Abstract

   Epidermal growth factor receptor (EGFR)-tyrosine kinase inhibitors
   (TKIs) are the most applied targeted therapy for EGFR-mutant lung
   adenocarcinoma (LUAD). The third-generation EGFR-TKI, osimertinib, is
   widely used throughout lung cancer treatment, with single or
   combination modes. One of the main barriers in osimertinib treatment is
   the acquired resistance and mechanisms are not fully understood. Gene
   expression other than genetic mutations might predict drug response and
   mediate resistance occurrence. We analyzed six datasets of
   osimertinib-resistant LUAD cells from the Gene Expression Omnibus (GEO)
   database and identified two hub genes, named MRAS and HEG1. We found
   that the expression mode of MRAS/HEG1 in LUAD was osimertinib-dependent
   and contributed to drug resistance. We also explored potential
   mechanisms of hub genes related osimertinib resistance and emphasized
   the M2 infiltration involved. Moreover, potential therapeutic agents
   conquering MRAS/HEG1-related resistance were also identified. In
   conclusion, MRAS and HEG1 might be responsible for osimertinib
   resistance and could be promising prognostic biomarkers for osimertinib
   response in LUAD, which might provide insights into therapeutic
   strategies.

Supplementary Information

   The online version contains supplementary material available at
   10.1007/s12672-024-01552-6.

   Keywords: Lung adenocarcinoma (LUAD), Osimertinib resistance,
   Biomarker, Muscle RAS (MRAS), Heart of glass 1 or EF-like domains 1
   (HEG1)

Introduction

   Lung cancer is the cancer with the highest morbidity and mortality in
   China and LUAD is the most common type of non-small cell lung cancer
   (NSCLC) [[36]1]. The epidermal growth factor receptor (EGFR) is one of
   the common alterations in NSCLC patients and its frequency ranges from
   11.9% (global) to 49.1% (Asia) [[37]2]. Studies from multiple centers
   in China show the frequency of EGFR mutation is 39.0–57.7% [[38]3]. The
   most frequent somatic mutations are in the kinase domain of EGFR and
   cause the constitutive activation independent of ligand, promoting the
   survival and proliferation of cancer cells [[39]4]. Most EGFR mutations
   comprise of exon 19 deletion and exon 21 L858R point mutation that is
   sensitive to first- (gefitinib and erlotinib) or second-generation
   (afatinib and dacomitinib) EGFR tyrosine kinase inhibitors (TKIs)
   [[40]5]. The third-generation EGFR TKI osimertinib showed superior
   efficacy in LUAD patients with EGFR mutation, especially acquired EGFR
   T790M-related resistance [[41]6]. However, the majority of LUAD
   patients who initially reported an osimertinib response will ultimately
   develop disease progression, which makes it crucial to identify
   biomarkers that predict resistance and guide treatment adjustments.

   The occurrence of acquired EGFR mutations (e.g., C797S) [[42]7] and
   loss or maintenance of the EGFR T790M mutation [[43]8] have been
   reported in tumors of patients with acquired osimertinib resistance.
   Additionally, off-target genetic alterations and amplification, as well
   as the activation of other signaling pathways have been discovered in a
   majority of LUAD patients with acquired osimertinib resistance [[44]2].
   Increasing evidence has shown that the high proportion of
   osimertinib-resistant LUAD don’t harbor the new occurrence of genetic
   alterations [[45]9]. Recent studies demonstrated that EGFR-independent
   mechanisms without gene mutation include the activation of insulin-like
   growth factor (IGF)-1 receptor and Src/Akt [[46]10, [47]11]. However,
   the detailed resistant mechanisms remain unclear. A comprehensive
   understanding of acquired osimertinib-resistant mechanisms is essential
   to inform future therapeutic decisions for patients with EGFR mutation
   advanced LUAD.

   In the present study, we aimed to identify the osimertinib-resistant
   biomarkers and explored related mechanisms. We first analyzed six
   datasets of osimertinib-resistant LUAD cells and identified the hub
   genes associated with osimertinib resistance. The expression levels of
   muscle RAS (MRAS) and heart of glass 1 or EF-like domains 1 (HEG1) were
   verified in the published RNA-seq data online and tumor tissues
   collected from the osimertinib-sensitive/resistant patients in our
   hospital. We also assessed the prognostic value of MRAS and HEG1 in
   predicting the survival of LUAD. Additionally, we established
   osimertinib-resistant (OR) clones and conducted a series of experiments
   in vitro. We found that the expressions of MRAS and HEG1 in LUAD were
   in a dose/time-dependent manner until the osimertinib resistance
   developed. Molecular elucidations of MRAS and HEG1 based on the KEGG
   pathway and immune analysis were essential for understanding the
   osimertinib resistance occurrence. Finally, we discussed the prospect
   of clinical agents conquering the MRAS/HEG1-related
   osimertinib-resistant LUAD by performing oncoPredict analysis and in
   vitro crystal violet viability assay.

