Long non-coding RNA HUMT hypomethylation promotes lymphangiogenesis and metastasis via activating FOXK1 transcription in triple-negative breast cancer

Clinical samples were collected from patients who underwent breast cancer resection in Sun Yat-sen University Cancer Center (SYSUCC). Detailed information for the sample collection was provided in Supplementary Material and Methods (Additional file 10).

Vectors for overexpressing, dCas9-CHIP, CRISPR were constructed, and details were provided in Supplementary Material and Methods. Primers, siRNAs, and sgRNAs were listed (Additional files 13 and 14).

Full detailed procedures of cell cultures, cell viability, migration and invasion assay, endothelial cell tube formation assay, western blot, quantitative real-time PCR (qRT-PCR), RNA immunoprecipitation (RIP), RNA pulldown, and chromatin immunoprecipitation (CHIP) are provided in Supplementary Material and Methods.

In situ hybridization (ISH), fluorescence in situ hybridization (FISH), and immunohistochemistry (IHC) were performed as previously described, and details are provided in Supplementary Material and Methods [14, 26].

All the in vivo experiments were performed following institutional and international guidelines and regulations. Detailed procedures were provided in Supplementary Material and Methods.

Corresponding methods and tools for the analysis of PAR-CLIP, CHIP-seq, and CIBERSORT are provided in supplementary tables (Additional files 15 and 16). Student’s t test and χ² test were used for comparison of the results between two groups, and the non-parametric test was adopted for data in abnormal distribution. One-way ANOVA was used for comparison in more than two groups. The Spearman test was adopted for analyzing the correlation of HUMT/YBX1/FOXK1. The Kaplan-Meier method with a log-rank test was used to compare the overall survival (OS) and disease-free survival (DFS) between the groups. The receiver operating characteristic (ROC) curve was adopted for analyzing the diagnostic value of HUMT in patients with TNBC. Statistical analysis was performed on GraphPad Prism 7 (GraphPad), MedCalc 18.2.1 software (MedCalc), and R 3.5.3 software (https://​www.​r-project.​org). The results were presented as mean ± SEM. A P value of < 0.05 was determined as statistically significant, and all P values were two-tailed.

To investigate the biological mechanism of TNBC lymph node metastasis, we grafted MDA-MB-231 and BT549 cells into the fat pads of nude mice and performed passages to screen highly lymph node-invasive cells in vivo (Fig. 1a, b). Cancer cells in metastatic lymph node from the third passage displayed a highly invasive nature to lymph nodes. Highly invasive cells in metastatic lymph nodes were isolated and termed 231LNM3 and 549LNM3.

To identify which biological change exerted an impact on triple-negative breast cancer lymph node metastasis, highly lymph node-invasive cells and parental cells were applied to genome-wide microarray in replicates (Fig. 1c). Representative upregulated and downregulated transcripts were shown in the heatmap (Fig. 1d) and volcano plot (Fig. 1e). HUMT was significantly upregulated in LNM3 cells compared with parentals and further validated by qRT-PCR (Fig. 1f).

We firstly evaluated the coding potential of HUMT using NCBI ORFfinder. No typical protein-coding open reading frame (ORF) (> 300 nt) was found on the sequence of HUMT, and there is a very low probability for the coding potential of HUMT (Additional file 1: Fig. S1a, b). The result was further validated by a PhyloCSF model and coding potential assessment tools (CPAT) (Additional file 1: Fig. S1c). The PhyloCSF value along the entire sequence of HUMT was calculated, and the negative value suggested a low coding potential. We further used the CPAT to confirm this result (Additional file 1: Fig. S1d). In summary, these observations confirmed that HUMT did not have a coding capacity.

The bioinformatic analysis also showed that HUMT was significantly upregulated in tumors compared with paired normal tissues of TNBC in the TCGA cohort. A pan-cancer analysis confirmed a higher level of HUMT in tumor tissue of various cancers (Additional file 2: Fig. S2).

