Elevated Ninj1 expression promotes lung tumorigenesis
To investigate the possible role of Ninj1 in the context of lung tumors, we performed immunohistochemical (IHC) analysis of Ninj1 expression in a tissue microarray composed of NSCLC tissues (n = 40) and normal lung tissues (n = 10). Significantly greater Ninj1 staining was observed in NSCLC tissues compared to that in normal lung tissues (P < 0.001) (Fig. 1a). Western blot analysis of tissue samples from the certified human bioresource bank in Korea [34] confirmed significantly elevated Ninj1 protein expression in lung tumor tissues (n = 10) compared to that in normal lung tissues (n = 8) (P = 0.0014) (Fig. 1b). Analyses of publicly available datasets from NSCLC patients further revealed that NINJ1 mRNA expression was significantly elevated in tumor tissues compared to levels in normal tissues (P < 0.001) (Fig. 1c) and was associated with poor overall and relapse-free survival in NSCLC patients (OS: P = 0.0105; RFS: P = 0.0172) (Fig. 1d). Next, we analyzed Ninj1 expression in mice harboring lung tumors caused by exposure to carcinogens such as urethane or the combination of 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone and benzo[a]pyrene (tobacco carcinogens, TC) [35, 36] (Fig. 1e, S1a) or caused by the oncogenic Kras mutation (KrasG12D/+) that is an established a genetic alteration that is characteristic of lung cancer [37](Fig. 1f). Consistent upregulation of Ninj1 expression was observed in these tumors compared to levels in normal lung tissues in control mice (P < 0.001).
To investigate the functional role of Ninj1 in lung tumorigenesis, we created a conditional transgenic (Tg) mouse model termed LoxP-stop-LoxP (LSL)-Ninj1Tg/+ and then crossed it with mice harboring the Scgb1a1-CreERTM recombinase-estrogen receptor fusion protein transgene [38] (Fig. S1b). We treated LSL-Scgb1a1-CreERTM;LSL-Ninj1Tg/+ mice with five doses of tamoxifen (TM) at 3 weeks of age and confirmed the presence of increased Ninj1 mRNA levels in the lungs through the use of PCR analysis (Fig. 1g). Immunofluorescence (IF) staining of lung tissues further revealed that approximately 80% of CCSP+ cells were Ninj1+ in Scgb1a1-CreERTM;L-Ninj1Tg/+ mice (Fig. 1h). No tumor nodules were observed in Scgb1a1-CreERTM;L-Ninj1Tg/+ mice at up to 1 year of age. Lung tumor initiation was induced by exposing Scgb1a1-CreERTM;L-Ninj1Tg/+ mice to urethane at 4 weeks of age. Bioluminescence imaging (Fig. 1i) and gross evaluation of the lung (Fig. 1j, k) revealed significantly more lung tumor nodules in Scgb1a1-CreERTM;L-Ninj1Tg/+ mice compared to that in Scgb1a1-CreERTM;LSL-Ninj1Tg/+ mice following exposure to urethane (P = 0.0011). Microscopic analyses revealed that the number of tumor nodules (P = 0.0259) (Fig. 1l), particularly those larger than 1 mm3 (Fig. 1m), was significantly higher in Scgb1a1-CreERTM;L-Ninj1Tg/+ mice than it was in Scgb1a1-CreERTM;LSL-Ninj1Tg/+ mice (1 mm3 ≤ volume < 5 mm3: P < 0.001; volume ≥ 5 mm3: P = 0.0154).
