Skip to main content
main-content
Top

26-11-2018 | EGFR inhibitors | Article

Aurora kinase A drives the evolution of resistance to third-generation EGFR inhibitors in lung cancer

Journal:
Nature Medicine

Authors: Khyati N. Shah, Roma Bhatt, Julia Rotow, Julia Rohrberg, Victor Olivas, Victoria E. Wang, Golzar Hemmati, Maria M. Martins, Ashley Maynard, Jonathan Kuhn, Jacqueline Galeas, Hayley J. Donnella, Swati Kaushik, Angel Ku, Sophie Dumont, Gregor Krings, Henry J. Haringsma, Liliane Robillard, Andrew D. Simmons, Thomas C. Harding, Frank McCormick, Andrei Goga, Collin M. Blakely, Trever G. Bivona, Sourav Bandyopadhyay

Publisher: Nature Publishing Group US

Abstract

Although targeted therapies often elicit profound initial patient responses, these effects are transient due to residual disease leading to acquired resistance. How tumors transition between drug responsiveness, tolerance and resistance, especially in the absence of preexisting subclones, remains unclear. In epidermal growth factor receptor (EGFR)-mutant lung adenocarcinoma cells, we demonstrate that residual disease and acquired resistance in response to EGFR inhibitors requires Aurora kinase A (AURKA) activity. Nongenetic resistance through the activation of AURKA by its coactivator TPX2 emerges in response to chronic EGFR inhibition where it mitigates drug-induced apoptosis. Aurora kinase inhibitors suppress this adaptive survival program, increasing the magnitude and duration of EGFR inhibitor response in preclinical models. Treatment-induced activation of AURKA is associated with resistance to EGFR inhibitors in vitro, in vivo and in most individuals with EGFR-mutant lung adenocarcinoma. These findings delineate a molecular path whereby drug resistance emerges from drug-tolerant cells and unveils a synthetic lethal strategy for enhancing responses to EGFR inhibitors by suppressing AURKA-driven residual disease and acquired resistance.