Results

Identification of the DEGs/hub genes by two bioinformatics in acquired
osimertinib-resistant LUAD cells

   To explore the hub genes associated with acquired osimertinib
   resistance in LUAD, six datasets of LUAD cells sensitive or resistant
   to osimertinib were downloaded from the GEO database and subsequently
   processed for DEGs analysis. The results showed that there were 437
   DEGs (GSE1033350), 2610 DEGs ([48]GSE153183), 125 DEGs ([49]GSE106765),
   491 DEGs ([50]GSE163913), 2539 DEGs ([51]GSE201608), and 743 DEGs
   ([52]GSE193258) (Fig. [53]1A and Table S1). The DEGs in the six
   datasets were further analyzed for GO enrichment and pathway enrichment
   analysis (GSEA). The top 10 GO terms and the top 15 pathways are shown
   in Fig. [54]1B and C, respectively. The DEGs were enriched in the
   pathways related to drug metabolism and resistance, such as the
   metabolism of xenobiotics by cytochrome P450 (hsa00980), drug
   metabolism-cytochrome P450 (hsa00982), and DNA replication (hsa03030)
   (Fig. [55]1C, Tables S2 and S3). Additionally, two genes (MRAS and
   HEG1) were identified as increasing in osimertinib-resistant cells by
   intersecting the DEGs in the six datasets using the Venn plot
   (Fig. [56]1D).

Fig. 1.

   [57]Fig. 1
   [58]Open in a new tab

   Identification and analyses of DEGs in six datasets of
   osimertinib-resistant LUAD cells. A The volcano plots of the
   differential gene expressions in six datasets of osimertinib-resistant
   NSCLC cells. The -log10 (P-values) (Y-axis) is plotted against the
   average log2 (Fold Change) (X-axis) in gene expressions. The log2 (FC)
   and P-value are set at > 1 and < 0.05, respec-tively. B GO enrichment
   analysis based on the DEGs of six GEO datasets. The corresponding terms
   with GO number are in Table S2. C GSEA analysis based on the DEGs of
   GEO six datasets. The corresponding terms with KEGG number are in Table
   S3. D The Venn diagram shows the intersection of the DEGs in the six
   GEO datasets. Two genes are spotted

   Six datasets were merged to create a combined dataset which was used
   for DEGs analysis. First, the batch effects were eliminated from the
   combined matrix (Fig. [59]2A). The analysis revealed that there were
   128 downregulated DEGs and 158 upregulated DEGs between osimertinib
   sensitive and resistant LUAD cells (Fig. [60]2B and C). The functional
   enrichment analysis showed that 158 DEGs related to osimertinib
   resistance were significantly associated with 195 GO terms and 66 KEGG
   pathways (Table S4 and S5). The top 10 GO terms and top 20 KEGG
   pathways are shown in Fig. [61]2D and E, respectively. The top five
   enriched pathways were cell cycle (hsa04110), spliceosome (hsa03040),
   ribosome biogenesis in eukaryotes (hsa03008), nucleocytoplasmic
   transport (hsa03013), and nucleotide metabolism (hsa03008). Genomic
   instability is known as one of the hallmarks of cancer. These findings
   may provide a new insight into osimertinib resistance in LUAD by
   examining DNA replication and repair.

Fig. 2.

   [62]Fig. 2
   [63]Open in a new tab

   Identification and analyses of DEGs from the combined datasets of
   osimertinib-resistant NSCLC cells. A Principal component analysis of
   six datasets before (upper panel) and after (low panel) elimination of
   batch effects. B The volcano plots of the differential gene expressions
   in the merged dataset of osimertinib-resistant NSCLC cells. C The
   -log10 (P-values) (Y-axis) is plotted against the average log2 (Fold
   Change) (X-axis) in gene expressions. The log2 (FC) and P-value are set
   at > 1 and < 0.05, respectively. D GO enrichment analysis based on 284
   DEGs of the merged dataset. The top 10 GO terms are shown. The
   corresponding terms with GO number are in Table S4. E GESA analysis
   based the concordant differences genes (P < 0.05) of the merged
   dataset. The top 20 terms are shown. The corresponding terms with KEGG
   number are in Table S5

MRAS and HEG1 are the hub genes in LUAD with acquired osimertinib-resistance

   As shown in Fig. [64]3A, MRAS and HEG1 were also identified as the DEGs
   in the merged dataset. We examined the expression of MRAS and HEG1 in
   the six GEO datasets. The heatmap analysis revealed that the expression
   levels of both genes in osimertinib-resistant LUAD cells were higher
   than those in osimertinib-sensitive LUAD cells (Fig. [65]3B). We also
   observed higher expression of MRAS and HEG1 in acquired TKI-resistant
   LUAD tumors by analyzing the online transcriptome sequencing data
   [[66]12] (Fig. [67]3C). To further confirm the clinical significance of
   the MRAS and HEG1 expression in predicting response to osimertinib, we
   conducted immunohistochemical (IHC) analysis on tumor tissues collected
   from LUAD patients. The patients were assigned to the sensitive
   subgroup and the resistant subgroup. The tumor biopsies of the
   sensitive subgroup were surgically taken from LUAD patients who carried
   EGFR L858R before any treatment. The tumor biopsies of resistant
   subgroup were collected from LUAD patients who had disease progression
   after achieving osimertinib treatment for 12–23 months. The
   clinicopathological information of eight patients was shown in Table
   S6. We found that LUAD patients with acquired osimertinib resistance
   showed significantly higher expression of MRAS and HEG1, compared to
   LUAD patients prior to treatment (Fig. [68]3D and E). Moreover,
   patients with high levels of MRAS had poor clinical results in LUAD
   (P = 0.027) (Fig. [69]3F). As shown in Fig. [70]3G, higher expression
   of HEG1 is prone to shorter lifetimes but with no statistical
   differences (P = 0.15), which needs larger sample sizes to verify
   further (Fig. [71]3G). The above results indicate that MRAS has a good
   prediction performance in osimertinib resistance and prognosis,
   however, HEG1 seems only can be an osimertinib resistance predictor in
   LUAD. More data about the MRAS and HEG1 expressions after osimertinib
   treatment in LUAD from the real world are demanded to validate the
   predictive accuracy.

