Aberrant RSPO3-LGR4 signaling in Keap1-deficient lung adenocarcinomas promotes tumor aggressiveness

Abstract

The four R-spondins (RSPO1-4) and their three related receptors LGR4, 5 and 6 (LGR4-6) have emerged as a major ligand-receptor system with critical roles in development and stem cell survival through modulation of Wnt signaling. Recurrent, gain-of-expression gene fusions of RSPO2 (to EIF3E) and RSPO3 (to PTPRK) occur in a subset of human colorectal cancer. However, the exact roles and mechanisms of the RSPO-LGR system in oncogenesis remain largely unknown. We found that RSPO3 is aberrantly expressed at high levels in approximately half of the Keap1-mutated lung adenocarcinomas. This high RSPO3 expression is driven by a combination of demethylation of its own promoter region and deficiency in Keap1 instead of gene fusion as in colon cancer. Patients with RSPO3-high tumors (~9%, 36/412) displayed much poorer survival than the rest of the cohorts (median survival of 28 vs. 163 months, logrank test p < 0.0001). Knockdown of RSPO3, LGR4, or their signaling mediator IQGAP1 in lung cancer cell lines with Keap1 deficiency and high RSPO3-LGR4 expression led to reduction in cell proliferation and migration in vitro, and knockdown of LGR4 or IQGAP1 resulted in decrease in tumor growth and metastasis in vivo. These findings suggest that aberrant RSPO3-LGR4 signaling potentially acts as a driving mechanism in the aggressiveness of Keap1-deficient lung adenocarcinomas.

Introduction

R-spondins are a group of four highly related secreted proteins (RSPO1-4) with critical roles in embryonic development and organogenesis as well as in the self-renewal and survival of adult stem cells.¹ In particular, loss of RSPO2 led to hypoplasia and reduced branching of the lung during mouse development.^{2, 3} Work from us and others demonstrated that RSPOs activate three related receptors LGR4-6 (leucine-rich repeat-containing, G protein-coupled receptor 4, 5, and 6) to potentiate Wnt signaling.^4-6 LGR4-6 contain a large extracellular domain with 17 leucine-rich repeats and a seven transmembrane (7TM) domain homologous to members of the rhodopsin family of G protein-coupled receptors.^7-9 LGR4-bound RSPOs directly interact with two membrane-bound E3 ligases (RNF43 and ZNRF3) which otherwise ubiquitinate Fzd receptors for degradation.¹⁰ Formation of the LGR4-RSPO-RNF43/ZNRF3 ternary complex induces the clearance of the E3 ligases, leading to reduced ubiquitination and eventually elevated levels of Wnt receptors on the cell surface and increased Wnt signaling.¹⁰ Just recently, we identified IQGAP1 as an LGR4-binding protein and showed that it plays an essential role in RSPO-LGR4-induced potentiation of Wnt signaling.¹¹ IQGAP1 is an intracellular scaffold protein that binds to and modulates the activities of a plethora of signaling molecules to regulate cell adhesion and migration.^{12, 13} We found that RSPO-LGR4 not only induces the clearance of RNF43/ZNRF3 but also increases the affinity of IQGAP1 for DVL bound to the Wnt signalosome. This leads to the formation of a supercomplex between RSPO-LGR4 and Wnt receptors. In this configuration, IQGAP1 brings in MEK1/2 to phosphorylate LRP5/6 for the β-catenin-dependent pathway and N-WASP/mDia1 to coordinate actin dynamics for the β-catenin-independent pathway.¹¹

Dysregulation of Wnt signaling occurs in nearly every major type of solid tumors. Gain-of-expression gene fusions of RSPO2 (to EIF3E) and RSPO3 (to PTPRK) were identified in 10% (7/68) of human colon cancer.¹⁴ The fusions were inferred to have a driving role in the carcinogenesis of the affected tumors due to their recurrent occurrence and exclusivity with Apc/β-catenin mutations.¹⁴ In MMTV-induced mouse models of breast and colon cancer, RSPO2 and RSPO3 were two of the most frequent viral integration sites, and ectopic expression of RSPO2/3 in mouse mammary epithelial cells increased tumor formation and metastasis.^15-17 Furthermore, knockout of LGR4 in mice led to profound hypoplasia and impaired tubulogenesis in multiple organs during development,^18-20 suggesting a critical role of LGR4 in the regulation of cell proliferation and migration. Intriguingly, LGR4 was found to be highly upregulated in both adenocarcinomas (AD) and squamous cell carcinomas (SqCC) of non-small cell lung cancer (NSCLC) despite low expression in normal adult lung.²¹ We found that RSPO3 was highly expressed in a subset of adenocarcinomas (ADs). Here we show that the aberrant RSPO3 expression in lung ADs was not driven by PTPRK fusion as in colon cancer, and that RSPO3-LGR4 signaling plays a major role in the aggressiveness of RSPO3-high tumors.

