Abstract
Lung cancer is the leading cause of cancer-related death worldwide, in large part due to its high propensity to metastasize and to develop therapy resistance. Adaptive responses to hypoxia and epithelial–mesenchymal transition (EMT) are linked to tumor metastasis and drug resistance, but little is known about how oxygen sensing and EMT intersect to control these hallmarks of cancer. Here, we show that the oxygen sensor PHD3 links hypoxic signaling and EMT regulation in the lung tumor microenvironment. PHD3 was repressed by signals that induce EMT and acted as a negative regulator of EMT, metastasis, and therapeutic resistance. PHD3 depletion in tumors, which can be caused by the EMT inducer TGFβ or by promoter methylation, enhanced EMT and spontaneous metastasis via HIF-dependent upregulation of the EGFR ligand TGFα. In turn, TGFα stimulated EGFR, which potentiated SMAD signaling, reinforcing EMT and metastasis. In clinical specimens of lung cancer, reduced PHD3 expression was linked to poor prognosis and to therapeutic resistance against EGFR inhibitors such as erlotinib. Reexpression of PHD3 in lung cancer cells suppressed EMT and metastasis and restored sensitivity to erlotinib. Taken together, our results establish a key function for PHD3 in metastasis and drug resistance and suggest opportunities to improve patient treatment by interfering with the feedforward signaling mechanisms activated by PHD3 silencing.
Significance: This study links the oxygen sensor PHD3 to metastasis and drug resistance in cancer, with implications for therapeutic improvement by targeting this system. Cancer Res; 78(7); 1805–19. ©2018 AACR.
Introduction
Metastatic dissemination and the development of therapy resistance are the two leading causes for the limited success of current cancer treatment. Tumor metastasis has been linked to the activation of an epithelial–mesenchymal transition (EMT)-like program, which involves dissociation of cell–cell contacts, enhanced migration and dissemination to distant sites (1). Multiple aspects of tumor malignancy, including metastasis and drug resistance, are commonly associated with tumor hypoxia, which acts primarily through the prolyl hydroxylase domain (PHD) proteins and the hypoxia-inducible factors (HIF; ref. 2). The role of PHDs in tumor growth is incompletely understood. While some studies have demonstrated that PHD deficiency can reduce tumor growth, metastasis, or therapy resistance (3–6), other reports highlight the role of PHDs as tumor suppressors (7–13).
Lung cancer, the leading cause of cancer-related death worldwide, is characterized by a high metastatic capacity and the rapid acquisition of resistance against existing treatments. The majority of lung cancers show an overactivation of EGFR signaling, which has been correlated with poor prognosis (14). Drugs targeting EGFR such as erlotinib and gefitinib have been approved for advanced non–small cell lung cancer, but acquired resistance usually develops within months of starting the treatment (15). In some cases, this resistance can be attributed to additional genetic mutations in EGFR, but in many patients alternative mechanisms are involved, which are not fully understood.
Here we show that the oxygen sensor PHD3 regulates the induction of EMT and the development of resistance against EGFR tyrosine kinase inhibitors in lung cancer. PHD3 is downregulated during EMT and its depletion potentiates EMT induction, as well as spontaneous metastasis. These effects are mediated by HIF-induced TGFα upregulation, which enhances SMAD signaling through activation of EGFR. Importantly, downregulation of PHD3 mediates the development of resistance against the EGFR inhibitor erlotinib. These findings highlight a novel mechanism through which PHD3 acts in a tumor-suppressive manner in lung cancer to promote metastasis and drug resistance.
Materials and Methods
Cell culture
Human lung carcinoma (A549, H441, H1299, HCC827, and HCC4006), breast carcinoma (MDA-MB-231), and colorectal carcinoma (HCT116) cell lines were purchased from ATCC. The NCH-604A cells, derived from a lung tumor metastasis to the brain were kindly provided by C. Herold-Mende (Department of Neurosurgery, University of Heidelberg, Germany). The G55 glioma cells were kindly provided by M. Westphal and K. Lamszus (University Medical Center Hamburg-Eppendorf, Hamburg, Germany). The cells were cultured in DMEM supplemented with 10% FBS, 100 U/mL penicillin, and 10 μg/mL streptomycin. For EMT induction by TGFα or TGFβ, cells were serum starved in DMEM without FBS overnight and incubated with 20–100 ng/mL TGFα or 0.5–10 ng/mL TGFβ1 in DMEM + 1% FBS for up to 3 days. For pharmacologic hydroxylase inhibition, cells were treated with 1 mmol/L DMOG (Merck Millipore) for 3 days. Erlotinib-resistant sublines were established by culturing parental cells with stepwise escalating concentrations of erlotinib from 2.5 μmol/L to 40 μmol/L over a period of 90 days. Blockade of TGFα in the cell culture supernatant was achieved using 10 μg/mL goat anti-TGFα antibody (AF-239-NA, R&D Systems); normal goat IgG (AB-108-C, R&D Systems) was used as control. Cells that showed unusual growth or morphology were tested for Mycoplasma contamination and only noncontaminated cells were used for experiments. Cell lines obtained from ATCC were authenticated by the manufacturer. No additional authentication was performed by the authors for any of the cell lines. The cell lines were used between passages 2 and 10.