Main

The approval and use of epidermal growth factor receptor (EGFR) inhibitors in EGFR-mutant non-small-cell lung cancer (NSCLC) has been a major clinical breakthrough, helping to define the paradigm of precision medicine. However, EGFR tyrosine kinase inhibitors (TKIs) often produce an incomplete response followed by progression and acquired resistance in 9–12 months, often a lethal event13. Disease progression occurs through tumor evolution on treatment, involving distinct genetic and nongenetic changes in cell state and signaling4. Furthermore, patient tumors develop acquired resistance via multiple mechanisms simultaneously58, and this polyclonal nature of resistance could limit the efficacy of approaches that target any single genetic driver of resistance. The heterogeneous nature of acquired resistance highlights the need to better understand and target residual disease, defined as the fraction of tumor cells that survive initial treatment and ultimately enable tumor progression in the presence of ongoing treatment9. Acquired resistance occurs through the selection of preexisting clones as well as the evolution of drug-tolerant (that is, persister) cells without genetic alterations that survive treatment through tumor cell adaptation that may involve the acquisition of genetic mutations later1012. Both genetic and nongenetic forms of resistance to EGFR TKIs have been identified, including secondary mutations in EGFR, amplification of various receptor tyrosine kinases, transformation to small-cell lung cancer and epithelial–mesenchymal transition (EMT)4,1316. Third-generation EGFR inhibitors, rociletinib and the US Food and Drug Administration–approved agent osimertinib, bind and inhibit mutant EGFR with and without the T790M mutation associated with resistance to previous-generation EGFR inhibitors1,2. For these drugs, approximately half of acquired resistance cases have unknown genetic drivers, and when genetic drivers exist, multiple drivers often co-occur in the same patient7,8,13. We identified a synthetic lethal interaction between EGFR TKIs and Aurora kinase inhibitors in acquired resistant cells that has important implications for the development of new treatment strategies aimed at preventing rather than intercepting acquired resistance.
We modeled acquired resistance to both osimertinib and rociletinib by deriving polyclonal acquired resistant cell lines on the basis of stepwise dose escalation over a period of 9 d followed by maintenance in 1 μM of drug over 6 weeks (osimertinib-resistant lines are denoted by OR and rociletinib-resistant lines by RR; see Fig. 1a). We generated eight acquired resistant models from four different EGFR-mutant NSCLC cell lines including PC9, HCC827 and HCC4006 expressing an EGFR exon 19 deletion and H1975 expressing a compound EGFR L858R and T790M mutation. There was a greater than tenfold change in half maximal inhibitory concentration (IC50) in each line compared to parental cells, and we also observed cross-resistance between drugs indicating a shared mechanism of resistance regardless of which EGFR inhibitor was used (Fig. 1b and Supplementary Fig. 1a). In response to TKI, resistant cells suppressed EGFR signaling, and we observed no activation of alternate receptor tyrosine kinases previously reported to facilitate bypass of EGFR inhibition (Supplementary Fig. 1b)17. In response to treatment, resistant cells demonstrated heightened extracellular signal-regulated kinase (ERK) and protein kinase B (PKB/Akt) signaling and reduced apoptosis, as measured by cleaved poly(ADP-ribose) polymerase (PARP), compared to parental cells (Fig. 1c). Exome sequencing revealed no recurrent mutations among independently derived acquired resistant lines, and no additional mutations in EGFR were detected (data not shown). We next sought to identify whether these cells harbored markers of cell states known to be associated with resistance to EGFR TKIs. Compared to parental cells, resistant cells had an increase in vimentin levels indicative of EMT, increased nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) signaling and minor changes in cancer cell stemness, all known to be associated with EGFR TKI resistance (Supplementary Fig. 1c)4,12,1720. Tumor protein p53 (TP53) and neuroblastoma RAS viral oncogene homolog (NRAS) signaling were not strongly associated with resistance (Supplementary Fig. 1d,e)21,22. Heritability analysis using single-cell clones indicated that the majority of cells derived from acquired resistant lines were resensitized to TKI after a period of drug withdrawal, indicating a nongenetic and reversible mechanism of drug resistance (Supplementary Fig. 1f).
On the basis of the absence of any obviously targetable driver of resistance, we sought to identify pathways revealed by drugs that synergistically inhibit growth when combined with EGFR TKIs. Across a 94-compound cancer-focused library, both Aurora kinase inhibitors in the panel, AZD1152 and VX680, were the top synergistic candidates when combined with 2 μM rociletinib in H1975-RR cells (Fig. 1d and Supplementary Table 1). The combination of these two agents as well as MLN8237, the most clinically advanced Aurora kinase inhibitor, with either osimertinib or rociletinib demonstrated synergistic reduction in cell growth in all models (Fig. 1e,f and Supplementary Fig. 2a,b). Aurora kinase inhibitors display significant cross-reactivity between AURKA, AURKB and AURKC23. Therefore, these data reveal a primary requirement for Aurora kinase signaling in models of acquired resistance to third-generation inhibitors of EGFR.
We sought to determine the relevant target of Aurora kinase inhibitors in driving drug synergy. We found approximately twofold messenger RNA upregulation but no increase in total protein levels for all three Aurora kinases in resistant cells compared to parental cells (Fig. 2a and Supplementary Fig. 3a,b). In contrast, we found significant activation of AURKA, but not AURKB or AURKC, in resistant models as indicated by increased autophosphorylation at T288 (Fig. 2a and Supplementary Fig. 3b). We next asked whether activation of AURKA is sufficient to confer resistance to EGFR TKIs. AURKA activity peaks during the G2/M phase of the cell cycle where it regulates chromosome alignment, mitotic spindle formation and chromosome segregation24. Parental PC9 cells synchronized using serum starvation or thymidine block into the G2/M phase had high levels of phosphorylated AURKA (phospho-AURKA), were more viable and had diminished apoptosis in comparison to parental cells that were treated when they were in the G1/S phase or asynchronously (Supplementary Fig. 3c–i). Transient AURKA overexpression, but not AURKB, caused resistance to EGFR TKIs at levels comparable to the Kirsten rat sarcoma viral oncogene homolog (KRAS) G12V mutant, a known driver of resistance (Fig. 2b and Supplementary Fig. 3j)25.
