High expression of NAC1 in breast cancer is associated with poor prognosis
In TCGA samples (N = 996) and Metabric samples (N = 1866), which are mostly primary tumor tissues, we found that NAC1 alterations were 7% and 6%, respectively (Fig. 1A), but in the metastatic tumor tissues (cBioportal database), NAC1 alterations were 25% and 12%, respectively (Fig. 1B). Also, we observed increased deep deletions of NAC1 mainly in breast metastatic tumors (BMT) as compared to breast primary tumors (BPT). Expression of NAC1 mRNA positively correlated with copy number alterations (CNAs) in breast cancer (Fig. 1C). Accordingly, the expression of NAC1 protein was substantially increased in tumor tissues in comparison to normal tissues (Fig. S1A). NAC1 expression was remarkably increased in the breast cancer (BC) tissues harboring alterations in TP53, RB, and c-MYC (Fig. S1B and S1C), the pathways known to be dysregulated in BC. Immunohistochemistry staining and whole slide machine learning analysis of patient tumor specimens from the University of Kentucky tissue bank found increased nuclear expression of NAC1 in stage 3 tumors compared to stage 1 and 2 tumors (Fig. 1D). As BC is a heterogeneous disease and its prognosis differs among different subtypes, we next performed bioinformatic analysis of NAC1 expression in different subtypes of BC. In this analysis, we subdivided the TCGA pan-cancer and Metabric datasets into six molecular subtypes: Luminal A, Luminal B, human epidermal growth factor receptor 2 (HER2), TNBC, claudin expressing tumors, and normal tissues. The TCGA-pan cancer dataset comprising 1084 samples has 171 basal tissues, 78 HER2-positive, 499 luminal A, 197 luminal B, and 36 normal samples; while the Metabric dataset consists of 2509 samples, including: 209 basal, 218 claudin-low, 224 HER2-positive, 700 luminal A, 475 luminal B, and 148 normal tissues) [18]. Our analysis found that NAC1 copy number and mRNA expression were increased in the most aggressive basal-subtype samples as compared to other subtypes or normal tissues (p < 0.05) (Fig. 1E and 1F). Additionally, we detected higher NAC1 protein expressions in TNBC cell lines than in luminal or normal epithelial cells (Fig. 1G). In the basal tissue samples but not in other subtypes of BC subtypes, NAC1 expression was associated with poor prognosis of patients (Fig. 1H and Fig. S2A-2C). Additionally, NAC1 promoter methylation was inversely correlated with the survival of patients’ survival in breast cancer (Fig. S2D). Analysis of the TNBC single cell Broad institute datasets[21] showed that NAC1 was not only expressed in tumor cells but also in various immune cells (Fig. 1I). In dividing basal cells, NAC1 expression positively correlated with the CSC markers CD44, ALDH1A3 and NOTCH2 (Fig. 1J). These results imply that the expression of NAC1 in TNBC may play an important role in driving the malignant phenotype of this disease.
High expression of tumoral NAC1 supports stemness and promotes malignant phenotype of TNBC
To determine whether there is a causal association between tumoral expression of NAC1 and tumor stemness, we silenced the expression of NAC1 and then examined the expression of CSC markers. Figure 2A-2C show that in the NAC1-deficient HCC1806 and BT549 TNBC cells, the expression of CD44, ALDH1A1/2, SOX2, OCT3/4, and Nanog were reduced. Flow cytometry analysis and confocal microscopy imaging revealed a reduced expression of CD44 and increased expression of CD24 in NAC1-knockdown cells (Fig. 2D and E, S2E). Consistently, ALDH1A1/2 activity, a key indicator of CSCs in TNBC, was lower in the tumor cells with depletion of NAC1 than the control cells (Fig. S2F). Furthermore, we show that knockdown of tumoral NAC1 expression reduced the in vitro mammosphere formation (Fig. 2F) and the in vivo tumorigenicity of the tumor cells (Fig. 2G). These observations suggest that tumoral NAC1 has a role in supporting enrichment of CSCs.