Results

RSPO3 is aberrantly expressed in a subset of lung ADs and its high expression is associated with poor survival

We mined the RNA-Seq data of LGR4-6, RSPO1-4, and other genes encoding Wnt ligands, receptors, and modulators in TCGA’s provisional lung AD cohort (LUAD, 230 samples) as provided by the cBio Portal website.²² Based on the RSEM values (RNA-Seq by Expectation-Maximization) ²³ of the transcripts, LGR4 was expressed at moderate to high levels in ~90% of the cases while LGR5 and LGR6 were expressed at much lower levels in nearly all samples (Supplementary Figure S1a). A similar pattern of expression for LGR4-6 was also found in the 186 lung cancer cell line collection characterized by the Cancer Cell Line Encyclopedia (CCLE) project ²⁴ (Supplementary Figure S1b). Using our recently characterized LGR4 monoclonal antibody 7E7²⁵, we analyzed the expression of LGR4 in normal and AD samples of the lung on tissue microarrays. LGR4 staining was detected in all AD samples with two representatives shown (Figure 1a-b). No expression of LGR4 was observed in respiratory epithelial or alveolar cells in normal adult lung (Figure 1c), consistent with previous report of LGR4 upregulation in lung ADs ²¹.

Figure 1.

High level of RSPO3 expression is correlated with poor survival in lung AD. (a-c), IHC staining of LGR4 in two lung AD samples (a-b) and normal adult lung (c). (d) Expression distribution of RSPO3 and LGR4 in 230 lung AD samples of TCGA’s LUAD cohort and RSPO3 in CCLE’s 186 lung cancer cell lines. Values in log2 of expression signal (RSEM for TCGA samples and RMA for CCLE samples) were sorted from low to high for each gene and plotted. (e) Kaplan-Meier survival plot of patients with tumors of high RSPO3 expression vs. the rest of the cohort in the TCGA LUAD set. RSPO3-high was defined as samples with log2[RSEM] values at 1 SD above the population mean (Z = +1) while the remaining samples were defined as RSPO-norm. (f) Scatter plot of RSPO3 expression among different tumor stages in TCGA’s LUAD cohort. (g) Kaplan-Meier survival plot of patients with RSPO3-high vs. RSPO3-norm tumors in Stages I-II of TCGA’s LUAD cohort. (h) Kaplan-Meier survival plot of patients with RSPO3-high vs. RSPO3-norm tumors in Stages III-IV of TCGA’s LUAD cohort. (I) Kaplan-Meier survival plot of data pooled from three lung cancer studies (TCGA LUAD, GSE3141 and GSE31210).

Of the potential RSPO ligands, RSPO3 was expressed at significant levels with a highly skewed pattern, i.e., high levels in ~10% (25/230) of the tumors with no to little expression in the rest of the cases (Figure 1d). The other three RSPOs were expressed at much lower, insignificant levels with two exceptions in RSPO2 (Supplementary Figure S1c). Comparison of the RSPO3 expression data between tumors and their matched normal tissues revealed that the high RSPO3 expression in tumors was due to upregulation (Supplementary Figure S2). Furthermore, the distributions of RSPO1-4 in CCLE’s lung cancer cell line collection are also similar to those in primary tumors (Figure 1d and Supplementary Figure S1d), i.e., RSPO3 is highly expressed in a subset (17%, 32/186) of the cell lines using a cutoff value of 4x above background whereas the other three RSPOs are rarely expressed. Taken together, these data indicate that LGR4 is abundantly expressed in the majority of primary lung ADs with a subset of them co-expressing high levels of RSPO3.

RSPO-LGR4-IQGAP1 signaling has potent effects on cell proliferation, adhesion, and migration¹¹. As metastasis is the major cause of cancer-related death, we asked if high RSPO3 expression is associated with patient survival. In univariate analysis of the TCGA LUAD cohort, patients with RSPO3-high tumors (defined as those with log2-RSEM values of RSPO3 at 1 standard deviation above the mean, i.e., Z = +1) showed much worse overall survival than the rest of the cohort (Figure 1e, median survival of 22 vs. 49 months, logrank test p = 0.008; hazard ratio (HR) = 3.8, 95% confidence interval (CI) = 1.7 to 8.8). In multivariate analysis using a continuous variable Cox proportional hazards model, RSPO3 was found to be predictive of outcomes independent of gender, age, and smoking status (p = 0.03, HR = 2.1, 95% CI = 1.1 – 4.3). The incidence of RSPO3-high tumors did not vary significantly among tumor stages (Figure 1f) and high RSPO3 expression was significantly correlated with poor survival in both early (I-II) and late (III-IV) stages of the cancer (Figure 1g-h), suggesting that upregulation of RSPO3 occurred as an early event during tumor progression. We then analyzed expression and survival data of two other independent lung cancer studies in the GSE database (GSE3141 and GSE31210).^{26, 27} Statistically significant association between high RSPO3 expression and poor survival was found in one cohort (GSE3141, Supplementary Figure S3a; median survival of 24 vs. 49 months, p = 0.001; HR = 14.7, 95% CI = 3.1 to 69.1) and a trend existed in the other study (Supplementary Figure S3b, logrank test p = 0.06, Gehan-Breslow-Wilcoxon Test p = 0.04). Of note, the frequency of RSPO3-high tumors and the overall survival time in the GSE3141 cohort are similar to those of the TCGA set whereas the GSE31210 cohort had lower frequency of RSPO3-high tumors (2% vs. 11%). Nevertheless, when the data from all three cohorts were pooled, high RSPO3 expression was found in ~9% (36/412) of the cases, and it was associated with poor survival (Figure 1i, median survival of 28.4 vs. 163 months, p < 0.0001; HR = 15.7, 95% CI = 6.8 to 36.2).

