Roles of hypoxia-inducible factor-1α (HIF-1α) versus HIF-2α in the survival of hepatocellular tumor spheroids

Authors


  • Potential conflict of interest: Nothing to report.

Abstract

Hypoxia-inducible factors (HIFs) provoke adaptation to hypoxic stress occurring in rapidly growing tumor tissues. Therefore, overexpression of HIF-1 or HIF-2 is a common feature in hepatocellular carcinoma but their specific function is still controversially discussed. To analyze HIF function in hypoxia-induced cell death we created a stable knockdown of HIF-1α and HIF-2α in HepG2 cells and generated tumor spheroids as an in vitro hepatocellular carcinoma model. Knockdown of HIF-1α enhanced expression of HIF-2α and vice versa. Unexpectedly, knockdown of HIF-1α or HIF-2α increased cell viability as well as spheroid size and decreased caspase-3 activity. Antiapoptotic Bcl-XL expression increased in both knockdown spheroids, whereas proapoptotic Bax was only reduced in HIF-1α-knockdown cells. Furthermore, an HIF-2α-knockdown significantly increased Bcl-2/adenovirus E1B 19 kDa-interacting protein 3 (BNIP3) expression in an HIF-1α-dependent manner. Concomitantly, electron microscopy revealed a substantial increase in autophagosomal structures in HIF-2α-knockdown spheroids and mito-/lysotracker costaining confirmed lysosomal activity of these autophagosomes. Blocking autophagosome maturation using 3-methyladenine restored cell death in HIF-2α-knockdown clones comparable to wildtype cells. Conclusion: An HIF-1α-knockdown increases HIF-2α expression and shifts the balance of Bcl-2 family members toward survival. The knockdown of HIF-2α raises autophagic activity and attenuates apoptosis by enhancing HIF-1α expression. Our data indicate that enhanced expression of one HIF-isoform causes a survival advantage in hepatocellular carcinoma development. HEPATOLOGY 2010

Hepatocellular carcinoma (HCC), one of the most frequent cancers worldwide, is characterized by rapid progression and bad prognosis.1 HCC develops as a consequence of chronic hepatitis or liver cirrhosis. Key to the malignant outcome is the hypoxic microenvironment, affecting cell death. In nontransformed cells hypoxia inhibits proliferation and severe hypoxia even promotes cell death.2 Tumor cells often adapt to hypoxia and hence are highly resistant toward hypoxia-induced cell death.3

Hypoxia-inducible factors (HIFs) are facilitating adaptation to hypoxia. HIFs are frequently up-regulated in HCC and control tumor progression and sensitivity to radiation therapy.4-6 They are heterodimers of various α-subunits and a common β-subunit. The α-subunits are hardly detectable at the protein level under normoxia because they are rapidly degraded in an oxygen-dependent manner. Three HIF-isoforms are identified as HIF-1, HIF-2, and HIF-3, referring to different α-subunits. Although HIF-1α is ubiquitously expressed, HIF-2α is restricted to specific cell types, including endothelial cells and hepatocytes. HIF-1 seems to be important for the “acute phase” of tumor development that is characterized by metabolic changes, decreased translation, and cell cycle arrest. However, HIF-2 appears to be relevant for the “chronic phase” to stimulate proliferation and erythropoiesis.6, 7 HIFs also control the balance of necrotic and apoptotic events. Although necrosis is an adenosine triphosphate-independent type of cell death, apoptosis is strictly regulated by proteins of the Bcl-2 family. They comprise antiapoptotic proteins, i.e., Bcl-2 and Bcl-XL and the proapoptotic proteins Bax and Bak, which were previously linked to HIF-signaling.8, 9 Thus, the balance of these factors is critical for regulation of cell demise. A further properly controlled process that interplays with apoptosis is autophagy, initially related to cell death. Recent approaches provided evidence that it is a continuously ongoing process needed for recycling of cytoplasmic components like mitochondria. Autophagy is characterized by the accumulation of multilamellar vesicles, which engulf cytoplasm and organelles, forming so-called autophagosomes. Their fusion with lysosomes generates autolysosomes, hydrolytically degrading vesicle contents for energy recovery. A key mediator of autophagy induction is Beclin-1, which interacts with the antiapoptotic proteins Bcl-2 and Bcl-XL, providing a link between autophagy and apoptosis.10, 11 During adaptation to hypoxia mitochondrial autophagy can be induced by Bcl-2/adenovirus E1B 19 kDa-interacting protein 3 (BNIP3) to prevent cell death by attenuating reactive oxygen species production.12, 13

