Transcriptional activation of cyclin D1 via HER2/HER3 contributes to cell survival and EGFR tyrosine kinase inhibitor resistance in non-small cell lung carcinoma

Several different mechanisms are implicated in the resistance of lung cancer cells to epidermal growth factor receptor-tyrosine kinase inhibitors (EGFR-TKIs), and only few have been functionally investigated. Here, using genetically knocked out EGFR and TKI-resistant lung cancer cells, we show that loss of wild-type EGFR attenuates cell proliferation, migration and 3D-spherid formation, whereas loss of mutant EGFR or resistance to TKIs reinforces those processes. Consistently, disruption of wild-type EGFR leads to suppression of HER2/HER3, while mutant EGFR ablation or resistance to TKIs increases HER2/HER3 expression, compensating for EGFR loss. Furthermore, HER2/HER3 nuclear translocation mediates overexpression of cyclin D1, leading to tumor cell survival and drug resistance. Cyclin D1/CDK4/6 inhibition resensitizes erlotinib-resistant (ER) cells to erlotinib. Analysis of cyclin D1 expression in patients with non-small cell lung carcinoma (NSCLC) showed that its expression is negatively associated with overall survival and disease-free survival. Our results provide biological and mechanistic insights into targeting EGFR and TKI resistance.


Introduction
Lung cancer is the leading cause of cancer mortality worldwide. In 2018, more than 2 million new cases and 1.7 million lung cancer-related deaths were reported 1 . Non-small cell lung carcinoma (NSCLC) is the main type of lung cancer, accounting for 85% of all lung cancer cases 2 .
Surgery or surgery followed by chemotherapy is the common choice for patients with resectable lung cancer. For those with advanced stage cancer, targeted therapy can be used to achieve a longer progression-free survival 2 .
Tyrosine kinase inhibitors (TKIs), such as erlotinib, gefitinib, and afatinib, are the commonly used targeted agents for NSCLC patients with certain EGFR mutations 3,4 . EGFR-TKIs bind to the tyrosine kinase domain of EGFR, blocking tyrosine kinase activity and downstream signaling.
Approximately 10% of NSCLC patients from Western countries and 30% of NSCLC patients from East Asia, harboring EGFR-activating mutations (exon 19 deletions and L858R point mutation), could benefit from TKIs treatment 5,6 . However, inevitably, almost all patients treated with TKIs develop resistance within 9-13 months after treatment initiation 3,7,8 .
Mechanisms of EGFR-TKI resistance in NSCLC are complicated. The gatekeeper mutation T790M contributes to almost 50% of all acquired resistance to elortinib and gefitinib. Activation of compensatory signaling pathways, such as c-MET, PI3K/AKT/mTOR, JAK2/STAT3, are other mechanisms of resistance against TKIs 3,9,10 . Furthermore, dysregulation of HER family receptors and EGFR nuclear trans-location are reported as mechanisms of resistance 11 . EGFR nuclear translocation is considered a transcriptional co-activator of several well-known transcription factors, such as STAT3/5 12,13 , E2F1 14 and RNA helicase A 15 , regulating the expression of many essential genes, including cyclin D1. Although the function of nuclear EGFR is relatively clear, functions of other nuclear HER family receptors, including HER2/3/4, remain unknown in NSCLC.
In this study, we aimed to characterize cellular and molecular changes of HER family receptors upon CRISPR/Cas9-mediated disruption of EGFR or acquisition of resistance to EGFR-TKIs. We showed that wild-type EGFR ablation results in a substantial inhibition of cell proliferation and migration, but that disruption of mutant EGFR or acquisition of TKI resistance promoted cell proliferation and migration resulting in cell survival. Furthermore, we found that upregulation of HER2 and HER3 mediated cyclin D1 overexpression, contributing to cell survival and proliferation upon EGFR loss or emergence of TKI-resistance.
A549 and H1299 contain wild-type EGFR and are resistant to the erlotinib (an EGFR-TKI), while H1650 and HCC827 harbor an EGFR exon 19 deletion, which is the most common EGFR mutation and makes tumor cells sensitive to TKIs. We established H1650-ER and HCC827-ER cell lines after long-term exposure to erlotinib (Supplementary Fig. 1B and 1C). For HCC827ER, IC50 increased dramatically from 0.1 ±0.18 μM to 47.6 ±0.17 μM. H1650ER cells were able to grow in high dose (50 μM) of erlotinib without loss of viability, although the parental cells had moderate response to erlotinib. Furthermore, we observed an aggressive pattern of hyper-progression when H1650ER cells were treated with erlotinib ( Supplementary Fig. 1C).