Lentiviral constructs and stable cell lines
Lentivirus was packaged by cotransfection of lentiviral vectors with the packaging plasmids pCI-VSVG and psPAX2 (Addgene plasmids 1733 and 12260) into 15-cm plates with HEK293T cells using Fugene HD (Roche). Medium was changed every 24 hours, and the 48- and 72-hour supernatants were pooled, filtered through a 0.45-μm filter, and ultra-centrifuged at 32,000 rpm, 4°C for 1 hour. Titers were determined by counting the number of GFP-positive colonies. The shPHD3, PHD3-V5 overexpressing, and corresponding control lines were generated by lentiviral transduction and selection with 4–10 μg/mL blasticidin. For the shRNA knockdown of HIF-1α, HIF-2α, and TGFα, shRNA clones were purchased from Open Biosystems (Thermo Scientific; see the Supplementary Materials and Methods for further details). Cells were subsequently transduced and selected with 4 μg/mL puromycin to create the stable shRNA pools.
Luciferase reporter assay
Cells were transiently transfected with constructs expressing firefly luciferase under a TGFα promoter (GeneCopoeia) or an SBE (SMAD-binding element) promoter (Addgene plasmid 16495; ref. 16) together with an SV40-Renilla luciferase construct (Promega) for normalization of transfection efficiency. Cells were cultured for 48 hours under normoxia or for 48–72 hours under hypoxia (for assays with the TGFα promoter–based reporter) and assayed for luciferase activity with the Dual-Luciferase Reporter-Assay System (Promega).
Real-time RT-PCR
Cell lines were cultured for 2 days before total RNA isolation using PureLink RNA Mini Kit (Life Technologies). The RT reaction was performed using RevertAid H Minus M-MuLV Reverse Transcriptase (Fermentas) at 42°C for 60 minutes, followed by inactivation of the enzyme by heating at 70°C for 10 minutes. The cDNAs were amplified using the primers listed in the Supplementary Materials and Methods, Absolute SYBR Green ROX mix (Thermo Scientific) and a thermal cycler StepOnePlus (Applied Biosystems); the reaction conditions were initial denaturation at 95°C for 15 minutes, 40 cycles of denaturation for 1 minute at 95°C, and annealing/extension for 1 minute at 60°C. The difference in the threshold number of cycles between the gene of interest and the housekeeping gene HPRT (hypoxanthine phosphoribosyltransferase 1) was then normalized relative to the standard chosen for each experiment and converted into fold difference.
Immunoblotting
Cells were harvested in lysis buffer [50 mmol/L Tris.HCl buffer (pH 7.5), 2% SDS, 2 mmol/L EGTA, and 20 mmol/L NaF] and protein concentrations were determined using the Bio-Rad DC Protein Assay. Homogenate supernatants (10–40 μg total protein per lane) were subjected to SDS-PAGE under reducing conditions, and the resolved proteins were transferred from the gels onto polyvinylidene difluoride membranes of 0.2-μm pore size. The membranes were incubated with anti-human PHD3 (1:10,000, Novus Biologicals, NB-100-303), E-cadherin (1:5,000, BD Biosciences, BD610181), N-cadherin (1:1,000, BD Biosciences, BD610920), vimentin (1:1,000, Dako M0725), HIF-1α (1:1,000, Cayman Chemical, 10006421), HIF-2α (1:1,000, Novus Biologicals, NB 100-122), Phospho-Smad2 (Ser465/467; 1:1,000, Cell Signaling Technology, 138D4), Smad2/3 (1:1,000, Cell Signaling Technology, 3102), phospho-EGFR (Tyr1068; 1:1,000, Cell Signaling Technology, 3777), EGFR (1:1,000, Cell Signaling Technology, 4267), phospho-AKT (Ser473; 1:1,000, Cell Signaling Technology, 9271), pan-AKT (1:1,000, Cell Signaling Technology, 9272), Phospho-ERK1/2 (1:1,000, Santa Cruz Biotechnology, sc-7383), ERK2 (1:5,000, Santa Cruz Biotechnology, sc-154), and tubulin (1:5,000, Dianova DLN09992). Antibody binding was detected with the use of horseradish peroxidase-conjugated secondary antibodies (1:2,500, Dianova). Immunoreactive bands were detected with ECL Western blotting reagents (Thermo Scientific).
Immunofluorescence
For immunostaining of cultured cells, 200,000 cells were seeded on poly-d-lysine/laminin-coated 8-well CultureSlides (BD Biosciences) and cultivated in medium with 1% FBS and 1 ng/mL TGFβ for 3 days. For immunostaining of tumor tissue, xenograft tumors were frozen in TissueTek OCT (Sakura) and sectioned into 5-μm-thick slices using a cryostat. Fixation was performed in methanol for 30 minutes at −20°C followed by rinsing in PBS (2 × 2 minutes). Blocking solution was applied (2% FCS, 2% BSA, 0.2% fish gelatin in PBS for cultured cells; M.O.M mouse-on-mouse, Vector Laboratories for tumor cryosections) for 30 minutes followed by incubation with anti-E-cadherin antibody (1:200 in blocking solution, BD 610181) and anti-vimentin (1:10,000 in blocking solution, Dako M0725) overnight at 4°C. Slides were washed three times with PBS and incubated with secondary anti-mouse Alexa Fluor 568 (diluted 1:200 in blocking solution) for 1 hour at room temperature. Cells were washed three times with PBS, incubated with a solution of DAPI (1:5,000), and TO-PRO (1:1,000) in water for 10 minutes to stain the nuclei, and mounted using Vectashield mounting medium (Vector Laboratories).
ELISA
A549 and H441 (50,000 cells/well, 24-well plates) were cultured for 48 hours in DMEM without growth factors. The amount of TGFα protein in the supernatant was determined by using TGFα ELISA kits according to the manufacturer's instructions (R&D Systems). The values were subtracted for background (DMEM without FBS) and the results were expressed as pg TGFα per mL growth medium.