Then, we conducted in vivo experiments on nude mice to validate the function of HUMT. The lymph node metastasis model using 231LNM3 and 549LNM3 with normal control (NC) or HUMT-KO (knock-out) was constructed. The HUMT-KO group exhibited improved lymph node metastasis-free survival (LNMFS) compared with the controls (Fig. 1g), a lower rate of lymph node metastasis, and smaller metastatic lymph node volume (Fig. 1h, i), indicating that HUMT promoted lymph node metastasis in TNBC.

Furthermore, the expression level of HUMT was detected in 228 cases of TNBC from SYSUCC. The patients were divided into two groups: HUMT-low and HUMT-high. The association between HUMT expression and clinicopathological characteristics was analyzed. A crosstabs analysis showed that HUMT expression was significantly correlated with the AJCC T stage and N stage (Additional file 11: Table S1).

The Kaplan-Meier method with a log-rank test showed a significantly poorer overall survival (OS) and disease-free survival (DFS) in the HUMT-high group (Fig. 1j). Moreover, the qRT-PCR analysis showed that a significantly higher level of HUMT in primary TNBC with lymph node metastasis compared with paracancerous tissues and those without lymph node metastasis, suggesting that HUMT might serve as a predictor for lymph node metastasis in TNBC (Fig. 1k).

Taken together, HUMT was correlated with lymph node metastasis and predicted tumor progression in TNBC.

We further investigated the RNA levels of HUMT in breast cancer cell lines and normal mammary epithelial cells using qRT-PCR. Interestingly, the expression level of HUMT was significantly higher in triple-negative breast cancer cell lines, especially MDA-MB-231 and BT549, than non-TNBC cell lines and normal epithelial cells (Fig. 2a). Further bioinformatic analysis confirmed a higher expression level of HUMT in basal-like patients than non-basal like ones, indicating HUMT might be responsible for the malignant characteristics of TNBC (Fig. 2b).

As the lymph node metastasis is always positively correlated with tumor cell proliferation and metastatic nature, we constructed the HUMT-KO MDA-MB-231 and BT549 cell lines using CRISPR methods to further investigate whether HUMT was involved in cell proliferation, metastasis, and lymphangiogenesis. CCK8 assay showed that cell proliferation was significantly suppressed in HUMT-KO cells than controls (Additional file 3: Fig. S3a). This was further validated in colony formation assay as the colony-forming capacity was significantly inhibited in the HUMT-KO group (Additional file 3: Fig. S3b). Wound healing assay, Transwell migration, and invasion assay indicated that HUMT silencing significantly suppressed the migration ability of cancer cells (Fig. 2c–e). Furthermore, a 3D invasion assay was conducted and showed a consistent result for invasion ability (Fig. 2f).

The migration activity of HLECs could be significantly impacted in the culture medium supernatant of HUMT-KO cells compared with controls, indicating a crucial role of HUMT in triple-negative breast cancer lymph node metastasis (Fig. 2g). Moreover, we examined whether HUMT promoted lymphangiogenesis by studying its effect on the tube formation ability of HLECs, which is a critical process during breast cancer lymph node metastasis. The culture medium supernatant of HUMT-KO cells significantly abolished the tube formation capacity of HLECs as the branch number and total length of tubes were decreased compared with the control groups (Fig. 2h).

In summary, these results indicated that HUMT promotes proliferation, lymphangiogenesis, and lymph node metastasis in TNBC in vitro.

To further elucidate the mechanism of HUMT upregulation, we analyzed the correlation between DNA methylation and the expression level of HUMT as DNA methylation was one of the most common ways of epigenetic modulation for gene expression [19, 27, 28]. Bioinformatic analysis indicated that basal-like breast cancer patients harbored the lowest level of DNA methylation at the promoter region in breast cancer subtypes, which was consistent with the highest RNA expression level mentioned above. Besides, two GC-enriched regions were also predicted and indicated CpG islands in the promoter region of HUMT (Fig. 3a). The analysis of CpG island-specific methylation revealed a significant negative correlation between methylation and expression levels in The Cancer Genome Atlas (TCGA) Program database (Fig. 3b). We further analyzed the whole-gene DNA methylation level of breast cancer patients in TCGA. The results showed a significant negative correlation between DNA methylation level and expression level of HUMT in all breast cancer (r = − 0.553, P < 0.001) and TNBC (r = − 0.563, P < 0.001) (Fig. 3c, d). Moreover, the methylation status of HUMT in Cancer Cell Line Encyclopedia (CCLE) was analyzed and revealed a significantly negative correlation in all cancer cell lines (r = − 0.816, P < 0.001) and breast cancer cell lines (r = − 0.812, P < 0.001) (Fig. 3e, f).