High RSPO3 expression is driven by a combination of Keap1 deficiency and demethylation of its own promoter

To identify the mechanisms of RSPO3 upregulation in a subset of lung ADs, we first searched for fusions to PTPRK, as was reported in colon cancer. To validate this approach, we analyzed the RNA-Seq data from TCGA’s colorectal cancer cohort²⁸, and found that high levels of RSPO3 occurred in 2% (4/220) of the tumors (Supplementary Figure S4). Importantly, transcripts from all four cases of RSPO3-high tumors contained recurrent fusions between exon-1 of PTPRK and exon-2 of RSPO3 (Supplementary Table S1) as described previously.¹⁴ Furthermore, none of the four RSPO3-high tumors contained mutations in Apc or β-catenin, consistent with their exclusivity with mutations in the Wnt pathway.¹⁴ When the same algorithm was applied to RNA-Seq data of the 11 lung AD samples with the highest RSPO3 expression, however, no fusion of RSPO3 to PTPRK or to any other gene was detected. These results suggest that in lung ADs, aberrant expression of RSPO3 was not due to fusion with PTPRK.

Next, we examined if RSPO3-high tumors were enriched for mutation of other genes commonly altered in lung AD. Initial analysis using the OncoPrint function of cBioPortal showed 100% correlation between upregulation of RSPO3 expression (defined those with Z > +1 by cBioPortal) and deficiency in Keap1 (Fisher’s exact test p = 0) (Supplementary Figure S5a). No association was found between high expression of RSPO3 and the mutation status of other cancer genes such as TP53, KRAS, and EGFR (Supplementary Figure S5a). Keap1 is a negative regulator of the transcription factor NRF2 (encoded by NFE2L2) and is mutated in NSCLC with an overall frequency of ~15%^29-31. Keap1 and NRF2 abnormalities were previously shown to be associated with poor outcome in NSCLC patients.^{32, 33} In this TCGA LUAD cohort, Keap1 was mutated or down-regulated in 17% (38/230, mutated = 24, low expression is estimated to be 14) of the tumors, whereas no mutation or upregulation of NFE2L2 was seen. Remarkably, 20 of the 25 RSPO3-high tumors had loss of Keap1 function due to either mutation or low expression (Figure 2a). The remaining five RSPO3-high tumors had no sequencing data on Keap1 at the time of this analysis, but all had high expression of Keap1/NRF2-regualted genes (Figure 2a), suggesting that gain of NRF2 activity occurred in these tumors. We also carried out a global analysis of genes that were associated with high RSPO3 expression in the TCGA LUAD set. The most prevalent finding is that RSPO3-high tumors were highly enriched in genes regulated by the Keap1-NRF2 pathway, such as ABCC2, AKR1C1, and AKR1C2 (Figure 2a). These data unequivocally show that high RSPO3 expression only occurred in a subset (approximately half) of Keap1-deficient tumors. We then asked if RSPO3 expression level differentiated survival of patients with Keap1-deficient tumors. Patients with tumors containing both high RSPO3 expression and Keap1 deficiency showed significantly shorter survival time than those with only Keap1 deficiency (Figure 2b, median survival of 22 vs. 49 months; logrank test p = 0.05, HR = 2.9, 95% CI = 1 to 8.8; Gehan-Breslow-Wilcoxon test p = 0.04). These results suggest that tumors with high RSPO3 expression largely account for the poor survival of patients with Keap1-deficienct tumors that was previously demonstrated in lung AD.^{32, 33}

Figure 2.

High RSPO3 expression in lung AD depends on Keap1 deficiency and RSPO3 promoter demethylation. (a) A heat map depicting the expression levels of RSPO3 and three NRF2-regulated genes vs. mutation and expression status of Keap1 in TCGA’s LUAD cohort. (b) Kaplan-Meier survival analysis of patients with tumors of Keap1 deficiency and high RSPO3 expression vs. those with tumors of only Keap1 deficiency and the rest of the cohort. (c) Representative WB of A549 cells stably expressing pLVX-GFP or pLVX-GFP-Keap1 with anti-GFP antibody (left panel), anti-Keap1 antibody (mid-panel) and anti-NRF-2 antibody (right panel). eKEap1: endogenous Keap1. The experiment was repeated twice. (d) RT-qPCR results of AKR1C2 and RSPO3 in GFP vector and GFP-Keap1-expressing cells. Expression data were normalized by 18S RNA and error bars are S.E.M. of three replicate experiments. (e) Representative ChIP-PCR results of three putative NRF2 binding sites located at -6824, -6043, and -5620 bp upstream of the TSS (transcription start site) of RSPO3. The antibodies used in the lanes are: 1 = anti-PolII, 2 = control IgG, 3 = anti-NRF2. Lanes 4 (no DNA) and 5 (input chromatin) are negative and positive controls, respectively. The experiments were repeated twice. (f) Comparison of methylation beta value of 22 CpG’s across the RSPO3 promoter region in RSPO3-high vs. RSPO3-norm tumors. (g) RT-qPCR results of RSPO3 expression in H2009 cells treated with vehicle, sulforaphane (SF, 10 μM) and azacytidine (AZ, 10 μM), and SF and AZ simultaneously (SF + AZ). Expression data were normalized by 18S RNA and error bars are S.E.M. of three replicate experiments.

