide, 100 mg/ml) to the incubation medium ( Fig. 7 and 8). Degradation of the HIF-1a/ FLuc fusion protein in reporter-transduced NIH3T3 and U87 cells under normoxic conditions had a bi-exponential clearance, readily separated into rapid (mean t 1/2 4-6 min) and slow (mean t 1/2 ,200 min) kinetic phase components ( Fig. 7A and 8A; Tables 1 and 2). A similar bi-exponential clearance profile of HIF-1a/ FLuc was observed in HEK293 reporter cells (Fig. 8C). In contrast, only a single exponential, reflecting a slow clearance phase (mean t 1/2 ,160 min) was identified in both NIH3T3 and U87 cells transduced with the HIF-1a(DODDD)/FLuc fusion construct. Similarly, only a single, very slow clearance component (mean t 1/2 ,400 min) was identified for the constitutively expressed, native FLuc protein under normoxic conditions in NIH3T3/FLuc and U87/FLuc reporter cells (Fig. 7A and 8A; Tables 1 and 2).
The presence of CoCl 2 (100 mM) together with cycloheximide (100 mg/ml) in the medium resulted in a loss of the rapid clearance component in both NIH3T3/HIF-1a/FLuc ( Fig. 7B; Table 1) and U87/HIF-1a/FLuc cells ( Fig. 8B; Table 2). The clearance profile could be described by a single exponential with a mean half-time of ,50 min after a 3 h or a 24 h exposure to CoCl 2 . The precise mechanism for regulating the HIF-1a levels is still not clear; both increased degradation [29] and decreased production of the protein have been proposed [30,31]. Reappearance of the bi-exponential clearance of the HIF-1a/FLuc fusion protein, reflecting rapid and slow degradative phase, was again observed after removal of CoCl 2 from the incubation medium in NIH3T3 reporter-transduced cells ( Fig. 7B; Table 2). The presence of residual, non-degraded HIF-1a/FLuc protein after 90 min of protein synthesis inhibition was demonstrated by Western blotting (Fig. 7C), and comparable levels and rates of degradation were observed by immunoblot for endogenous HIF-1a and HIF-1a/ FLuc. Thus, partial degradation of the fusion protein with a persistence of a functional Firefly luciferase component is unlikely.
To demonstrate involvement of the proteasome in the degradation of the HIF-1a, cell lines bearing the HIF-1a/FLuc reporter cells were pretreated with a proteasome inhibitor, MG132 (10 mM for 3 h), followed by the addition of cycloheximide to the medium. MG132 prevented HIF-1a/FLuc protein degradation in all cell lines ( Fig. 7A and 8A; Tables 1 and 2).
To determine if the observed rapid and slow clearance components reflect different rates of cytoplasmic and nuclear HIF-1a/FLuc protein degradation, NIH3T3/HIF-1a/FLuc reporter cells were analyzed by immunofluorescence microscopy (Fig. 7D). Reporter cells were pretreated with CoCl 2 for 3 h, followed by its removal from the media and the addition of cycloheximide for 0, 7 and 90 min. HIF-1a/FLuc distribution within cells at 7 min after addition of cycloheximide showed slightly more nuclear intensity and a few cells visualized by DAPI staining and GFP expression were not visualized by anti-HIF-1a fluorescent antibody. By 90 minutes, approximately half the reporter cells were not visualized by anti-HIF-1a immunofluorescence (Fig. 7D), suggesting that some cells clear the HIF-1a/FLuc fusion protein more rapidly then other cells. GFP immunofluorescence remained unchanged under cycloheximide treatment and was used as an internal control to identify reporter-transduced cells (Fig. 7D).