Invasion assay
Matrigel basement membrane matrix (BD Biosciences, concentration approx. 10 mg/mL) was mixed 1:10 with DMEM containing 100,000 cells and was allowed to polymerize in Transwell inserts with a pore size of 8.0-μm (Corning) for at least 1 hour at 37°C. The inserts were then placed in wells containing medium supplemented with 10% FBS and 0.25 ng/mL TGFβ1; serum-free medium was added on top of the Matrigel to generate a chemotactic gradient. Eighteen hours after seeding, invading cells were fixed with 3.7% PFA and stained with DAPI for 10 minutes. The inserts were first imaged by fluorescence microscopy to quantify the total number of cells, the cells that did not cross the filter were then removed by wiping the upper surface of the filter with a tissue and the remaining cells were reimaged. The invasion index was calculated as the number of invading cells divided by the total number of cells, quantified using ImageJ. At least three independent experiments in triplicate were performed for each sample.
Quantification of cell viability following erlotinib treatment
Cells were incubated for two weeks with different erlotinib concentrations, as indicated in the figures. The supernatants were discarded and the remaining viable adherent cells were stained with 0.05% crystal violet in 2% methanol for 1 minutes. The 6-well plate was then rinsed with water and 2 mL of 10% acetic acid was added to each well to solubilize the stained cells. The absorbance of each well as a measure of cell density was read at 590 nm with a micro plate spectrophotometer. For quantification of cell number, the cells were trypsinized for 10 minutes and collected in DMEM + 10% FBS. Subsequently, living cells were counted with a CASY Cell Counter (Model TT, Roche).
Colony formation assay
Five-hundred cells were seeded in 6-well plates and incubated for two weeks in the presence of the indicated amount of erlotinib. The supernatants were discarded and the remaining viable adherent cells were stained with 0.05% crystal violet in 2% methanol for 15 minutes. The 6-well plate was then rinsed with water and the macroscopically observable colonies were counted at erlotinib concentrations at which well-resolved single colonies were detectable (20 μmol/L for A549, 40 μmol/L for H1299, 5 μmol/L for HCC4006, and 1 nmol/L for HCC827 cells).
MTT assay
Five-thousand cells were seeded in 96-well plates. After cell attachment, erlotinib was added as indicated in a volume of 200 μL. Subsequently, the medium was exchanged with 200 μL of fresh medium, 50 μL of a 5 mg/mL solution of MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide, Sigma-Aldrich) in PBS was added, the plates were wrapped in aluminium foil and incubated for 4 hours in a humidified incubator at 37°C. Afterwards, the MTT medium was removed from the wells and the MTT-formazan crystals were solubilized by addition of 200 μL DMSO, followed by addition of 25 μL/well of glycine buffer (0.1 mol/L glycine, 0.1 mol/L NaCl adjusted to, pH 10.5, with 1 mol/L NaOH) and measurement of absorbance at 540 nm. Wells without cells but with medium and erlotinib were used as a blank.
Chromatin immunoprecipitation
Two 15-cm dishes of confluent A549 cells (corresponding to 2.5–5.0 × 107 cells), grown under hypoxia (1% O2) for 3 days, were used for each condition. Proteins bound to DNA were cross-linked with 1% formaldehyde for 10 minutes at room temperature, after which 0.1 mol/L glycine was added for 5 minutes to stop the cross-linking. Cells were collected and centrifuged at 1,500 × g (5 minutes, 4°C), washed in cold PBS containing A-PMSF, and centrifuged again at 1,500 × g (5 minutes, 4°C). Cells were lysed for 10 minutes on ice in 1 mL chromatin immunoprecipitation (ChIP) lysis buffer (1% SDS, 10 mmol/L EDTA, 50 mmol/L Tris pH 8.1, 100 μmol/L A-PMSF, Roche protease inhibitor mix). DNA was sheared by sonication (30-second on/30-second off, 28 cycles; Bioruptor, Diagenode) and the lysate was centrifuged at 16,100 × g at 4°C for 10 minutes. Supernatants were collected and stored at −80°C. For determination of DNA concentration, 20 μL of sheared lysate was diluted with 100 μL TE buffer containing 10 μg RNase A. After 30 minutes at 37°C, 3.8 μL of proteinase K (20 mg/mL) and 7.5 μL of 10% SDS was added and incubated for 2 hours at 37°C, followed by incubation at 65°C overnight. Samples were purified using NucleoSpin columns (Macherey-Nagel), according to the manufacturer's instructions. DNA was eluted with 50 μL 5 mmol/L Tris pH 8.5 and concentration was determined by NanoDrop. For ChIP, lysate volumes corresponding to 25 μg DNA were diluted in 900 μL of dilution buffer (0.01% SDS, 1.1% Triton X-100, 1.2 mmol/L EDTA, 167 mmol/L NaCl, 16.7 mmol/L Tris/HCl pH 8.1) and precleared by