To confirm the correlation between high RSPO3 expression and Keap1 deficiency, we tested if the Keap1-NRF2 pathway directly regulates RSPO3 transcription. Analysis of the microarray data of CCLE’s 186 lung cancer cell lines revealed that nearly all the genes that are best correlated with RSPO3 expression are in the Keap1-NRF2 pathway (Supplementary Figure S5b). A549 , one of the most commonly used cell lines for lung cancer studies, is mutated in Keap1,²⁹ and expressed high levels of RSPO3 and LGR4 (Supplementary Table S2). To test if RPSO3 expression is regulated by the Keap1-NRF2 pathway, GFP-Keap1 was expressed in A549 cells, which, as expected, reduced the level of NRF2 (Figure 2c). RT-qPCR analysis revealed that the mRNA level of RSPO3 in cells expressing GFP-Keap1 cells was reduced by ~80% compared to GFP control cells (Figure 2d). The expression of AKR1C2, a well-known target of NRF2, was nearly completely repressed in GFP-Keap1 cells (Figure 2d). To further confirm these results, we used the H1944 NSCLC cell line which expresses high levels of RSPO3 and other NRF2-regulated genes based on the gene expression data in CCLE. Expression of Keap1 in H1944 cells led to reduction in NRF2 level with commensurate decreases in mRNA levels of AKR1C2 and RSPO3 when compared to cells expressing GFP control (Supplementary Figure S6a-c). While these results suggest that overexpression of Keap1 led to reduced expression of RSPO3 in Keap1-deficient cells, it is worth noting that the level of recombinant Keap1-GFP was much higher than that of the endogenous form (Fig. 2c), which may have non-specific effect. Previously, genome-wide analysis of gene expression change following knockdown of NRF2 in A549 cells was carried out by others³⁴. We examined this dataset (GSE38332) and found that RSPO3 expression was decreased by 75% following the knockdown of NRF2 in A549 cells. Taken together, these results indicate that the high expression of RSPO3 in A549 and H1944 cells was sensitive to loss of NRF2, consistent with the correlation between high RSPO3 levels and Keap1 deficiency in tumor samples.

To find if the promoter region of RSPO3 contains NRF2 binding sites,³⁵ we screened a 50 kb genomic sequence surrounding the transcription start site (TSS) of RSPO3 and identified nine motifs that could potentially be bound by NRF2 (Supplementary Figure S7). ChIP (chromatin immunoprecipitation)-PCR was carried out for each of the nines sites and three consecutive sites (−6824, −6043, and −5620, Supplementary Figure S7) were found to be bound by NRF2 in A549 cells (Figure 2e). We also examined the Encode Genome data using the UCSC Genome Browser (http://genome.ucsc.edu/index.html) and found that there is a DNaseI hypersensitivity site encompassing the −6820 NRF2 binding site of RSPO3 in A549 cells. Overall, these bioinformatic and experimental data suggest that NRF2 directly binds to an upstream region of the RSPO3 promoter, which may play a major role in driving RSPO3 expression to aberrantly high levels in Keap1-deficient lung ADs.

Despite the clear relationship between Keap1 and RSPO3, deficiency of Keap1 alone was insufficient to drive RSPO3 expression as only approximately half of Keap1-deficient lung ADs had high levels of RSPO3 (Figure 2a). We then searched for other potential mechanism such as gene amplification, mutation burden, fraction of copy number altered genome, and DNA methylation. No significant difference was found between RSPO3-high and –norm tumors in RSPO3 copy number, mutation burden, or fraction copy number altered genome (Supplementary Figure S8a-d). Strikingly, RSPO3-high tumors had a significant decrease in the level of methylation, particularly so in a region around 1500 bp downstream of the TSS of RSPO3 (Figure 2f). Parallel analysis of mutation, expression, and methylation data of RSPO3 and Keap1 revealed that not all tumors with Keap1-deficiency and RSPO3 low methylation had high levels of RSPO3 (Supplementary Figure S9a-c), indicating that additional factors are required to achieve maximal expression for RSPO3. Nevertheless, these findings suggest that deficiency in Keap1 and demethylation of the RSPO3 promoter are necessary but not sufficient for high RSPO3 expression in lung ADs. To test this mechanism, we used the lung adenocarcinoma cell line H2009 which has neither Keap1 mutation (based on the lack of expression of NRF2-regulated genes) nor RSPO3 expression (Supplementary Table S2). H2009 cells were treated with either the demethylating agent azacytidine or the NRF2 activator sulforaphane, or both. RT-qPCR analysis showed that demethylation slightly increased RSPO3 expression while NRF2 activation alone had no effect (Figure 2g). Importantly, combination of demethylation and NRF2 activation led to much higher increase (~150-fold) in expression of RSPO3 (Figure 2g), suggesting that H2009 cells provide other factor(s) necessary for high RSPO3 expression. Overall, these results support the model where NRF2 activation and RSPO3 promoter demethylation are the major drivers of RSPO3 expression in lung ADs.

