Targeting GSTP1 as Therapeutic Strategy against Lung Adenocarcinoma Stemness and Resistance to Tyrosine Kinase Inhibitors

Abstract

Glutathione S-transferase pi (GSTP1), a phase II detoxification enzyme, is known to be overexpressed and mediates chemotherapeutic resistance in lung cancer. However, whether GSTP1 supports cancer stem cells (CSCs) and the underlying mechanisms in lung adenocarcinoma (LUAD) remain largely unknown. This study unveiled that GSTP1 is upregulated in lung CSCs and supports tumor self-renewal, metastasis, and resistance to targeted tyrosine kinase inhibitors of LUAD both in vitro and in vivo. Mechanistically, CaMK2A (calcium/calmodulin-dependent protein kinase 2 isoform A)/NRF2 (nuclear factor erythroid 2-related factor 2)/GSTP1 is uncovered as a regulatory axis under hypoxia. CaMK2A increased GSTP1 expression through phosphorylating the Sersine558 residue of NRF2 and promoting its nuclear translocation, a novel mechanism for NRF2 activation apart from conventional oxidization-dependent activation. Upregulation of GSTP1 in turn suppressed reactive oxygen species levels and supported CSC phenotypes. Clinically, GSTP1 analyzed by immunohistochemistry is upregulated in a proportion of LUAD and serves as a prognostic marker for survival. Using patient-derived organoids from an ALK-translocated LUAD, the therapeutic potential of a specific GSTP1 inhibitor ezatiostat in combination treatment with the ALK inhibitor crizotinib is demonstrated. This study demonstrates GSTP1 to be a promising therapeutic target for long-term control of LUAD through targeting CSCs.

Keywords: calcium/calmodulin-dependent protein kinase 2 isoform A; cancer stem cells; glutathione S-transferase pi; lung adenocarcinoma; nuclear factor erythroid 2-related factor 2.

Figure 1

GSTP1 was upregulated in CSC and supported stemness and metastasis properties in LUAD. A,B) Correlation of GSTP1 mRNA level with ALDH⁺/CD44⁺ (A) and CD133⁺ (B) CSC populations, respectively, in LUAD cell lines analyzed by Pearson correlation test. C) Western blot analysis of GSTP1 expression in tumorspheres and corresponding monolayers derived from HKULC4, HCC827, and H1299 cells. D) Western blot analysis of GSTP1 expression in ALDH⁺/CD44⁺ CSC and ALDH⁻/CD44⁻ non‐CSC fractions sorted from HKULC4, HCC827, H1299, A549, and H1975 cells. E,F) Tumorspheres serially passaged for two generations using HCC827 cells with GSTP1‐KD (E), and A549 cells with GSTP1‐OE (F), showing representative images of tumorspheres (left) and histograms of sphere numbers (right). G,H) Proportions of ALDH⁺/CD44⁺ subsets in HCC827 cells with or without GSTP1‐KD (G), and A549 cells with or without GSTP1‐OE (H). I,J) In vivo tumorigenicity in SCID mice evaluated by subcutaneous injection of HCC827 cells with or without GSTP1‐KD (I), and A549 cells with or without GSTP1‐OE (J), showing representative images of xenografts (left) and the corresponding tumor growth curves (right). K,L) In vivo limiting dilution assays for frequency of CSCs in HCC827 cells with GSTP1‐KD (K) and A549 cells with GSTP1‐OE (L). CSC frequencies and p values were calculated using the ELDA online tool (https://bioinf.wehi.edu.au/software/elda/). M) In vitro limiting dilution assay for frequency of CSCs in HCC827 ALDH⁺/CD44⁺ CSC and ALDH⁻/CD44⁻ non‐CSC fractions with or without GSTP1‐KD, respectively. N) In vivo limiting dilution assay for frequency of CSCs in HCC827 ALDH⁺/CD44⁺‐CSC and ALDH⁻/CD44⁻ non‐CSC fractions with or without GSTP1‐KD, respectively. Representative images of xenografts (left) and the corresponding CSC frequencies (right) are shown. O) Kaplan–Meier curve and log‐rank test showed tumor‐free survival of SCID mice injected with 2500 and 1000 cells, respectively. Data are presented as mean ± SD of triplicate measurements. * p < 0.05, ** p < 0.001, *** p < 0.005 versus respective control by Student's t‐test.