AURKA can be activated by upstream factors that facilitate its autophosphorylation, including targeting protein for Xklp2 (TPX2), neural precursor cell expressed developmentally down-regulated protein 9 (NEDD9), protein ajuba (AJUBA) and serine/threonine-protein kinase PAK 1 (PAK-1)26. We investigated each one and observed a consistent increase in TPX2 protein, and to a lesser extent transcription, and an increase in phospho-AURKA in all resistant models (Fig. 2c and Supplementary Fig. 4a,b). TPX2 activates AURKA by locking it in an active conformation and protecting it from protein phosphatases24. Overexpression of TPX2 activated AURKA and caused EGFR TKI resistance (Fig. 2b), whereas expression of other reported AURKA activators did not (Supplementary Fig. 4c,d). TPX2 is degraded by the ubiquitin E3 ligase anaphase-promoting complex (APC) bound to the specific activator cadherin-1 (CDH1) during mitotic exit and the G1 phase27. Using subcellular fractionation, we confirmed that CDH1 was nuclear in parental and resistant cells as previously reported28. In contrast, TPX2 was cytosolic in cells with acquired resistance, whereas in parental cells it was more likely to be nuclear (Supplementary Fig. 4e). Hence, TPX2 is not colocalized with the complex responsible for its degradation in resistant cells. Together, these data suggest that altered TPX2 localization in cells with acquired resistance contributes to AURKA activation during the interphase and promotes acquired resistance to third-generation EGFR TKIs.
We next sought to determine the mechanism of synergy focusing on MLN8237, the most advanced AURKA inhibitor (Fig. 1e,f). Combination treatment resulted in a reduction in cell proliferation and an increase in cell death measured by YO-PRO-1 positivity in models with acquired resistance (Fig. 2d,e). In PC9-RR mouse xenografts, rociletinib partially abrogated tumor growth and led to rapid tumor progression, combination of rociletinib and MLN8237 led to a stronger initial reduction in tumor growth, which was sustained for 70 d (P = 2.2 × 10−11) (Fig. 2f). We observed no apparent toxicity on the basis of body weight (Supplementary Fig. 5a). We observed similar results with osimertinib, which in combination with MLN8237 resulted in decreased tumor growth in nine out of ten tumors derived from PC9-OR cells (P = 0.001) (Fig. 2g and Supplementary Fig. 5b). To understand how AURKA might regulate apoptosis and proliferation, we probed several signaling pathways known to be associated with resistance to EGFR inhibitors18,19,29. Combination treatment caused a decrease in ERK and NF-κB signaling, indicating multiple potentially overlapping routes through which AURKA signaling contributes to cell growth (Fig. 2h and Supplementary Fig. 6). Therefore, the combination of EGFR and Aurora kinase inhibitors induces apoptosis and acts synergistically in suppressing the growth of acquired resistant cells in vitro and in vivo.
We next sought to understand the molecular mechanisms underlying how this combination engaged the apoptotic machinery, focusing on proapoptotic factor Bcl-2-like protein 11 (BIM) because its induction is essential for cell death induced by EGFR TKIs30,31. BIM and its splice variant BIM extra long (BIM-EL) are regulated by phosphorylation leading to proteasomal degradation32. EGFR inhibition alone in parental PC9 cells suppressed phospho-BIM, resulting in the accumulation of BIM-EL, and induced PARP cleavage consistent with previous reports (Fig. 2h)30,31. In contrast, in resistant cells the combination of EGFR TKI and MLN8237 was necessary to suppress phospho-BIM, leading to the accumulation of BIM-EL, its primary effector apoptosis regulator BAX (BAX) and cleaved PARP (Fig. 2h). In normal cells, AURKA suppresses apoptosis during mitosis by suppressing BIM24,33. Since acquired resistant cells expressed high levels of phospho-AURKA throughout the cell cycle in contrast to parental cells (Supplementary Fig. 3d), these data indicate that resistant cells co-opt this natural function of AURKA for use throughout the cell cycle, thereby exploiting a redundancy in the control of BIM that is normally temporally segregated. Therefore, cells with acquired resistance escape from EGFR inhibition through a shift in the control of the proapoptotic machinery from EGFR alone to both EGFR and AURKA (Fig. 2i).
We next sought to elucidate the temporal processes leading to AURKA activation in cells with acquired resistance. On the basis of our modeling after treatment, we divided the period of drug response into three distinct phases: (i) a sensitive phase; (ii) a drug-tolerant phase in which remaining cells persist; and (iii) a proliferative acquired-resistance phase. We measured signaling dynamics across a time course of 9 d in H1975 cells treated with osimertinib and compared this with H1975-OR cells that were exposed for more than 6 weeks (Fig. 3a). While osimertinib treatment inhibited EGFR throughout, we observed BIM-mediated apoptosis for 2 days, after which it was suppressed indicating that the remaining cells were drug-tolerant. In the sensitive phase, we observed a gradual increase in TPX2 followed by activation of AURKA peaking during the establishment of drug tolerance and maintained into acquired resistance (Fig. 3b). These data indicate that AURKA activation emerges after chronic EGFR inhibition and is maintained in drug-tolerant cells and those with acquired resistance.
Since high levels of TPX2 and AURKA cause mitotic errors and polyploidy34,35, we hypothesized that its abnormal levels should also leave a signature of defects associated with mitotic stress. We surveyed for mitotic defects induced by EGFR TKI treatment and in cells with acquired resistance. EGFR TKI treatment for 72 h resulted in an accumulation of errors in centrosome biogenesis, spindle assembly and chromosome segregation in parental models and those with acquired resistance (Supplementary Fig. 7a,b and Supplementary Table 2), indicating that mitotic stress is a feature of EGFR inhibition. Upon EGFR inhibition, errors in mitosis led to the generation of polyploid cells (Supplementary Fig. 7c), which were phenocopied by overexpression of AURKA or TPX2 in parental cells, implying causation (Supplementary Fig. 7d,e). These data indicate that cells with acquired resistance emerge from drug-tolerant cells through the AURKA-dependent suppression of BIM, coincident with mitotic stress driven by abnormal levels of TPX2 and AURKA.
Because AURKA became activated during drug tolerance, we hypothesized that AURKA might also be necessary for the formation and survival of drug-tolerant cells. While osimertinib and rociletinib drug-tolerant persister (DTP) cells12 maintained EGFR inhibition and suppressed BIM-mediated apoptosis, they also displayed increased levels of phospho-AURKA and TPX2 (Fig. 3c). We observed this mechanism of tolerance to EGFR inhibition with other generations of EGFR inhibitors, including erlotinib and afatinib (Fig. 3d). To determine whether AURKA inhibition blocks the emergence of acquired resistance in EGFR TKI-naive NSCLC cells, we treated single-cell derived PC9 and H1975 cells with either single-agent EGFR TKI or MLN8237 or their combination over a period of 13 weeks and measured the rate of outgrowth of resistant clones. Combination treatment enhanced the magnitude of the response and delayed the emergence of resistance compared to monotherapy (Fig. 3e,f). This was because the combination increased the proportion of cells displaying evidence of apoptosis over the course of treatment, leading to a reduction in the formation of drug-tolerant, residual cells and was independent of which EGFR or Aurora kinase inhibitor was used (Fig. 3g and Supplementary Fig. 8a–c). Combination treatment was also effective in eradicating previously formed DTP cells with near complete elimination within 1 week, indicating that AURKA activity is necessary for their survival (Fig. 3h). These data suggest that the combination of EGFR TKI and Aurora kinase inhibitors administered simultaneously or in sequence at the time of residual disease may be an effective means to enhance the initial response and forestall acquired resistance.