To further investigate the role of NAC1 in promoting TNBC progression, we performed bulk RNA sequencing analysis on TNBC cells with or without knockdown of NAC1. Our analysis found 856 differentially expressed genes (DEGs) in the NAC1-deficient cells. Out of these DEGs, 576 were downregulated, and 280 were upregulated (adjusted p-value < 0.05, log2FoldChange > 1) (Fig. S3A). Kyoto Analysis of the Encyclopedia of Genes and Genomes (KEGG) also showed the alterations of some cancer-associated pathways in the NAC1-deficient samples in comparison to the controls (Fig. S3B). Analysis of the sequencing data revealed the enrichment of the genes essential for epithelial-mesenchymal transition (EMT) in tumor cells expressing NAC1 (Fig. 3A), and the downregulations of stemness and EMT-associated genes such as MUC5B family genes, L1CAM, MMP14, MMP1, ADAM17 and SDC4 [22–25] in the NAC1-deficient tumor cells (Fig. 3B, Fig. S3A, Fig. S3C-E). E-cadherin expression increased in the NAC1-deficient tumor cells (Fig. 3C). The tumor stemness marker ALDH1A3, a member of the aldehyde dehydrogenase family and an aldolase uniquely expressed in MDA-MB-231 cells, was significantly downregulated in the NAC1-deficient cells, as compared to the control cells (Fig. S3F). Under hypoxia, cancer stem cells orchestrate the reprogramming of the TME to promote tumor progression[26]. Indeed, the level of NAC1 protein was elevated in the hypoxic tumor cells (Fig. 3D), and the Gene set enrichment analysis (GSEA) showed that the hypoxia response-associated pathways were downregulated in NAC1-deficient TNBC cells (Fig. 3E). Also, the mRNA expressions of the hypoxia marker CA9 and tumor vascularization VEGFA were reduced in the tumor cells deficient in NAC1 (Fig. 3F and G). These data also suggest the role of NAC1 in promoting tumor progression.
Our experiments using MDA-MB-231 and HCC1806 cell lines showed that tumoral expression of NAC1 had a role in bolstering the proliferation, migration, and invasion of tumor cells. Figure 4A-C show that knockdown of NAC1 expression significantly decreased the proliferation of the tumor cells, reduced their colony formation (Fig. 4D), and inhibited their migration ability (Fig. 4E and F). In addition, the hanging drop assay demonstrated that the sphere size, sphere number and migration ability of the tumor cells subjected to knockdown of NAC1 were significantly decreased (Fig. 4G), suggesting that the expression of NAC1 confers tumor cell resistance to anoikis, a cellular feature that contributes to cancer aggressiveness. Expression of NAC1 in tumor cells also affects their metastatic ability. In C57BL/6J syngeneic mice, the tail vein injection of NAC1-expressing EO771 tumor cells led to increased lung colonization of tumor cells, but few colonies in the lung were observed in the C57BL/6J mice injected with the NAC1-deficient EO771 tumor cells (Fig. S4A). The similar difference in lymph nodes metastasis between NAC1-expressing and NAC1-deficient MDA-MB-231 cells was observed in nude mice (Fig. S4B). Additionally, significantly fewer lung metastases were found in NAC1−/− C57BL/6J mice than in wild-type mice (Fig. S4C-D). Nevertheless, orthotopic injection of NAC1-deficient MDA-MB-231 tumor cells to NSG mice resulted in more tumor cell colonization in the lung, as compared with the injection of the NAC1-expressing cells (Fig. S4E). Because C57BL/6J, nude and NSG mouse have distinct genetic background and immune system, the discrepancy in tumor cell dissemination observed may be attributed to the difference in the host immune status of these mice.