Fig. 3.

   [72]Fig. 3
   [73]Open in a new tab

   MRAS and HEG1 are highly expressed in osimertinib-resistant cells and
   tissues. A The Venn plot of DEGs from the merged dataset and six
   datasets. B The heatmap of MRAS and HEG1 in osimertinib-sensitive and
   resistant LUAD cells from six GEO datasets. C Bare mean of counts of
   MRAS and HEG1 from the online RNA-seq data [[74]9]. D IHC analysis of
   MRAS in tumor biopsies from LUAD patients (left panel). The DAB
   positive signals were analyzed by the Halo data analysis system (right
   panel). E IHC analysis of HEG1 in tumor biopsies from LUAD patients
   (left panel). The DAB positive signals were analyzed by the Halo data
   analysis system (right panel). F-G The Kaplan–Meier plots of the
   survival analysis of LUAD patient data from the [75]GSE14814
   dataset (N = 265)

MRAS and HEG1 expressions in LUAD are in osimertinib used-dependent manner
until the resistance develops

   Next, we investigated the relationship between MRAS and HEG1
   expressions and osimertinib sensitivity. We analyzed the expression in
   the tumor and adjacent normal tissues from the TCGA cohort. As shown in
   Fig. [76]4A and B, adjacent normal tissues showed higher expression
   levels of both MRAS and HEG1 compared to that in tumor tissues, which
   hinted the baseline levels of MRAS and HEG1 expressions in LUAD were
   low. We further examined the expression levels of MRAS and HEG1 in
   NCI-H1975 cells treated with osimertinib. We found that both mRNA and
   protein expression of MRAS and HEG1 were elevated in a dose-dependent
   manner (Fig. [77]4C–E). NCI-H1975 was treated with or without 10 nM of
   osimertinib at different time points. The qRT-PCR results showed that
   osimertinib induced MRAS and HEG1 expression in a time-dependent way,
   starting at two days after treatment (Fig. [78]4F and G).
   Immunoblotting data showed that MRAS and HEG1 proteins were
   significantly increased three days after treatment (Fig. [79]4H).

Fig. 4.

   [80]Fig. 4
   [81]Open in a new tab

   MRAS and HEG1 are in osimertinib use-dependent manner in LUAD. A The
   paired expression of MRAS in the LUAD cohort from the TCGA dataset. B
   The paired expression of HEG1 in the LUAD cohort from the TCGA dataset.
   C NCI-H1975 cells were treated with indicated concentations of
   osimertinib for 72 h. QRT-PCR was conducted to analyze MRAS mRNA. D
   NCI-H1975 cells were treated with indicated concentations of
   osimertinib for 72 h. QRT-PCR was conducted to analyze HEG1 mRNA. E
   NCI-H1975 cells were treated with indicated concentrations of
   osimertinib for 72 h. WB was conducted to analyze MRAS and HEG1
   proteins. F NCI-H1975 cells were treated with 10 nM of osimertinib for
   the indicated time. QRT-PCR was conducted to analyze MRAS mRNA. G
   NCI-H1975 cells were treated with 10 nM of osimertinib for the
   indicated time. QRT-PCR was conducted to analyze HEG1 mRNA. H NCI-H1975
   cells were treated with 10 nM of osimertinib for the indicated time. WB
   was conducted to analyze MRAS and HEG1 protein. I The schematic figure
   of establishing osimertinib-resistant (OR) clones. J Crystal violet
   cytotoxic assay was conducted to assess IC[50] of osimertinib in
   NCI-H1975 parental (Par) and OR clones. K QRT-PCR (top) and WB (bottom)
   analysis of MRAS in NCI-H1975 Par and OR cells. L. QRT-PCR (top) and WB
   (bottom) analysis of HEG1 in NCI-H1975 Par and OR cells. *, P < 0.05;
   ***, P < 0.001; ****, P < 0.0001

   To further explore the roles of MRAS and HEG1 in acquired osimertinib
   resistance, we generated osimertinib-resistant NCI-H1975 clones
   (osimertinib-resistant-1/2, OR-1/2) as shown in the schematic figure
   (F[82]ig. [83]4I). The sensitivities of parental and resistant
   NCI-H1975 cells to osimertinib were evaluated using a crystal violet
   cytotoxic assay. As shown in Fig. [84]4J, OR-1/2 clones displayed
   remarkably enhanced resistance to osimertinib, with the IC[50]
   increasing by approximately two hundred thousand-fold. We found that
   mRNA and protein expressions of MRAS and HEG1 were remarkably
   upregulated in OR-1/2 clones compared to their parental counterparts
   (Fig. [85]4K and L).

   Combining all these data above, MRAS and HEG1 were proven to play
   important roles in the osimertinib-resistant LUAD but the mechanism
   remained unknown.

Roles of MRAS and HEG1 in osimertinib resistant LUAD

   We further investigated the potential resistant mechanisms of MRAS and
   HEG1 by performing the KEGG pathway analysis in the TCGA cohort. The
   median MRAS and HEG1 expression were used as the cut-off value to
   assign samples with higher expression to the MRAS^high and HEG1^high
   subgroups (Fig. [86]5A and C), respectively. The EGFR mutation status
   were also analyzed in the TCGA cohort (Table S9). The MRAS^high
   subgroup showed highly enriched pathways in viral protein interaction
   with cytokine and receptor, cytokine-cytokine receptor interaction,
   cell adhesion molecules, malaria, and protein digestion and absorption
   (Fig. [87]5B). On the other hand, the MRAS^low subgroup had highly
   enriched pathways in maturity-onset diabetes of the young, arginine and
   proline metabolism, complement and coagulation cascades, gastric
   cancer, and metabolism of xenobiotics by cytochrome P450 (Fig. [88]5B).
   The most highly enriched pathways in the HEG1^high subgroup were
   ECM-receptor interaction, protein digestion and absorption, PI3K-Akt
   signaling pathway, focal adhesion, and neuroactive lig-and-receptor
   interaction; the most enriched pathways in the HEG1^low subgroup were
   ascorbate and aldarate metabolism, biosynthesis of nucleotide sugars,
   aldosterone-regulated sodium reabsorption, and inositol phosphate
   metabolism (Fig. [89]5D).

