Methods
Cell culture
LY2, MCF7, CAMA-1, MDA-MB-175, MDA-MB-361, MDA-MB-231 and MDA-MB-435 cell lines were cultured in DMEM. BT474, T47D, ZR75B, MPE600, HCC1428 cell lines were cultured in RPMI. All cell lines were obtained as a kind gift from Dr. Joe Gray, (Oregon Health Sciences University, USA), and maintained with 10% Fetal Bovine Serum (FBS) at 37 °C in 5% CO2 and 21% O2. All cell lines were verified by short tandem repeat (STR) genotyping. Genomic DNA was extracted by Wizard SV Genomic DNA purification system (Promega). STR profiles were compared with publically available profiles using Promega Powerplex 1.2.
Reagents
Antibodies used in this report are as follows: ER-α (clone HC-20, Santa Cruz), HIF-1α (BD Bioscience), HIF-2α (NB100-132, Novus Biologicals), phospho-p70 S6 kinase (p-p70-S6K; 9205, Cell Signaling Technology), phospho-4E-BP1 (2855, Cell Signaling Technology), β-Actin (clone AC-15, Sigma-Aldrich), Alexa Fluor- 488 (A-11008, Invitrogen), Alexa Fluor-594 (A11012, Invitrogen), HRP-anti-Mouse IgG (NA931V, GE Healthcare), and HRP-anti-Rabbit IgG (NA934V, GE Healthcare). Reagents used in this study are estrogen (17-β-estradiol) used at 10 nM (Sigma-Aldrich), MG132 used at 10 uM (Cayman Chemical).
Western blot
Western blots were performed as previously [
34]. Briefly, cells were lysed with urea lysis buffer (9 M urea, 150 mM, β-mercaptoethanol and 75 mM Tris pH 7.4), or RIPA buffer (Cell Signaling #9806), sonicated for 30 s and centrifuge at 15,00 rpm at 4C for 30 min. Protein quantification was performed by Bradford (BioRad Cat.500-0205), and 20–100 μg of protein were loaded in each well of a polyacrylamide gel. PVDF membranes were blocked with 5% milk in TBS-T/0.05% (Tris Buffered Saline with Tween 20 to 0.05%) at room temperature (RT) for 30 min, primary antibodies were incubated overnight at 4C in TBS-T/0.05%. After washing with TBS-T/0.05%, secondary incubation was performed at RT for 45 min followed by TBST/0.05% washes. Western blot signal was detected using Enhanced Chemiluminescent (ECL) substrate (Pierce 32106, or GE Healthcare RPN2235) in a FlourChemE machine. Exposures were chosen to provide maximum visual information about the changes in band intensity without causing overexposures that would obscure faint signals in neighboring lanes. Each western blot was repeated from 3 to 6 times, and the averages and standard deviations for the intensities of the western blot replicates for each figure are graphed and presented in tabular form in the Additional files, as indicated throughout the manuscript.
Immune fluorescence
Cells were cultured on glass coverslips in six well plates, using complete media (RPMI or DMEM) with phenol red, supplemented with 10% FBS. Hypoxic samples were placed into the HypOxygen H35 Workstation for 48 h. Coverslips were washed twice in Phosphate Buffered Saline (PBS), fixed in Acetone 10 min at -20C, PBS washed, and nonspecific antibody binding blocked with 10% Bovine Serum Albumen (BSA) and 5% goat serum. Antibodies specific for the estrogen receptor alpha chain (Santa Cruz Biotechnology, HC-20 sc-543) were used at 1:100, anti- rabbit-488 (Molecular Probes/Invitrogen) at 1:1000. Nuclei were counterstained with DAPI (4',6-diamidino-2-phenylindole).
Plasmids
Plasmid transfection was performed using Lipofectamine (Invitrogen) and Plus Reagent (Invitrogen) in Opti-MEM. ShScramble, sh
HIF1A and sh
HIF2A plasmids were previously described [
34], HIF-1αODD was used to produce stabilize HIF-1α at normoxia [
23] pCMV-hER-α was used to overexpress ER-α [
35]. Reporter assays used ALT-4, a plasmid encoding canonical Estrogen Response DNA binding sequences (Estrogen Response Elements, ERE) controlling expression of firefly luciferase [
35].
