EGFR blockade enriches for lung cancer stem-like cells through Notch3-dependent signaling

Cells and transfections

HCC2429, HCC827, and HCC4006 cells were maintained in RPMI with 10% fetal bovine serum. H358, HCC827 and HCC4006 cells were obtained from ATCC within 6 months of the experiments reported, and were identity-verified by STR analysis and certified as mycoplasma-free. Transfections were performed with Lipofectamine 2000 (Invitrogen) reagent according to the manufacturer’s instructions.

Ligands and inhibitors

EGF was purchased from R & D Systems. Erlotinib was a generous gift from Dr. William Pao at Vanderbilt University. Gamma secretase inhibitor (PF-03084014) was kindly provided by Pfizer Global Research and Development, La Jolla Laboratories (San Diego, CA) and was described previously(23, 24).

Antibodies

Following antibodies were used in this study: EGFR (1005) Santa Cruz Biotechnology, EGFR (Ab12) and EGFR (Ab15) are from Neomarker, Notch1 (5B5), Notch3 (8G5), and Notch3 (D11B8) and EGFR (pY1173) obtained from Cell Signaling Technology. Mouse-anti phosphotyrosine is from BD Transduction Laboratories. β-tubulin antibodies were obtained from Sigma.

Plasmid constructs

The pCDNA-EGFR and pCDNA-EGFR (D816A) and Renilla luciferase constructs were provided by Graham Carpenter (Vanderbilt University). Dr. Thao P. Dang provided pCMV-FLAG-N3DA, and pHES1-luciferase constructs. The TP1-luc reporter construct contains 12 tandem repeats of CSL binding sites upstream of luciferase.

Co-immunoprecipitation, immunoprecipitation and western blotting

Cells were washed twice in ice-cold phosphate buffered saline, harvested and lysed with NP40 buffer (10 mM phosphate buffer, 120 mM NaCl, 2.7 mM KCl, 1% Nonidet P40, 10% glycerol) for co-immunoprecipitation experiments or lysed with RIPA buffer (10 mM phosphate buffer, 120 mM NaCl, 2.7 mM KCl, 1% Nonidet P-40, 0.5% DOC, 0.1% SDS) supplemented with complete mini-EDTA free protease inhibitor mixture (Roche) and phosphatase inhibitor mixture cocktails 2 and 3 (sigma), 2 mM NaF and pervanadate for immunoprecipitation for detection of phosphorylation. Equal amount of lysates were precipitated using appropriate antibodies and protein G magnetic beads, or equal amounts of protein were mixed with SDS sample buffer and separated on SDS-PAGE prior to Western analysis.

Aldefluor assay and Flow cytometry

The aldefluor assay kit (Stem cell Technologies) was used to determine the ALDH+ cells. The assay was performed according to manufacturer’s instructions with modifications. Cells were suspended in aldefluor assay buffer and divided into two groups. One group was pretreated for 10 min with ALDH-specific inhibitor Diethylaminobenzaldehyde (DEAB) before incubation with ALDH enzyme substrate Bodipy-Aminoacetaldehyde (BAA) for 45 minutes at 37° C. Cells were centrifuged and re-suspended in a fresh aldefluor assay buffer to remove the unutilized substrate. Cells were analyzed on a FACSCalibur (BD Biosciences) Flow Cytometer. For the analysis of ALDH+ cells, DEAB treated sample was used as a negative control and ALDH activity in presence of DEAB was considered as a baseline.

Pulmosphere formation assay

To study the stem-like cell phenotype, sphere formation assays were performed as described previously (25) with modifications. HCC827 cells treated with vehicle control or erlotinib were trypsinized and counted using Luna automated cell counter. Cells were seeded in 96-well plates at 1000 cells per well in RPMI supplemented with 10% fetal bovine serum, 35 μg/ml bovine pituitary extract (Life Technologies), N2 supplement (Invitrogen), 20 ng/ml EGF, 20 ng/m (Life Technologies), basic fibroblast growth factor (Roche) and 50% Geltrex LDEV-free, hESC-qualified, reduced growth factor basement membrane matrix (Geltrex) (Life Technologies). Cells containing the semi solid medium were seeded in triplicate. The culture was allowed to solidify at 37°C for 30 min followed by layering of 200 μl of similar growth medium without 50% geltrex and incubated for one to 3 weeks. Pulmosphere number was determined using the GelCount mammalian cell colony counter (Oxford Optronix).