To study cell death regulation in HCC, we knocked down HIF-1α, HIF-2α, and HIF-1β in HepG2 cells and generated tumor spheroids as an in vitro tumor model. Spheroids lacking one HIF-isoform up-regulate the other and thereby elicit survival advantages by dysregulating autophagy and apoptosis.

Abbreviations:

BNIP3, Bcl-2/adenovirus E1B 19 kDa-interacting protein 3; HCC, hepatocellular carcinoma; HIF, hypoxia-inducible factor; 3-MA, 3-methyladenine; siRNA, small interfering RNA; TEM, transmission electron microscopy.

Materials and Methods

Materials.

All chemicals were purchased from Sigma-Aldrich (Schnelldorf, Germany), if not indicated otherwise. For details, see Supporting Information.

Tumor Spheroid Cell Culture.

HepG2 spheroids were grown from 1 × 106 cells in coated (Sigmacote) 250-mL spinner flasks (Integra Biosciences, Fernwald, Germany) with 250 mL complete medium in a Cell Spin stirrer system (22.5 rotations/min, Integra Biosciences). The 125 mL cell culture medium was replaced every second day. Because spheroids develop a natural oxygen gradient, they were only incubated at normoxia. Ten spheroids per clone were analyzed to determine growth. For details, see Supporting Information.

Immunohistochemical Analysis.

Stainings of HIF-1α, -2α, -1β, and cleaved caspase-3 were performed as described in the Supporting Information. Peroxidase and Mayer's hemalum solution (Merck, Darmstadt, Germany) stainings were monitored with Axioskop 40. For analyzing cell viability and cleaved caspase-3-positive cells, 10 sections per experiment were stained with 4′,6-diamidino-2-phenylindol (DAPI) for 30 minutes and mounted with Fluromount-G (Biozol, Eching, Germany). Cell nuclei were captured by Axioskop 40. Viable cells were defined as cells with intact nuclei. For the calculation of cell viability, the number of viable cells was compared to the total cell number. For analysis of cleaved caspase-3-positive cells 20 spheroid sections per experiment were counted for peroxidase positive signals and compared to viable cells. Definition and counting of cell nuclei was performed by the TissueQuest Cell Analysis Software Beta v. 2.2 (TissueGnostics, Wien, Austria).

Mito- and LysoTracker Costaining.

HepG2 cells were seeded 1 day prior to experiments. The following day medium was changed and cells were stained for 30 minutes with 250 nM MitoTracker (Invitrogen, Karlsruhe, Germany), followed by a 3-day incubation at 21% or 1% O2. Afterwards, cells were stained with LysoTracker (30 μM; Invitrogen) for 30 minutes. Fluorescence was monitored with Axiovert 200M (Zeiss, Goettingen, Germany).

Statistical Analysis.

Each experiment was performed at least three times. Data in bar graphs are given as mean values ± standard error of the mean (SEM). Statistically significant differences were calculated by a one-way analysis of variance (ANOVA) followed by Tukey tests. P values are given in the figure legends.

For further experimental details, see Supporting Information.

Results

Compensatory Increase of HIF-1α in HIF-2α-Knockdown Cells and Vice Versa.

Initially, we determined the sensitivity of HCC cells toward hypoxia-induced cell death. Propidium iodide staining after 6 days hypoxia revealed a similar increase of necrosis in HepG2 (23.0 ± 4.4% versus 46.5 ± 3.5%) and HUH-7 (17.6 ± 2.7% versus 42.4 ± 8.9%) monolayer cells under normoxia versus hypoxia. Not observing a difference in sensitivity between cells, further experiments were performed only with HepG2 cells.

