Validation of osimertinib resistance in HCC827/OR and H1975/OR cell lines
Two NSCLC cell lines with acquired resistance to osimertinib, HCC827/OR and H1975/OR, were established from their parental cells. CCK-8 assay showed that after 72h of osimertinib exposure, the IC50 values were 0.0032μM and 0.0036μM in HCC827 and H1975 cells, respectively. As expected, HCC827/OR and H1975/OR cells showed significantly decreased sensitivity to osimertinib, with the IC50 values of 3.91μM and 5.13μM, respectively (Figure 1A and 1B). Similarly, colony formation assay also indicated that the colony formation abilities of HCC827/OR and H1975/OR cells were significantly higher than their parental sensitive cells (Figure 1C and 1D). After osimertinib exposure, the HCC827 and H1975 cells exhibited both apoptosis and cell cycle arrest, while HCC827/OR and H1975/OR cells were not affected (Figure 1E and 1F). Moreover, unlike their parental cells, osimertinib-resistant cells maintained high phosphorylation levels of EGFR and downstream signaling components even following treatment with osimertinib (Figure 1G). These results indicate that the HCC827/OR and H1975/OR cells are more resistant to osimertinib than their parental cell lines.
Identification and validation of the differentially expressed lncRNAs between osimertinib-sensitive and -resistant NSCLC cell lines
To explore the role of lncRNAs in osimertinib resistance in NSCLC, we first examined the differentially expressed lncRNAs between osimertinib-resistant cells and their parental cells using RNA-Seq (Figure 2A). A total of 64678 lncRNAs were identified from our NSCLC cells, and the majority of them were from intergenic regions (~42%), exonic of protein-coding genes (~21%), and introns of protein-coding genes (~20%) (Figure 2B). Our results showed that 427 lncRNAs were differentially expressed between HCC827/OR and HCC827 cells, consisting of 212 upregulated and 215 downregulated lncRNAs. While 471 lncRNAs were differentially expressed between H1975/OR and H1975 cells, consisting of 218 upregulated and 253 downregulated lncRNAs. Of these, 17 lncRNAs were consistently upregulated and 15 lncRNAs were consistently downregulated in both osimertinib-resistant cells compared with their parental cells (Figure 2C, Table 1). As previously reported, chromosomal abnormality often occurs in NSCLC, and some lncRNAs have been recognized as important regulators of chromosome stability (15,16). Therefore, we investigated the chromosomal localization of these differentially expressed lncRNAs. As shown in Figure 2D, these lncRNAs were significantly enriched in chromosome 1 and 12.
To validate the RNA-Seq results, we randomly selected 20 differentially expressed lncRNAs for qPCR, and found seven of them were consistent with the RNA-Seq data. Compared with the parental sensitive cells, four lncRNAs (lnc-TMEM132D-AS1, lnc-XLOC_003074, lnc-AC006064.5, and lnc-ZNF213-AS1) were significantly upregulated both in HCC827/OR and H1975/OR cells, whereas three lncRNAs (lnc-GAS5, lnc-AC007952.4, and lnc-SNHG29) were markedly downregulated in two resistant cell lines (Figure 2E).
Upregulated lnc-TMEM132D-AS1 reduces sensitivity to osimertinib in NSCLC cells
According to the results mentioned above, lnc-TMEM132D-AS1 was the most significantly upregulated lncRNA after acquired resistance to osimertinib in NSCLC cells. To preliminary explore the clinical significance of lnc-TMEM132D-AS1, we analyzed its expression in lung cancer tissues, as well as the relationship between its expression and clinical outcomes in lung cancer patients using The Cancer Genome Atlas (TCGA) database. The results showed that the expression of lnc-TMEM132D-AS1 was significantly higher in lung cancer tissues than in normal tissues (P<0.001). The patients with high lnc-TMEM132D-AS1 expression showed decreased overall survival (OS) and progression-free survival (PFS) (both P<0.001) (Figure 3A and 3B). Additionally, we found that lnc-TMEM132D-AS1 was located on chromosome 12. Therefore, we selected lnc-TMEM132D-AS1 for further investigation. Firstly, we determined the expression of lnc-TMEM132D-AS1 in cultured osimertinib-sensitive and -resistant NSCLC cells, as well as in the plasma samples of osimertinib-sensitive and -resistant NSCLC patients. Compared to the parental sensitive cells, significantly elevated level of lnc-TMEM132D-AS1 was observed not only in the whole-cell lysates, but also in the culture supernatants of HCC827/OR and H1975/OR cells (Figure 3C and 3D). Similarly, the level of lnc-TMEM132D-AS1 in the plasma of NSCLC patients with acquired resistance to osimertinib was about 8-fold higher than that of osimertinib-sensitive patients (Figure 3E). These data provide evidence that lnc-TMEM132D-AS1 is upregulated in NSCLC cells with acquired resistance to osimertinib both in vitro and in vivo, and is a potential biomarker for resistance to osimertinib therapy.