Soft agar assay

To measure in vitro tumorigenicity due to erlotinib treatment treated and untreated H358 cells at a density of 10,000 cells per well in 6-well plate were plated in soft agar, in triplicate. The assay was performed using 0.5% and 0.35% agar in RPMI1640 supplemented with 10% FBS as the base and top layers, respectively. Cells were incubated for 21 days and medium was refreshed twice per week. Colonies were counted using GelCount (Oxford Optronix). The colony efficiency was calculated as proportion of colonies per total number of seeded cells. The data was analyzed using GelCount software.

Pharmacological inhibition of EGFR increases the fraction of ALDH+ cells in lung cancer cell lines

We tested whether there is a relationship between EGFR inhibition and the fraction and number of stem-like cells in NSCLC cell lines. EGFR mutated lung adenocarcinoma cells were treated with DMSO or 0.1μM erlotinib and the medium was changed every day with fresh drug. Equal numbers of cells were subjected to ALDH enzymatic activity assays with the ALDH+ and ALDH− cell populations quantified using flow cytometry. We found that HCC4006 and HCC827 NSCLC cell lines, each with an EGFR activating mutation (EGFR ΔE746–A750) (Fig. 1a and b), showed dramatic increases in the fraction of ALDH+ cells upon treatment with erlotinib compared to DMSO treated cells (Table 1). Interestingly, for HCC4006 cells, which have very low basal ALDH activity, while treatment with erlotinib killed much of the ALDH− population, it appeared to maintain or even increase the total number of ALDH+ cells by a small percentage. However, unlike HCC4006 cells, HCC827 cells appear to have a large fraction of cells with low to moderate baseline ALDH activity (3.5% for HCC4006 vs 17.7% for HCC827). This basal activity may not accurately reflect the stem-like cell population. When the analysis is based on those cells with the highest ALDH activity, by gating on the DMSO treated cells without DEAB, the total number of ALDH+ cells with very high activity increased in the HCC827 cells from 491,200 to 782,100 (Supplementary Fig. 1A and 1B; Supplementary Table 1). This demonstrates that while erlotinib treatment causes a large reduction in the total cell numbers for each cell lines, the total number of ALDH high+ cellsis increased. Supplementary Fig. 1a and b shows that both low positive and high positive fractions increase between 3 and 5 days of erlotinib treatment in HCC827 cells.

We further wanted to explore if this phenomenon was also present in H1650 cells, which are also EGFR mutant but resistant to erlotinib due to loss of PTEN(27). Interestingly, erlotinib treatment also had no effect on the ALDH populations in these cells (Supplementary Fig. 2). Additional studies with A549 and H358 cells carrying wild type EGFR and a K-RAS mutation (28) (Supplementary Fig. 3a and b) also showed increased ALDH activity upon exposure to erlotinib. Unlike H1650, which are completely insensitive to erlotinib, the H358 cell line and A549 cell line are somewhat sensitive to erlotinib (H358 more so than A549) (29) and also show increases in the fraction of ALDH+ cells with erlotinib treatment. These data suggest that this phenomenon is not just restricted to cells with an EGFR activating mutation, but limited to those with some EGFR-signaling dependence, and that EGFR activity is coupled with ALDH activity.

Erlotinib treatment enhances pulmosphere-forming potential in EGFR mutated lung cancer cells

Previous studies have demonstrated that ALDH+ cells possess features similar to cancer stem cells such as increased pulmosphere-forming ability. To determine if the residual cell population after erlotinib treatment is more stem-like we performed the pulmosphere formation assay. HCC827 cells were treated with 0.1 μM erlotinib for five days. Remaining cells were allowed to recover under the regular growth conditions for four days and subjected to sphere formation culture assay. As expected from the ALDH+ data, compared to DMSO, erlotinib treated cells showed an increased number and size of the pulmospheres in semisolid matrix (Fig. 2a and b). We also performed a soft agar colony formation assay, which measures anchorage-independent growth and is an indicator for cell transformation using H358 cells. Cells treated with erlotinib showed significant increase in number of colonies formed compared to DMSO control (Fig. 2c and d). These data clearly suggest that erlotinib treatment increases the clonogenic potential in the surviving population of cells.