Hypoxia caused HIF-1α protein expression after 4 hours in HepG2 wildtype (WT) cells (Fig. 1A). This response was enhanced in HIF-2α-knockdown cells (sh2α), whereas the knockdown of HIF-1α (sh1α) completely suppressed protein expression. HIF-2α protein expression was also enhanced in WT cells exposed to hypoxia (Fig. 1B). Although expression was strongly reduced in the sh2α cells, sh1α cells displayed increased HIF-2α protein expression under hypoxia compared with WT cells. Thus, the knockdown of one α-subunit enhanced expression of the remaining α-subunit in HepG2 cells under hypoxia. In contrast, HIF-1β expression was neither altered in sh1α nor in sh2α cells compared to WT cells, but was completely lost in the HIF-1β-knockdown clone (sh1β, Fig. 1C). However, HIF-1α as well as HIF-2α expressions were reduced in response to hypoxia in the sh1β clone compared to WT cells (Fig. 1D,E).

Figure 1.

Increased HIF-1α expression in sh2α cells and vice versa. (A-E) WT, sh1α, sh2α, and sh1β cells were incubated at 21% O2 or 1% O2 for 4 hours. Expression of HIF-1α, -2α -1β, and actin was followed by western analysis. (F) Peroxidase (brown) detection of HIF-1α, -2α, or -1β expression and Mayer's hemalum (blue) stainings in sections of WT, sh1a, sh2α, and sh1β spheroids. Scale bars = 100 μm.

Additionally, we used spheroids similar in size to follow expression of the HIF-subunits. We found a gradual distribution of HIF-1α, HIF-2α, and HIF-1β (Fig. 1F). Necrotic cores were surrounded by cells without detectable HIF expression presumably because of a general regression of protein biosynthesis. The area adjacent to the necrotic region showed high HIF expression that decreased toward the periphery. Interestingly, HIF localization of the remaining subunit reached less hypoxic regions in the sh1α and sh2α clone, confirming enhanced expression of the respective isoform. Protein expression patterns in monolayer culture and spheroids implied that a knockdown of one HIF-isoform is compensated by enhanced expression of the other one in HepG2 cells.

Growth and Cell Viability Were Enhanced in sh1α and sh2α Tumor Spheroids.

To characterize the influence of the HIF-knockdown on spheroid growth we determined the diameters of WT, sh1α, sh2α, and sh1β spheroids up to 15 days of culture. Interestingly, sh1α and sh2α spheroids revealed a significant growth benefit compared to WT and sh1β spheroids. sh1α and sh2α clones reached ∼700 μm (Fig. 2A,B), whereas WT and sh1β spheroids grew to ∼500 μm under identical conditions (Fig. 2C). To determine whether this growth advantage resulted from differences in cell viability, Trypan blue staining with complete spheroids was performed in parallel to growth measurements. Comparing 15-day-old spheroids of the same size revealed that the necrotic core was smaller in sh1α and in sh2α clones compared to WT or sh1β spheroids (Fig. 2D). In addition, we analyzed the ratio of living to total cell numbers in sections of 12-day-old spheroids. We observed that sh2α and sh1α clones contained (3.03 ± 0.59-fold, 2.37 ± 0.46-fold) more viable cells than WT spheroids (Fig. 2E). Viability was not significantly altered in the sh1β clone (1.24 ± 0.09-fold). Interestingly, HIF expression was detectable when spheroids reached a diameter of ∼300 μm, which was paralleled by growth advantage of the knockdown clones and occurrence of the necrotic core. These data indicate that spheroid growth is enhanced in sh1α and sh2α clones and correlates with reduced hypoxic cell death.

Figure 2.

Enhanced growth and cell viability in sh1α and sh2α tumor spheroids. (A-C) WT, sh1α, sh2α, and sh1β spheroids were grown for 15 days. Sizes were determined after times indicated. (D) 15-day-old whole spheroids were stained with trypan blue. Scale bars = 200 μm. (E) Intact nuclei of 12-day-old WT, sh1α, sh2α, and sh1β spheroids were calculated in comparison to all nuclei detectable. *P < 0.05 compared to control; #P < 0.05 compared to the sh2α clone.