High RSPO3-LGR4 signaling in lung cancer cells promotes cell growth and migration in vitro

Recently, we showed that shRNA-mediated knockdown (KD) of LGR4 or RSPO3 in A549 and H460 led to impaired Wnt signaling¹¹. Here we compared the growth rates of A549 expressing either an effective LGR4 shRNA (shLGR4-40) or one of three ineffective ones (shLGR4-39, 41, and 42). Cell growth in real time was measured using the xCELLigence assay³⁶. Only the cell line with the effective LGR4 shRNA (#40) grew ~50% slower than the parental cells (Figure 3a). To confirm these results, we screened additional LGR4 shRNAs (a total of 15) and only found one (shLGR4-43) that gave a partial KD of LGR4. Cells expressing shLGR4-43 showed reduced Wnt signaling to a lesser extent than those with shLGR4-40 (Supplementary Figure S10a-b). Importantly, cells expressing shLGR4-43 displayed a significant decrease in cell growth, migration and invasion when compared to parental and control shRNA cells, but not to the same extent as cells of shLGR4-40 (Supplementary Figure S10c-e). Furthermore, A549 cell lines with complete KD of RSPO3 by two independent shRNAs (#63 and 67) which we described previously ¹¹ displayed a similar extent of decrease in the rate of cell growth (Figure 3b). The decrease in cell growth was further confirmed using the MTT cell proliferation assay for both LGR4 and RSPO3 KD cells (Supplementary Figure S11a-b). KD of LGR4 in H460 cells which are also mutated in Keap1 and express high levels of RSPO3 and LGR4 (Supplementary Table S2) resulted in a similar level of decrease in cell growth (Figure 3c). Repeated efforts (twice) failed to obtain H460 cells with stable KD of RSPO3 using either shRSPO3-63 or -67 whereas plenty of cells survived with vector control. These results suggest that endogenous RSPO3-LGR4 signaling plays an important role in the regulation of cell growth in both A549 and H460 cells.

Figure 3.

RSPO3-LGR4-IQGAP1 signaling regulates cell growth and migration in lung cancer cells. (a) Real time growth curves of A549 cells with KD of LGR4 (shLGR4-40) vs. parental A549 cells and three other cell lines expressing ineffective LGR4 shRNAs (#39, 41, and 42). (b) Real time growth curves of A549 cells with KD of RSPO3 (#63 and 67) vs. parental A549 cells and those expressing a control shRNA. (c) Real time growth curves of H460 cells with KD of LGR4 (shLGR4-40) vs. parental H460 cells and those expressing a control shRNA. (d) Migration results of H2009 cells treated with vehicle and RSPO3 at 10 and 100 ng/ml. *p < 0.05 vs. control, Dunnett’s test following One-way ANOVA (p < 0.0001). (e) WB results of LGR4 (top panel) in parental H2009 cells (P) and those expressing a control shRNA (C) or shLGR4-40 (KD) using the antibody 7E7, and WB results of IQGAP1 (lower panel) of parental (P) and H2009 cells expressing control shRNA (C) or shIQGAP1-85 (KD). Actin was probed as loading control. (f) Migration results of H2009 cells with LGR4- or IQGAP1-KD in response to RSPO3 treatment (10 ng/ml). *p < 0.05 vs. parental cells, Dunnett’s test following One-Way ANOVA (p < 0.0001). All the experiments were repeated twice with representative data shown here. Error bars are S.E.M.

IQGAP1 plays pleotropic roles in the regulation of cell adhesion and migration in normal and cancer cells^{12, 13, 37}. We found that KD of RSPO3, LGR4, or IQGAP1 in A549 cells led to significant reduction in cell migration and invasion.¹¹ Next, we tested if RSPO3 stimulation of H2009 lung cancer cells which express LGR4 and IQGAPs, but not RSPO3 (Supplementary Table S2), would enhance cell migration. Treatment of H2009 cells with RSPO3 increased their migration in a dose-dependent fashion (Figure 3d). H2009 cell lines stably expressing shRNA targeting LGR4 (shLGR4-40), or IQGAP1 (shIQGAP1-85), and control shRNA were generated, and their KD effect was confirmed by WB analysis (Figure 3e and Supplementary Figure S12a-b). KD of LGR4 led to a total loss of response to RSPO3 treatment in the migration assay whereas parental or control shRNA showed increased migration as expected (Figure 3f and Supplementary Figure S13). KD of IQGAP1 led to ~70% reduction in RSPO3-induced increase in migration (Figure 3f). The residual response may be due to incomplete IQGAP1 KD (Figure 3e, lower panel) or redundant function of IQGAP3 in these cells. Overall, these results indicate that RSPO3 is able to promote cell migration through the LGR4-IQGAP1 axis in lung cancer cells with normal Keap1 function.

RSPO3-LGR4 signaling is crucial to the regulation of EMT of RSPO3-high lung cancer cells