We next explored the contribution of AURKA activation in clinical residual disease and progression. We tested an EGFR-L858R-positive and EGFR-T790M-negative patient-derived xenograft (PDX) tumor model from a residual mass obtained from a patient demonstrating an incomplete response to erlotinib19. Rociletinib treatment only modestly impaired tumor growth, indicating cross-resistance between erlotinib and rociletinib in this model. In contrast, the combination robustly decreased tumor growth compared to rociletinib alone (P = 3.4 × 10−4) and in most cases induced tumor regression with no observed toxicity on the basis of weight (Fig. 4a and Supplementary Fig. 9a,b). Combination treatment induced apoptosis as evidenced by increased staining for cleaved caspase-3 as well as loss of Ki-67 staining in tumor tissue (Supplementary Fig. 9c,d). We also observed efficacy and lack of toxicity using osimertinib in combination with MLN8237 (P = 6.6 × 10−5) (Fig. 4b and Supplementary Fig. 9e,f).
Consistent with the adaptive resistance observed in vitro, tumors treated with rociletinib for 30 d had lower phospho-EGFR levels and an increase in phospho-AURKA and TPX2 levels compared to vehicle alone (Fig. 4c). Enhanced TPX2 levels in rociletinib-treated tumors were also evident through immunohistochemistry (IHC) (Fig. 4d,e). EGFR suppression also induced mitotic stress in vivo, and there was a significant increase in the number of abnormal mitoses quantified via H&E staining (Supplementary Fig. 9g,h). The combination also induced apoptosis as evidenced by suppression of phospho-BIM, increased total BIM and cleaved PARP compared to either single agent (Fig. 4c). These results establish mechanistic equivalence between processes occurring in vitro with those occurring in patient residual disease and indicate that pharmacologically targeting AURKA at the point of maximal response or residual disease may be a viable clinical strategy to deepen responses.
We next sought to establish the clinical relevance of heightened TPX2/AURKA signaling in mediating acquired resistance to EGFR inhibition in NSCLC. Staining and automated quantification of TPX2 levels in matched diagnosis and relapse samples from nine patients with advanced-stage EGFR-mutant NSCLC who underwent treatment with erlotinib revealed a significant increase in TPX2 levels after acquired resistance compared to pretreatment (P = 0.003) (Fig. 4f). Using a threshold TPX2 score of 2, which was higher than what we observed in all nine pretreatment samples, we observed TPX2 positivity in six out of nine cases (Fig. 4g and Supplementary Fig. 10a). Interestingly, we observed increased TPX2 levels in three cases that also displayed EGFR-T790M-positivity or MET amplification upon relapse, suggesting that nongenetic AURKA activation as a driver may co-occur with other genetic drivers of acquired resistance. In three acquired resistance cases to third-generation inhibitors (two osimertinib and one rociletinib), all three were TPX2-positive and were increased compared to pretreatment (P = 0.02) (Fig. 4h and Supplementary Fig. 10b). Together, resistance was associated with an increase in TPX2 regardless of the EGFR TKI used (positivity in 75% of cases, P = 0.00002) (Fig. 4i). These data suggest that AURKA activation via TPX2 is a feature of most acquired resistant EGFR-mutant lung cancers regardless of therapy. We propose that TPX2 could be used as a biomarker to select patients for combination therapy with an EGFR TKI and an Aurora kinase inhibitor in EGFR-mutant lung adenocarcinoma.
In summary, our findings have important implications for intercepting EGFR TKI resistance in patients with NSCLC, offering an alternative approach to combat the emergence of resistance. In light of observations that multiple distinct mechanisms of drug resistance can co-occur within the same patient at the time of relapse58, AURKA activation might co-occur with other factors driving resistance or could even provide a mechanism on which such resistance-causing mutations could appear, giving rise to multiple genetically distinct clones11. The maintenance of residual disease by AURKA may provide a fertile ground for the formation of such resistance-causing mutations, potentially through a late-emerging resistance model10,11. For example, mitotic abnormalities catalyzed by AURKA hyperactivity may give rise to gene amplifications that have been observed in patients progressing on EGFR inhibitors. Since mitotic errors lead to chromosomal instability contributing to disease progression and drug resistance36, resistance driven by AURKA may contribute to tumor heterogeneity and promote the generation of distinct clones harboring different genetic drivers of drug resistance. If correct, this adaptive response to EGFR inhibition could actually enhance tumor heterogeneity. Data from our study and others demonstrate that AURKA contributes to a number of pathways and processes previously associated with resistance to EGFR inhibition, including NF-κB, ERK and EMT26,37. Therefore, it appears that AURKA is associated with a number of seemingly disparate mechanisms of acquired resistance, warranting further investigation.
Our results call for clinical trials testing the combination of Aurora kinase and EGFR inhibitors in EGFR-mutant lung adenocarcinoma, up front, at the point of residual disease and after acquired resistance in tumors harboring high levels of TPX2. Patients progressing on first- and third-generation EGFR TKIs often have high levels of TPX2, indicating therapeutic relevance in a significant fraction of acquired resistant, immunotherapy-refractory38 lung cancers. As single-agent therapies, Aurora kinase inhibitors have reached phase 3 clinical trials39,40 and have nonoverlapping toxicity profiles with EGFR inhibitors. We propose that the most effective use of this combination should be directed toward eliminating residual cancer cells before they acquire genetic mechanisms of resistance that could be polyclonal and heterogeneous in nature. While clinical studies are necessary to determine the degree to which this combination strategy can delay the onset of resistance in patients, these results call to action a proactive paradigm aimed at preventing resistance rather than the current reactive paradigm of intercepting and treating drug resistance incrementally.