Activation of STAT3 is involved in the NAC1-mediated oncogenic roles
To explore the molecular mechanism by which NAC1 promotes TNBC progression, we used the BART platform (http://bartweb.org) to analyze the transcription factors (TFs) and regulators likely associated with the altered gene expressions through comparing the RNA sequencing data between the tumor cells with or without depletion of NAC1. Analysis of TFs found that the downregulated genes associated with loss of NAC1 were strongly associated with STAT3 transcriptional activity (p < 0.00001, AUC = 0.74) (Fig. 5A, Fig. S5A), while the upregulated genes were highly associated with chromatin modifier EZH2 (p < 0.00001, AUC = 0.842) (Fig. S5B and S5C). These results are consistent with our previous analysis, showing that combining the expressions of both NAC1 and EZH2 in clinical samples could predict the outcome of immunotherapy better than either alone [17].
Comparing the differentially expressed genes in the tumor cells with or without depletion of NAC1 via use of the gene set enrichment analysis (GSEA), we observed reduced expressions of the JAK/STAT pathway-associated genes in the downregulated gene set, as compared to the upregulated gene set in the tumor cells with NAC1 depletion (Fig. 5B). We further analyzed the expression profile of STAT3 transcriptome and protein in the NAC1-deficient tumor cells. We found that the transcription of STAT3 was similar in the NAC1-deficient cells and control cells, as analyzed by qPCR, RNA sequencing data and UALCAN-TCGA tumor tissue samples (p > 0.05) (Fig. 5C, D and E), but STAT3 protein expression increased significantly in the TCGA-tumor samples in comparison to the normal tissues (Fig. 5F). To further interrogate the correlation and clinical relevance of NAC1 protein expression and STAT3 activity, we performed IHC staining for NAC1, phospho-STAT3, proliferation marker Ki67 and apoptosis marker caspase3 in TNBC samples from the tissue bank of University of Kentucky. Figure 5G shows that the level of phospho-STAT3 protein positively correlated with NAC1 and Ki67 expression in TNBC samples; conversely, phospho-STAT3 protein level was negatively correlated with caspase-3 expression. Depletion of NAC1 in TNBC cells caused a reduction of phospho-STAT3 protein and slightly affected the expression of STAT3 (Fig. 5H). These results suggest that NAC1 is involved in activation of STAT3. Analysis of the RNA sequencing data revealed that depletion of NAC1 led to a significant reduction of canonical JAK/STAT3 regulator, JAK1 (Fig. 5I). Protein analysis showed similar results, i.e., depletion of NAC1 led to a reduction in expression of JAK1 (Fig. 5J). Notably, we observed reduced expression of CD44 mRNA expression in NAC1-deficient tumor cells (Fig. 5K), and CRISPR knockout of CD44 resulted in a decrease of the expression of JAK1 protein, suggesting that CD44 is an upstream regulator of JAK1 (Fig. 5L). These results imply that the CD44-JAK1-STAT3 axis plays a role in the oncogenic function of NAC1 in TNBC.
Expression of tumoral NAC1 is associated with activation of immunosuppressive signaling.