Then, we observed the the influence of lnc-TMEM132D-AS1 on osimertinib sensitivity in NSCLC cells. After silencing the expression of lnc-TMEM132D-AS1 in HCC827/OR cells, we found that the cell survival rate and colony formation capacity after osimertinib exposure were significantly decreased, the osimertinib-induced apoptosis was markedly increased, and cell cycle was obviously arrested in G2/M phase (Figure 3F-I). These data suggest that lnc-TMEM132D-AS1 can significantly reduce the osimertinib sensitivity and is a potential target for overcoming osimertinib resistance.
The potential mechanisms of lnc-TMEM132D-AS1 in osimertinib resistance
In order to explore the molecular mechanism of lnc-TMEM132D-AS1 in osimertinib resistance, we firstly determined its subcellular localization, because the functional mechanisms of lncRNAs are dependent on their subcellular distribution (17). We found that lnc-TMEM132D-AS1 was mainly expressed in the cytoplasm of NSCLC cells (Figure 4A), indicating that it might play important roles in modulating mRNA stability, translation and signalling pathways through acting as a sponge of miRNAs (18). Therefore, we set out to explore the potential miRNAs interacting with lnc-TMEM132D-AS1, as well as their potential mRNA targets.
Firstly, we screened the differentially expressed mRNAs between osimertinib-sensitive and -resistant NSCLC cells using RNA-Seq, and selected 576 mRNAs which showing similar expression tendencies in both paired cell lines, then calculated the Pearson correlation coefficient (PCC) between these mRNAs and lnc-TMEM132D-AS1 to evaluate the co-expression relationships. A total of 215 mRNAs with PCC > 0.8 and a correlation P value < 0.05 were chosen to build the co-expression network (Figure 4B). Subsequently, using DIANA TOOLS, Targetscan, and miRDB databases, we found 148 miRNAs with binding capacity to lnc-TMEM132D-AS1, as well as their 756 potential target mRNAs. Accordingly, we took the intersection of the selected 215 differentially expressed mRNAs with the predicted 756 potential target mRNAs. As a result, we constructed the lnc-TMEM132D-AS1-miRNA-mRNA interaction network, including 125 miRNAs and 162 target mRNAs (Figure 4C).
Then, we performed GO and KEGG analyses to explore the putative functions of these 162 target mRNAs. We analyzed all three GO aspects: biological processes, cellular composition, and molecular function. The top 15 dysregulated GO processes are shown in Figure 4D. In the biological processes subgroup, the top three dysregulated GO processes were immune response, inflammatory response, and positive regulation of cell proliferation. In the cellular composition subgroup, the top three dysregulated GO processes were extracellular environment, extracellular region, and integral component of plasma membrane. While in the molecular function subgroup, the top three dysregulated GO processes were ATP binding, identical protein binding, and extracellular matrix structural constituent. Interestingly, one of these target mRNAs, ENTPD1, was included in three GO terms, which were immune response, extracellular environment, as well as ATP binding. For KEGG analysis, the top 20 dysregulated pathways are shown in Figure 4E. Of these, the environmental information processing, ascorbate and aldarate metabolism, and pentose and glucuronate interconversions were the most enriched pathways.
To further mine the resistance-associated genes, we performed PPI network analysis based on these 162 target mRNAs, and identified some hub genes, of which the top 10 were CDH1, TIMP3, FBN1, DLJ4, CDH2, GJA1, NGFR, ENTPD1, IGF2, and NKX2-5. Most of these hub genes have been shown to take part in chemoresistance. For example, FBN1, GJA1, and IGF2 were related to cisplatin resistance (18-20), while TIMP3, CDH2, NGFR, and ENTPD1 were involved in resistance to sorafenib, 5-fluorouracil, icotinib and cytarabine, respectively (21-24). These findings support that, at least partially, lnc-TMEM132D-AS1 might play a role in osimertinib resistance in NSCLC.