Methods

Cell culture and compounds

H1975, HCC827 and HCC4006 cells were obtained from ATCC. PC9 cells were a gift from F. Koizumi (National Cancer Center Research Institute and Shien-Lab, Tokyo, Japan). PC9 parental cell line identity was confirmed by short tandem repeat analysis (Genetica). Cells were used for no longer than 12 months before being replaced and were routinely tested for Mycoplasma to ensure the accuracy of experimental data. Rociletinib was obtained from Clovis Oncology. Erlotinib, afatinib and osimertinib (AZD9291) were purchased from Selleck Chemicals.

Generation of cells with acquired resistance and drug-tolerant cells

Cells with acquired resistance were derived by treating individual cell lines with increasing concentrations of rociletinib or osimertinib starting at 50 nM, followed by a stepwise dose escalation every 48 h up to 1 μM. Cell lines with acquired resistance that were derived from rociletinib treatment (PC9-RR, H1975-RR, HCC4006-RR, HCC827-RR) or osimertinib treatment (PC9-OR, H1975-OR, HCC4006-OR, HCC827-OR) were maintained in 1 μM of the respective drug. To generate DTP cells, parental PC9 and H1975 were treated with 1 μM of erlotinib, afatinib, rociletinib and osimertinib for 9 d, according to protocols described previously12. Cells were washed and replenished with fresh drug every 48 h.