EGFR signaling down-regulates Notch-mediated transcriptional activity in a kinase-dependent manner

To assess if the erlotinib-mediated stem-like cell phenotype is due to modulation of transcriptional activity of Notch, we examined the Notch transcriptional activity in the presence of wild type- or kinase inactive-EGFR. HEK293 cells were transiently transfected with EGFR and Notch3-ICD along with Notch inducible CSL-synthetic-(Fig. 3a) or full-length Hes1 promoter-driven luciferase reporters (Fig. 3b). Two days after transfection transcriptional activity was measured using a luciferase assay. The coexpression of EGFR with Notch3-ICD decreased Notch3-ICD-mediated CSL and Hes1 reporter activities in a dose-dependent manner. We further determined if kinase activity is essential for EGFR-mediated negative regulation of Notch3. Notch3-ICD was co-expressed with a kinase-inactive mutant of EGFR and co-expression of this kinase-inactive EGFR did not result in a decrease in Notch-mediated transcriptional activity, demonstrating that EGFR negatively regulates Notch activity through its tyrosine kinase activity. Interestingly, there was a slight increase in reporter activity compared to baseline when EGFR (KD) was transfected Fig. 3A and 3B), which may have been due to interference with endogenous EGFR activity. To make certain that the EGFR transfection was not having any other effect on the cell populations or the reporter in the absence of Notch, EGFR and EGFR (KD) were transfected with reporter alone (Supplementary Figure 4). We did not observe any EGFR mediated Notch reporter activity in the absence of Notch overexpression. Additionally, we tested if pharmacological inhibition of EGFR using erlotinib can stimulate the expression of Notch target genes in lung cancer cells. The qRT-PCR analysis of Hes1 expression showed a significant increase after 24 or 48 hours of treatment with erlotinib in HCC827 cells (Fig. 3c). This demonstrates EGFR kinase activity inhibits Notch signaling, and thus erlotinib treatment relieves this inhibition resulting in Notch transcriptional activation.

Gamma secretase inhibitor treatment eliminates erlotinib-induced stem-like cells by decreasing Notch activity

The major Notch pathway inhibitors in clinical testing are gamma secretase inhibitors (GSIs) (30). These compounds block the final activation (S3 cleavage) step in canonical Notch signaling. GSIs inhibit the cleavage of the Notch receptor preventing release of the intracellular domain into the cytoplasm and subsequent translocation to the nucleus, thus abrogating canonical signal activation for all the Notch receptors. Since we find that EGFR inhibition increases ALDH+ cells and also activates Notch transcriptional activity, we sought to determine if the erlotinib-induced increase of ALDH+ cells is inhibited by treatment with a GSI. Concomitant treatment with GSI PF-03084014 and erlotinib reduced the number of ALDH+ cells in both HCC4006 and HCC827 cells observed with erlotinib alone (Fig. 4a and b). Next we sought to identify if erlotinib-stimulated Notch transcriptional activity is sensitive to GSI. Consistent with the GSI sensitivity of erlotinib-induced stem-like cells, both HCC827 and HCC4006 showed increased expression of Hes1 and Hey1 with erlotinib treatment, which was sensitive to PF-0308401 following 3 days of treatment (Fig. 4c). These data support the hypothesis that erlotinib increases ALDH positivity through activation of Notch signaling and suggests a potential clinical strategy for overcoming this effect through combined EGFR and Notch inhibition.

The erlotinib-induced increase in the ALDH+ fraction in lung cancer cells is dependent on Notch3, but not Notch1

To determine whether erlotinib mediated expansion of ALDH+ cells is mediated by the Notch1 and/or Notch3 receptors, we evaluated the role of Notch1 and 3 in the erlotinib-induced increase of ALDH+ cells using the EGFR-mutated HCC4006 and HCC827 cell lines. Cells were transfected with non-targeting control (NTC) or three separate Notch1 or Notch3 siRNAs and treated with erlotinib. Knockdown for both Notch1 and Notch3 receptors was efficient(Supplementary Fig. 5a and b). As expected, erlotinib treatment stimulated a fivefold and threefold increase in ALDH+ cells in HCC4006 and HCC827, respectively. While knockdown of Notch1 had no effect on erlotinib-stimulated cells, the Notch3 knockdown completely abolished this effect to baseline levels, (Fig. 5a and b) suggesting that erlotinib induces stem-like cells through Notch3 and not Notch1.