Wild-type EGFR ablation attenuates cell proliferation, migration and 3D spheroid formation, but disruption of EGFR 19del or acquired resistance accelerates them
To elucidate the effect of EGFR ablation and TKI-resistance on proliferation and migration of lung cancer cells with different EGFR mutational statuses, we performed colony formation, trans-well, wound healing and 3D spheroid tumor formation assays.
In EGFR wt/wt cell lines (A549 and H1299), EGFR ablation significantly inhibited cell proliferation, vertical and horizontal migration and tumor formation. Cell proliferation of EGFR-KO cells decreased by 50% compared to the parental cells (Fig. 1A). Similarly, impaired ability of migration was observed in trans-well and wound healing assays ( Fig. 1B and 1C). In 3D-spheroid tumor formation assay, original A549 and H1299 cells formed larger tumors and showed stronger invasion ability than corresponding EGFR -/cells (Fig. 1D). In contrast, we observed opposite results in EGFR 19del and erlotinib-resistant cells. In H1650 and HCC827, EGFR 19del ablation or acquired erlotinib resistance led to acceleration of cell proliferation, vertical and horizontal migration, and 3D-spherid formation (Fig. 1A-D).

Distinct changes in pERK(1/2) and pAKT levels upon EGFR knockout and acquired TKI
resistance EGFR status can significantly affect two main downstream pathways, i.e. the MAPK/pERK(1/2) and PI3K/AKT. EGFR-activating mutations, overexpression or inhibition can dramatically affect these two pathways and substantially influence cancer cell growth and survival. To precisely investigate the effect of different EGFR status on these two pathways, we performed a panel of western blot to detect changes in the levels of phosphorylated ERK (pERK)(1/2) and phosphorylated AKT (pAKT) in different NSCLC cell lines.
We found significant changes in pERK(1/2) and pAKT levels upon EGFR ablation or TKI treatment.

EGFR wt/wt disrupted cells
Dysregulation of HER family receptors can occur in patients with NSCLC upon TKIs treatment 11 .
Precise evaluation of HER family receptor expression levels upon and during treatment with TKIs can provide deeper understanding of mechanisms of TKI resistance. Therefore, we investigated alterations in HER2 and HER3 expression levels upon EGFR ablation and EGFR-TKI resistance.
In EGFR wt/wt cell lines (A549 and H1299), EGFR disruption resulted in a slight decrease in HER2 and HER3 levels ( Fig. 3A and 3B). However, cells with mutant EGFR, i.e. H1650 and HCC827, showed a different pattern. HER2 and HER3 were dramatically upregulated in H1650 EGFR -/and H1650-ER cells at both mRNA and protein level compared to the parental cells ( Fig. 3A and 3B).
In HCC827-ER cells, only HER2 protein level was upregulated (Fig. 3B). Although HER3 was significantly upregulated in HCC827-ER at RNA level, this was not reflected at protein level, suggesting that post-transcriptional regulation mechanisms may be involved.
In addition, we evaluated pSTAT3 levels, since it is an important downstream protein of HER family receptors. In EGFR wt/wt cell lines, pSTAT3 was undetectable (Fig. 3B), while pSTAT3 level was significantly increased in HCC827ER compared to HCC827 cells (Fig. 3B). Together, these results suggest that upregulation of HER2 and HER3 occurs only in mutant EGFR knockout and TKI-resistant NSCLC cell lines, but not in wild-type EGFR knockout cells.