Fig. 5.

   [90]Fig. 5
   [91]Open in a new tab

   KEGG analysis based on the hub genes. A The volcano plot of DEGs in
   MRAS^high and MRAS^low subgroups from TCGA dataset. The
   -log10 (P-values) (Y-axis) is plotted against the average log2 (Fold
   Change) (X-axis) in gene expressions. The log2 (FC) and P-value are set
   at > 1 and < 0.05, respectively. B KEGG analysis based on the MRAS^high
   and MRAS^low subgroups. C The volcano plot of DEGs in HEG1^high and
   HEG1^low subgroups from TCGA dataset. The -log10 (P-values) (Y-axis) is
   plotted against the average log2 (Fold Change) (X-axis) in gene
   expressions. The log2 (FC) and P-value are set at > 1 and < 0.05,
   respectively. D KEGG analysis based on the HEG1^high and HEG1^low
   subgroups

   Evidence has shown that acquired resistance to osimertinib in NSCLC
   patients is associated with the mutual effects between the tumor and
   its surrounding microenvironment [[92]13]. We tried to clarify whether
   MRAS and HEG1 contributed to building the immunosuppressive environment
   after osimertinib treatment. The CIBERSOR method was used to calculate
   the degree of immune cell infiltration in each sample from the TCGA
   cohort and analyze the correlation between MRAS and HEG1 and immune
   infiltration (Fig. [93]6A and B). Of note, both MRAS and HEG1 had
   positive relations with M2 infiltration. We also explored the
   expression relation between two hub genes and 60 representative genes
   in the immune checkpoint pathway [[94]14], which showed closely
   positive correlations (Fig. [95]6C and D).

Fig. 6.

   [96]Fig. 6
   [97]Open in a new tab

   Relations between the hub genes and immune-related cells and genes. A-B
   The relationship between MRAS/HEG1 expressions and immune cell
   infiltration. C-D The relationship between MRAS/HEG1 expressions and
   immune markers

   We also analyzed the expression of MRAS and HEG1 in LUAD patients from
   the TCGA database who had the most common clinical mutations linked to
   EGFR TKI resistance [[98]15, [99]16] (Figure S1 and S2). These findings
   indicate that the bypassing pathway activating, and immune
   microenvironment remodeling contribute to MRAS and
   HEG1-related-acquired resistance to osimertinib in LUAD patients.

Identification of potential therapeutic agents for the MRAS/HEG1-related
osimertinib resistance

   Upregulation of MRAS and HEG1 in LUAD is closely related to osimertinib
   resistance but there is no existing strategy using MRAS or HEG1
   inhibition to conquer the resistance. The signaling pathways
   accompanied by overexpressing MRAS and HEG1 might be potential targets
   in LUAD. We used the oncoPredict model to identify potential
   therapeutic agents for the MRAS/HEG1-related osimertinib resistance.
   Our findings predicted that six agents, including AZD8186, AZD1332,
   cerdiranib, dasatinib, doramapimod, and WIKI4 were effective against
   osimertinib-resistant NSCLC cells, MRAS^high LUAD patients, and
   HEG1^high LUAD patients (Fig. [100]7). The sensitivities of parental
   NCI-H1975 and resistant OR-1/2 cells to these agents were subsequently
   examined using a crystal violet cytotoxic assay (Fig. [101]8). OR-1/2
   cells showed relatively higher sensitivity to doramapimod compared to
   parental NCI-H1975 cells (Fig. [102]8C). However, compared to parental
   NCI-H1975 cells, OR-1 and OR-2 displayed reduced and enhanced
   sensitivities to WIKI4, respectively (Fig. [103]8D).

Fig. 7.

   [104]Fig. 7
   [105]Open in a new tab

   OncoPredict analysis of osimertinib-resistant LUAD cells and patients
   with differential expression of MRAS and HEG1. A OncoPredict analysis
   of osimertinib-sensitive or resistant LUAD cells to six agents. B
   OncoPredict analysis of LUAD patients with low and high levels of MRAS
   to six agents. C OncoPredict analysis of LUAD patients with low and
   high levels of HEG1 to six agents

Fig. 8.

   [106]Fig. 8
   [107]Open in a new tab

   Crystal violet cytotoxic assay in parental NCI-H1975 and OR-1/2 clones.
   A Crystal violet cytotoxic assay was conducted to assess IC[50] of
   AZD1332 in NCI-H1975 parental (Par) and osimertinib-resistant (OR)
   clones. B Crystal violet cytotoxic assay was conducted to assess IC[50]
   of dasatinib in NCI-H1975 Par and OR clones. C Crystal violet cytotoxic
   assay was conducted to assess IC[50] of doramapimod in NCI-H1975 Par
   and OR clones. D Crystal violet cytotoxic assay was conducted to assess
   IC[50] of WIKI4 in NCI-H1975 Par and OR clones. E Crystal violet
   cytotoxic assay was conducted to assess IC[50] of AZD8186 in NCI-H1975
   Par and OR clones. F Crystal violet cytotoxic assay was conducted to
   assess IC[50] of cediranib in NCI-H1975 Par and OR clones