Cell proliferation and apoptosis assays

Cell lines were seeded in 384-well assay microplates at a density of 1,000 cells per well in a total volume of 40 μl per well and incubated at 37 °C, 5% CO2 overnight. Following drug exposure, proliferation was measured by staining with Hoechst 33342 (Thermo Fisher Scientific) nuclear dye; apoptosis was measured using YO-PRO-1 early apoptosis dye (Thermo Fisher Scientific) and analyzed using a CellInsight High-Content Microscope (Thermo Fisher Scientific) for the indicated time. IC50 values were determined using Prism version 6.0 (GraphPad). For drug synergy, fixed-dose ratios were used to determine five different drug combinations. Following 72 h of drug exposure, proliferation and cell death were measured by staining with Hoechst 33342 nuclear dye and YO-PRO-1, respectively, and analyzed using a CellInsight High-Content Microscope. Synergistic, additive or antagonistic effects were determined using the combination index method devised by Chou and Talalay41.

Combination drug screen

H1975-RR cells were seeded in 384-well microplates at a density of 1,000 cells per well in the presence of 2 μM rociletinib or vehicle; after 24 h, they were exposed to three different doses of compounds from a 90-drug library for 72 h. At the end of this period, nuclei were stained with Hoechst 33342 and counted using a CellInsight High-Content Microscope. The screen was repeated three times using varying library concentrations of 5 μg ml−1, 500 ng ml1 and 50 ng ml−1, and each combination was measured in quadruplicate. Raw cell numbers were median-normalized on a per-plate basis. For each compound in the library, the relative cell number in the DMSO plate was compared with the number in the rociletinib plate using a two-tailed Student’s t-test. A synergy score was developed based on the −log10 of the P value of this t-test and was signed to indicate synergistic inhibition of growth (positive score) or antagonism (negative score). The reported synergy score is based on the average of scores over three different library concentrations.

Clonogenic growth assay

Colony outgrowth assays were performed using crystal violet staining and quantification. Briefly, cells were seeded in 12-well microplates at a density of 1,000 cells per well. Appropriate drugs were added after an additional 24 h. Cells were exposed to drug or DMSO for 9–10 d, with medium change and fresh drug added every 3 d. Cells were fixed with 4% formaldehyde and stained with 0.5% crystal violet. Pictures of stained cells were taken using an EPSON Perfection V600 scanner. Growth was quantified by dissolving crystal violet in 0.1% SDS and absorbance was quantified at 590 nm using a spectrophotometer and normalized to DMSO treatment.

Immunoblot

Cells for immunoblots were collected and lysed in lysis buffer containing 1 mol l−1 Tris-HCl buffer, pH 7.6, 0.5 mol l−1 EDTA, 5 mol l−1 NaCl, 1% NP-40 and 1% Triton X-100, supplemented with protease and phosphatase inhibitors (Calbiochem). Samples were sonicated and then centrifuged at 14,000 r.p.m. for 10 min at 4 °C. Protein concentrations were determined by Bradford assay (Bio-Rad). Equal amounts of protein (10–40 μg) were loaded onto SDS–polyacrylamide gel electrophoresis gels, transferred to a polyvinylidene difluoride membrane (Bio-Rad) and incubated with the indicated primary antibodies. Proteins were detected via incubation with horseradish peroxidase–conjugated secondary antibodies, Clarity Western ECL Blotting Substrates (Bio-Rad) or SuperSignal West Femto Maximum Sensitivity Chemiluminescent Substrate (Thermo Fisher Scientific). Antibodies for phospho-EGFR (Y1068), phospho-ERK1/2 (T202/Y204), ERK1/2, phospho-Akt (S473), Akt, cleaved PARP, phospho-AURKA (T288), AURKA, phospho-Rb (S780), BIM, phospho-BIM (S69), BAX, vimentin, phospho-NF-κB p65 (S536), NF-κB p65, pan phospho-AURKA/B/C (T288/T232/T198), CD44 and histone H3 (9715) were purchased from Cell Signaling Technology; TPX2 and pan total AURKA/B/C were purchased from Sigma-Aldrich; EGFR, NEDD9/Cas-L, AJUBA, PAK-1, CD24, CD133, β-tubulin and p53 were purchased from Santa Cruz Biotechnology; FZR1/CDH1 was purchased from Abcam; and V5 tag was purchased from Thermo Fisher Scientific. Band intensities were quantified using Adobe Photoshop CS3. Phospho-receptor tyrosine kinase arrays were performed according to the manufacturer’s protocol (R&D Systems).

Cell cycle synchronization and analysis

For double thymidine block, cells were seeded and 2 mM thymidine was added; later, thymidine was released from this block by washing the cells three times with PBS and adding complete media followed by a second thymidine block and release. For synchronization using serum starvation and release, cells were kept in serum-free media for 48 h. To synchronize cells in the G1/S phase, cells were released for 2 h; to synchronize cells in the G2/M phase, cells were released for 8 h followed by drug treatment. In all cases, the expected cell cycle was validated using fluorescence-activated cell sorting. For cell cycle analysis, cells were fixed in cold ethanol and resuspended in propidium iodide/RNase staining solution (Cell Signaling Technology). After incubation for 15 min at room temperature in the dark, flow cytometric analysis was performed on a FACS Aria II Flow Cytometer (BD Biosciences). Flow cytometry data were analyzed with the FlowJo software to measure polyploidy (>4 N).