Pan-HER family inhibitor, afatinib, suppresses pAKT via inhibition of HER2 and HER3
HER2 and HER2-containing heterodimers have the strongest kinase activity compared to other HER family receptors, which makes them an important therapeutic target in patients with cancer. Afatinib not only can block EGFR, but also inhibits the tyrosine kinase activity of HER2 and other HER family receptors. Since we showed that HER2 and HER3 were upregulated in EGFR 19del disrupted and ER cells, we speculated that suppression of HER2 and HER3 may further inhibit tyrosine kinase activity and tumor proliferation. Therefore, we treated H1650 EGFR -/and HCC827ER cells with afatinib to block HER2-and HER3-mediated signaling. We observed a decreased level of pAKT upon afatinib treatment in H1650 EGFR -/and HCC827-ER cells ( Fig. 4A and 4B). However, we did not observe significant inhibition of cell proliferation, although the downstream pAKT signaling was decreased by afatinib. Thus, these results suggest that overexpression of HER2 and HER3 may play other functions than tyrosine kinase activity in H1650 EGFR -/and HCC827-ER cells.

HER2 and HER3 are overexpressed in the nuclei of EGFR 19del ablated and ER cells
EGFR nuclear translocation is one of the mechanisms of resistance to TKI in NSCLC. Whether nuclear translocation of other HER family receptors is also involved in EGFR-TKIs resistance is unclear. Hence, we assessed HER2 and HER3 expression levels in non-nuclear and nuclear fraction in our NSCLC cell line panel. We found that HER2 and HER3 were mainly expressed in the nuclei in all tested NSCLC cell lines ( Fig. 5A-E). Ablation of EGFR wt/wt influenced neither HER2 and HER3 expression, nor their nuclear-membrane distribution in A549 cells (Fig. 5A). Similarly, we did not observe significant changes of HER2 and HER3 in H1299 EGFR wt/wt and EGFR -/cells due to the relatively low expression of protein (Fig. 5A). Interestingly, HER2 and HER3 were both significantly overexpressed in the nuclei of EGFR 19del disrupted and ER cells compared to the parental cells H1650 and HCC827 ( Fig. 5C-E).

Transcriptional complex of HER2/HER3/pSTAT3 increases in nuclei, enriches in the cyclin D1
promoter region and reinforces cyclin D1 expression in EGFR 19del ablated and ER cell lines HER2/HER3/Stat3 can form a transcriptional complex and stimulate cyclin D1 expression in breast cancer 20,21 . However, this mechanism has not yet been confirmed in lung cancer.
Therefore, we performed co-immunoprecipitation assays to explore this mechanism. Our results showed that in EGFR wt/wt cell lines, EGFR deletion did not influence HER2/HER3/pSTAT3 heterotrimer formation in the nucleus ( Fig. 6A and 6B). However, in EGFR-mutant cell lines, EGFR ablation or ER led to HER2/HER3/pSTAT3 nuclear enrichment ( Fig. 6 C and 6D).
In the next step, we performed sequential ChIP assay to detect the interactions between HER2/HER3/pSTAT3 and the cyclin D1 (CCND1) promoter. We found that HER2/HER3 complex bound more to the CCND1 promoter in HCC827ER and H1650 EGFR -/cells than in original cell lines (Fig. 7A). Conversely, less binding was observed in EGFR wt/wt ablated cells than in parental cell lines (Fig. 7A). Consequently, higher cyclin D1 mRNA levels were observed in EGFR 19del disrupted and ER cells compared to both parental and wild-type EGFR cells (Fig. 7B).
Furthermore, western blotting showed that EGFR wt/wt ablation downregulated cyclin D1, while disruption of EGFR 19del or erlotinib resistance increased its expression (Fig. 7C). Collectively, deletion of EGFR reduced affinity of HER2/HER3 complex to cyclin D1 promoter, decreased cyclin D1 expression in EGFR wt/wt cells, but reinforced it in cells with mutant EGFR.

Ribociclib resensitizes cells to erlotinib
Although no drug is available to directly and specifically inhibit cyclin D1, an inhibitor of cyclin D1/CDK4/6 (ribociclib) has been approved by FDA in March 2017 for breast cancer treatment.
We asked whether ribociclib can resensitize ER cells to erlotinib treatment. We treated EGFR 19del cells (H1650ER and HCC827ER) with either erlotinib or ribociclib or combinations thereof. We observed a significant combinatorial effect of ribociclib and elortinib in H1650ER and HCC827ER cells colony formation ( Fig. 8A and 8B).
To mimic in vivo tumor growth, we further treated H1650ER and HCC827ER cells using a 3D spheroid model. Consistently, a significant combinatorial effect of ribociclib and elortinib was observed in H1650ER 3D-spheroid tumor (Fig. 8C), whereas HCC827ER cells are not able to form 3D spheroids with treatment.