Methods

Data download

   We downloaded six datasets related to the EGFR-TKI resistance from the
   Gene Expression Omnibus (GEO, [108]https://www.ncbi.nlm.nih.gov/gds),
   including [109]GSE103350, [110]GSE106765, [111]GSE153183,
   [112]GSE163913, [113]GSE201608 and [114]GSE193258. Only osimertinib
   sensitive and resistant samples could be included in the study. The
   specific dataset taken for LUAD resistant or sensitive cells for
   osimertinib are listed as follows:

   Osimertinib sensitive cells datasets:

   [115]GSE103350 dataset: [116]GSM2768998

   [117]GSE193258 dataset: HCC2935_DMSO_1, HCC2935_DMSO_2, HCC2935_DMSO_3,
   HCC827_DMSO_1, HCC827_DMSO_2, HCC827_DMSO_3, H1975_DMSO_1,
   H1975_DMSO_2, H1975_DMSO_3, PC9_DMSO_1, PC9_DMSO_2 and PC9_DMSO_3.

   [118]GSE106765 dataset: [119]GSM2850068 and [120]GSM2850071.

   [121]GSE153183 dataset: [122]GSM4635290 and [123]GSM4635291.

   [124]GSE163913 dataset: [125]GSM4990622 and [126]GSM4990625.

   [127]GSE201608 dataset: [128]GSM6068510.

   Osimertinib resistant cells datasets:

   [129]GSE103350 dataset: [130]GSM2769000 and [131]GSM2769003.

   [132]GSE193258 dataset: HCC2935_osi_DTP_1, HCC2935_osi_DTP_2,
   HCC2935_osi_DTP_3, HCC827_osi_DTP_1, HCC827_osi_DTP_2,
   HCC827_osi_DTP_3, H1975_osi_DTP_1, H1975_osi_DTP_2, H1975_osi_DTP_3,
   PC9_osi_DTP_1, PC9_osi_DTP_2 and PC9_osi_DTP_3.

   [133]GSE106765 dataset: [134]GSM2850070 and [135]GSM2850073.

   [136]GSE153183 dataset: GSM4635292and [137]GSM4635293.

   [138]GSE163913 dataset: [139]GSM4990624 and [140]GSM4990627.

   [141]GSE201608 dataset: [142]GSM6068511 and [143]GSM6068512.

   The survival analysis was conducted in [144]GSE14814. Besides, the
   RNA-seq data, mutation data and clinical pathological data of LUAD
   samples were downloaded from the TCGA Genomic Data Commons.

Differentially expressed genes (DEGs) analysis

   We used two methods to select DEGs associated with osimertinib
   resistance. The first method uses the edgeR R package to identify DEGs
   [[145]17]. The genes satisfying |log2(FC)|> 1 and P-value < 0.05 genes
   were considered to have significant differences. The second method
   combined the datasets with the R software package named inSilicoMerging
   [[146]18]. We used Empirical Bayes to eliminate batch effects
   [[147]19]. The DESeq2 R package was finally used to find out the DEGs
   of the combined dataset. When |log2(FC)|> 1 and P-value < 0.05 were
   satisfied, the genes were statistically significant in the combined
   matrix [[148]20].

Function enrichment analysis

   The DEGs screened from the methods above were subjected to Gene
   Ontology (GO) enrichment analysis, Kyoto Encyclopedia of Genes and
   Genomes (KEGG) pathway analysis, and Gene Set Enrichment Analysis
   (GSEA). A P-value < 0.05 was considered significant. The final
   visualizations of function enrichment analysis were performed by the
   ggplot2 R package.

Drug sensitivity prediction

   To predict potential therapeutic agents for the acquired
   osimertinib-resistant patients, we used the "oncoPredict" R package to
   predict the semi-maximum inhibitory concentration (IC[50]) of
   therapeutic drugs in osimertinib-sensitive and resistant cells, as well
   as MRAS^high/MRAS^low and HEG1^high/HEG1^low NSCLC patients [[149]21].
   The package was designed by using the cell lines expression and the
   TCGA gene expression profiles on the Genomics of Cancer drug
   sensitivity in Cancer (GDSC, [150]https://www.cancerrxgene.org) to
   construct Ridge regression models.

Cell lines and reagents

   NCI-H1975 cell and HEK-293T were cultured in DMEM (Cat. 11995500BT,
   Gibco, USA) and 1640 (Cat. C11875500BT, Gibco, USA), 10% fetal bovine
   serum (FBS, Vazyme, China), and 1% penicillin–streptomycin (Cat.
   150063, Invitrogen, USA). HEK-293T and NCI-H1975 cells were purchased
   from BeNa Culture Collection (BNCC, China) and authenticated
   periodically (at least every 6 months) via STR profiling (Beijing
   Tsingke Biotech, China). NCI-H1975 cells were treated with 2 μM of
   osimertinib for seven days and surviving clones were selected and
   continually maintained in the dose of 2 μM. Osimertinib (Cat. S7297)
   and AZD1332 (Cat. A-495) were purchased from Selleck (USA) and Alomone
   Labs (Israel), respectively. AZD8186 (Cat. HY-12330), cerdiranib (Cat.
   HY-10205), dasatinib (Cat. HY-10181), doramapimod (Cat. HY-10320), and
   WIKI4 (Cat. HY-16910) were purchased from MedChem Express (USA).