Plasmid transfections

LacZ, TPX2, AURKA, AURKB, AJUBA, NEDD9, PAK-1 and KRAS-G12V were obtained in a pLX304 backbone from Addgene; 1 μg per well of plasmids were transfected with 0.1% FuGENE HD transfection reagent (Promega) for 48 h before further analysis.

Quantification of mitotic defects

Cells were plated overnight on a tissue culture–treated 8-well chamber slide (Thermo Fisher Scientific). After 72 h of drug treatment, cells were washed with PBS and fixed by with ice-cold methanol at −20 °C for 3 min. Following fixation, cells were permeabilized with PBS and 0.1% Triton X-100 for 3 min, blocked in PBS with Tween 20 (PBS, 5% BSA, 0.1% Triton X-100) for 30 min and then incubated with primary antibodies in blocking buffer for 90 min. Cells were washed with PBS and incubated with species-specific fluorescent secondary antibodies (Alexa-conjugated; Thermo Fisher Scientific). DNA was stained with Hoechst 33342 (1:5,000) for 5 min in PBS. Coverslips were mounted with ProLong Antifade mountant (Thermo Fisher Scientific). Antibodies were anti-α-tubulin (1:500; Sigma-Aldrich) and anti-γ-tubulin (1:500; Sigma-Aldrich). Images were collected with a ZEISS Cell Observer using the 404, 488 and 561 nm laser.

Human tissue and IHC

All patient tumor samples analyzed were obtained under institutional review board–approved protocols with informed consent obtained from each patient under the guidance of the University of California, San Francisco (UCSF). All relevant ethical regulations were followed. The mutational status of EGFR or other known drivers or resistance was determined using either FoundationOne (Foundation Medicine) or UCSF pathology. Tissues were fixed in 10% formalin overnight and embedded in paraffin. Tissue sections of PDX and patient samples were sectioned on slides with 4-μm thickness. The paraffin sections were deparaffinized in xylene and rehydrated in a graded alcohol series, boiled with 10 mmol l−1 of citrate buffer (pH 6) for 10 min and treated with 0.3% H2O2 for 10 min. The steps were performed using the Envision two-step method using the Envision and DAB Color kit (Gene Tech). The TPX2 antibody (1:200 dilution; Sigma-Aldrich), cleaved caspase-3 (1:200 dilution; Cell Signaling Technology) and mouse monoclonal antibody Ki67 (1:100 dilution; Leica Biosystems) were used; PBS was used as the negative control. Images were captured using a ZEISS Axio Imager M1, and immunoreactivity was evaluated with IHC Profiler42 as an Image J plug-in in a blinded manner. The evaluation was based on staining intensity and the extent of staining. The staining area was scored using the following scale: 0, 0–10% of tissue stained positive; 1, 10–20% stained positive; 2, 20–40% stained positive; 3, 40–70% stained positive; and 4, >70% positive cells. The sum of staining score index (intensity + extent) was designated as follows: 0–2, negative expression; 3–4, strong expression. The IHC score was generated from three different areas of the slides and the average score was calculated for each sample.

Mouse xenograft studies

Cell line xenograft experiments were performed in female C.B-17 SCID mice aged 8 weeks by injecting 5 × 106 PC9-RR, PC9-OR tumor cells within 50% Matrigel gelatinous protein mixture (Corning). Tumors were allowed to grow until they reached a minimum volume of 150 mm3, and mice were randomized to receive treatment by oral gavage 7 d per week for 71 days. Rociletinib was formulated using 5% DMSO, 15% Solutol HS 15 (Sigma-Aldrich) in 80% water; osimertinib was formulated in 1% DMSO, 30% polyethylene glycol 300 and 69% water. MLN8237 was formulated using 10% (2-hydroxypropyl)-β-cyclodextrin in water. Tumor growth was assessed twice weekly by caliper measurements. A minimum of ten tumors per treatment group were assessed for the duration of the study. For PDX, patient-derived tumor cells were engrafted subcutaneously into the flank of C.B-17 SCID mice. Tumors were allowed to grow until they reached a minimum volume of 200 mm3; then, animals were randomly placed into control or treatment groups. Animals were treated daily for 30 d via oral gavage, and tumor volume was calculated daily using caliper measurements. The percentage change in tumor growth was based on volumes calculated from the size on day 1 at the beginning of treatment. All animal studies were conducted in accordance with the UCSF Institutional Animal Care and Use Committee, and all relevant ethical regulations were followed.