Reduced Caspase-3 Activity, Dysregulated Bcl-XL and Bax Expression in sh1α and sh2α Tumor Spheroids.

To analyze the pathway of cell demise, expression of cleaved caspase-3 was determined as an indicator of apoptosis. An average of 4.11 ± 0.77% of the cells in WT spheroids showed caspase-3 activity in 10-day-old spheroids, which was normalized to 1. In sh1α and sh2α spheroids caspase-3 activity was significantly reduced to 0.36 ± 0.10-fold and 0.32 ± 0.12-fold, respectively (Fig. 3A). However, caspase-3 activity was insignificantly reduced to 0.67 ± 0.24-fold in sh1β spheroids. To elucidate underlying mechanisms we examined the balance between pro- and antiapoptotic Bcl-2 family members. Bcl-2 protein was not detectable in the spheroids (data not shown). Antiapoptotic Bcl-XL protein expression was increased in sh1α and sh2α spheroids compared to WT or sh1β clones (Fig. 3B). At the same time, proapoptotic Bax expression was diminished in sh1α spheroids and completely absent in the sh1β clone, but remained unaltered in sh2α spheroids (Fig. 3C). These data suggest that sh1α cells reduce hypoxia-induced apoptosis by decreasing Bax and increasing Bcl-XL, whereas sh2α spheroids might require additional mechanisms to overcome proapoptotic responses.

Figure 3.

Attenuated apoptosis in sh1α and sh2α cells. (A) Cleaved caspase-3-positive cells in sections of 10-day-old WT, sh1α, sh2α, and sh1β spheroids were calculated relative to intact nuclei. *P < 0.05 compared to control. (B,C) Whole cell protein was isolated from 12-day-old WT, sh1α, sh2α, and sh1β spheroids. Bcl-XL, Bax, and actin expression were detected by western analysis.

Increased BNIP3 Expression in sh2α Cells.

To further elucidate mechanisms accounting for the survival advantage of sh2α spheroids we examined the expression of the BH3-only and Bcl-2 family member BNIP3. Although BNIP3 was classified as an HIF-1-dependent inducer of apoptosis,9 recent work allocated survival mechanisms by inducing autophagy.12 Because BNIP3 protein is predominantly stabilized under acidic conditions present in whole spheroids,14 it became rapidly degraded in spheroid-suspensions during cell isolation. Therefore, BNIP3 mRNA as well as protein levels were analyzed in HepG2 monolayer cells incubated under hypoxia for 16 hours. In WT cells, hypoxia induced BNIP3 mRNA (2.37 ± 0.17-fold) and protein expression (Fig. 4A,B). In sh2α cells, BNIP3 mRNA induction was significantly enhanced (4.23 ± 0.37-fold), also evident at the protein level. sh1α and sh1β cells failed to induce BNIP3 expression in response to hypoxia. These data corroborate BNIP3 as an HIF-1 target in HepG2 cells.

Figure 4.

BNIP3 mRNA and protein in sh2α cells. WT, sh1α, sh2α, and sh1β monolayer cells were incubated at 21% or 1% O2 for 16 hours. (A) Total mRNA was analyzed for BNIP3 and actin expression by quantitative real-time polymerase chain reaction (PCR). The ratio of BNIP3 versus actin under control conditions was set to 1. *P < 0.05 compared to hypoxia of WT cells. (B) BNIP3 and actin expression were followed by western analysis.

Increased Autophagy in sh2α Tumor Spheroids.