Crystal violet viability assay

   Cells were seeded in 96-well plates followed by drug treatment for
   72 h. Cells were then stained with crystal violet (Cat. A600331, Sangon
   Biotech, China) for 30 min. The crystal violet solution was washed with
   running water and the plate was dried. Crystal violet was dissolved
   with fresh 100 mM sodium citrate solution for one hour and the
   absorbance was determined at 595 nm using a colorimetric plate reader
   (VersaMax™, Molecular Devices).

Western blot (WB)

   Cells were washed with cold phosphate-buffered saline (PBS, pH 7.4)
   three times and lysed with radioimmunoprecipitation assay buffer (RIPA
   buffer, Cat. 89900, Thermo Fisher, USA) supplemented with protease and
   phosphatase inhibitor (Cat. A32959, Thermo Fisher, USA). Protein was
   quantified using Bradford assay (Cat. 5000001, Bio-Rad, USA). At least
   40 μg of lysates were resolved by SDS-PAGE gel and transferred to
   0.45 μm PVDF (Cat. IPVH00010, Millipore, USA). The blots were blocked
   with 5% non-fat milk dissolved in tris-buffered saline (TBS)
   supplemented with 0.1% Tween® 20 detergent (TBST) at room temperature
   for one hour and processed for the incubation of indicated primary
   antibodies at 4 ℃ overnight. The blots were washed with TBST and
   incubated with appropriated horseradish peroxidase-conjugated secondary
   antibodies (Cat. 31430 and 31460, Thermo Fisher, USA) at room
   temperature for one hour followed by visualization of immunoreactive
   bands by chemiluminescence (Cat. E423-01, Vazyme, China). Antibodies
   are as follows: MRAS (Cat. PA5-112613, Thermo Fisher, USA), HEG1 (Cat.
   Bs-15449R, Bioss Biotech, China), and GAPDH (Cat. 301341, Zenbio,
   China).

Quantitative reverse transcription polymerase chain reaction (qRT-PCR)

   Messenger RNA (mRNA) was isolated from cells using TRIzol™ (Cat.
   15596026, Invitrogen, USA), followed by cDNA synthesis using a qPCR
   cDNA synthesis kit (Cat. R323-01, Vazyme, China). PCR reactions were
   conducted using ChamQ SYBR qPCR Master Mix (Cat. Q321-02, Vazyme,
   China) and the Bio-Rad CFX96 deep cycler. Relative mRNA expression was
   calculated by the comparative CT method. Primers are listed below:

   Gapdh-forward: 5′-GTCTCCTCTGACTTCAACAGCG-3′,

   Gapdh-reverse: 5′-ACCACCCTGTTGCTGTAGCCAA-3′.

   MRAS-forward: 5′-CCACCATTGAAGACTCCTACCTG-3′,

   MRAS-reverse: 5′-ACGGAGTAGACGATGAGGAAGC-3′.

   HEG1-forward: 5′-CTGCCACCTTTGCTGTTCAGAG-3′,

   HEG1-reverse: 5′-CTGGTGTTGTCTGCGACGCATT-3′.

Patients and tissue specimen

   The tumor slices were obtained from Sichuan Cancer Hospital &
   Institute, Sichuan Cancer Center with approval from the Institutional
   Ethics Committee (protocol code: SCCHEC-04-2022-002). All patients
   signed the informed consent form, and clinical information is provided
   in Supplementary Table 6.

Immunohistochemistry (IHC)

   Immunohistochemical staining was performed on tumor tissue slides as
   previously described [[151]22]. The slides were incubated with rabbit
   polyclonal anti-HEG1 (1:100, Cat. Bs-15449R, Bioss, China) and
   anti-MRAS (1:100, Cat. PA5-112613, Thermo Fisher, USA) antibodies at 4℃
   overnight. HEG1 and MRAS expression were evaluated by using the Halo
   data analysis system to calculate the 3’,3’-doaminobenzidine
   tetrahydrochloride (DBA) positive area proportion of high-power images
   (40X).

Statistics

   The significance between treatment groups for the in vitro studies was
   determined by a Student t-test (two-tailed) or a One-way ANOVA
   (two-tailed) with Dunnett’s post-hoc test. The significance in IHC
   analysis was determined by an independent samples t-test. P values less
   than 0.05 were categorized as statistically significant. Error bars for
   all figures represent SD unless otherwise stated. Figures of in vitro
   experiments are representative of at least three independent
   experiments.

Discussion

   EGFR TKIs are commonly administered to treat NSCLC patients with EGFR
   mutations [[152]23–[153]26]. However, in more than half of the patients
   receiving first- or second-generation EGFR TKIs, the occurrence of a
   new EGFR mutation, T790M, contributes to acquired resistance [[154]27].
   The third-generation EGFR-TKI, osimertinib, was initially developed to
   target T790M mutation and has shown selective activity in targeting
   L858R or Exon19 del combined T790M [[155]23, [156]28–[157]30]. However,
   most NSCLC patients eventually become refractory to osimertinib through
   on-target or off-target mechanisms, which refer to the occurrence of
   new genetic alterations at the EGFR genes or genetic alterations and
   activations of other signaling pathways, respectively [[158]31].
   According to studies in clinical trials and real-world registries, the
   resistant mechanisms vary when osimertinib is administered as
   first-line or second-line treatment. Studies have shown that in cases
   where osimertinib is administered as first- or second-line therapy,
   off-target mechanisms account for approximately 20% to 50%. These
   mechanisms include MET amplification, HER2 amplification, PIK3CA
   amplification, BRAF mutation, KRAS mutation, PIK3CA mutation, PTEN
   loss, oncogenic fusions, and histologic transformation [[159]2]. With
   emerging literature highlighting the impact of signaling pathways on
   acquired osimertinib resistance [[160]32–[161]34], we sought to
   determine biomarkers for osimertinib resistance.