normal rabbit IgG antibody (Cell Signaling Technology, #2729) and protein A/G Sepharose mixture. Subsequently, 1–3 μg HIF-1α (Novus Biologicals, NB-100-134) and HIF-2α (Novus Biologicals, NB-100-122) antibodies were added and the samples were rotated at 4°C overnight. Thirty microliters of a protein A/G sepharose mixture, preequilibrated in dilution buffer, was added to the lysates and incubation continued for at least 2 hours at 4°C. Beads were collected by centrifugation, washed once in low salt buffer (0.1% SDS, 1% Triton X-100, 2 mmol/L EDTA, 20 mmol/L Tris pH 8.1, 150 mmol/L NaCl), once in high salt buffer (0.1% SDS, 1% Triton X-100, 2 mmol/L EDTA, 20 mmol/L Tris pH 8.1, 500 mmol/L NaCl), once in LiCl buffer (0.25 mol/L LiCl, 1% NP40, 1% deoxycholate, 1 mmol/L EDTA, 10 mmol/L Tris pH 8.1), and twice in TE buffer (10 mmol/L Tris pH 8.1, 1 mmol/L EDTA) for 5 minutes on a rotator at 4°C. Beads were finally resuspended in 100 μL TE buffer containing 10 μg RNase A. In parallel, 1/10 volume (2.5 μg) of the initial lysate (10% input samples) was diluted with 100 μL TE buffer containing 10 μg RNase A. After 30 minutes at 37°C, 3.8 μL of proteinase K (20 mg/mL) and 7.5 μL of 10% SDS were added and both input and immunoprecipitates were incubated for 2 hours at 37°C followed by incubation at 65°C overnight. Samples were purified using NucleoSpin columns (Macherey-Nagel) according to the manufacturer's instructions. DNA was eluted with 50 μL 5 mmol/L Tris pH 8.5 for quantification by real-time PCR. The reaction mixture contained 2 μL of ChIP DNA or 1% input DNA, primers at 0.25 μmol/L and 10 μL of Fast SYBR Green Master Mix (2×, Applied Biosystems) in a total volume of 20 μL. The PCR program was as follows: 95°C (20 seconds), 40× (95°C (3 s), 60°C (30 s)). The ChIP-PCR primers for the HRE in the promoter of the human TGFA gene were: HRE for, ACTCACAGGTCCCTTTCCTG; HRE rev, CGCCTGACTTCAGACACCAC. Melting curve analysis revealed a single PCR product. The amount of precipitated chromatin was calculated as fold enrichment over IgG according to the equation: fold enrichment = 2−(CtIP-CtIgG).
In vivo tumor transplantation
Animal experiments were approved by the veterinary department of the regional council in Darmstadt Germany (#V54-19c20/15-F42/17). To assess spontaneous metastasis, A549 cells transduced with shSIMA/shScrambled, shSIMA/shTGFα, shPHD3/shScrambled, and shPHD3/shTGFα were transplanted subcutaneously in the dorsal right flanks of 6- to 8-week-old NMRI female mice (8 × 106 cells per mouse; n = 10 mice per group). Spontaneous lung metastases were scored macroscopically on isolated lungs after 36 days.
Experimental lung metastasis
A549 cells and H1299 cells (4 × 106 cells, in two independent injections of 2 × 106 in 200-μL PBS) were injected in the tail vein of 6- to 8-week-old female NMRI nude (A549) or CD1 nude (H1299) mice. Six weeks (A549) or 8 weeks (H1299) later, the mice were sacrificed and macroscopic metastases were scored on isolated lungs.
Bioinformatic analysis
Datasets of The Cancer Genome Atlas Research Network (TCGA; ref. 17) were retrieved from the cBio portal (http://www.cbioportal.org/public-portal/index.do; ref. 18). PHD3 promoter methylation (Illumina Infinium Human DNA Methylation 450 array) and expression (RNA-seq z-scores) data were available for a total of 358 lung squamous cell carcinoma samples. For PHD3 promoter methylation analysis tumors with a beta-value > 0.1 were considered as methylated, tumors with a beta-value ≤ 0.1 were considered as unmethylated. Overall survival and PHD3 expression data (U133 microarray z-scores) were available for 132 lung squamous cell carcinoma samples.
Gene expression and clinical data for non–small cell lung carcinoma patients (n = 25) treated with erlotinib (19) were retrieved from the Gene Expression Omnibus. Gene expression data for erlotinib-sensitive and resistant non–small cell lung carcinoma cell lines (n = 42; ref. 20) were retrieved from Oncomine. Gene expression and drug resistance data for lung carcinoma cell lines (26 lines with expression and resistance data; ref. 21) were retrieved from the Cancer Cell Line Encyclopedia (http://www.broadinstitute.org/ccle/home); the cutoff between erlotinib-sensitive and resistant cells was set at EC50 of 4 μmol/L erlotinib.
Statistical analysis
Results are presented as mean + or ± SEM. Animals that unexpectedly died before tumors were collected from the remaining animals were excluded from the analysis. Statistical analysis was performed using the two-tailed Student t test or the log-rank test for survival analysis. Statistical significance was defined as P < 0.05 (*, P < 0.05; **, P < 0.01; ***, P < 0.001).
Additional experimental procedures are described in the Supplementary Data.