Discussion
The two HIF-1a chimeric reporter systems that were developed in this study provide an opportunity to investigate HIF-1a stabilization/degradation process in different cell lines, both in culture and in xenografts. Several groups have studied and employed the fusion between the HIF-1a-ODD domain and Firefly Luciferase. The resulting fusion protein (ODD-Luc) is responsive to hypoxia and hypoxia mimetics in live cells and can be used for imaging HIF-1 oxygen/VHL-regulated activity in realtime under different conditions [32,33,34]. A mouse that ubiquitously expresses the ODD-Luc reporter has been successfully used to study the action of small molecule inhibitors of HIF prolyl hydroxylase activity [35]. However, the ODD-Luc reporter is useful only for studying the O 2 -ODD-VHL-dependent mech-anism of regulating HIF-1a expression. HIF-1a protein levels are also regulated by oxygen-independent mechanisms that reflect genetic alterations in signaling pathways or regulatory factors, and result in constitutive high levels of HIF-1a and HIF-1 transcriptional activity [36,37]. Two PAS domains, A and B, of the HIF-1a subunit in the N-terminal region frequently mediate proteinprotein interactions. One or both of the HIF-1a PAS domains have been functionally implicated in heterodimer formation, nuclear translocation and HIF-1a stabilization via HSP90 association [38]. The C-terminal part of the HIF-1 a is involved Cells were prepared for immunofluorescence as described in Materials and Methods and were incubated with the anti-HIF-1a antibody followed by an Alexa-568-conjugated secondary antibody. Cells were also visualized for GFP expression. Fluorescence images were acquired using the same acquisition parameters at 640 magnification. All panels represent the magnified images after contrast and brightness adjustment to visualize the subcellular localization of the protein. in protein transactivation. That is why the presence of the entire length of HIF-1a in reporter protein is essential for understanding HIF-1a biology and regulation of HIF-1a stability, since this occurs at multiple levels and involves more than the ODD domain.
A comparison between HIF-1a/Fluc and HIF-1a(DODDD)/ Fluc expression levels, as measured by BLI, demonstrated important differences between non-tumorigenic NIH3T3 and HEK293 reporter cells and tumorigenic PTEN-defective U87 reporter cells. Non-tumorigenic NIH3T3 and HEK293 cells had low basal normoxic levels of the HIF-1a/FLuc expression that were readily detectable by BLI, but not by immunoblotting. The higher sensitivity of the bioluminescence reporter in comparison to the immunoblot readout in living cells provides the ability to detect low basal levels of HIF-1a expression without exposing cells to hypoxia or hypoxia-mimetics agents. In contrast, U87 reporter cells had high basal levels of HIF-1a/FLuc expression. Similar results have been shown for other cancer cells lines using an immunoblotting assessment of endogenous HIF-1a. For example, HIF-1a overexpression under standard growth conditions was detected by immunoblot in metastatic breast cancer MDA-MB-231 cells and DU145 prostate cancer cell lines [39]. Interestingly, other cells (including MCF-7, HT-29 colon, MiaPaCa pancreatic, A549 lung, and BX-PC3 prostate cancer cells) do not show measurable HIF-1a levels under normoxia [39]. In addition, our reporter systems demonstrated that various cell lines responded differently to inhibition of a prolyl hydroxylases, either by hypoxia or CoCl 2 . The reporter-transduced U87 cells showed a weaker response to CoCl 2 and hypoxia, relative to normoxic baseline values, when compared to that observed in NIH3T3 reportertransduced cells.
Further validation of the reporter systems was obtained with NIH3T3 and U87 cells bearing a mutated HIF-1a(DODDD)/Fluc fusion gene. Both cell lines demonstrated similar high levels of  reporter expression (BLI and immunoblot), reflecting the effect of ODDD deletion on the stability of the fusion protein [35]. In addition, we demonstrated a significant BLI response in vivo when animals bearing U87/HIF-1a/FLuc xenografts were given an i.p injection of CoCl 2 , but not in animals bearing U87/HIF-1a(DODDD)/FLuc or native FLuc expressing (control) xenografts.