Figure 2

GSTP1 enhanced metastasis and LUAD resistance to anti‐cancer drugs. A,B) Transwell migration (upper panel) and invasion (lower panel) assays performed in HCC827 (A) and A549 (B) cells with or without GSTP1 manipulation. Histograms showed the relative migrated and invaded proportions of the seeded cells compared to the control group. C) In vivo tail vein injection model for the evaluation of tumor metastasis with or without GSTP1‐OE. Luciferase labeled A549 cells (2 × 10^6) were injected into nude mice through tail veins, and bioluminescence imaging was performed after 10 weeks. In vivo bioluminescence images of mice injected with GSTP1‐OE or empty vector (EV) control cells (left). Scatter plot for the quantitative comparison of bioluminescence signals in respective groups (right). D) Representative images of lung harvested at the 10th week after injection of A549 cells with GSTP1‐OE or EV control, featuring gross images of harvested lung with white arrows indicating tumor nodules (left), H&E stained histology of random areas of control lung (EV) and tumor nodules of GSTP1‐OE cells (middle), and high power view of the marked areas in the middle panel (right), respectively. E,F) Cell viability by MTT assay of scramble control and GSTP1‐KD cells treated with cisplatin, using HCC827 (E), and H1975 (F) cells. G) Cell viability assay for cisplatin treatment by MTT comparing A549 cells with GSTP1‐OE and EV control. H,I) Cell viability assay by MTT comparing effects of HCC827 control and GSTP1‐KD cells treated with gefitinib (H) or erlotinib (I). J) Effects of GSTP1‐KD on afatinib sensitivity of H1975 cells by MTT assay. Data represented mean ± SD of triplicate measurements. * p < 0.05, ** p < 0.001, *** p < 0.005 versus respective control by Student's t‐test.

Figure 3

GSTP1 upregulation was mediated by CaMK2A/NRF2 S558 axis. A) Western blot analyses of protein levels in LUAD cells with or without CaMK2A manipulation, including CaMK2A, pCaMK2A T286, NRF2, GSTP1, and nuclear NRF2. B) Relative luciferase activities of GSTP1 reporters with wild type (WT) or mutant NRF2‐binding sites in H1299 cells with or without CaMK2A‐OE. C–E) Effects of GSTP1‐OE in HCC827 cells with or without CaMK2A‐KD, with respect to tumorspheres (C), cell viability under cisplatin treatment (D), and in vivo tumorigenesis (E). F) Correlation of pCaMK2A and GSTP1 level in 178 clinical LUAD cases analyzed by Pearson correlation, and the significance was tested by the Chi‐square test. G) Co‐immunoprecipitation (Co‐IP) of NRF2 in HEK293T cells with CaMK2A‐OE and/or NRF2‐OE, using CaMK2A as bait. H) Co‐IP of NRF2 in A549 and H1299 cells with or without stable CaMK2A‐OE using CaMK2A as bait. I) Venn diagram of prediction of NRF2 phosphorylation sites by CaMK2A using on‐line databases of NETWORKIN, NetPhos3.1 and Scansite4 showed that NRF2 T267, S558, and T586 were the candidates. J) Time‐dependent CaMK2A kinase assay with NRF2 WT or mutant protein precipitated from total cell lysate of HEK293T cells with exogenously overexpressed empty vector, NRF2 WT, NRF2 T267A, NRF2 S558A, or NRF2 T586A, respectively. K) Relative luciferase activities of GSTP1 reporter with wild type NRF2‐binding sites in H1299 cells exogenously forced to overexpress empty vector, NRF2 WT, NRF2 T267A, NRF2 S558A, or NRF2 T586A, respectively. L) Co‐IP of NRF2 in HEK293T cells overexpressing CaMK2A concurrent with either NRF2 WT or NRF2 S558A using CaMK2A as bait. Data represented mean ± SD of triplicate measurements. * p < 0.05, ** p < 0.001, *** p < 0.005 versus respective control by Student's t‐test.