High expression of CCND1 is correlated with lower overall survival and disease-free survival rates in patients with lung cancer
To validate the clinical relevance of our results, we analyzed gene expression data from patients with lung cancer retrieved from The Cancer Genome Atlas (TCGA) database. Based on Kaplan-Meier survival analysis, NSCLC patients with higher cyclin D1 had a significantly lower overall survival (OS) and disease-free survival rates (DFS) (Fig. 9A). The median DFS and median OS were significantly shorter in patients with higher CCND1 expression than those with lower gene expression (2.58 vs 3.55 years, p=0.013; 3.52 vs 5.16 years, p=0.038).
Furthermore, there was a significant positive correlation between HER2 and cyclin D1 expression ( Fig. 9 B), which was stronger than EGFR/ cyclin D1 or HER3/cyclin D1 correlation.
Taken together, these results suggest that cyclin D1 may be a promising therapeutic target for patients with NSCLC.

Discussion
Epidermal growth factor receptor is a key therapeutic target for patients with Non-Small Cell Lung Cancer. Although the majority of patients with NSCLC harboring EGFR activating mutations benefit from EGFR-TKI treatment, acquired resistance inevitably arises. Extensive efforts have led us to the identification of several resistance mechanisms, including secondary and tertiary mutations, alternative pathway activation, cMET amplification, HER2 amplification, EGFR nuclear trans-localization, oncogenic lncRNA regulation, and tumor microenvironment alterations 3,8,25,26,27 . The underlying mechanisms of EGFR-TKI resistance are not fully understood, mainly due to lack of approaches for complete elimination of EGFR in cells. Although EGFR nuclear translocation is one of the resistance mechanisms, whether HER2/3/4 nuclear translocation contributes to TKI resistance remains largely unknown. Therefore, we aimed to characterize a panel of EGFR knockout NSCLC cell lines and further to investigate other HER family receptors upon EGFR loss or acquired resistance. We show that cells with mutant EGFR behave differently than wild-type cells after EGFR disruption. We demonstrate that HER2 and HER3 overexpression and transcriptional activation of cyclin D1 are other resistance mechanisms to TKIs in NSCLC (Fig. 10).
Our results show that wild-type EGFR ablation resulted in a significant inhibition of cell proliferation, migration and 3D spheroid formation, but EGFR 19del ablation increased those processes. Wildtype EGFR plays an essential role in cell growth by a ligand-dependent mechanism. Therefore, knockout of EGFR wt/wt leading to the downregulation of those abilities is not a surprise. Although EGFR exon 19 deletion is an activating mutation that can promote cell growth through a ligand-independent activation of tyrosine kinase domain 28 , disruption of mutant EGFR in H1650 cell line promoted cell growth and migration abilities. We indeed observed that the acute inhibition of EGFR using TKIs in mutant cells resulted in a dramatic suppression of cell proliferation. However, EGFR 19del cell lines with long-term exposure to elortinib (HCC827-ER and H1650-ER) developed an even stronger proliferative ability and aggressiveness. These data suggest that EGFR 19del knockout may simulate EGFR-TKI resistance. This is in agreement with the fact that drug resistance leads to tumor hyper-progression 29 . A recent study provided a proof-of-concept of therapeutic benefits of EGFR knockout in EGFRmutant NSCLC 30 . However, our study brings some concerns whether it is safe to remove mutant EGFR from tumor cells.
We observed significant changes in the levels of pERK(1/2) and pAKT upon EGFR ablation, inhibition, or acquired resistance. Firstly, pAKT levels were higher in A549 EGFR -/-, H1650 EGFR -/and HCC827-ER than in parental cells. Our data suggest that activation of the PI3K/AKT pathway may be an important compensation mechanism for tumor cell survival after EGFR loss. In agreement with our findings, a recent study showed that Akt activation drives acquired resistance to EGFR-TKI 31 . However, our study further revealed that activated Akt induced by EGFR ablation can be further decreased by elortinib treatment. In addition, pERK level significantly increased only in A549 EGFR -/cells upon elortinib treatment, which might be due to off-target effects of erlotinib.
Our data revealed that HER2 and HER3 were significantly overexpressed in EGFR 19del ablated and ER cells, but downregulated in EGFR wt/wt disrupted cells. Dysregulation of HER family receptors has been reported in patients treated with EGFR-TKIs and monoclonal antibodies, but the underlying mechanisms are not fully understood 32 . We showed that upregulation of HER family receptors upon TKIs treatment or EGFR ablation only occurred in EGFR-mutant tumor cells, which might be of importance in a clinical setting. Moreover, our findings, to our knowledge for the first time, further revealed a significant nuclear overexpression of HER2 and HER3 in NSCLC EGFR 19del ablated or ER cells. Due to lack of an efficient approach for targeting nuclear protein, these findings may partially explain ineffectiveness of combination therapies by anti-EGFR and other HER family receptors.
We found a new mechanism of tumor cell survival through HER2/HER3/STAT3 complex binding and transcriptional activation of cyclin D1 in EGFR targeting in NSCLC (Fig. 10). Although cyclin D1 overexpression mediated by cyclinD1/CDK4/6 signaling and HER2/HER3/STAT3 nuclear translocation has been identified as an important resistance mechanism and therapeutic target in breast cancers 21,[33][34][35] , whether they also occur in lung cancer was unknown. Firstly, we showed that HER2 and HER3 have nuclear overexpression in mutant EGFR knockout and ER cell lines, which may compensate for EGFR loss. Secondly, we observed that HER2, HER3 and pSTAT3 form more heterotrimers in nuclei upon mutant EGFR ablation and erlotinib resistance.
Thirdly, HER2/3/pSTAT3 complexes were enriched and bound to cyclin D1 promoter and subsequently activated cyclin D1 expression only after mutant EGFR knockout and erlotinib resistance. This resulted in an increased in S and G2/M phase cell population. However, we observed an opposite result in cells with wild-type EGFR. Furthermore, ribociclib was able to resensitize cells to erlotinib. Thus, our findings demonstrate that cyclin D1 overexpression mediated by HER2/3 nuclear translocation is a mechanism of TKI resistance in NSCLC. In addition, cyclin D1 high overexpression is associated with shorter overall survival and diseasefree survival. Cyclin D1 might be a promising therapeutic target for patients with resistant NSCLC. Concomitant inhibition of EGFR and cyclin D1 may result in more prominent outcomes.