Next, LSL-Ninj1Tg/+;Sftpc-CreERT2;KrasG12D/+ mice were established by crossing LSL-Ninj1Tg/+;Sftpc-CreERT2 mice with KrasG12D/+ Tg mice. TM exposure-induced Sftpc-CreERT2;LSL-Ninj1Tg/+;KrasG12D/+ mice exhibited increased Ninj1 mRNA levels in the lung (Fig. 1n) and elevated Ninj1 protein expression in approximately 80% of SPC+ alveolar type II epithelial cells (AT2s) (Fig. 1o). Sftpc-CreERT2;L-Ninj1Tg/+;KrasG12D/+ mice exhibited significantly increased lung tumor formation compared to that in Sftpc-CreERT2;LSL-Ninj1Tg/+;KrasG12D/+ mice according to bioluminescence imaging (Fig. 1p) and gross analysis of lung tumor nodules (P = 0.0056) (Fig. 1q, r). Microscopic analyses revealed that Sftpc-CreERT2;L-Ninj1Tg/+;KrasG12D/+ mice possessed significantly higher amounts of tumor nodules (P = 0.001) (Fig. 1s), particularly those larger than 1 mm3 (volume < 1 mm3: p = 0.0406; 1 mm3 ≤ volume < 5 mm3: P = 0.0088; volume ≥ 5 mm3, P = 0.0157) (Fig. 1t), and exhibited worse overall survival (P < 0.001) (Fig. 1u) compared to these characteristics in control mice. These results collectively indicate the role of Ninj1 as a driver of lung tumorigenesis and suggest that this protein may serve as a clinically useful biomarker for poor prognosis in patients with NSCLC.
Ninj1 induces CSC phenotypes and survival potential against diverse cell death inducers, thus promoting lung tumor formation
We investigated mechanism by which Ninj1 promotes the growth of lung tumors. As NINJ1 has been reported as a p53 target gene [21], we analyzed Ninj1 expression in normal human bronchial epithelial (NHBE) cell lines (HB56B, BEAS-2B, and HBE), 13 human NSCLC cell lines carrying wild-type (WT) TP53 that encodes the p53 protein (H1944, H226B, H460, H292, and A549), or mutant/null TP53 cell lines (Calu-1, H1975, HCC827, H522, HCC15, H1299, H226Br, and PC-9) (Table S1). As anticipated, most NSCLC cell lines possessed higher levels of Ninj1 expression compared to that in the three NHBE cell lines (Fig. 2a). However, neither TP53 mutations nor KRAS or epidermal growth factor receptor (EGFR) mutations exhibited an obvious correlation with basal levels of Ninj1 expression in these NSCLC cells (Table S1). As diverse pyroptotic, necrotic, and apoptotic cell death inducers have been demonstrated to induce p53-dependent increases in Ninj1 expression in macrophages [22], we next analyzed the response of NSCLC cell lines carrying WT or mutant/null TP53 to pyroptotic (i.e., hypoxia, paclitaxel, and cisplatin) [39, 40], necrotic (i.e., hypoxia, glucose deprivation, paclitaxel, and cisplatin) [41,42,43,44], and apoptotic (i.e., hypoxia, serum or glucose deprivation, paclitaxel, and cisplatin) stimuli [41, 45]. Indeed, NSCLC cell lines carrying WT TP53 (H460 and A549) and not those carrying null/mutant TP53 (H1299 and H226Br) exhibited marked increases in the mRNA (Fig. 2b) and protein expression (Fig. 2c) of Ninj1.
We also investigated how Ninj1 promotes the growth of lung tumors by employing Ninj1high and Ninj1low NSCLC cell subpopulations that were sorted using flow cytometry (Fig. 2d). Compared to the Ninj1low subpopulations (lower 28.6% of live cells) from H460 and A549 cells, Ninj1high cells (upper 4.5% of live cells) derived from the corresponding parental cells exhibited similar Ki67 expression (Fig. 2e). In contrast, the results from a clonogenic assay (i.e., an anchorage-dependent colony formation assay that evaluates the survival of a single cell and proliferation into a colony) [46] (Fig. 2f, h) and from a soft-agar colony formation assay (an anchorage-independent formation assay that evaluates the survival and proliferation of cells within a harsh environment under unattached conditions) [47] (Fig. 2g, i) revealed more prominent colony-forming capacities in the Ninj1high subpopulations than were observed in the Ninj1low subpopulations when cultured under normal culture conditions (Fig. 2f, g) or in the presence of diverse cell death inducers (Fig. 2h, i).