Results
PHD3 is silenced during EMT and its depletion promotes a mesenchymal phenotype in lung tumors
EMT is induced by microenvironmental factors such as TGFβ (22) and hypoxia (23–26). Previous work has suggested a cross talk between these key stimuli (27–29). To investigate the mechanisms through which EMT and oxygen sensing interact in lung cancer, we assessed the effect of TGFβ-induced EMT on the main oxygen sensors PHD1–3. TGFβ induced a pronounced EMT in lung adenocarcinoma cells (Fig. 1A and B). Strikingly, there was a profound reduction in the level of PHD3 protein following TGFβ treatment, which was mirrored by a downregulation of the PHD3 transcript in the lung carcinoma cell lines A549, H441, and NCH-604A (Fig. 1B and C; Supplementary Fig. S1A and S1B), whereas the expression of PHD1 and PHD2 was more variable or did not change. As PHD-dependent hydroxylation initiates the targeting of HIF proteins for degradation, we next examined whether TGFβ affects the abundance of HIFs. In line with the reduced PHD3 expression, we saw an upregulation, particularly of HIF-2α protein levels, following TGFβ treatment (Fig. 1D). We next assessed whether PHD3 plays a functional role in EMT. Indeed, increased PHD3 levels in A549 cells abrogated the induction of EMT and partially attenuated the downregulation of E-cadherin and the increase in N-cadherin and vimentin following TGFβ treatment (Fig. 1E). Conversely, silencing of PHD3 promoted the induction of EMT. PHD3 loss was associated with a partial induction of a mesenchymal phenotype already in the absence of exogenous EMT stimuli and it potentiated the induction of EMT by TGFβ treatment (Fig. 1F and G; Supplementary Fig. S1C–S1G). Correspondingly, treatment with the prolyl hydroxylase inhibitor dimethyloxalylglycine (DMOG) enhanced EMT induction (Supplementary Fig. S1H), similar to PHD3 silencing. Silencing of PHD3 significantly promoted cell invasion, the principal functional effect of EMT (Fig. 1H). Taken together, these data establish PHD3 as a critical regulator of EMT and show that PHD3 loss promotes EMT in tumor cells.
PHD3 silencing promotes EMT through upregulation of TGFα
Stimulation of TGFβ receptors by cognate ligands induces the phosphorylation of receptor bound SMAD proteins and the association of SMADs with cytoplasmic common mediator SMAD (co-SMAD) to activate the transcription of multiple genes linked to EMT (30). An additional EMT-regulatory mechanism involves the induction of EGFR signaling, which can promote EMT, for example, via downstream activation of Ras/ERK or PI3K/AKT signaling (31). Therefore, we next examined whether silencing of PHD3 could lead to the deregulation of these two central EMT inducing pathways, indicating a molecular crosstalk between PHD3 and SMAD/EGFR signaling. We detected a marked enhancement of SMAD activation in PHD3 knockdown A549 cells (Fig. 2A; Supplementary Fig. S2A). In addition, PHD3-depleted cells showed an increase in EGFR phosphorylation, alongside activation of the downstream effectors ERK and AKT (Fig. 2B). Importantly, the potentiation of SMAD signaling following PHD3 knockdown was dependent on EGFR activity, as it was abrogated by treatment with the specific EGFR tyrosine kinase inhibitor erlotinib (Fig. 2C; Supplementary Fig. S2A), suggesting a complex PHD3-dependent interplay between the EGFR and TGFβ pathways. To examine how PHD3 loss affects EGFR activity, we assessed whether the silencing of PHD3 could alter the levels of EGFR ligands. Strikingly, PHD3 silencing led to a marked upregulation of TGFα, a ligand specific for EGFR (Supplementary Fig. S2B). In contrast, the expression of other key EGFR ligands including EGF, heparin-binding EGF-like growth factor (HB-EGF), or amphiregulin were only modestly increased or variable between cell lines (Supplementary Fig. S2B). Similarly, both TGFβ and DMOG treatment, which lead to PHD3 downregulation or inhibition, upregulated TGFα expression (Supplementary Fig. S2C). Interestingly, while strong upregulation of TGFα upon PHD3 silencing was seen in all lung cancer cell lines examined, this was not observed in other types of cancer, including glioblastoma (G55), mammary carcinoma (MDA-MB-231), and colorectal carcinoma (HCT 116; Supplementary Fig. S2D and S2E; ref. 11). We next determined whether TGFα treatment could mimic the effect of PHD3 silencing. Indeed, TGFα not only activated EGFR phosphorylation and signaling, but also induced pronounced EMT in A549 cells (Fig. 2D–F). To directly assess whether PHD3 disruption induces EMT through increased TGFα levels, we silenced TGFα in PHD3-silenced A549 cells (Supplementary Fig. S3A). Importantly, knock down of TGFα with two separate shRNA constructs partially suppressed the EMT-promoting effect of PHD3 depletion (Fig. 2G), similar to pharmacologic inhibition of EGFR signaling (Supplementary Fig. S3B). Furthermore, inhibition of EGFR with erlotinib suppressed the invasive capacity of A549 cells; this effect could be reverted by silencing of PHD3 at intermediate erlotinib concentrations, whereas at higher concentrations, PHD3 depletion could not outcompete erlotinib (Supplementary Fig. S3C). Consistently, the activation of EGFR and the increase in SMAD activity elicited by PHD3 loss were reversed following TGFα silencing (Fig. 2G and H). Furthermore, the expression of the SMAD target genes SMAD7, PAI-1, and TGFβ (32) was increased following PHD3 silencing, and this effect was reversed upon cosilencing of TGFα (Fig. 2I). Similarly, blocking of secreted TGFα with an anti-TGFα antibody, but not treatment with control immunoglobulin, suppressed the increase in SMAD activity induced by depletion of PHD3 (Fig. 2J). Collectively, these experiments demonstrate that PHD3 loss promotes EMT in lung cancer through the elevated production of TGFα, which activates EGFR and augments SMAD signaling.