Hypoxia treatment
Cells were subjected to 1% O2 for the specified time (HypOxygen H35 Workstation). Cells were passaged under normoxic conditions but cultured and harvested inside the hypoxia chamber.
ER-α reporter assays and analysis of media estrogenic effects
To test the estrogenic effects of our normal culture media, cells were transfected with a plasmid encoding canonical Estrogen Response DNA binding sequences (Estrogen Response Elements, ERE) controlling expression of firefly luciferase [
35]. The media we tested included: 1) “standard growth media” containing phenol red, (DMEM for MCF7 or RPMI for BT474, T47D and ZR75B) with 10% FBS; or 2) “E
2 (−), estradiol-free” composed of phenol red-free DMEM with 10% charcoal stripped FBS; or 3) “E
2 (+), defined estrogen media”, composed of phenol red-free DMEM, 10% charcoal stripped FBS supplemented with 10 nM estradiol. Cells incubated with either standard growth media or cells grown in defined estrogen medium had similarly high levels of luciferase signal (Additional file
2). Since there was no difference between the standard media and the estrogen-defined, phenol red-free media, all experiments, ER-α transcriptional activity assays and proliferation assays were performed in standard media.
For experiments in Fig.
4, cells were transfected with a plasmid encoding canonical Estrogen Response DNA binding sequences (Estrogen Response Elements, ERE) controlling expression of firefly luciferase [
35] and incubated at normoxia or at hypoxia for 48 h.
All luciferase activity was determined by Dual-Glo Luciferase assay reagent (Promega) measured in a Monolight 2010 Luminometer (Promega). Firefly luciferase was normalized to protein concentration.
In vitro growth curves
One hundred thousand cells were plated on 6-well plates in triplicate. The following day, cells were washed with PBS, and phenol red free DMEM with 10% charcoal stripped FBS was added. 10 nM estradiol was added for E2 (+) media or ethanol carrier for the E2 (−) media. Every 3 to 4 days, cells were counted.
Cell cycle analysis
Cells were plated into 6-well plates and cultured under appropriate conditions for 3 days (MCF7, T47D, ZR-75-B) or 6 days (BT474). Cells were harvested, fixed and permeabilized in suspension in 70% ethanol, nuclei were stained with 40 μg/ml propidium iodide and 100 μg/ml RNAse A. DNA content was analyzed using an Accuri C6 Fluorescence Activated Cell Sorter (BD Biosciences). At least 150,000 cells were analyzed per sample. Cell cycle fractions were determined using the cell cycle analysis component in the FLOWJO software package (FLOWJO Enterprise).
Real-time quantitative reverse transcriptase PCR
Total RNA was extracted from cells using RNeasy Mini Kit (Qiagen) following the manufacturer’s instructions. cDNA was generated from 1.5 μg of RNA using iScript (BioRad) following the manufacturer’s instruction. Power SYBR Green PCR reactions were performed in triplicate for each sample and analyzed using the AB Step One Plus sequence detection system. Data were normalized to TBP levels.
Statistical analysis
Student’s
t-test was used to determine significance. All error bars represent the standard error of the mean. Two way ANOVA was performed for the growth curve experiments in Fig.
5 using the Graphpad Software package. Student’s
t-test was also used to analyze whether our data would reveal enhanced proteolysis of the estrogen receptor alpha under hypoxic conditions reported in Fig.
3d and Additional file
3D . Although we may not have thoroughly inhibited proteolysis in these experiments, the MCF7 sample showed a statistically clear increase in protein stability with MG132 treatment in hypoxic conditions versus normoxic conditions, (significance level (alpha) = 0.05,
p = 0.044), Additional file
3D. Data for the other three cell lines did not reach this significance, with p values for BT474 of 0.765, for T47D of 0.98, and for ZR-75-B of 0.69.
Discussion
The recent move towards precision medicine, which aims to prospectively identify clinical drug responders versus non-responders, requires a parallel preclinical move to the experimental use of panels of tumor-derived cell lines to more accurately assess possible therapeutic responses and frequencies. Here, we bridge that gap to define the effects of hypoxia on ER-α expression and activity in a panel of breast cancer cell lines that represent common genomic variations seen in human ER-α positive tumors. Previous work with exemplar cell lines, primarily MCF7 or ZR75, has suggested a role for hypoxia in reducing ER-α expression [
28,
29,
31,
44]. However the literature taken as a whole contains multiple, conflicting results particularly with regard to mechanisms and with no consideration given to the potential influence of genetic background variations that commonly occur in ER-α positive tumors. Thus, common responses that might be relevant for improving hormonal therapies remain unclear.