Given the traits of CSCs in regard to resistance to hazardous microenvironments [4], we hypothesized that Ninj1 endows NSCLC cells with CSC phenotypes, including survival capacity against hazardous environments. Ninj1high subpopulations in H460 and A549 cells exhibited increased CSC properties, including CSC-associated marker gene expression (i.e., ALDH1A1, POU5F1, NANOG, and SOX2) (Fig. 2j), tumorsphere formation [8] (Fig. 2k), and tumorigenicity in the limiting dilution assay (P = 0.0171) (Fig. 2l) compared to these characteristics the Ninj1low subpopulations and their corresponding NSCLC cells. The H460 and A549 subpopulations obtained from sphere-forming culture conditions also possessed increased Ninj1 expression (Fig. 2m) and CSC-associated marker gene expression (Fig. 2n) compared to that of their corresponding NSCLC cells cultured under monolayer conditions. Moreover, subpopulations from H460, A549, and H226B cells were obtained by increasing the activity of aldehyde dehydrogenase (ALDH) [48] (Fig. 2o), another general property of CSCs, and these cells consistently possessed upregulated Ninj1 protein (Fig. 2p) and mRNA (Fig. 2q) expression levels of Ninj1 compared to levels in ALDHlow subpopulations and their corresponding NSCLC cells.
To provide direct evidence for the functional role of Ninj1 in CSC phenotypes and tumorigenic activities in NSCLC cells, we selected Ninj1low (H1299 and H226Br cells) and Ninj1high (A549 and H460) expression cells and established their sublines that were stably transfected with an expression vector that was empty (EV) or carrying either human Ninj1 or control or Ninj1-specific shRNA, respectively. The established cells that exhibited forced overexpression (H1299-Ninj1 and H226Br-Ninj1) or downregulation (A549-shNinj1 and H460-shNinj1) of Ninj1 expression and their corresponding control cells (H1299-EV, H226Br-EV, A549-shCon, or H460-shCon)(Fig. 3a) possessed similar proliferation rates (Fig. 3b). In agreement with the role of Ninj1 as a cohesion molecule [17], the established cells possessing upregulation or downregulation of Ninj1 expression exhibited significantly increased or decreased cell-cell cohesion, respectively, without any detectable changes in their adhesion to extracellular matrix (ECM) components such as type I collagen (Col) and fibronectin (Fn) (Fig. 3c).
When cultured under normal conditions, H1299-Ninj1 and H226Br-Ninj1 cells exhibited significantly greater capacities for colony formation, while A549-shNinj1 and H460-shNinj1 cells possessed decreased colony formation capacities compared to those of the corresponding control cells (Fig. 3d). When exposed to the diverse cell death inducers described above, H1299-Ninj1 cells exhibited significantly increased anchorage-dependent (Fig. 3e) and anchorage-independent (Fig. 3f) colony-forming capacity and decreased caspase 1 and caspase 3 cleavage (markers for pyroptotic or apoptotic cell death, respectively [49]) (Fig. 3g) compared to that of their control cells. Conversely, resistance to these cell death inducers was significantly decreased in H460-shNinj1 cells compared to that in H460-shCon cells (Fig. 3e-g). Furthermore, compared to their control cells, CSC-associated phenotypes, including protein (Fig. 3h) and mRNA (Fig. 3i) expression of CSC markers and ALDH activity (Fig. 3j), were higher in H1299-Ninj1 cells and attenuated in H460-shNinj1 cells. Consistently, sphere-forming activities were significantly increased by Ninj1 expression and were attenuated by Ninj1 silencing (Fig. 3k).