Notch and EGFR receptors co-precipitate and this interaction is dependent on EGFR kinase activity

Next, we sought to determine if there is an EGFR activation-dependent functional association between EGFR and Notch receptors in HCC2429 cells, a cell line that expresses wild-type EGFR and both Notch1 and 3 receptors and multiple Notch ligands, and in which the combination of Notch and EGFR-targeted therapies show an increased efficacy(10, 31). Cells were stimulated with EGF for increasing amounts of time and total EGFR was precipitated, and blotted for Notch3. These analyses identified clear EGF-dependent co-precipitation of Notch3 with EGFR (Fig. 6a). In the absence of ligand, little or no association was observed between EGFR and Notch receptors, indicating that the association is ligand (EGF)-dependent. To further validate the interaction between EGFR and Notch3, co-immunoprecipitations were performed using three different antibodies that target distinct epitopes on the EGFR protein. Western blot analyses showed that Notch3 co-immunoprecipitates with all three EGFR antibodies, and this association was further enhanced when stimulated with EGF (Supplementary Fig. 6).

We further tested if the association between EGFR and Notch receptor is dependent on the kinase activity of EGFR. HCC4006 and HCC827 cells expressing constitutively active EGFR were treated with vehicle control or erlotinib and subjected to co-immunoprecipitation. As expected, control cells showed strong interaction between Notch and EGFR, and erlotinib treatment completely abolished this interaction(Fig. 6b and c). Reciprocal immunoprecipitation revealed the same result (Fig. 6b and c). These experiments demonstrate that EGFR associates with Notch3 receptor in a kinase-dependent fashion.

EGFR mediated tyrosine phosphorylation of Notch receptors

Little is known about post-translational modifications of the Notch receptor and their role in regulating Notch activity(32). The co-immunoprecipitation and functional associations above led us to speculate that Notch3 could be a novel direct or indirect substrate for EGFR mediated tyrosine phosphorylation, which may mediate its role as a negative regulator. To demonstrate tyrosine phosphorylation of the Notch receptor, HCC2429 cells that overexpress Notch3 due to translocation, and wild-type EGFR were stimulated with EGF for increasing durations of exposure. Total protein lysates were immunoprecipitated with a Notch3 antibody and analyzed with either phosphotyrosine or Notch3 antibody (Fig. 7a) or immunoprecipitated with phosphotyrosine and subjected to western analysis with Notch3 (Fig. 7b). These data clearly show that following the addition of EGF, Notch3 is phosphorylated in a time-dependent manner. These data are the first to demonstrate tyrosine phosphorylation of Notch and to identify Notch as a novel direct or indirect substrate for the EGFR tyrosine kinase.

To confirm that the observed tyrosine phosphorylation of Notch3 is dependent on EGFR kinase activity, cells were treated with vehicle or erlotinib for 1 or 24 hours. Treatment with erlotinib reduced the tyrosine phosphorylation of Notch3 in a time-dependent manner (Fig. 8a and b). Furthermore, the phosphorylation was detected reciprocally, demonstrating that Notch3 phosphorylation is due to EGFR kinase activation. A similar pattern was observed for tyrosine-1173 auto phosphorylation of EGFR, further confirming that the Notch3 receptor is tyrosine phosphorylated in an EGFR kinase-dependent fashion.

To further address if EGFR mediated Notch3 tyrosine phosphorylation is indeed a key regulatory event that modulates the stem-like phenotype in lung cancer cells we used H1650 cells, which do not show an increase in ALDH activity in response to erlotinib treatment (Supplementary Fig. 2), and also show a continuously active EGFR in the presence of erlotinib (Supplementary Fig. 7). Phosphotyrosine analysis of Notch3 revealed that under basal conditions Notch3 is tyrosine phosphorylated. Treatment with erlotinib did not prevent the Notch3 tyrosine phosphorylation (Supplementary Fig. 7) suggesting that inhibition of EGFR is a key event required for the Notch3 activation.