The EMT-promoting effect of PHD3 loss is HIF-dependent
We next addressed the mechanism by which PHD3 disruption leads to TGFα upregulation and EMT induction. Our experiments confirmed that loss of PHD3 increases the levels of HIFs, especially HIF-2α in lung cancer (Supplementary Fig. S1C–S1E). Notably, TGFα has been described as a HIF target gene (25). Indeed, hypoxic treatment markedly increased TGFα mRNA levels and TGFα promoter activity in various lung cancer cell lines (Supplementary Fig. S4A and S4B). Furthermore, ChIP confirmed that HIF-1α and HIF-2α bind to a hypoxia response element in the promoter of the TGFA gene (Fig. 3A; Supplementary Fig. S4C). Interestingly, while HIF-1α bound similarly in control and shPHD3 cells, HIF-2α only showed binding in PHD3-silenced cells (Fig. 3A), in line with the more potent upregulation of HIF-2α following PHD3 depletion (Supplementary Fig. S1C–S1E). Importantly, silencing of HIF-1α and HIF-2α (Supplementary Fig. S4D–S4F) blocked the increase in TGFα elicited by downregulation of PHD3 (Fig. 3B). In line with these findings, the activation of both EGFR and SMAD signaling following silencing of PHD3 was reduced by the knockdown of the HIF subunits and most potently by HIF-2α disruption (Fig. 3C and D). Importantly, the silencing of HIF-1/2α also abrogated the induction of EMT elicited by PHD3 depletion (Fig. 3E). To corroborate our findings that PHD3 regulates EMT through the HIF-dependent control of TGFα, requiring PHD3 enzymatic activity (Supplementary Figs. S1H and S2C), we used wild-type PHD3 or the PHD3-H196A mutant, which lacks hydroxylase activity (12). Reexpression of wild-type PHD3 in PHD3-deficient cells restored SMAD activity to control levels (Fig. 3F). In contrast, the enzymatically inactive PHD3 mutant did not reduce SMAD activity, indicating that the regulation of EMT by PHD3 requires its hydroxylase activity. Consistently, only wild-type PHD3, but not the PHD3-H196A mutant, reverted the induction of EMT caused by PHD3 silencing (Fig. 3G). To further corroborate the function of PHD3 and its hydroxylase activity in the control of the EMT phenotype in lung cancer, we expressed wild-type and mutant PHD3 in the lung carcinoma cell line H1299, which has very low levels of endogenous PHD3 (Supplementary Fig. S4G) and displays a mesenchymal phenotype. Expression of wild-type, but not mutant PHD3, substantially reduced TGFα expression and EGFR activation (Fig. 3H; Supplementary Fig. S4H). Importantly, wild-type, but not the hydroxylase-deficient PHD3 mutant, suppressed the invasiveness of H1299 cells (Fig. 3I). Collectively, these series of experiments demonstrate that PHD3 controls EMT through its hydroxylase activity and that PHD3 loss drives EMT through the HIF-dependent upregulation of TGFα.
PHD3 loss promotes tumor metastasis through TGFα
The activation of EMT during carcinogenesis is linked to the acquisition of an invasive and metastatic tumor phenotype (1). We therefore next assessed whether the EMT phenotype induced by downregulation of PHD3 would be sufficient to drive tumor metastasis, using different in vivo models. First, in an experimental metastasis model of lung cancer, control and shPHD3 A549 cells were transplanted intravenously in nude mice. We observed an over 5-fold increase of metastatic nodules in the lungs of mice injected with cells deficient for PHD3 (Fig. 4A). In line with the crucial role of TGFα in the induction of the EMT phenotype following PHD3 downregulation, cosilencing of TGFα in the PHD3 knockdown background fully reversed the increased metastatic capacity of PHD3 (Fig. 4A). Importantly, these findings were corroborated in a lung cancer model of spontaneous metastasis. PHD3 silencing prominently increased the spontaneous metastatic capacity of A549 cell to the lung, whereas cosilencing of TGFα fully reverted this phenotype (Fig. 4B). Consistent with the increased metastatic capacity, the PHD3-silenced tumors exhibited a profound reduction of E-cadherin levels compared with controls, which was fully reverted by cosilencing of TGFα (Fig. 4C). To functionally corroborate the role of PHD3 as a suppressor of metastasis, we reintroduced PHD3 in the lung carcinoma cell line H1299, which has very low endogenous PHD3 levels. This significantly suppressed the metastatic capacity of the H1299 lung cancer cells and reduced the occurrence of metastatic nodules by more than 9-fold (Fig. 4D). These results show that PHD3 functions as a suppressor of tumor metastasis in lung cancer, and that its loss markedly promotes EMT and the dissemination to distant tissues through TGFα.
Reduced expression of PHD3 is associated with poor prognosis in lung cancer patients
To assess the clinical relevance of PHD3 silencing in lung cancer patients we first assessed whether PHD3 can be epigenetically silenced by promoter hypermethylation in lung tumors, as has been reported in other cancers (11, 33, 34). Analysis of data from the TCGA lung cancer cohort revealed that the PHD3 promoter was methylated in over one third of the patients, and lung cancers with increased PHD3 CpG methylation exhibited significantly lower PHD3 levels (Fig. 5A). We next compared PHD3 expression between lung carcinomas (35) of early stages (stage I and II) and later stages (stage III + IV). Importantly, PHD3 expression was markedly reduced at later stages of lung tumor progression (Fig. 5B), consistent with an involvement of PHD3 deficiency in metastatic spread. This was confirmed by survival analysis of lung cancer patients, demonstrating that lower PHD3 expression was associated with significantly worse prognosis (Fig. 5C). These findings indicate that PHD3 is frequently silenced by promoter methylation in lung cancer and that a reduction in PHD3 expression is associated with a worse clinical prognosis.