Next, we analyzed the tumorigenic capacity of these established NSCLC cells using an in vivo limiting dilution assay. H1299-Ninj1 and H226Br-Ninj1 cells possessed significantly greater tumorigenicity than did their corresponding control cells, while A549-shNinj1 and H460-shNinj1 cells exhibited significantly decreased tumorigenicity (Fig. 3l). Once developed, xenograft tumors from H1299-Ninj1 and H226Br-Ninj1 cells displayed significantly faster growth than did their control tumors, while those from A549-shNinj1 and H460-shNinj1 cells exhibited significantly slower growth compared to that of their control tumors (Fig. 3m). The expression of CSC markers (Oct4 and Nanog) was increased in H1299-Ninj1 xenograft tumors and was attenuated in H460-shNinj1 xenograft tumors compared to that in their corresponding control tumors (Fig. S2a). An associated elevation in Ninj1 and Nanog expression was also observed in tumor nodules in Scgb1a1-CreERTM;L-Ninj1Tg/+ and Sftpc-CreERT2;L-Ninj1Tg/+;KrasG12D/+ mice (Fig. S2b). These findings suggest that Ninj1 expression that is increased either through innate mechanisms or through the action of cell death inducers protects NSCLC cells from various environmental insults in the tumor, thus promoting tumor development and growth.
Ninj1high subpopulations in human NSCLC exhibit increased CSC traits and survival potential against pyroptotic, necrotic and apoptotic cell death inducers
To assess the clinical relevance of these findings, we analyzed the role of Ninj1 expression in the functional features of CSCs in NSCLC cells obtained from patient-derived tumors (Fig. 4a). IF staining of the ALDHhigh subpopulation (Fig. 4b) and western blot analysis of the sphere-forming subpopulation (Fig. 4c) within primary cultured patient-derived NSCLC cells revealed elevated Ninj1 protein levels compared to levels in their corresponding controls. Additionally, increased mRNA levels of Ninj1 and CSC marker genes were observed in the ALDHhigh (Fig. 4d) and sphere-forming (Fig. 4e) subpopulations compared to levels in the controls. Ninj1high subpopulations within the tumors also possessed significantly increased capacities for sphere formation and CSC marker gene expression compared to that of their corresponding Ninj1low subpopulations (Fig. 4f). Analysis of publicly available datasets from patients with NSCLC (GSE77803) further revealed positive correlations between the expression levels of NINJ1 and CSC markers (Fig. 4g). When Ninj1 expression in primary cultured cells was depleted using siRNAs, the expression of CSC marker genes (Fig. 4h) and ALDH activity (Fig. 4i) were significantly decreased.
Next, we analyzed the role of Ninj1 in the resistance of patient-derived NSCLC cells to various cell death inducers. Similar to the results from NSCLC cell lines, the aforementioned cell death inducers caused marked increases in Ninj1 expression without consistent changes in caspase 1 and caspase 3 cleavage events in the primarily cultured cells (Fig. S3a). Compared to their corresponding Ninj1low subpopulations, Ninj1high subpopulations also possessed similar Ki67 expression but greater anchorage-dependent and anchorage-independent colony formation when cultured in normal culture conditions or in the presence of various cell death inducers (Fig. S3b-f).
We further analyzed mice harboring patient-derived xenograft (PDX) tumors that had been treated with three cycles of a clinically relevant combinatorial chemotherapeutic regimen (i.e., a 7-day regimen comprising paclitaxel and cisplatin for 1 d, followed by a 6-day drug holiday) [33] (Fig. 4j). The Ninj1high populations within the tumors were monitored before and after chemotherapy. The PDX tumors shrank to < 50% of their original volume after treatment (P < 0.001) (Fig. 4k). IF staining revealed increased numbers of Ninj1+ and Nanog+ cells within the residual tumors after chemotherapy compared to those in untreated control tumors, and there was a positive correlation between Ninj1 and Nanog expression (Fig. 4l, S3g). Taken together, these results indicate the presence of distinct Ninj1high subpopulations possessing CSC phenotypes in human NSCLC.