It was observed that A549 cells with KD of LGR4 displayed an obvious change in cell morphology, from spread-out, mesenchymal-like to cobble-stone, epithelial-like cells (Supplementary Figure S14). Staining of β-catenin, a key component of cell-cell adhesion, revealed that LGR4-KD cells displayed increased levels of membrane-bound β-catenin (Figure 4a, left panel) when compared to control shRNA cells (Figure 4a, right panel). These observations suggest that loss of RSPO3-LGR4 signaling transformed A549 cells from mesenchymal-like to epithelial-like cells, i.e., reversal of EMT. As the EMT process is critical to the development of tumor metastasis³⁸, we then examined a series of epithelial and mesenchymal markers by WB analysis. KD of LGR4 in either A549 or H460 cells led to a significant decrease in the levels of mesenchymal markers (fibronectin, N-cadherin and vimentin) with commensurate increase in the level of the epithelial marker E-cadherin (Figure 4b and Supplementary Figure S15a). The reversal of EMT markers was even more pronounced in cells with KD of RSPO3 (Figure 4c and Supplementary Figure S15b), consistent with the more severe decrease in migration and invasion of these cells.¹¹ We then queried if levels of RSPO3 mRNA were correlated with those of EMT markers in primary tumors by analyzing the expression data of the TCGA LUAD set. A positive correlation was found between expression of RSPO3 and the mesenchymal marker Snail (Figure 4d, R = 0.46, p < 0.0001, Spearmen test). In contrast, a clear negative correlation was found between the level of RSPO3 and that of E-cadherin (Figure 4e, R = −0.20, p = 0.002, Spearman test). Taken together, these results indicate that endogenous RSPO3-LGR4 signaling not only enhances cell migration and invasion but also promotes the EMT process in lung AD cells, which may be the underlying mechanisms for the high aggressiveness of RSPO3-high lung ADs.

Figure 4.

RSPO3-LGR4 signaling regulates EMT in lung cancer cells. (a) Representative micrographs of confocal immunofluorescence microscopy of β-catenin in A549 cells with LGR4-KD (shLGR4-40) or control shRNA. β-catenin was stained with a mouse anti-β-catenin antibody followed by Alexa488-labeled anti-mouse antisera. (b) WB analysis of EMT markers in parental A549 and H460 cells (P) and those with KD of LGR4 (KD) or control shRNA (C). (c) WB analysis of EMT markers in parental A549 cells (P) and cells expressing vector control (V), RSPO3-shRNA #63, or #67. All the experiments were repeated twice with representative data shown here. (d) Plot of the expression of RSPO3 vs. that of Snail in TCGA’s LUAD set (230 samples). (e) Plot of the expression of RSPO3 vs. that of E-cadherin in TCGA’s LUAD set (230 samples).

Knockdown of LGR4 or IQGAP1 reduced tumor growth and metastasis in vivo

We then evaluated the effect of knocking down LGR4 and IQGAP1 on tumor growth and metastasis using xenograft models of A549 cells. In the subcutaneous model, tumors from LGR4-KD cells showed ~70% reduction in weight at the end of the study when compared to those of parental or control-shRNA cells (Figure 5a and Supplementary Figure S16a; p = 0.001, one-way ANOVA; p <0.05, LGR4-KD vs. parental or control shRNA cells, Dunnett’s test). Reduction in levels of LGR4 protein in LGR4-KD tumors was confirmed by WB (Supplementary Figure S16b). Immunohistochemical staining showed that LGR4-KD tumors had increased levels of E-cadherin but decreased levels of vimentin (Figure 5b), consistent with the reversal of EMT in LGR4-KD cells. To determine if KD of LGR4 also affects tumor metastasis, A549 cells with control or LGR4-KD shRNA were tested in the tail vain injection model using athymic nude mice. Loss of LGR4 led to significant decrease in tumor formation in the lung (1/5 for LGR-KD cells vs. 5/5 for control cells, p = 0.02, Fisher’s exact test). The tumors from control shRNA cells were much larger with two animals dead before sacrificing (Figure 5c and Supplementary Figure 17a) while the only tumor formed from LGR4-KD cells was microscopic (Figure 5c and Supplementary Figure S17b). None of the mice from either group had tumors in the brain. We also tested the effect of IQGAP1 KD on tumor metastasis in luciferase-expressing A549 cells by in vivo imaging which was previously validated in SCID-Beige mice³⁹. Loss of IQGAP1 led to significant decrease in lung colonization (Figure 5d). Overall, 3/7 animals with control shRNA had tumors in the lung and 1/7 animal had tumors in both the brain and lung while 0/6 animals of IQGAP1-shRNA had detectable tumors (p = 0.049, Fisher’s exact test, one-tailed). Lack of lung colonization by IQGAP1-KD cells were confirmed at necropsy. All together, these in vivo data provide further support that LGR4 and IQGAP1 have functions in the regulation of tumor growth and metastasis in lung cancer cells.

Figure 5.

KD of LGR4 or IQGAP1 in A549 cells reduced tumor growth and metastasis in vivo. (a) Growth curve of subcutaneous tumors of A549 cell with KD of LGR4 vs. parental A549 cells and those expressing control shRNA. (b) Representative micrographs of IHC of E-cadherin and vimentin in subcutaneous tumors derived from A549 cells with LGR4 KD or control shRNA. The stromal (S) areas are demarcated by dotted lines. (c) Representative micrographs of H&E histology of lung tumors formed by A549 cells with KD Of LGR4 or control shRNA. The tumor areas (T) are demarcated by dotted lines. (d) In vivo imaging results of lung metastasis of A549-luciferase cells with KD of IQGAP1 or control shRNA.