Real-time PCR

RNA was isolated according to the manufacturer’s instructions (TRIzol; Thermo Fisher Scientific); 1 μg of total RNA from each sample was subjected to first-strand complementary DNA synthesis according to the manufacturer’s recommendations (Promega). Quantitative PCR was performed on a CFX96 Real-Time PCR Detection System (Bio-Rad) with a PrimeTime Gene Expression Master Mix (Integrated DNA Technologies) according to the manufacturer’s protocol. TPX2 was amplified with the following primers: 5′-AGGGGCCCTTTGAACTCTTA-3′ (forward primer) and 5′-TGCTCTAAACAAGCCCCATT-3′ (reverse primer). 60S ribosomal protein L13a (L13aRPL13A) was used as an endogenous control with the following primers: 5′-CGGATTTGGTCGTATTGG-3′ (forward primer) and 5′-TCCTGGAAGATGGTGATG-3′ (reverse primer). AURKA was amplified with AGTTGGCAAACGCTCTGTCT (forward primer) and GTGCCACACATTGTGGTTCT (reverse primer). AURKB was amplified with TCCCTGTTCGCATTCAACCT (forward primer) and GTCCCACTGCTATTCTCCATCAC (reverse primer). AURKC was amplified with ACAACACCGGAACATCCTTC (forward primer) and TGCTGGTCCAACTTCTGATG (reverse primer). The cycling conditions were as follows: one cycle at 95 °C for 3 min; 40 cycles at 95 °C for 15 s and 60 °C for 60 s. The specificity of the PCR amplification was validated by the presence of a single peak in the melting curve analyses. Each real-time quantitative PCR experiment was repeated three times.

In vitro resistance assay

Single-cell-expanded PC9 and H1975 cells were plated in 96-well microplates at a density of 500 cells per well (~10% confluency), and drug treatment began the following day. Each treatment group had 12 replicates, and drug was replaced every 72–96 h. Each microplate was harvested at the end of day 1, 3, 5, 10, 15, 20, 30, 45, 60, 75 and 90 and cell proliferation was measured.

Reversibility of resistance

To test the reversibility of resistance, we seeded single-cell derived clones using a limited dilution method into 384-well microplates. Single-cell derived clones were allowed to expand in the absence or presence of drugs for 14 d. Once these single cells achieved about 80% confluence at the end of 14 d, a subset that had expanded in the absence of drug were tested for sensitization by adding 1 μM of respective EGFR TKI for 72 h. For each cell line, 96 single-cell clones were analyzed in each treatment condition.

Ras-guanosine triphosphate pull-down assay

Cells were washed twice in ice-cold PBS and lysed in 1% TX100-TNM lysis buffer (20 mmol /l−1 Tris, pH 7.5, 5 mmol l−1 MgCl2, 150 mmol l−1 NaCl, 1% Triton X-100) supplemented with 1 mmol l−1 dithiothreitol and protease and phosphatase inhibitors (Sigma-Aldrich). Equal amounts of protein from each sample were added to 10 µl of packed GST-coupled Ras binding domain of Raf (GST-Raf-RBD) or Ral GDS-Rap binding domain (Ral-GDS-RBD) beads in 300–500 µl of 1% TX100-TNM lysis buffer and rotated at 4 °C for 1–2 h. Beads were washed three times with 1 ml cold lysis buffer and boiled in lithium dodecyl sulfate sample buffer (Thermo Fisher Scientific).

Subcellular fractionation

Nuclear and cytoplasmic fractions were prepared using the NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Fisher Scientific) according to the manufacturer’s instructions. Protein concentration was quantitated using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific). Equal amounts of protein were loaded into each lane and separated on a 4–12% Bis-Tris gel (Thermo Fisher Scientific), then transferred onto a nitrocellulose membrane.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Electronic supplementary material

Supplementary information is available for this paper at https://​doi.​org/​10.​1038/​s41591-018-0264-7.

Acknowledgements

We thank members of the Bandyopadhyay laboratory for helpful discussions and technical assistance. We also thank J. Gordon from the LCA microscopy core for technical assistance and reagents. This work was supported by National Cancer Institute grant nos. U01CA168370 (S.B.), NIGMS R01GM107671 (S.B.), R01CA169338 (T.G.B) and U54CA224081 (S.B., T.G.B).

Competing interests

H.J.H., L.R., A.D.S. and T.C.H. are employees of Clovis Oncology. S.B. recieves funding and/or has a consultancy relationship with Ideaya Biosciences and Pfizer.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Literature

Related topics

Image Credits