Ninj1-mediated activation of the canonical Wnt/β-catenin signaling pathway
To elucidate the mechanism by which Ninj1 mediates the acquisition of CSC phenotypes, we investigated the effects of Ninj1 expression on the Wnt/β-catenin, Notch, and Hedgehog pathways that play critical roles in stem cell function [12, 13]. The expression of the representative target genes of the Wnt/β-catenin signaling pathway was significantly upregulated in H1299-Ninj1 and H226Br-Ninj1 cells and was decreased in H460-shNinj1 and A549-shNinj1 cells compared to levels in their respective control cells, while the expression levels of the Notch, Hedgehog, and Hippo pathways did not exhibit consistent changes in these cells (Fig. 5a). Notably, Ninj1 induced a decrease in the level of active β-catenin (β-cateninact) that functions to mediate canonical Wnt signaling [9], while no detectable changes were observed in non-canonical Wnt signaling mediators, including phosphorylated forms of c-Jun N-terminal kinase (JNK), protein kinase C (PKC), and c-Jun, thus indicating Ninj1-mediated regulation of the canonical Wnt signaling pathway (Fig. 5b). We then explored the direct evidence supporting the functional involvement of Ninj1 in the activation of the Wnt/β-catenin signaling pathway. The TOPFlash luciferase reporter assay (Fig. 5c, left), a common tool used to measure the activation of the Wnt/β-catenin signaling pathway [50], and real-time PCR analysis of AXIN2 and MYC expression (Fig. 5d, left) as representative target genes of the pathway [51] revealed increased activation of the Wnt/β-catenin pathway in H1299-Ninj1 and H226Br-Ninj1 cells compared to that in their corresponding control cells. Ninj1-mediated Wnt/β-catenin signaling is further enhanced by the addition of exogenous Wnt3a, a Wnt ligand that activates the canonical Wnt signaling pathway and promotes lung cancer progression [16]. In contrast, A549-shNinj1 and H460-shNinj1 cells exhibited significant attenuation in Wnt/β-catenin signaling events compared to that of their corresponding control cells (Fig. 5b-d, right; Fig. S4a). Additionally, western blot (Fig. 5e) and IF (Fig. 5f) analyses revealed that the Wnt3a-mediated nuclear localization of β-catenin (Fig. 5e) and β-catenin (Fig. 5f) was markedly suppressed in H460-shNinj1 cells compared to that in their corresponding control cells. We further observed significantly greater levels of nuclear β-catenin expression in the FACS-sorted Ninj1high populations from H460 cells (P < 0.001) (Fig. 5g) and three different PDXs (P < 0.001) (Fig. 5h, S4b) compared to that observed in their corresponding Ninj1low populations. CCSP+ club cells in lung tumors from urethane-exposed Scgb1a1-CreERTM;L-Ninj1Tg/+ mice and SPC+AT2s in lung tumors from Sftpc-CreERT2;L-Ninj1Tg/+;KrasG12D/+ mice also possessed significantly higher nuclear β-catenin expression compared to that in the corresponding control cells from urethane-exposed Scgb1a1-CreERTM;LSL-Ninj1Tg/+ mice (P = 0.0138) and Sftpc-CreERT2;LSL-Ninj1Tg/+;KrasG12D/+ mice (P = 0.0107), respectively (Fig. 5i, S4c). Using IF analyses of an NSCLC tissue microarray (n = 40), we further demonstrated significant increases in the nuclear β-catenin+Ninj1+ populations in tumors compared to those in normal tissues (P < 0.001) (Fig. 5j). These results suggested that Ninj1 is involved in the activation of the canonical Wnt/β-catenin signaling pathway in NSCLC.
Given the general role of the Wnt/β-catenin signaling pathway in human cancers [9, 10, 14], we assessed the pathological role of Ninj1 in histologically distinct epithelial tumors, including breast and colon cancers. Public data analysis revealed that NINJ1 expression was a poor prognostic factor in these cancers (Fig. S5a). Positive correlations between the expression levels of NINJ1 and SOX2 were observed in these cancers (Fig. S5b). We further validated the significantly increased CSC marker gene expression in the Ninj1high population from patients with breast and colorectal cancers compared to that in the Ninj1low population from the corresponding tumors (Fig. S5c). Hence, Ninj1 may be implicated in the development and progression of various human cancers.