We surprisingly find that reduced oxygen availability produces rapid reduction of ER-α protein levels in all of our genetically-diverse cell lines (Fig.
1a, b). This is largely effected by enhanced proteolysis, although a theoretical role for altered ER-α -specific translation attenuation cannot be entirely ruled out (Fig.
3d, Additional file
3D). Hypoxia reduces ER-α levels in essentially all of the cells within the population (Fig.
1c). We also uniquely find that the mechanisms that enhance proteolysis of ER-α during hypoxia can be activated even in breast tumor derivatives that are naturally ER-α negative and very distantly related, for example the claudin low molecular phenotype represented by MDA-MB-231 (Fig.
3b). Thus, we propose that the molecular mechanisms that promote ER-α proteolysis are likely to be fundamental to the hypoxic response, and that ER-α may be one of a potentially large suite of proteins that undergo proteolytic regulation during hypoxic adaptation. In support of this hypothesis, others have shown hypoxia-induced proteolysis of the α-secretases ADAM10 and TACE without alteration in mRNA levels in neuroblastoma [
45], and of the MYC oncogene in human colon carcinoma cells and primary human keratinocytes, involving the ubiquitin ligases FBXW7 and DDB1, and cathepsins D and S [
46]. Similarly, the ubiquitin ligase Siah2 has been implicated in hypoxic proteolysis of the HIF-1α prolyl hydroxylases PHD1 and PHD3 [
47] and of the E1 subunit of α-ketogluterate dehydrogenase complex [
48].
There is significant controversy regarding hypoxia-induced alterations in
ESR1 mRNA levels. We find that none of the 10 cell lines in our panel show an obvious reduction in
ESR1 mRNA abundance after 24 h of hypoxic culture, although they exhibit significant, rapid reduction in the level of ER-α protein (Fig.
1a versus Fig.
3a). This concurs with the findings in MCF7 [
28] and ZR75 [
31] which assessed mRNA at a similar time point (24 h). In other reports, a reduction in the level of
ESR1 mRNA involving ERK kinase activity [
29] were determined after 72 h of hypoxic culture, which may explain their discrepant results. Finally, reports of early (8–24 h), HIF-1α-dependent decreases in the levels of
ESR1 mRNA [
30] in MCF7 and T47D, contrast directly with our findings. These studies differed in that the level of
ESR1 mRNA was analyzed in estrogen-starved cells in the presence versus absence of hypoxia. We predict that our results are most likely to represent the acute, common behavior of
bona fide breast tumors undergoing hypoxic adaptation in an estrogenic environment such as the breast or metastatic site in the female body.
Using shRNAs targeting
HIF1A and
HIF2A we directly demonstrate that HIF-1α, but not HIF-2α commonly decreases ER-α protein levels (Fig.
2c versus Additional file
7B) in hypoxic culture conditions. However, these shRNA studies only decreased the levels of HIF-1α mRNA and protein without entirely eliminating expression, which may explain why ER-α protein levels were not more robustly restored in hypoxia in these cell lines (Fig.
2c). Similarly, transient overexpression of a stable HIF-1α in normoxic conditions did not entirely eliminate ER-α expression (Fig.
2d), which also might be explained as a technical failure to transfect all cells in the population. Alternatively, other hypoxia-induced factors that function independently of HIF-1α expression may also influence ER-α stability. For example, the weak HIF-2α expression seen under normoxic conditions might in fact be stabilizing ER-α in an unknown fashion, and the reduction in both molecules might be somehow linked in hypoxia. Proteases activated by hypoxia in other systems including cathepsins B, D, and S [
46,
49] and calpains [
50‐
52] as well as inducible E3 ubiquitin ligases such as SIAH2 [
48,
53] may also provide HIF-1α - independent ER-α degradation in hypoxia. However it is clear from our studies that HIF-1α plays an important role in ER-α regulation.
Since mRNA levels of
ESR1 do not decrease in response to hypoxia in any of the cell lines we examined (Fig.