PHD3 silencing confers resistance to EGFR inhibitors
The induction of an EMT phenotype in tumor cells has been tightly linked to the development of drug resistance, including resistance against EGFR-targeting therapeutics (36–38). Furthermore, our data demonstrate that the induction of EMT following PHD3 loss is mediated through the TGFα-induced activation of EGFR signaling. Therefore, we assessed whether silencing of PHD3 is linked to the acquisition of resistance against EGFR inhibitors. Indeed, PHD3-silenced A549 cells displayed a dramatically increased resistance against high concentrations of the EGFR-specific inhibitor erlotinib, whereas the growth of controls cells was already inhibited at low erlotinib concentrations (Fig. 6A). Similarly, the ability of A549 tumor cells to form colonies in the presence of erlotinib was increased following disruption of PHD3 (Fig. 6B). Erlotinib has been used for the treatment of lung cancer patients both with wild-type EGFR and with activating EGFR mutations, but it is most effective in the latter group (39). Therefore, we next examined whether PHD3 also controls the erlotinib sensitivity of lung cancer cell lines carrying activating EGFR mutations, such as HCC827 (with EGFR E746-A750 deletion) and HCC4006 (with EGFR L747 - E749 deletion and A750P substitution; ref. 40). Indeed, we found that silencing of PHD3 in these cell lines also promoted their resistance to erlotinib (Supplementary Fig. S5A–S5D), indicating that PHD3 loss is a common mechanism that can lead to enhanced erlotinib resistance in both EGFR wild-type and EGFR-mutant lung cancer cells. Consistently, overexpression of wild-type PHD3, but not the enzymatically inactive H196A mutant reduced the ability of H1299 cells to form colonies in the presence of erlotinib (Fig. 6C). Importantly, depletion of TGFα reversed the increased cell survival of PHD3 silenced tumor cells in the presence of erlotinib (Fig. 6D). Conversely, addition of TGFα alone to control A549 cells mimicked the effect of PHD3 loss and mediated resistance to erlotinib (Fig. 6E). This can be explained by the fact that TGFα was able to partially overcome the inhibitory effect of erlotinib, inducing EGFR autophosphorylation and downstream signaling (Supplementary Fig. S6). These results show that increased TGFα levels underlie the enhanced resistance to therapeutic EGFR inhibition conferred by PHD3 loss and are in line with previous studies showing that EGFR ligands are capable of activating EGFR tyrosine kinase activity and signaling even in the presence of erlotinib (41) and that higher TGFα levels in erlotinib-treated patients with metastatic cancer correlate with worse prognosis (42).
As a next step, we addressed whether loss of PHD3 can represent an important event in the spontaneous acquisition of resistance against EGFR inhibitors. To this end, we selected A549 cells resistant to erlotinib by culturing the cells in the presence of escalating concentrations of the drug. Intriguingly, the erlotinib-resistant cells resulting from this selection had a markedly reduced expression of PHD3 (Fig. 7A; Supplementary Fig. S7A–S7C). Moreover, the erlotinib resistant A549 cells also exhibited pronounced EMT, along with increased TGFα levels (Fig. 7B). To assess whether the downregulation of PHD3 causally underlies the acquisition of spontaneous erlotinib resistance, we reexpressed PHD3 in the erlotinib-resistant A549 and H441 cells. Notably, PHD3 reexpression in erlotinib-resistant cells markedly increased their sensitivity to the drug (Fig. 7C and D) and reversed the EMT phenotype induced by the acquisition of erlotinib resistance (Fig. 7E). To test whether PHD3 loss is a common event in erlotinib-resistant lung cancer cells, we analyzed publically available datasets containing gene expression and drug resistance data for multiple lung cancer cell lines (20, 21). Indeed, erlotinib-resistant lung cancer cell lines exhibited a marked reduction of PHD3 compared with erlotinib-sensitive ones (Fig. 7F; Supplementary Fig. S7D). Furthermore, erlotinib-resistant and low PHD3–expressing cells had a pronounced shift towards a mesenchymal phenotype, as evidenced by a decrease in E-cadherin and an increase in vimentin levels (Fig. 7F; Supplementary Fig. S7E). Finally, we analyzed the therapeutic response as a function of PHD3 expression in a cohort of lung cancer patients that had received erlotinib treatment (19). Intriguingly, patients bearing lung tumors with high PHD3 levels showed a significantly better survival response (Fig. 7G), indicating that low PHD3 expression is linked with erlotinib resistance in lung cancer patients. Taken together, these findings demonstrate that PHD3 is commonly downregulated during the acquisition of resistance against EGFR inhibitors in lung cancer and that the loss of PHD3 functionally underlies and is sufficient to induce resistance to erlotinib.
Discussion
Understanding the molecular mechanisms regulating tumor metastasis and drug resistance is of central importance for the design of effective cancer therapies. Our work provides novel insights into these processes by uncovering an intricate link between the microenvironmental control of oxygen sensing and the regulation of EMT. We identify the oxygen sensor PHD3 as a key regulator of EMT induction, metastasis, and therapeutic resistance in lung cancer through the control of TGFα. Importantly, our clinical data demonstrate that low levels of PHD3 expression are linked to poor prognosis and resistance to anti-EGFR therapy, providing a mechanism through which lung tumor cells attain central cancer hallmarks by PHD3 inactivation.