Ninj1-mediated promotion of the assembly of the LRP6-FZD2 signalosome
We investigated the mechanism by which Ninj1 activates the Wnt/β-catenin signaling pathway. The mRNA levels of Wnt1-8a, seven-span FZD1–10, single-span LRP5 and LRP6, cytosolic effectors Dvl2 and Dvl3, and β-catenin were either decreased or remained unchanged in H1299-Ninj1 cells compared to those in H1299-EV cells (Fig. 6a, S6a), thus suggesting that Ninj1-mediated Wnt/β-catenin signaling occurs through post-transcriptional mechanisms. As Wnt binding to two cell surface receptors (LRP5/6 and FZD) is known to induce LRP5/6 phosphorylation through the intermediation of Disheveled (Dsh; Dvl in mammals), Axin, and its associated kinase GSK-3 [9, 10] to thus release β-catenin from the multiprotein destruction complex [52], we hypothesized that Ninj1 may regulate β-catenin stability. Indeed, upon treatment with the protein synthesis inhibitor cycloheximide [53], the half-life of β-catenin was significantly increased in H1299-Ninj1 cells and was decreased in H460-shNinj1 cells and PDX-derived primary tumor cells transfected with Ninj1 siRNAs (PDX-siNinj1) compared to that in the corresponding control cells (Fig. 6b, S6b, S6c). Moreover, pretreatment with MG132 increased β-catenin levels in H1299-EV, H460-shNinj1, and PDX-siNinj1 cells compared to levels in H1299-Ninj1, H460-shCon, and PDX-Scr cells, respectively (Fig. 6c, S6d). Notably, compared to control cells, H1299-Ninj1, H460-shNinj1, and PDX-siNinj1 cells exhibited increased and decreased LRP6 phosphorylation, respectively, without detectable changes in LRP6, FZD2, Dvl3, Axin1, and GSK-3β protein expression (Fig. 6d, S6e). Moreover, Wnt3-mediated LRP6 phosphorylation was markedly enhanced in H1299-Ninj1 cells and was attenuated in H460-shNinj1 cells (Fig. S6f). In contrast, the phosphorylation of other receptor tyrosine kinases, including epidermal growth factor receptor (EGFR) and insulin-like growth factor receptor (IGF-1R), was not affected by the modulation of Ninj1 expression (Fig. 6d, S6e). We then assessed if Ninj1 can induce ligand-independent activation of the LRP6/β-catenin signaling cascade by utilizing IWP-2, a Wnt/β-catenin inhibitor that blocks Porcn-mediated Wnt palmitoylation [54]. IWP-2 treatment effectively blocked LRP6 phosphorylation and β-catenin activation in H1299-EV cells (Fig. 6e). In contrast, Ninj1-mediated LRP6 phosphorylation and β-catenin activation remained unchanged in H1299-Ninj1 cells following IWP-2 treatment. Thus, Ninj1 appears to possess the capacity to activate LRP6 in a ligand-independent manner.
We then determined if Ninj1 interacts with LRP signaling components. Co-immunoprecipitation analyses revealed an association between Ninj1 and LRP6, FZD2, Dvl3, Axin1, and GSK-3β in PDX-Scr and H460 cells, and this association did not occur in PDX-siNinj1 and H460-shNinj1 cells (Fig. 6f, S6g). LRP6 co-immunoprecipitation with Ninj1, FZD2, Dvl3, Axin1, and GSK-3β was also observed in H460 cells but not in H460-shNinj1 cells (Fig. 6g). Co-immunoprecipitation of Ninj1, LRP6, FZD2, Dvl3, Axin1, and GSK-3β was also observed in H1299-Ninj1 cells and not in H1299-EV cells (Fig. 6h). ROR1, another Wnt receptor [9], was observed to interact with LRP6 [55]. Indeed, LRP6 and not Ninj1 co-immunoprecipitated with ROR1 (Fig. 6f-h). A pull-down assay using GST-tagged Ninj1 recombinant proteins revealed that Ninj1 as the full-length (FL) protein or N-terminal (NT) domain was associated with LRP6, FZD2, Dvl3, Axin1, and GSK-3β (Fig. 6i). These findings suggest that the Ninj1-mediated Wnt/β-catenin signaling pathway occurs through the assembly of the LRP6-FZD2 signalosome.