As NAC1 showed a role in sustaining tumor stemness (Fig. 2), and CSCs can interact with immune cells in tumor microenvironment (TME) and contribute to immune evasion [27], we next wanted to know whether tumoral expression of NAC1 has any effects on immunosuppressive pathways. Using the gene set enrichment analysis (GSEA), we found the alteration of the genes associated with immune response such as TGFβ1, TGF-α signaling, interferon-gamma and inflammation-associated genes (Fig. 6A), and the downregulations of the innate immunity-associated genes in the NAC1-deficient cells (Fig. 6B). Analysis of the RNA sequencing data showed a decrease of the myeloid-derived cells granulation-linked factors in the NAC1-deficient tumor cells (Fig. 6C). Also, the levels of G-CSF, CCL2, SOD2, and IL6 were downregulated in the NAC1-deficient TNBC cells (Fig. 6D-G). Notably, analysis of gene ontology (GO) enrichment demonstrated the genes associated with secretion pathways, such as secretory granules, secretory vesicles and Golgi apparatus, were downregulated in NAC1 KD cells (Fig. S5D). Since EMT activates the Rab6A-mediated exocytotic process to promote immunosuppressive cytokines secretion in cancer and NAC1 promotes EMT (Fig. 3), we examined the effect of NAC1 on Rab6A. we found that NAC1 deficiency significantly reduced the expression of Rab6A (Fig. S5E). Because IL6 is a major ligand involved in regulation of STAT3 activity and cooperates with G-CSF to polarize myeloid cells towards immunosuppression, we then determined the effect of NAC1 on IL6 expression, using qPCR and enzyme-linked immunosorbent assay. Figure 6F shows that expression of IL6 was significantly decreased in the NAC1-deficient tumor cells. Further, we showed that depletion of tumoral NAC1 led to significant reduction of soluble G-CSF (Fig. 6G). In contrast, EO771 cells subjected to forced expression of NAC1 had a significantly increased amount of soluble IL6 (Fig. 6H). Additionally, analysis of clinical samples showed a positive correlation between NAC1 and TGFβ in metastatic tumor samples (Fig. 7A) but not in the primary tumor tissues (Fig. 7B). The RNA-seq analysis showed that the expressions of TGFβ1 (Fig. 7C), the TGFβ-interacting proteins SMAD3and SMAD5 (Fig. 7D and E), and the TGFβ ligands BMP1 and BMP4 were all decreased (Fig. 7F-G) in the tumor cells subjected to knockdown of NAC1. These results suggest that the tumoral NAC1 may have an important role in regulating immunosuppression-associated pathways.
Role of tumoral NAC1 in tumor initiation and progression is determined by immune status of the host.
As NAC1 was shown to contribute to tumor stemness and immunosuppressive pathways including IL6, G-CSF, and TGF-alpha (Fig. 6&7) and, both of which can affect tumor development and progression, we next compared tumor initiation of MDA-MB-231 cells with or without knockdown of NAC1 expression. In these experiments, nude mice or NSG mice were inoculated s.c. with tumor cells (1x106 cells/mouse), and then the tumor growth was closely monitored (Fig. 8A). Figure 8B, and 8C show that in NK cell-competent nude mice, the tumor cells transfected with non-targeting shRNA caused apparent tumor growths; by contrast, the tumor cells subjected to RNAi-mediated depletion of NAC1 barely induced tumor growth. Interestingly, in NSG mice deficient in NK cell we observed that the tumor cells with depletion of NAC1 produced larger tumors than the tumor cells expressing NAC1 (Fig. 8D and E). We further examined and compared the tumor cell proliferation in vivo in the nude and NSG mice and observed a similar discrepancy to that of tumor growth (Fig. S6A and 6B). To explore the cause underlying the different pattern of tumor initiation and development between nude mice and NSG mice, we performed immunofluorescence analyses on MDSCs and NK cells of the resected tumors using the respective marker, as both nude mice and NSG mice possess myeloid derived cells [28]. Figure 8F shows that in nude mice, tumor infiltration of MDSCs was substantially reduced in NAC1 depleted tumors as compared with that in NAC1-expressing tumors; however, in NSG mice, MDSCs infiltration was increased in NAC1 depleted tumors compared with that in NAC1-expressing tumors. Immunofluorescence analyses of NK cells in the tumor specimens from nude mice showed that there were more infiltrations of NK1.1+/CD16+ double positive cells (mature and activated NK cells) in NAC1-deficient tumors than in NAC1-expressing tumors; NK1.1+/ CD16− cells, which are inactive NK cells, were detected in the NAC1-expressing tumors but barely detected in the NAC1-deficient tumors (Fig. 8G). Consistent with the observation shown in Fig. 8D and 8E, the in vivo limiting-dilution assay demonstrated that the tumorigenicity of the NAC1-depleted tumor cells in NSG mice was higher than that of the NAC1-expressing tumor cells (Fig. S6C). These observations imply a possible interaction between MDSCs and NK cells in the NAC1-expressing tumors.