Figure 4

Hypoxia activated the CaMK2A/NRF2/GSTP1 axis. A) Western blot analysis of HIF1A, NRF2, CaMK2A, pCaMK2A T286, GSTP1, and nuclear NRF2 expressions in LUAD cell lines incubated under normoxic (20% O₂) or hypoxic (1% O₂) condition. B) Representative images of multiplex IHC staining for pCaMK2A, GSTP1, and CA9 in LUAD‐OG1‐derived xenografts, comparing relatively hypoxic and non‐hypoxic regions at low power (left‐most photomicrograph). High power views (other panels) of marked areas, comprising relatively hypoxic regions with CA9^high staining (upper), and relatively non‐hypoxic regions with CA9^low staining (lower). C) Histogram comparison of normalized fluorescence intensities of CA9, pCaMK2A, and GSTP1 signals of multiplex IHC staining from CA9 high and CA9 low regions of LUAD‐OG1‐derived xenografts. D) Representative images of multiplex IHC staining of xenografts from HCC827 control (sh‐ctrl, upper panel) or CaMK2A‐KD cells (sh3, lower panel), comparing relatively hypoxic (CA9^high) and non‐hypoxic (CA9^low) regions. E,F) Normalized fluorescence intensities of CA9 and GSTP1 signal of multiplex IHC staining performed in FFPE xenograft tissue derived from HCC827 cells with or without CaMK2A‐KD. Color scheme: pCaMK2A (Cyan; Opal670), GSTP1 (Red; Opal570), CA9 (Green; Opal520), and DAPI (Blue; DAPI). ** p < 0.001, *** p < 0.005 versus corresponding control by Student's t‐test.

Figure 5

CaMK2A/GSTP1 axis enhanced LUAD stemness and drug resistance through ROS suppression. A) Western blot analysis of pCaMK2A and GSTP1 expressions in cells with or without 100 µm CoCl₂ and/or 5 mm NAC treatment. B,C) Effect of NAC (1 mm) on CoCl₂‐induced intracellular Ca²⁺ and oscillations measured by Fura‐2 in HCC827 cells. D) Number of measurements and average intensity of oscillation was shown in each column. E) Western blot analysis of pCaMK2A and GSTP1 expressions in cells with or without 100 µm CoCl₂ and/or 1 µm of the calcium chelator BAPTA/AM. F–I) Relative intracellular ROS levels of HCC827 cells with or without CaMK2A‐KD (F) or GSTP1‐KD (G), and A549 cells with or without CaMK2A‐OE (H) or GSTP1‐OE (I) detected by CellRox dye and flow cytometry. J–M) Relative mitochondrial ROS levels of HCC827 cells with or without CaMK2A‐KD (J) and GSTP1‐KD (K) detected by mitoSOX dye and flow cytometry. Relative intracellular ROS levels detected by CellROX dye (L) and mitochondrial ROS levels detected by mitoSOX dye (M) in A549 cells with or without CaMK2A and GSTP1 manipulation. N) The effects of the ROS scavenger NAC on tumorspheres formation of HCC827 cells with or without GSTP1‐KD. The left panel showed representative images of tumorspheres (left) and histograms of sphere numbers (right). O,P) The effects of NAC, with or without GSTP1‐KD, on cell viability by MTT assay of HCC827 cells treated with cisplatin (O), or gefitinib (P). Data represented mean ± SD of triplicate measurements. * p < 0.05, ** p < 0.001, *** p < 0.005 versus respective control by Student's t‐test.