The immunofluorescence analysis of the HIF-1a/FLuc subcellular localization and trafficking in reporter-transduced cell lines compared well with that of endogenous HIF-1a in wild-type cells. However, significant differences were observed between NIH3T3, HEK293 and U87 cells. These observations suggest a difference in cytoplasm-nuclear trafficking of the HIF-1a and its protein fusion between normal and tumor cells. It was previously shown under normal oxygen tension that pVHL is engaged in a constitutive shuttling between the nucleus and cytoplasm [13,40]. Furthermore, it is generally accepted that subcellular localization of the HIF-a subunit in mammalian cells depends on the tumor suppressor factor pVHL, which mediates ubiquitination of HIF-1a. However, it has not been fully established whether proteasomal degradation of HIF-1a occurs in the nucleus, in the cytoplasm, or if it can take place in both compartments. Prior reports are somewhat contradictory. According to some reports [41,42] and the results of our immunofluorescence microscopic studies, the HIF-1a (and HIF-1a/FLuc) translocation to the nucleus appears to be constitutive and independent of oxygen tension, pVHL status and the presence of a functional ODD domain in HIF-1a [13,14]. In many cell lines HIF-1a-pVHL interaction is abrogated in response to hypoxia or hypoxiamimetic agents, and HIF-1a accumulates in the nucleus [43,44,45]. This was clearly demonstrated for the HIF-1a/FLuc fusion in U87 and HEK293 reporter cells, and for endogenous HIF-1a in U87 and HEK293 wild-type cells, and may be occurring to a lesser degree in NIH3T3 cells as well. The presence of HIF-1a/FLuc and HIF-1a in the cytoplasm and nucleus of NIH3T3 reporter and wild-type cells exposed to CoCl 2 and hypoxia indicates continuous shuttling of the protein between the nucleus and the cytoplasm. We speculate that the degradation of HIF-1a in this cell type can take place in both compartments at different rates, possibly due to a residual hydroxylation of HIF-1a or to other non-oxygen-dependent process. Figure 7. Assessment of HIF-1a/FLuc degradation in NIH3T3 cells under different conditions. The degradation rate of the reporter proteins (HIF-1a/FLuc, HIF-1a(DODDD)/FLuc and FLuc) was determined by sequential bioluminescence measurements following addition of a protein synthesis inhibitor (CHX 100 mg/ml) to the medium. The data were fitted to either a single or double exponential equation to estimate the protein degradation rate (half-life). The graphic plots and fitted curves are representative experiments (A and B); NIH3T3 cells lysates were used for immunoblotting with antibodies to HIF-1a to assess HIF-1a/FLuc fusion protein and endogenous HIF-1a protein degradation after CoCl 2 -containing media was removed and replaced with CHX containing media for 0, 7 and 90 min (C). Under the same conditions as that described for panel (C), the cells were analyzed by fluorescence imaging using the same procedure, described in the legend to The predominant nuclear localization of HIF-1a/FLuc and HIF-1a in U87 cells under normoxic conditions indicates that these proteins are either imported or sequestered to the nucleus in U87 reporter cells to a greater extent than that in NIH3T3 cells, or that their nucleus-to-cytoplasm shuttling of HIF-1a is reduced in U87 compared to NIH3T3 cells. This is consistent with the 8-  10 fold difference observed in bioluminescence between the two cell lines under normoxic conditions. Reduced pVHL trafficking was recently demonstrated in clear cell renal cell carcinoma, and was associated with strong nuclear localization of HIF-1a and shorter patient survival [46]. Although the mechanism of HIF-1a shuttling between nucleus and cytoplasm is poorly understood, it is becoming evident, that HIF-1a subcellular distribution and compartmentalization of HIF-1a degradation are regulated in a cell-specific manner [47] and that there are significant differences between normal cells and cancer cells. This may reflect pVHL malfunction in cancer cells or activation of the other transduction pathways in cancer cells.
For example, activation of the PI3K-AKT/mTOR pathway impacts on HIF-1 through an increase in HIF-1a protein levels [37], caused by augmentation of the protein translation [9], [48] or by stabilization of the protein [36]. In addition, it was reported that activation of the PI3K-AKT/mTOR pathway also causes a major increase in nuclear localization of HIF-a/Sima from Drosophila melanogaster, which is a functional homologue of mammalian HIF-1a [49]. U87 human glioma cells are PTENnegative with increased PI3K-AKT/mTOR pathway activity [24,50]. The difference in nuclear localization of HIF-a/Fluc and wild-type HIF-a in this cell line compare with a normal cells is a new observation and requires further investigation.