Our study demonstrates an intricate connection between the hypoxic response and EMT at the level of the oxygen-sensing machinery. Importantly, we reveal that the downregulation of PHD3 is a central step in EMT. External EMT triggers such as TGFβ prominently decreased PHD3, and PHD3 silencing sensitized cells to EMT induction. A functional role for PHDs in the control of EMT and invasion has been suggested by previous reports (27, 29, 43). Notably, our data uncover the molecular mechanism behind the function of PHD3 in EMT, showing that HIF and TGFα mediate the induction of EMT following PHD3 suppression. From a tumor biology point of view, the coregulation of EMT and the HIF pathway through PHD3 may allow invading and metastasizing tumor cells to activate and use the prosurvival features of HIF, enabling them to rapidly adapt to the microenvironment encountered at metastatic sites. Mechanistically, we show that TGFβ-mediated PHD3 inactivation promotes EMT and metastasis through the cooption of TGFα/EGFR signaling in lung cancer cells. We identify PHD3 as a crucial molecular node mediating the bidirectional interplay between the TGFβR and EGFR pathways (Fig. 7H). Activation of TGFβR/SMAD signaling following TGFβ stimulation potently downregulates PHD3, which stabilizes HIF-1/2α and upregulates the transcription and secretion of the HIF target gene TGFα. TGFα, in turn, acts in an auto/paracrine manner to stimulate EGFR activity, which enhances SMAD signaling and EMT. Notably, this mechanism comprises a positive feedback loop that potentiates the TGFβ signal and may lower the threshold for efficient EMT also in neighboring cells through the paracrine function of TGFα. While our data indicate involvement of both HIF-1α and HIF-2α in mediating the effects of PHD3 loss, HIF-2α generally showed stronger and more robust activity in this context (Fig. 3A–E). This is consistent with our findings (Supplementary Fig. S1C–S1E) and previous reports (11, 44, 45), demonstrating that PHD3 predominantly regulates HIF-2α levels. It is interesting to note that the PHD3-mediated regulation of EMT through TGFα appears to be characteristic for lung cancer cells, but not for other tumor entities, as the TGFα upregulation following PHD3 silencing was not observed in glioblastoma, mammary, or colorectal carcinoma cells, although PHD3 was expressed in these cells and was efficiently silenced by transduction of shRNA constructs (Supplementary Fig. S2D; ref. 11). These findings are in line with previous studies showing that lung carcinomas often co-overexpress EGFR together with TGFα in an autocrine loop to sustain EGFR hyperactivation (46). Interestingly, we have previously shown that PHD3 is a central regulator of EGFR activity through the control of EGFR internalization in glioma (10, 11). This endocytic adaptor function of PHD3 in gliomas is independent of the PHD3 hydroxylase activity and HIFs, indicating that PHD3 may impact on the EGFR pathway at different levels and in a tumor entity–specific manner.
Previous work has demonstrated that resistance against EGFR tyrosine kinase inhibitors and blocking antibodies correlates with a shift from an epithelial to a mesenchymal state and that factors that induce EMT simultaneously promote the acquisition of resistance to EGFR inhibitors (19, 20, 47–49). In line with these findings, we show that depletion of PHD3 prominently coinduced EMT and resistance against erlotinib, whereas overexpression of PHD3-sensitized lung cancer cells to this drug. Importantly, we also found that cells rendered resistant through exposure to escalating concentrations of erlotinib had reduced PHD3 levels and could be resensitized to erlotinib through expression of PHD3. Our data indicate that silencing of PHD3 constitutes a novel important mechanism responsible for the acquisition of resistance to EGFR inhibitors, which complements previously established mechanisms of resistance development, including mutations in EGFR or activation of alternative receptor tyrosine kinases such as HER2, HER3, MET, IGF-1R, and AXL (19, 49, 50). The clinical relevance of these findings is underscored by the fact that lung cancer patients with a better clinical outcome following erlotinib treatment have higher PHD3 levels. This could point to novel strategies for developing biomarkers to predict therapy response to EGFR inhibitors or for overcoming resistance, for example, by targeting TGFα-dependent EGFR activation.
In summary, our results highlight a crucial role for PHD3 in the control of lung tumor EMT, metastasis, and drug resistance. We uncover a novel mechanism involving the TGFα-dependent activation of EGFR signaling, through which PHD3 coordinately regulates these key processes in tumor progression. This provides novel insights into lung tumor pathophysiology and could form the basis for improved patient prognosis and treatment.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Authors' Contributions
Conception and design: H. Dopeso, A. Acker-Palmer, B.K. Garvalov, T. Acker
Development of methodology: H. Dopeso, H.-K. Jiao, T. Acker
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Dopeso, H.-K. Jiao, A.M. Cuesta, A.-T. Henze, L. Jurida, M. Kracht, B.K. Garvalov, T. Acker
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Dopeso, H.-K. Jiao, L. Jurida, B.K. Garvalov, T. Acker
Writing, review, and/or revision of the manuscript: H. Dopeso, M. Kracht, A. Acker-Palmer, B.K. Garvalov, T. Acker
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Dopeso, H.-K. Jiao, B.K. Garvalov, T. Acker
Study supervision: B.K. Garvalov, T. Acker
Acknowledgments
We would like to thank Sabine Gräf for excellent technical assistance and Christel Herold-Mende, Department of Neurosurgery, University of Heidelberg for providing the NCH-604A cell line. This work was supported by the DFG KFO210 (to T. Acker and B.K. Garvalov), the Deutsche Krebshilfe (to T. Acker and B.K. Garvalov), the Behring-Röntgen Foundation (to T. Acker and B.K. Garvalov), DFG SFB 834 (to A. Acker-Palmer), Gutenberg Research College (GRC) at Johannes Gutenberg University Mainz (to A. Acker-Palmer), the Clusters of Excellence Cardio-Pulmonary System (ECCPS; EXC 147) at the Universities of Giessen and Frankfurt (to T. Acker and A. Acker-Palmer) and Macromolecular Complexes (CEF; EXC 115) at the University Frankfurt (to A. Acker-Palmer), and an EMBO Long-Term Fellowship EMBO ALTF 1181-2011 (to H. Dopeso).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.