As BNIP3 was recently associated with hypoxia-induced autophagy,12 we used transmission electron microscopy (TEM) to identify autophagosomal structures in the spheroids. Cultivating WT spheroids for 12 days revealed an increased number of autophagic vacuoles in the outer, HIF-expressing spheroid cell layers (Fig. 5A, left panels). sh2α spheroids showed increased formation of autophagic vacuoles compared to WT spheroids, whereas the number of autophagic vacuoles was strongly attenuated in sh1α and sh1β spheroids. Interestingly, autophagosome formation was found in all clones in the perinecrotic regions of the spheroids, an area with obvious loss of cell integrity (data not shown). Furthermore, TEM pictures showed no influence of the HIF-1α-, -2α-, or -1β-knockdown on single cell size and cell density. To analyze whether autophagic vacuoles were lysosomally active, we stained HepG2 monolayer cells with the mitochondrial specific dye MitoTracker (green). Afterwards cells were incubated for 3 days under normoxia or hypoxia. Finally, LysoTracker (red) was added to mark lysosomally active vesicles. Under normoxia there was little or no autolysosomal activity, irrespective of HIF-1α, -2α, or -1β expression (data not shown). Hypoxia enhanced colocalization of condensed mitochondria and acidic lysosomes (yellow), predominantly in the sh2α clone compared to WT cells (Fig. 5A, right panel). In sh1α and sh1β cells the formation of acidic vesicles or colocalization with mitochondria was barely detectable. To exclude an impact of HIFs by directly altering Beclin-1 expression, we performed immunohistochemical stainings in spheroids but noticed no differences (Supporting Information Fig. S1A). Apparently, formation of autolysosomal structures as well as BNIP3 expression is enhanced in the sh2α clone, but attenuated in sh1α and sh1β clones compared to WT cells.

Figure 5.

Enhanced autophagy in sh2α tumor spheroids. (A, left panel) Areas between 2nd and 5th outer cell layer of 12-day-old WT, sh2α, sh1α, sh1β spheroids were analyzed by TEM (original magnification ×3,000). White arrows represent autophagic vacuoles; n, nucleus. (A, right panel) WT, sh1α, sh2α, and sh1β monolayer cells were stained with MitoTracker (green) before incubation at 1% O2 for 3 days followed by LysoTracker (red) staining. Colocalization is shown in merged images (yellow). Scale bars = 20 μm. (B) 8-day-old sh2α spheroids were cultured for 4 days with or without 5 mM 3-MA. Trypan blue staining was performed with whole spheroids. Scale bars = 200 μm.

To determine whether increased autophagy accounts for attenuated cell death in sh2α spheroids, autophagy was inhibited with 3-methyladenine (3-MA) (5 mM), which specifically inhibits autophagosome formation.15 Trypan blue staining revealed that 3-MA increased cell death in sh2α spheroids (Fig. 5B). Our data provide evidence that a decrease in Bax and an increase in Bcl-XL expression reduced apoptosis in sh1α spheroids. sh2α spheroids overexpress Bcl-XL and BNIP3 to diminish apoptosis and to induce autophagy.

Knockdown of HIF-1α Reversed Survival Advantages in sh2α Tumor Spheroids.

To analyze whether the survival advantage of sh2α spheroids was due to up-regulation of HIF-1α, we knocked down HIF-1α by small interfering RNA (siRNA) treatment (si1α) in 7-day-old sh2α spheroids prior to the occurrence of specific size differences (Fig. 2A). HIF-1α mRNA was significantly reduced (0.55 ± 0.08-fold) compared to untreated or siControl (siC)-treated spheroids (Fig. 6A). Furthermore, si1α spheroids reduced HIF-1α protein expression in the outer, less hypoxic cell layers (Fig. 6B). Determination of spheroid size revealed reduced growth at day 11 comparable to sh1β spheroids (Fig. 6C) and an increase in caspase-3-positive cells compared to untreated or siC spheroids (Fig. 6D). Furthermore, Bax protein and BNIP3 mRNA expression were lowered in si1α spheroids (Fig. 6E,F). Our data suggest that up-regulation of HIF-1α in sh2α spheroids accounts for their survival advantage.

Figure 6.

HIF-1α-knockdown reversed the survival advantage in sh2α tumor spheroids. (A) Total mRNA was analyzed for HIF-1α and 18S ribosomal RNA expression by quantitative real-time PCR. The ratio of HIF-1α versus 18S ribosomal RNA of sh2α spheroids was set to 1. (B) Peroxidase (brown) detection of HIF-1α expression and Mayer's hemalum (blue) staining in sections of siC and si1α spheroids. Scale bars = 200 μm. (C) Sizes of sh2α, sh2α + siC, sh2α + si1α, and sh1β spheroids were determined after times indicated. (D) Cleaved caspase-3-positive cells in sections of sh2α, sh2α + siC, sh2α + si1α spheroids were calculated in proportion to intact nuclei. (E) Whole-cell protein was isolated from sh2α, sh2α + siC, sh2α + si1α spheroids. Bax and actin expression were detected by western analysis. (F) Total mRNA was analyzed for BNIP3 and 18S ribosomal RNA expression by quantitative real-time PCR. The ratio of BNIP3 versus 18S ribosomal RNA of sh2α spheroids was set to 1. *P < 0.05 compared to control.