Deletion of the ODDD in the HIF-1a/Fluc fusion protein prevents hydroxylation by prolyl-hydroxylases and abrogates HIF-1a(ODDD)-pVHL interaction. The HIF-1a(ODDD)/Fluc fusion was shown to have markedly reduced nucleus-to-cytoplasm shuttling in NIH3T3 and U87 reporter cells, and increased accumulation of the mutant fusion protein in the nucleus. This led to the generation of high bioluminescence signals from the stabilized protein in the HIF-1a(ODDD)/Fluc reporter cell lines. These results suggest that nuclear export plays a major role in the control of HIF-1a level in cells, adding a new level of complexity to the regulation of this transcriptional factor.
Our immunofluorescence observations prompted us to study the dynamics of HIF-1a/FLuc protein degradation in greater detail. The question being addressed is whether the difference in the level of HIF-1a/FLuc expression under baseline normoxic conditions, as well as the ability of this fusion protein to accumulate following exposure to CoCl 2 and to moderate hypoxia, is due to a difference in the pVHL-ODDD-dependent degradation process of HIF-1a/ FLuc in different cell lines. A bi-exponential bioluminescence profile of HIF-1a/FLuc protein degradation was observed in NIH3T3, HEK293 and U87 reporter cells, indicating that both ''rapid'' and ''slow'' clearance mechanisms were operative. The half-time of the rapid clearance in these cells was ,3-6 min and consistent with the currently accepted half-life of HIF-1a (,5 min) under normal nonhypoxic conditions [51,52,53]. Since the rapid phase HIF-1a/FLuc degradation was abolished by CoCl 2 treatment and by deletion of the oxygen-dependent degradation domain from HIF-1a in NIH3T3 and U87 reporter cells, it is likely that the rapid clearance component reflects the dynamics of the oxygen-dependent pathway of HIF-1a degradation in these cells. A second, slow clearance component of the HIF-1a/FLuc degradation under normoxic conditions in these cell lines was also revealed in our kinetic analysis of the bioluminescence clearance profile. This component does not reflect degradation of native FLuc protein since immunoblots of cells taken 7, 30 and 90 min after addition of cycloheximide in the clearance studies showed the presence of both non-degraded HIF-1a/FLuc and non-degraded HIF-1a bands.
The dynamic profile and kinetic analysis of the HIF-1a/FLuc degradation in NIH3T3, HEK293 and U87 cells suggests that the rapid and slow components of degradation are compartmentalized. To determine, whether the observed kinetic and anatomic compartmentalization in the cell are related, we imaged the changes in subcellular distribution during cycloheximide-induced clearance of the HIF-1a/FLuc protein in NIH3T3 reportertransduced cells by immunofluorescence imaging. Some of reporter cells cleared their fluorescence (both cytoplasmic and nuclear) more quickly than others. This observation suggests that the observed kinetic compartmentalization may reflect a heterogeneity and difference in HIF-1a/FLuc protein degradation rate within two or more functionally different populations of NIH3T3/ HIF-1a/FLuc cells. Despite a similar bi-phasic degradation profile of HIF-1a/FLuc in NIH3T3 and U87 reporter cells, the basal expression level and subcellular localization of the fusion protein in these cell lines is very different. This suggests the involvement of other regulatory mechanisms, which may account for differences in the spacial distribution and stabilization of HIF-1a in cells with different genetic backgrounds.