Discussion

Dysregulation of Wnt signaling occurs in nearly every major type of solid tumor with colon cancer being most frequently affected^{40, 41}. RSPOs and LGR4-6 were recently found to form a ligand-receptor system with critical roles in normal development and stem cell survival through modulation of Wnt signaling^{1, 4-6}. The finding of recurrent gene fusions of RSPO2 and RSPO3 in a subset of colon cancer clearly established a role of RSPO2/3 in human cancer¹⁴, though clinicopathological characteristics of RSPO2/3-altered colon tumors have yet to be characterized. In NSCLCs, genetic alterations in APC and β-catenin of the Wnt pathway have also been detected, albeit at a low frequency⁴². Here we found that RSPO3 was expressed at exceedingly high levels in a subset (~9%) of lung ADs that showed much less favorable outcome. In contrast to the mechanism of gene fusion in colon cancer, the high RSPO3 expression in lung ADs were derived from a combination of deficiency in Keap1 and demethylation of the RSPO3 promoter. This is consistent with a recent publication that reported that RSPO2-EIF3E and RSPO3-PTPRK fusions were found at a frequency of 4% (3/75) in a colon cancer cohort but missing in 121 primary lung cancers⁴³. Furthermore, we found that LGR4 function promotes cell growth, migration, and invasion of lung cancer cells both in vitro and in vivo. These data suggest that poor survival of lung AD patients with RSPO3-high tumors was potentially due to, at least in part, aberrant RSPO3-LGR4 signaling.

Activation of RSPO3 transcription by NRF2 in a subset of lung ADs is a surprise. Expression of RSPOs is tightly controlled during development with low levels in normal adult tissues⁴⁴. During lung development, only RSPO2 is expressed and its function is essential for laryngeal-tracheal and lung morphogenesis^{2, 3, 44}. In the normal adult lung, RSPO2 is no longer expressed while RSPO3 is found at very low levels based on analysis of gene expression data from multiple databases, including UniGene (http://www.ncbi.nlm.nih.gov/unigene) and GEO (Gene Expression Omnibus, http://www.ncbi.nlm.nih.gov/geo). Oncogenic functions of RSPO2 and RSPO3 invariably depend on aberrant expression, either due to MMTV integration as found in mouse models of breast and colon cancer or fusion with a neighboring gene that is highly transcribed as in human colorectal cancer^{15-17, 45}. The mechanism of RSPO3 upregulation in lung ADs is unique for its dependence on NRF2 activation and demethylation of its own promoter. The three NRF2 binding site sequences upstream of the RSPO3 promoter are not conserved in the mouse. Activation of NRF2 by treatment with xenobiotics alone in normal cells did not upregulate RSPO3 based on the analysis of microarray data deposited in GEO. Furthermore, Keap1/NRF2 is mutated in 27% (48/178) of lung squamous cell carcinomas in the TCGA cohort. However, none of the Keap1/NRF2-mutated samples displayed aberrant expression of RSPO3. Moreover, only a fraction of lung AD tumors with Keap1 deficiency and RSPO3 promoter demethylation did not show high levels of RSPO3 (Supplementary Figure 8). These observations indicate that the aberrant expression of RSPO3 in selected Keap1-deficient lung ADs depends on NRF2 activation, demethylation, and other factor(s) found in lung ADs that are yet to be identified.

The much reduced survival of patients with RSPO3-high tumors suggests that RSPO3-LGR4 signaling enhances tumor metastasis. Previously, it was reported that elevated Wnt/TCF signaling promoted aggressive metastasis of human lung cancer cells into the brain and bone in mouse xenograft models⁴⁶. In the K-ras mouse model of lung adenocarcinoma, however, concomitant activation of β-catenin with K-ras only led to increase in tumor initiation and growth, but not in metastasis⁴⁷. Intriguingly, in the TCGA LUAD set, high levels of Axin2, the most validated indicator of Wnt/TCF signaling, were not associated with patient outcome (Supplementary Figure S18a), and expression of RSPO3 was not correlated with those of Axin2 (Supplementary Figure S18b). These observations imply that RSPO3-LGR4 most likely promotes the aggressiveness of lung ADs through enhancement of the β-catenin-independent pathway. This pathway, also called the non-canonical pathway, plays a critical role in the control of cell polarity and migration of normal cells and the invasion and metastasis of cancer cells⁴⁸. Recently, we uncovered that RSPO3-LGR4 is coupled to IQGAP1 to potentiate both the canonical and non-canonical Wnt pathways. IQGAP1 plays major roles in the regulation of cell adhesion and migration of normal and cancer cells through direct modulation of actin polymerization proteins^{11-13, 37}. Aberrant IQGAP1 activity was suggested to promote tumor metastasis in multiple types of solid tumors, including those of the lung^49-51. High RSPO3-LGR4 activity in lung AD cells may preferentially enhance the β-catenin-independent Wnt pathway to promote cell migration and invasion through modulating the roles of IQGAP1 in regulating actin dynamics. Further studies will be required to dissect the exact mechanisms of how RSPO3-LGR4 functions to modulate the two distinct pathways of Wnt signaling in lung AD cells.

In conclusion, we propose that RSPO3 expression is aberrantly activated in a subset of lung ADs as a result of Keap1 deficiency, demethylation of its own promoter and other unknown factor (s). High levels of RSPO3 activate LGR4-IQGAP1 to potentiate Wnt signaling, leading to increases in cell growth, motility and metastasis, and therefore poor patient survival. Inhibition of RSPO3-LGR4 signaling may provide a therapeutic approach to attenuate or even block the progression of lung ADs with high RSPO3 expression.