Conclusions
In conclusion, our results show a distinct expression pattern of HER family receptors after EGFR loss in different NSCLC cells and it appears to be a mechanism of resistance to TKIs in NSCLC.

Moreover, inhibition of wild-type and mutant EGFR showed distinguished impacts on other HER
family receptors as well as tumor survival. In addition, cyclin D1 may be a potential therapeutic target for patients with resistant NSCLC.

Cell culture and establishment of EGFR knockout (EGFR -/-) cell lines by CRISPR/Cas9
Four lung cancer cell lines (A549, H1299, H1650 and HCC827) were used in this study. A549 (EGFR wt/wt ) was purchased from ATCC. H1299 (EGFR wt/wt ) and H1650 (EGFR 19del  CO2. All human cell lines have been authenticated using STR (or SNP) profiling within the last three years. All experiments were performed with mycoplasma-free cells.

Anti-tumor agents
Erlotinib and afatinib were purchased from Selleckchem (Munich, Germany). Ribociclib was purchased from LC laboratories (Woburn, USA). All drugs were dissolved in dimethyl sulfoxide (DMSO) and stored at -20°C.

Establishment of EGFR-TKI resistant cell lines
EGFR-TKI resistant cell lines were established by long-term exposure to erlotinib (3 months). To set up EGFR TKI-resistant cell lines, HCC827 EGFR 19del and H1650 EGFR 19del were treated with erlotinib starting at 50 nM, followed by a stepwise dosage increase every 5 days up to 20 μM and 50 μM, respectively. The erlotinib-resistant (ER) cell lines HCC827ER and H1650ER were maintained in 2 μM erlotinib for long-term culture.