The expression of HIF-1a protein is modulated through different pathways that alter HIF-1-dependent transcriptional activity and makes the HIF-1 transcription factor an attractive target for new drug development [4,54,55]. Several examples of HIF-1 inhibitors that target pathways associated with HIF-1 activation have been described [55]. For example, a number of anticancer drugs have been shown to inhibit HIF-1, but none of these drugs target HIF-1 directly and specifically [4,54,56,57,58,59,60,61,62,63], [64,65], [66], [63]. Some of these inhibitors are associated with signaling pathways that include mTOR [36,67], AKT [68], Her2/Neu [9]. However, most inhibitors affect multiple signaling pathways and only indirectly target the HIF-1 transcription factor. Therefore, the ability to identify, visualize and validate changes in the dynamics and stability of the HIF-1a protein is likely to be useful as a pharmacodynamic end point and read-out of inhibition in the development of new anticancer drugs that target HIF-1a in the future. In summary, the introduction of a HIF-1a/FLuc fusion gene into different cell lines allowed us: 1) to study the time-dependent accumulation and degradation of the fusion protein with greater sensitivity and to better quantify the rate of degradation using BLI, compared to that obtained in previous studies based on immunoblot assays; 2) to better assess the differences in subcellular localization and clearance of the fusion protein in normal and tumor cells using immunofluorescence imaging, providing spatial information to better understand HIF-1a trafficking and regulation in different cells; 3) to image HIF-1a/FLuc fusion protein stabilization and accumulation over time in animals bearing reporter-transduced xenografts. Finally, the possibility to visualize the modulation of HIF-1a using the chimeric reporter systems developed in the current study, in combination with transcriptional reporters based on the binding of an active HIF-1 complex to hypoxia-response elements (HRE) which drive reporter genes [28,69,70,71,72,73], provides the opportunity to utilize them as tools for high-throughput screening of compounds affecting HIF-1 activity, both in vitro and in vivo. Figure S1 FACS analysis of reporter-transduced cells. GFP expression in developed reporter cell lines was used for FACS sorting and transduction normalization. Found at: doi:10.1371/journal.pone.0005077.s001 (0.91 MB TIF) Figure S2 Immunofluorescence analysis of HIF-1a and its fusions expression in NIH3T3 cells. The NIH3T3/HIF-1a/Fluc, NIH3T3/HIF-1a(DODDD)/Fluc and NIH3T3 non-tansduced cells were cultured under normoxia and hypoxia for 6 h. Cells were prepared for immunofluorescence as described in Materials and Methods and then were incubated with the anti-HIF-1a antibody followed by a Alexa-568-conjugated secondary antibody. Cells were also visualized for GFP expression. Fluorescence images were acquired at 640 magnification. All panels represent the magnified images after adjusted contrast and brightness to visualize the subcellular protein localization. Found at: doi:10.1371/journal.pone.0005077.s002 (4.01 MB TIF) Figure S3 Immunofluorescence analysis of HIF-1a and its fusions expression in U87 cells. The U87/HIF-1a/Fluc, U87/ HIF-1a(DODDD)/Fluc and U87 non-tansduced cells were cultured under normoxia and hypoxia for 6 h. Cells were prepared for immunofluorescence as described in Materials and Methods and then were incubated with the anti-HIF-1a antibody followed by a Alexa-568-conjugated secondary antibody. Cells were also visualized for GFP expression. Fluorescence images were acquired at 640 magnification. All panels represent the magnified images after adjusted contrast and brightness to visualize the subcellular protein localization. Found at: doi:10.1371/journal.pone.0005077.s003 (3.46 MB TIF) Figure S4 Immunofluorescence analysis of HIF-1a and HIF-1a/FLuc expression in HEK293 cells. The reporter-transuced HEK293/HIF-1a/Fluc and non-transduced HEK 293 cells were cultured under normoxia or hypoxia for 6 h. Cells were prepared for immunofluorescence as described in Materials and Methods and then were incubated with the anti-HIF-1a antibody followed by a Alexa-568-conjugated secondary antibody. The reporter cells were also visualized for GFP expression. Fluorescence images were acquired at 640 magnification. All panels represent the magnified images after adjusted contrast and brightness to visualize the subcellular protein localization. Found at: doi:10.1371/journal.pone.0005077.s004 (3.10 MB TIF)