Presence or absence of NK cells alters the effect of tumoral NAC1 on MDSCs
To further demonstrate the impact of NAC1 on MDSCs and the influence of NK cells, we depleted myeloid cells or NK cells of nude mice using Ly6G antibody and NK1.1 antibody, respectively, and then monitored tumor growth in mice inoculated with MDA-MB-231 cells with or without depletion of NAC1 (Fig. 9A). Consistent with what we observed in NSG mice (Fig. 8D and E), in the nude mice depleted of NK cells, NAC1-depleted tumor cells exhibited a substantially enhanced tumorigenicity as compared with the control tumor cells (Fig. 9B-F); and depletion of myeloid cells (Fig. 9G) caused a decreased growth of NAC1-expressing tumors but led to an enhanced growth of NAC1-depleted tumors in nude mice (Fig. 9H-L). These results demonstrate that in the presence of tumor NAC1, myeloid-derived cells have negative effect on NK cells, and loss of NAC1 decreases the tumor-promoting activity of MDSCs but increases the tumor-inhibiting activity of MDSCs as well as NK cells. When NK cells are depleted, the NAC1-expressing and NAC1-depleted tumor cells both show enhanced tumorigenicity; however, when myeloid-derived cells are deprived, the tumorigenicity of NAC1-expressing cells is reduced, but NAC1-deficient tumor cells show enhanced tumor growth, suggesting that absence of the tumor-inhibiting myeloid-derived cells may diminish the activity of NK cells in NAC1-depleted tumors. These results indicate that the effect of NAC1 on myeloid-derived cells is NK cell-dependent. Also, as deficiency of tumoral NAC1 caused downregulation of IL6, G-CSF, and TGFβ1 (Fig. 6&7), all of which have immunosuppressive effects, NAC1 may control the status of MDSCs through modulating the levels of these cytokines, thereby impacting tumor initiation and development.
MDSCs with high expression of NAC1 support stemness of TNBC
Next, we determined the expression of NAC1 in MDSCs and its effect on CSCs, as the interaction of immune cells with CSCs in TME has crucial roles in tumor growth and metastasis [29, 30]. We inoculated BALB/c mice with 4T1 tumor cells and then compared NAC1 expression in MDSCs from tumor-bearing mice with that in MDSCs from tumor-free mice. Figure 10A shows that in addition to the high level of NAC1 in the tumor infiltrated MDSCs, NAC1 was also up-regulated in MDSCs from the spleen, blood, and bone marrow of the tumor-bearing mice, compared with those from the tumor-free animals. To assess the effect of NAC1 in MDSCs on their activity, we co-cultured EO771 tumor cells with MDSCs from the wild-type mice or NAC1 knockout mice (Fig. 10B), then analyzed and compared the levels of CD44 and aldolase activity in the tumor cells. Figure 10C and 10D show that CD44 expression and aldolase activity in the EO771 tumor cells co-cultured with Gr1+/CD11b+ cells from NAC1−/− mice were reduced, as compared with that in the EO771 cells co-cultured with Gr1+/CD11b+ cells from the wild-type mice, suggesting a role of NAC1-expressing MDSCs in maintaining CSCs. Also, the tumor initiation of EO771 cells was lower in NAC1−/− C57BL/6J mice than that in wild-type mice (Fig. 10E). These results imply that NAC1 expression in MDSCs affects their tumor-promoting function as well as tumor stemness. Additionally, Gr1+/CD11b+ cells from the tumor-bearing NAC1−/− C57BL/6J mice showed stronger cytocidal effect than those from wild-type mice when co-cultured with tumor cells (Fig. 10G and 10H), suggesting that expression of NAC1 in MDSCs controls their tumor-inhibitory or tumor-promotive activity.