Materials and Methods

Data downloading and bioinformatic analysis

All data of gene expression, mutation, methylation, copy number, and overall survival of TCGA’s Lung Adenocarcinoma cohort (LUAD, Provisional, 230 samples) were downloaded from the cBioPortal for Cancer Genomics (http://www.cbioportal.org/public-portal/)²², the CCLE portal (http://www.broadinstitute.org/ccle/home)²⁴, or the Broad Firehose database. Detection of gene fusions were performed with the software TopHat-Fusion on UCSC genome assembly hg19 using default parameters ^{52, 53}. Screening of NRF2-binding sites in the promoter region of RSPO3 was performed using a published position weight matrix for NRF2³⁵.

Cell lines, plasmids, shRNA constructs, stable cell line generation, recombinant proteins, and tissue microarrays

A549, H460, and H2009 cells were obtained from Dr. Bingliang Fang at the M.D. Anderson cancer center. The cell lines were originally given by Dr. John Minna at the University of Texas Southwestern Medical Center. H1944 cells were from Dr. Joseph Nevins’ laboratory then at Duke University. A human Keap1 plasmid was purchased from Addgene⁵⁴ and Keap1 was subcloned into the vector pLVX-Ac-GFP to express Keap1 as a GFP-Keap1 fusion. shRNA constructs of LGR4, IQGAP1, and RSPO3 and their stable lines in A549 and H460 cells were described previously¹¹. Cells were grown in DMEM + 10% FBS with penicillin and streptomycin. Tissue microarrays of lung ADs and normal lungs were purchased from US Biomax. IHC staining with LGR4 antibody 7E7 was performed as before²⁵.

RT-qPCR, ChIP- PCR, Immunofluorescence, and WB analysis

Total RNA isolation, RT-qPCR of RSPO3 and AKR1C2, and immunofluorescence analysis were carried as we described previously⁴. For the RSPO3 induction experiment, H2009 cells were treated with azacytidine for 4 days and then induced with sulforaphane for 6 hrs. All qPCR probes were purchased from Life Technologies. ChIP-PCR of the NRF2 binding sites were carried out as described previously³⁵. Protein extraction and WB were carried out using RIPA buffer plus protease inhibitors (Roche)⁵⁵. Primary antibodies used are: GFP (Life Technologies, #A6455), Keap1 (Santa Cruz, #sc-15246), NRF2 (abcam, #ab62352), E-cadherin (Cell Signaling, #3195), β-catenin (Cell Signaling, #9562), N-cadherin (BD Transduction Laboratories, #610920), Vimentin (Cell Signaling, #5741), Fibronectin (BD Transduction Laboratories, #610077), Zeb1 (Cell Signaling, #3396), IQGAP1 (BD Transduction Laboratories, #610611), and β-actin (Cell Signaling, #4970).

Cell growth and migration assays in vitro

The xCELLigence RTCA system (Roche, Mannheim, Germany) was used to monitor cell growth in real time. Briefly, 2,000 cells were seeded into a 96-well E-plate (Roche) in DMEM media with 10% FBS (with puromycin for stable cells) and the cell growth was monitored for 6 days continuously. For the MTT assay (Cayman Chemical Company, Ann Arbor, Michigan), cells were plated into 96-well plates at a density of 1000 cells/well with medium mentioned above and cell numbers were determinate daily for up to 10 days. Migration and invasion assays of A549 and H2009 cells were carried out as described previously¹¹.

In vivo studies

Six-week-old female BALB/c athymic nude mice were purchased from Charles River Laboratories. For tumor growth, mice (n = 5 in each group) were injected subcutaneously with 4×10⁶ cells per animal in 1:1 DMEM/Matrigel (BD Biosciences) into the right dorsal flanks. Tumor measurements were carried out blindly with a caliper at least once per week, and volumes were calculated using the formula: V = 0.52 × (length)² × width. Mice were euthanized when they met the institutional euthanasia criteria for tumor size and overall health condition. The tumors were then removed and photographed. Immunohistochemistry (IHC) analysis of vimentin (Cell Signaling #5741) and E-cadherin (Cell Signaling #3195) was performed on paraffin sections with the Vectastain ABC-Elite kit according to the manufacturer’s instructions. Experimental metastasis to the lung was induced by injecting 1×10⁶ cells in 100 µl PBS through the tail vein of BALB/c athymic nude mice. Seven weeks following the injection, the survival animals were sacrificed and their lungs were removed. Tissue samples were fixed in 10% buffered formalin overnight and then washed with PBS, transferred to 70% ethanol and then embedded in paraffin, sectioned and stained with H&E. The entire lungs of the LGR4-KD group were scanned by H&E staining since no macroscopic lesions could be seen. For the study of A549-luciferase cells with IQGAP1-KD, 11-week-old Fox Chase SCID® Beige mice were injected with 1×10⁶ cells through the tail veil at n = 7 for the control shRNA group and n = 6 for the IQGAP1-KD group. The mice were imaged one per week for six weeks starting on week 2 post injection using an IVIS-II imaging system as described³⁹. All mouse experiments were approved by the Institutional Animal Care and Use Committee (IACUC) of the University of Texas Health Science Center at Houston.

Statistical methods

All statistical analyses were performed using the software GraphPad Prism5 unless otherwise noted. Data are presented as means ± S.E.M.