Figure 6

GSTP1 was upregulated and correlated with poor prognosis of human LUAD. A) Scatter plot of GSTP1 relative mRNA levels in 58 paired LUAD (T) and corresponding non‐tumor (NT) lung from TCGA database. B,C) Kaplan–Meier curves comparing progression free (PFS)sh (B) and overall survival (OS) (C) of GSTP1 expressions stratified by median level, based on the Kaplan–Meier Plotter database (http://kmplot.com/analysis/index.php?p=service&cancer=lung). D) Kaplan–Meier curves comparing 5‐year overall survival (OS) stratified by GSTP1 expressions analyzed by the TIMER2.0 database (http://timer.cistrome.org). E) Scatter dot (left) and waterfall (right) plot of GSTP1 mRNA levels in paired LUAD and non‐tumor (NT) lung of 45 local patients. F,G) Immunohistochemistry (IHC) images of high (F) and low (G) GSTP1 expression (brown staining) in LUAD. H,I) Kaplan–Meier curves and log‐rank tests of PFS (H), or OS (I), stratified by GSTP1 IHC expression categories in 197 LUAD of local patients. J) Cox regression analysis of PFS using GSTP1 expression level, pathological stage, gender, and smoking history as variables.

Figure 7

Patient‐derived tumor organoids of ALK‐translocated LUAD showed GSTP1 regulated CSC. A) Images of cancer organoids under bright‐field illumination (left) and by H&E stained histology (right). B) Identification of ALK‐translocation variant of LUAD‐OG1 by PCR. C) Sequencing results of patient‐derived primary tissue and LUAD‐OG1 verified the EML4‐ALK variant 1. D) Tumorsphere formation assay of LUAD‐OG1 with or without GSTP1‐KD. Representative bright field images of tumorsphere (left) and histograms of sphere numbers (right) were shown. E) In vitro limiting dilution assay for frequency of CSCs LUAD‐OG1 tumorspheres with or without GSTP1‐KD. CSC frequency was estimated using public ELDA online tool (http://bioinf.wehi.edu.au/software/elda/). F) Transwell migration assay of LUAD‐OG1 with or without GSTP1 downregulation. Histograms showed relative migrated proportion of the respective cells. G) In vitro effects of treatment with the GSTP1 inhibitor ezatiostat on crizotinib sensitivity in LUAD‐OG1 evaluated by Cell titer–Glo cell viability assay. H) Schematic diagram of the treatment regimen with 1% Tween 80/saline (Group 1), crizotinib (10 mg kg⁻¹; Group2), ezatiostat (25 mg kg⁻¹; Group 3), or the combination of ezatiostat with crizotinib (Group 4) in NOD‐SCID mice. n = 4 mice per group. I) Representative image of LUAD‐OG1 xenografts from the four groups at the endpoint are shown. J) Graph of tumor growth curve during 24 days treatment. K) Graph of animal body weight monitored twice a week. L) Ex vivo effects of GSTP1 inhibitor ezatiostat treatment on crizotinib sensitivity in crizotinib‐resistant LUAD‐OG1 evaluated by Cell titer–Glo cell viability assay. M) Schematic diagram of oral gavage of the treatment regimen with 1% Tween 80/saline (Group 1), crizotinib (25 mg kg⁻¹; Group2), ezatiostat (25 mg kg⁻¹; Group 3), or the combination of ezatiostat with crizotinib (Group 4) in SCID mice. n = 6–7 mice per group. N) Graph of tumor growth curve until the first death occurred (Day 31). O) Kaplan–Meier survival curves showing the tumor‐free survival rate of each group. Tumor volume exceeding 500 mm² was considered as human endpoint according to the guideline of animal ethics. * p < 0.05, ** p < 0.001, *** p < 0.005 versus respective control by Student's t‐test.

Figure 8

Schematic depiction of molecular mechanism of CSC regulation through hypoxia‐induced CaMK2A/NRF2/GSTP1 pathway. Hypoxia induced ROS stress triggers the increase of intracellular calcium level, and further activates CaMK2A by phosphorylation at T286. CaMK2A activation leads to an upregulated GSTP1 via direct phosphorylation of NRF2 at S558 residue in a KEAP1 independent manner. In CSC where the CaMK2A/NRF2/GSTP1 axis is activated, ROS homeostasis is maintained by GSTP1 upregulation, enabling self‐renewal, cell survival, drug resistance, tumor longevity, and propagation.