3a), the decreased protein levels cannot result from direct transcriptional repression by HIF heterodimers or of HIF-induced transcriptional repressors. Thus HIF activity must indirectly influence factors that enhance proteolysis of ER-α. This may simply involve transcriptional induction of ubiquitin ligases or proteases, or more indirectly by induction of a known HIF-dependent transcriptional target such as miRNA-155 [
54,
55] or a methyltransferase such as WDR5 [
56]. Alternatively, HIF-1α and ER-α might compete for transcriptional co-activators, for example P300/CBP (reviewed in [
2]). In this scenario, increased hypoxic levels of HIF-1α could result in ER-α degradation, due to the lack of co-activators that prevent proteasome targeting. Finally, ER-α may compete with HIF-1α for VHL-mediated ubiquitination and proteasome targeting, as reported in renal cell carcinomas [
57]. Others have implicated HIF-1α in ER-α regulation by overexpression of chimeric HIF-1α-VP16 [
44] and by HIF-1α silencing by siRNA in a single cell line, MCF7 [
30]. However, our results demonstrate that this activity is more common among breast tumors than the results in MCF7 would suggest, and reveal that HIF-1α inhibitors should be further explored in preclinical studies as co-therapeutics in the endocrine therapy setting.
Finally, we find that the hypoxia-induced reduction in ER-α levels decreases ER-α -directed transcription (Fig.
4a-d) and significantly reduces proliferation (Fig.
5c-f) without inducing overt cell death in each cell line tested. Unexpectedly, this lack of estrogen signaling is not overcome by switching to other proliferative signaling pathways in any of our cells over the time courses we assessed (Fig.
5, 6–16 days). This leads us to speculate that chronic residence in hypoxic environments may be one explanation for the later disease reoccurrence seen in some ER-α positive breast cancer patients. We hypothesize that a dormant, non-proliferating phenotype may allow tumor cells to persist undetected, until conditions change to oxygenate the environment or until cells eventually switch to proliferation via other signaling pathways. We note that hypoxia can induce protein phosphatase 2A, which preserves viability without proliferation in glioblastoma multiforme-derived cells [
58] and we speculate that a similar mechanism may be at play in our system. Furthers studies involving chronic hypoxia are ongoing in our laboratory. The hypoxia sensitivity of ER-α positive, but not ER-α negative, basal-like cell lines reported here may also provide one explanation for clinical observations that ER-α positive tumors grow more slowly than ER-α negative tumors, providing patients with better 5 year survival statistics. We speculate that ER-α negative tumors will grow in vivo regardless of environmental oxygen levels, whereas growth of ER-α positive tumors will be limited to that proportion of the tumor that is oxygenated. Finally, accounting for differential expression of ER-α and other hypoxia sensitive molecules in hypoxic regions of tumors offers one explanation for the intra-tumor heterogeneity of ER-α expression seen in many clinical breast tumor sections.
Several publications have shown that ER-α levels are reduced in regions of potentially hypoxic tissue in human clinical specimens, although in each case the authors were unable to determine whether this resulted from a lack of nutrients, oxygen, or an increase in cellular waste products. For example, Cooper et al. [
28] examined cells adjacent to necrotic breast tumor cores by immuno-histochemical methods, and found that ER-α protein levels were decreased, HIF-1α levels increased, and expression of the HIF-1α target gene CA-IX was increased. Similarly, Kronblad et al. [
29] also demonstrated reduced ER-α and increased HIF-1α halos of tumor cells surrounding necrotic cores. Finally Lloyd, et al. [
59] measured the spatial distribution of ER-α reactivity in relationship to vascularity in breast tissue sections by immuno-histochemical staining. Each study found reduced ER-α levels either adjacent to necrotic cores or distal from vasculature. Unlike clinical results, we can definitely say that hypoxia alone is enough to produce reduced ERα protein levels in ER-α positive tumor cells despite variation in other accompanying genetic mutations.
Acknowledgements
The authors thank Dr. Joe Gray for the panel of breast cancer cell lines and Dr David Shapiro (University of Illinois) for the ER-α and ER-luciferase reporter plasmids. We also thank Ricardo A. Mejia and Gloria E. Reynolds for technical assistance and Drs. Amato J. Giaccia, John P Murnane, Ester M. Hammond, Christine Janson, Lara Cobler, and Trent A. Watkins for helpful discussions.