Discussion

Targeting HIF emerges as a rational approach to treat solid tumors. HCCs are known to overexpress both HIF-α-subunits correlating with poor patient prognosis.5, 16 Taking into account that HIF isoforms differentially regulate gene expression during tumor progression, an analysis of the specific functions of each subunit is needed. Therefore, we established HepG2 tumor spheroids deficient for HIF-1α, HIF-2α, or HIF-1β as an HCC in vitro tumor model to elucidate cell death regulation.

We observed that HIF-1β-knockdown cells decreased HIF-1α as well as HIF-2α protein expression. As HIF-1α and HIF-2α mRNA expression was induced under hypoxia in HepG2 cells and the HIF-1α promotor contains hypoxia-responsive elements,17 we suggest that both isoforms regulate each other. Furthermore, HIF-1β expression was enhanced in perinecrotic areas of spheroids, presumably as a result of glucose deficiency, which accompanies hypoxia in spheroids and is known to induce HIF-1β expression.18

Importantly, the knockdown of one α-subunit increased the expression of the other α-subunit. Spheroids lacking one α-subunit displayed growth and survival advantages as compared to both WT and β-knockdown spheroids. This was accompanied by a shift in the expression of Bcl-2-family members toward antiapoptotic proteins and induction of autophagy. Specifically, HIF-1α-depleted spheroids gained survival benefits by decreasing proapoptotic Bax but increasing antiapoptotic Bcl-XL expression. In contrast, HIF-2α-knockdown spheroids escaped apoptosis by up-regulation of antiapoptotic Bcl-XL and BNIP3-dependent autophagy (Fig. 7).

Figure 7.

HIF-α-depletion and tumor spheroid survival. In hypoxic WT cells Bcl-XL binds Beclin-1 to attenuate autophagy. However, autophagy is induced because of the interference of hypoxia-induced BNIP3 with the Beclin-1-Bcl-XL-complex, thereby releasing Beclin-1. Additionally, apoptosis is controlled by mitochondrial Bax and Bcl-XL. In hypoxic sh2α cells BNIP3-Bcl-XL complexing is amplified and Beclin-1 is released to induce autophagy. Moreover, enhanced Bcl-XL expression decreases apoptosis. In hypoxic sh1α cells autophagy is inhibited because BNIP3 expression is lost. Enhanced Bcl-XL expression inhibits Beclin-1 release. Apoptosis is repressed by increased Bcl-XL and decreased Bax expression.

Our finding that the loss of one HIF α-subunit was compensated by up-regulation of the respective remaining isoform in HepG2 cells was previously noticed in glioma cells and chondrocytes.19, 20 Compensation might circumvent the loss of target gene expression of, e.g., vascular endothelial growth factor, adrenomedullin, or glucose transporter 1, which are regulated by both HIF isoforms.7 Compensatory mechanisms may result from increased reactive oxygen species formation in HIF-2α-knockdown cells. HIF-2 up-regulates superoxide dismutase and catalase20, 21 and their loss may increase reactive oxygen species, thereby inducing HIF-1α expression. As there are currently no studies to explain up-regulation of HIF-2α with a HIF-1α-knockdown, we suspect indirect signaling-pathways. Further investigations will be needed to analyze details.