Cell viability
Cell viability was measured using the MTS assay. Cells were seeded in flat bottom 96-well plates at a density of 1x10 3 cells per well. After 24 hr, cells were treated with appropriate drugs at different doses for another 72 hr. Subsequently, CellTiter 96 Aqueous One Solution reagent (Promega, Madison, USA) was added to each well, according to manufacturer's instruction.
Plates were incubated at 37°C for 1hr. The optical density (OD) was determined at 490 mm wavelength using a Synergy H1 plate reader (BioTek, Winooski, USA).

Cell proliferation
The colony formation assay was used to determine cell proliferation. Cells were seeded in 6-well plates at a density of 1x10 4 cells per well and were allowed to grow for 4-5 days. Subsequently, cells were fixed with 4% (v/v) paraformaldehyde for 20 min at room temperature (RT) in the dark. After washing with phosphate-buffered saline (PBS), cells were stained with 0.5% (w/v) crystal violet for 15 min. Then, cells were washed three times with demineralized water and plates were dried at room temperature. Crystal violet was eluted by 10% (v/v) acetic acid and OD was measured at 590 mm wavelength using a Synergy H1 plate reader.

Cell migration
To measure lateral migration, cells were seeded in a 6-well plate at a density of 7x10 5 cell per well. After 24 hr incubation, a wound was gently scraped using a sterile 200 μl pipette tip. The detached cells were washed away with PBS. Pictures were taken at 0 hr, 12 hr, 24 hr, 48 hr using a CK2 inverted microscope (Olympus, Tokyo, Japan). The pictures were analyzed by ImageJ (National Institute of Health, USA).
To measure vertical migration, cells were seeded in the top trans-well inserts at a density of 1x10 4 cells per insert. The inserts were gently put onto 24-well plates filled with 750 μl cell culture medium. Culture medium and remaining cells in the inserts were removed after 24 hr incubation at 37°C with 5% CO2. The insert membrane was fixed with 4% (v/v) paraformaldehyde and stained with 0.5% (w/v) crystal violet. Pictures of migrated cells were randomly taken using a CK2 inverted microscope. Subsequently, crystal violet was eluted with 10% (v/v) acetic acid and the OD590 value was measured using a Synergy H1 plate reader.

Generation of a 3D-spheroid tumor model
To mimic actual tumor microenvironment, a 3D-spheroid tumor model was constructed, which has been detailed previously 17 . The U-bottom ultra-low attachment culture plate was used for the generation of a 3D-spheroid tumor model. Approximately 2x10 3 cells per well were seeded into U-bottom ultra-low attachment 96-well plates. Plates were centrifuged at 300 g for 5 min to form a preliminary 3D spheroid tumor. After 5 days of incubation, 3D spheroid tumors were generated. The pictures were taken under the CKX41 microscope (Olympus, Tokyo, Japan) controlled by Leica Application Suite software. The image data were analyzed by ImageJ.

RNA isolation, cDNA synthesis and qRT-PCR
Cells were harvested by trypsinization and RNA was isolated using Maxwell LEV simply RNA

Western blot analysis
Cells were harvested by trypsinization and lysed on ice in RIPA buffer containing 1x PhosSTOP and protease inhibitor (PI) cocktail (Roche, Mannheim, Germany) for 20 min. The lysates were sonicated for 10-20 seconds to reduce the viscosity. Next, samples were centrifuged at 16,000 RCF, 4°C for 20 min to remove insoluble residues. Protein concentration was determined using the BCA Protein Assay Kit (Pierce, Rockford IL, USA), according to the manufacturer's instruction.
Thirty micrograms denatured protein of each sample was mixed with SDS buffer and loaded onto a pre-cast 4-12% NuPAGE Bis-Tris gel (Invitrogen, USA). Later, proteins were transferred to a polyvinylidene difluoride (PVDF) membrane. After blocking with 5% (w/v) skimmed milk in PBST (Phosphate Buffered Saline with Tween 20) at room temperature for 1 hr, the membrane was incubated overnight at 4°C with appropriate primary antibody. Then, the membrane was treated with HRP-conjugated secondary antibody for 1 hr at RT. After each step the membrane was washed three times with PBST. Blots were lighted up with enhanced chemiluminescence (ECL) solution (GE Healthcare, Amersham, UK). Pictures were captured using a chemi genius II bio-imaging system.