   Our study analyzed six online GEO datasets and discovered that MRAS and
   HEG1 were significantly upregulated in osimertinib-resistant LUAD
   cells. Higher expression levels of MRAS and HEG1 were further confirmed
   in online data and tumor biopsies from LUAD patients with resistance to
   osimertinib in our hospital. We generated osimertinib resistant LUAD
   cell lines with elevated mRNA and protein levels of both MRAS and HEG1.
   We also proved a dose/time-dependent manner of MRAS and HEG1.
   Otherwise, lower expression of MRAS and HEG1 in tumor tissues than its
   paired paracancerous tissues in LUAD. These results above imply that
   MRAS and HEG1 expressions in LUAD are in osimertinib used-dependent
   manner until the resistance develops, which prompts us to conduct deep
   mechanism exploration.

   We still investigated the potential of MRAS and HEG1 in predicting the
   survival of LUAD patients. The LUAD patients with higher MRAS
   expression showed poorer survival probability (P < 0.05). On the other
   hand, the LUAD patients with higher HEG1 expression showed poorer
   survival probability than those with lower HEG1 expression, but the
   difference was not significant (P > 0.05). These indicate that MRAS
   might be both an effective osimertinib response and survival biomarker
   in LUAD. Although HEG1 is osimertinib resistance biomarker, the
   efficiency of HEG1 in predicting survival needs more studies to prove.

   It has been reported that MRAS and HEG1 are associated with cancers.
   MRAS belongs to the Ras subfamily of GTP-binding proteins. Its
   expression, gene polymorphisms, and mutation have been linked to
   various diseases, including coronary artery disease, painful
   temporomandibular disorders, Noonan syndrome, and cancers
   [[162]35–[163]39]. A previous study has shown that activated MRAS can
   initiate the epithelial-mesenchymal transition (EMT) and tumorigenesis
   of normal epithelial cells [[164]40]. MRAS interacts with various
   proteins, including Ras/Rap1 (RA)-guanine nucleotide exchange factor
   (GEF)-2, SCRIB, or SHOC2, to promote differentiation, aggregation,
   proliferation, and migration through MAPK and AKT signaling cascades
   [[165]41–[166]44]. MRAS also promotes estrogen-independent growth of
   breast cancer cells through the interaction with Rlf and the following
   Ral/JNK activation in an EGFR-dependent manner [[167]45]. However, MRAS
   coupled with G-protein gamma subunit 2 (GCN2) retards breast cancer
   cell proliferation by inhibiting ERK/Akt signaling [[168]46]. HEG1
   regulates the concentric growth of the heart and recruits
   RAS-interacting protein 1 (Rasip1) to cell–cell junctions through
   direct binding, which maintains the stabilization of epithelial cell
   (EC) cell–cell junctions [[169]47, [170]48]. In hepatocellular
   carcinoma (HCC) cells, HEG1 promotes intrahepatic metastasis, lung
   metastasis, and EMT by inducing the expression and stability of
   β-catenin [[171]49].

   The most recent studies have revealed the impact of MRAS and HEG1 on
   therapeutic resistance in cancers. In models of KRAS mutant lung and
   colorectal cancer models, KRAS G12C inhibitors suppress the Hippo/
   yes-associated protein (YAP) signaling pathway by mislocalizing SCRIB
   in the cytoplasm. The activated YAP transports into the nucleus and
   initiates the transcription of MRAS. The increasing MRAS interacts with
   SHOC2 and PP1 and subsequently reactivates MEK/ERK signaling by
   dephosphorylating CRAF at Ser259, which ultimately drives adaptive
   resistance to KRAS G12C inhibitors [[172]39]. In HCC cells, the protein
   arginine methyltransferase 3 (PRMT3)/IGF2BP1 axis triggers the
   resistance to oxaliplatin by stabilizing the mRNA of HEG1 [[173]50].

   Our study suggested that MRAS and HEG1 might be drivers of osimertinib
   resistance by activating signal pathways or altering the immune
   environment. It has been known that the regulation of the PI3K-Akt
   signaling pathway is a promising therapeutic tactic for osimertinib
   resistance, which may be the key point in HEG1-related resistance
   [[174]16]. An immunosuppressive environment was also identified at
   osimertinib resistance with decreased T cell infiltration and
   activation and increased macrophage infiltration and M2 polarization
   [[175]13]. Interestingly, our results also revealed that MRAS and HEG1
   expressions interaction with M2 infiltration might result in
   osimertinib resistance. Moreover, it was found that MRAS expression is
   higher in LUAD patients with ALK mutation. A rare ALK rearrangement has
   been reported as a might incentive for acquired osimertinib resistance
   in a lung cancer patient [[176]51], which implies a positive
   association between ALK mutation and the upregulation of MRAS.