The expression of ENTPD1, a target mRNA of lnc-TMEM132D-AS1, correlates with tumor-infiltrating immunosuppressive cells and poor prognosis in patients with NSCLC
Given that GO analysis showed that one of the lnc-TMEM132D-AS1 target mRNAs, ENTPD1, were similarly related to three GO terms, and PPI analysis also indicated that ENTPD1 was one of the hub genes, so we selected it for further study. ENTPD1 also referred to as CD39 and NTPDase, is a Ca2+ and Mg2+-dependent integral cell membrane molecule, which phosphohydrolases extracellular ATP to produce AMP. Subsequently, these AMP can be further hydrolyzed into adenosine by CD73(25). Previous studies have indicated that ENTPD1 is overexpressed in some blood and solid tumor cells, such as acute B lymphoblastic leukemia (B-ALL), acute myeloid leukemia (AML), NSCLC, and hepatocellular carcinoma, promoting tumor cell growth (26-28). Moreover, ENTPD1 has been shown to be overexpressed in cytarabine resistant AML cells (24). Consistent with previous studies, we also observed that ENTPD1 was upregulated in osimertinib-resistant cells compared to sensitive cells (Figure 5A). Using TCGA data, we found that higher ENTPD1 expression was significantly associated with poorer OS and PFS in lung cancer patients (Figure 5B).
Besides tumor cell growth-promoting activity, ENTPD1 also plays a vital role in regulating the anti-tumor immune response. Extracellular adenosine produced by ENTPD1 can bind to adenosine receptors, which are expressed in some immune cells, including regulatory T cells (Tregs) and myeloid-derived suppressor cells (MDSCs). Acting through adenosine receptors, extracellular adenosine can help tumor cells to escape from host immunosurveillance by promoting the generation of Tregs and MDSCs, and contributing to their immunosuppressive properties (29,30). As previous studies suggested, the densities of tumor-infiltrating Tregs and MDSCs are correlated with tumor stage, and can be used to predict the patient outcomes in various types of cancer (31,32). Therefore, we used TISIDB database to analyze the correlation between ENTPD1 level and the abundance of tumor-infiltrating Tregs and MDSCs. Our results showed that both Tregs and MDSCs abundances were positively related to ENTPD1 expression in both LUAD and LUSC samples (Figure 5C and 5D), which implicates that ENTPD1 can serve as an important tumor immunosurppressive molecule in NSCLC. To verify whether ENTPD1 expression could affect patient outcomes because of tumor-infiltrating Tregs and MDSCs, we conducted an analysis of the prognostic value of ENTPD1 expression in lung cancer patients based on the tumor infiltration of Tregs and MDSCs. As a result, high ENTPD1 expression in both enriched Tregs cohort and enriched MDSCs cohort was significantly associated with a poor prognosis (all P < 0.05) (Figure 5E and 5F). Thus, ENTPD1 may affect the prognosis of NSCLC patients at least through two coordinated mechanisms, one is directly promoting the growth of NSCLC cells, the other is regulating tumor infiltration of Tregs and MDSCs.
To verify whether lnc-TMEM132D-AS1 can modulate the expression of ENTPD1, we silenced the lnc-TMEM132D-AS1 level in osimertinib-resistant HCC827/OR cells, and found that lnc-TMEM132D-AS1 knockdown could significantly decrease the mRNA level of ENTPD1 (Figure 5G). According to our lnc-TMEM132D-AS1-miRNA-mRNA network, ENTPD1 was a target of four miRNAs, hsa-miR-16-5p, hsa-miR-195-5p, hsa-miR-424-5p, and hsa-miR-497-5p, which potentially binding to lnc-TMEM132D-AS1. After lnc-TMEM132D-AS1 knockdown, we found that the expression of hsa-miR-16-5p and hsa-miR-195-5P markedly increased, indicating that lnc-TMEM132D-AS1 might regulate the expression of ENTPD1 by interacting with hsa-miR-16-5p and hsa-miR-195-5P (Figure 5H). However, the interactions among these molecule need further investigation.