Nuclear isolation and co-immunoprecipitation (co-IP)
The nuclear isolation and co-immunoprecipitation (co-IP) assay has been detailed elsewhere 18,19 .
Briefly, 3x10 8 Cells were harvested under non-denaturing conditions for nuclear isolation, followed by washing with PBS. Cells were resuspended in 400 μl hypotonic lysis buffer and incubated on ice for 20 min. Nuclei were pelleted by centrifugation at 500 g, 4°C for 10 min.
Supernatant contained the non-nuclear fraction and the pellet was the nuclear fraction.
For co-IP, samples were gently washed three times with clod PBS containing 0.1% CA-630 and 1x protease inhibitor (PI) to minimize cytoplasmic contamination. Nuclear protein samples were extracted using 300 μl RIPA buffer containing 1x PhosSTOP and PI. After 20 min incubation on ice, samples were sonicated and centrifuged at 16,000 RCF, 4°C for 20 min. The concentration of non-nuclear and nuclear protein samples was determined by BCA Protein Assay Kit, according to the manufacturer's instruction. Approximately 300 μg of cell lysate at 1-1.5 mg/ml was then pre-cleaned by rotation incubation with 20 μl 50% Protein A/G PLUS-Agarose (Santa Cruz Biotechnology, Texas, USA) at 4°C for 60 min. Subsequently, appropriate antibody was added to pre-cleaned samples, followed by overnight rotation incubation at 4°C. Next, 30 ul 50% Protein A/G PLUS-Agarose was added to each sample. The mixtures were incubated at 4°C for overnight with rotation. The pellets were collected by centrifugation at 1000 g for 3 min at 4°C and washed five times with 500 μl RIPA buffer. The pellets were well mixed with 20 μl 4x SDS buffer and denatured by heating at 95°C for 10 min. Denatured protein samples were analyzed by western blot.

Kaplan-Meier survival and correlation analysis of patient data
Data for survival analysis was retrieved from The Cancer Genome Atlas (TCGA) using a standard processing pipeline. Cumulative disease free survival of NSCLC (LUAD, LUSC) was analyzed using the Kaplan-Meier method with log-rank test. Correlation between the expressions of different genes was calculated using Spearman correlation (non-log scale) based on Gene Expression Profiling Interactive Analysis (GEPIA) 23 . A two-sided p-value less than 0.05 was considered statistically significant.

Data availability
The data that support the findings of this study are available from the corresponding author upon reasonable request.

Consent for publication
Not applicable.      Co-Immunoprecipitation analysis of HER2/3/pSTAT3 heterotrimer from A549 and A549 EGFR -/nuclear lysates. B. Co-Immunoprecipitation analysis of HER2/3/pSTAT3 heterotrimer from H1299 and H1299 EGFR -/nuclear lysates. C. Co-Immunoprecipitation analysis of HER2/3/pSTAT3 heterotrimer from H1650 and H1650 EGFR -/nuclear lysates. D. Co-Immunoprecipitation analysis of HER2/3/pSTAT3 heterotrimer from HCC827 and HCC827ER nuclear lysates. Fig. 7. HER2 and HER3 complex binding to cyclin D1 promoter and cyclin D1 expression levels upon EGFR loss or erlotinib resistance in different NSCLC cell lines. A. Enrichments of HER2 and HER3 complex binding to cyclin D1 promoter detected by sequential ChIP and qRT-PCR. Chromatins were first immunoprecipitated with HER2 antibody and were reimmunoprecipitated using HER3 antibody. Primer set for sequential ChIP was designed to amplify the fragment of cyclin D1 promoter including the binding site of HER2/3/pSTAT3 complex. B. Relative mRNA expression of cyclin D1. C. Western blot analysis of cyclin D1 expression in different NSCLC cell lines.

Fig. 8. Effects of the treatment with Elortinib and/or Ribociclib in H1650ER and HCC827ER cells.
A. erlotinib or ribociclib or the combinations thereof for HCC827ER. B. erlotinib or ribociclib or the combinations thereof for H1650ER. D. erlotinib or ribociclib or the combinations thereof for H1650ER-3D spherid tumor.