In recent studies HIF-2α promoted tumor growth by increasing proliferation, whereas HIF-1α antagonized this function.22 We propose that HepG2 spheroids lacking either HIF-1α or -2α and, thus, increased expression of the remaining isoform, are characterized by reduced apoptosis and increased spheroid size without changes in tumor cell proliferation (Supporting Information Fig. S1B). In line, we were able to reverse the survival advantage of spheroids lacking HIF-2α and, thus, increased HIF-1α expression, by HIF-1α siRNA-treatment. Reduced apoptosis was attributed to changes in the expression of various Bcl-2 family proteins such Bcl-XL, Bax, and BNIP3. Although Chen et al8 recently proposed Bcl-XL as an HIF-1-specific target in prostate cancer cells, our data suggest that it is regulated by both HIF-1 and HIF-2 in HepG2 cells. Bcl-XL exerts its antiapoptotic function by antagonizing Bax-induced apoptosis (for review, see Ref. 23). In agreement with previous work, Bax expression was reduced in spheroids with a HIF-1α-knockdown.24 Conclusively, HIF-1α-knockdown spheroids develop survival advantages by increasing Bcl-XL and decreasing Bax expression, thus shifting the apoptotic balance toward antiapoptotic family members. Although we found down-regulation of Bax in HIF-1β- or HIF-2α-knockdown spheroids with an additional HIF-1α-knockdown, these clones did not develop survival advantages. As the anti-apoptotic player Bcl-XL and autophagy inducing BNIP3 were also reduced, we propose a new balance in Bcl-2 family members to account for these responses.

HIF-2α-knockdown cells up-regulated HIF-1α and enhanced Bcl-XL and BNIP3 expression (Fig. 7). This corroborates previous reports characterizing BNIP3 as an HIF-1-specific target.25 This is further supported by the observation that HIF-1α siRNA-treatment of HIF-2α-knockdown cells down-regulated BNIP3 expression. Recent work supports a role of BNIP3 in autophagy, eliciting prosurvival effects.12 BNIP3 binds to the Beclin-1/Bcl-XL complex, thereby releasing Beclin-1, which subsequently induces autophagosome formation.26 Further crosstalk between autophagy and apoptosis is provided by localization-dependent effects of Bcl-XL. When located at mitochondria, Bcl-XL has antiapoptotic functions. However, localization at the endoplasmic reticulum favors the interaction between Bcl-XL and Beclin-1 to inhibit the autophagic activity of the latter,27, 28 which can be overcome by increasing BNIP3 expression. In line, the increased expression of Bcl-XL in both HIF-1α and -2α-knockdown spheroids was accompanied by enhanced BNIP3 expression in HIF-2α-deficient cells only, whereas BNIP3 was completely lost in both HIF-1α- and -1β-knockdown cells. This was correlated with increased autophagy in the outer hypoxic cell layers of the spheroids and in hypoxic monolayer cells deficient in HIF-2α. Loss of BNIP3 attenuated autophagy in HIF-1α- and -1β-knockdown cells. Because Beclin-1 expression remained unaltered in all clones we exclude an impact of HIFs on autophagy by directly altering Beclin-1 expression. Interestingly, perinecrotic autophagy occurred in all spheroids independent of the HIF-status. We suppose that this HIF-independent autophagy resulted from anoxic conditions in that particular area of the spheroid pointing to a complete loss of cell integrity.29 Blocking autophagosome formation downstream of the proposed HIF/Beclin-1/Bcl-XL/BNIP3 signaling circuits diminished survival advantages of HIF-2α-knockdown spheroids, thus reinforcing autophagy as an important mechanism to support survival of HIF-2α-deficient cells under hypoxic conditions.

We established a tumor spheroid model with HepG2 cells deficient for HIF-1α, HIF-2α, or HIF-1β to analyze the different roles of HIFs in cell viability. We further provide evidence that HIF-1α-deficient tumor cells shift the Bcl-2 family proteins toward an antiapoptotic expression profile, up-regulating Bcl-XL and down-regulating Bax. In contrast, tumors lacking HIF-2α develop a survival advantage by increasing antiapoptotic Bcl-XL expression and enhancing autophagy. Because the knockdown of one HIF-subunit is compensated by a survival advantage induced by the remaining subunit, targeting just HIF-1α or HIF-2α for HCC treatment should be carefully considered.

Acknowledgements

We thank Tanja Keppler and Anke Biczysko for excellent technical support.

Ancillary