   Additionally, oncoPredict analysis showed that osimertinib-resistant
   LUAD with high levels of MRAS and HEG1 might respond to certain agents,
   including AZD1332 (a selective inhibitor of the tropomyosin receptor
   kinase receptors (TrKs)), dasatinib (a Src family kinases inhibitor),
   doramapimod (a pan-p38/MAPK inhibitor), WIKI4 (a Wnt/β-catenin
   inhibitor), AZD8186 (a selective PI3Kβ inhibitor), and cediranib (a
   selective VEGF signaling inhibitor). A study has reported that
   AXL/Cubomain-containing protein 1 (CDCP1)/SRC/AKT signaling axis is
   augmented upon osimertinib treatment and eventually confers to
   osimertinib resistance in LUAD cells [[177]52]. Consistently, our in
   vitro crystal violet cytotoxic assay showed that both parental
   NCI-H1975 and OR-1/2 clones, especially OR-2 clone, are comparatively
   susceptible to dasatinib, whose efficacy in unselected LUAD patients
   has been evaluated [[178]53]. Furthermore, Nie M et al. reported that
   acetylcholine (Ach) binds to muscarinic Ach receptors (mAchRs) and
   subsequently promotes the transcription of β-catenin-mediated WNT
   target genes, which ultimately causes cell survival and drug tolerance
   of drug-tolerant persister (DTP) NSCLC cells [[179]54]. In this study,
   OR-2 clone was more sensitive to the Wnt/β-catenin inhibitor WIKI4 than
   the parental NCI-H1975, whereas OR-1 clone exhibited insensitivity.
   RAS/MAPK signaling has been found hyperactivated in osimertinib and
   other EGFR TIK-resistant NSCLC cells [[180]55] and they key downstream
   of MRAS [[181]56].

   Although IHC of MRAS and HEG1 showed that samples from LUAD patients
   with osimertinib resistance had higher expression levels of MRAS and
   HEG1 compared to samples from LUAD treatment naïve patients, the sample
   size is too small, and more clinical samples are needed to further
   verify. Additionally, the findings are mainly based on the existing
   data sets and few in vitro experiments, for example, overexpression of
   MRAS or HEG1 should be performed to further assess the association of
   MRAS or HEG1 with sensitivity to osimertinib in our future
   investigation. The cell line/patient-derived LUAD mouse models are also
   needed to strengthen the rationale and understanding of mechanisms and
   potential for targeting MRAS and HEG1.

   Overall, it is implied that elevating MRAS and HEG1 might predict the
   development of acquired osimertinib resistance, and the underlying
   mechanisms require further investigation. These findings suggest that
   certain drugs may be effective in treating lung cancer patients with
   high levels of specific proteins, and further research is needed to
   explore these treatments in more detail.

Conclusion

   In summary, MRAS and HEG1 showed good prognostic value in predicting
   resistance to osimertinib and might play important roles in the
   resistance development in LUAD. More data from the real world are
   demanded to validate the accuracy and further studies are required to
   strengthen the rationale.

Supplementary Information

   [182]12672_2024_1552_MOESM1_ESM.docx^ (57.6KB, docx)

   Supplementary Material 1. Table S8. List of HEG1 differential gene KEGG
   enrichment analyses.
   [183]12672_2024_1552_MOESM2_ESM.xlsx^ (35.4KB, xlsx)

   Supplementary Material 2. List of EGFR mutation status in the TCGA 454
   dataset.
   [184]12672_2024_1552_MOESM3_ESM.tif^ (2.7MB, tif)

   Supplementary Material 3. Figure S1: The MRAS expression in LUAD
   patients with different gene mutations from the TCGA dataset;
   [185]12672_2024_1552_MOESM4_ESM.tif^ (2MB, tif)

   Supplementary Material 4. Figure S2: The HEG1 expression in LUAD
   patients with different gene mutations from the TCGA dataset.
   [186]12672_2024_1552_MOESM5_ESM.docx^ (506.5KB, docx)

   Supplementary Material 5. Table S1. List of genes analyzed individually
   for differences in the six datasets.
   [187]12672_2024_1552_MOESM6_ESM.docx^ (49.7KB, docx)

   Supplementary Material 6. Table S2. List of six data GO enrichment
   analyses.
   [188]12672_2024_1552_MOESM7_ESM.docx^ (33.7KB, docx)

   Supplementary Material 7. Table S3. List of six data GSEA enrichment
   analyses.
   [189]12672_2024_1552_MOESM8_ESM.docx^ (50.6KB, docx)

   Supplementary Material 8. Table S4. List of six data sets and
   concurrent analysis of variance for GO enrichment analysis.
   [190]12672_2024_1552_MOESM9_ESM.docx^ (48.8KB, docx)

   Supplementary Material 9. Table S5. List of GSEA enrichment analyses
   [191]12672_2024_1552_MOESM10_ESM.docx^ (19.9KB, docx)

   Supplementary Material 10. Table S6. Clinical information of NSCLC
   patients received tumor surgery/biopsies
   [192]12672_2024_1552_MOESM11_ESM.docx^ (78.4KB, docx)

   Supplementary Material 11. Table S7. List of MRAS differential gene
   KEGG enrichment analyses.

Author contributions

   ML, BT, and LH conducted the experiments and drafted the manuscript. RX
   and PHH participated in the analysis and interpretation of data. LH, CX
   and QL designed the work. All authors read and approved the final
   manuscript.

Funding

   This research was funded by Sichuan Province Science and Technology
   Support Program (No.2022NSFSC0691) and National Natural Science
   Foundation of China (No. 922591002).

Data availability

   Data is provided within the manuscript or supplementary information
   files.

Declarations

Ethics approval and consent to participate

   The study was conducted in accordance with the Declaration of Helsinki
   and approved by the Ethics Committee for Medical Research and New
   Medical Technology of Sichuan Cancer Hospital (protocol code:
   SCCHEC-04-2022-002).

Competing interests

   The authors declare no competing interests.

Footnotes

   Publisher's Note

   Springer Nature remains neutral with regard to jurisdictional claims in
   published maps and institutional affiliations.

Contributor Information

   Lanlin Hu, Email: hulanlin@jflab.ac.cn.

   Qiang Li, Email: liqiang@sichuancancer.org.

References