BT483, MCF7, and ZR-75-1 cells were identified as the cell lines with hypermethylation status at the promoter region of HUMT (Fig. 3g). The qRT-PCR analysis was conducted after decitabine (DAC) treatment. In vitro experiments showed that decitabine, a methyltransferase inhibitor, could significantly upregulate the expression of HUMT in MCF7, ZR-75-1, and BT483 cells, indicating that HUMT was regulated in an epigenetic way (Fig. 3h). Pan-cancer analysis of TCGA datasets also revealed a significant correlation between HUMT expression and methylation level of CpG island in the promoter region (Additional file 4: Fig S4).

Taken together, HUMT was upregulated in highly invasive cells due to the hypomethylation at the promoter region.

We next explored the mechanism by which HUMT regulates cancer progression in TNBC. As the functions of lncRNAs were tightly associated with subcellular distribution, bioinformatic analysis by lncATLAS was used to predict the location. Nearly half of HUMT was predicted to be located in the nuclei in cell lines (Additional file 5: Fig. S5a). We further extracted the RNA in the cytoplasm and nuclei. qRT-PCR showed that about 40% of HUMT was distributed in the nuclei, which was further confirmed by FISH (Additional file 5: Fig. S5b-c).

Previous studies have reported the important role of lncRNAs in transcription by recruiting corresponding proteins [29, 30]. A considerable part of HUMT was located in the nucleus of the cancer cells, indicating that HUMT might exert its function by recruiting transcription complexes and further enhance or inhibit gene transcription as previously reported. To verify this hypothesis and identify HUMT-interacting proteins, we utilized a GEO dataset and forecasted that a DNA binding protein YBX1 might bind to HUMT (Fig. 4a). The interaction of YBX1 protein with HUMT was validated using RNA immunoprecipitation (RIP) assay (Fig. 4b). RNA pulldown using a biotinylated RNA probe of HUMT was used as a reverse proof (Fig. 4c).

Previous studies reported that YBX1 played a crucial role in transcription [31, 32]. To further investigate whether the HUMT-YBX1 complex exerted its function by transcription regulation, we analyzed the CHIP-seq result of YBX1 protein in three independent cell lines from the ENCODE database to identify candidate DNA with which YBX1 might interact (Additional file 6: Fig. S6a-c). FOXK1, a transcript well known for functions related to extensive cancer progression activity, drew our attention. The binding motif of YBX1 on the FOXK1 promoter region could be predicted (Additional file 6: Fig. S6d). Further analysis revealed the YBX1-binding peaks about 800 bp upstream TSS, and specific primers for different promoter segments of FOXK1 were designed (Fig. 4d, e). To determine which site(s) YBX1 bound within the FOXK1 promoter, ChIP was performed. Our results revealed that YBX1 protein bound to FOXK1 at the − 700 to approximately − 850-bp promoter region of FOXK1, and the combination was significantly downregulated when HUMT was knocked out (Fig. 4f). Next, we conducted a dCas9-CHIP assay to further validate the occupancy of HUMT at FOXK1 promoter region (Fig. 4g). HUMT was significantly enriched in the group using FOXK1 promoter-targeted gRNAs than IgG or control gRNA group (Fig. 4h).

Next, we investigated the effect of the HUMT-YBX1 complex on the FOXK1 expression and cell proliferation. Bioinformatic analysis showed that the expression of HUMT and YBX1 positively correlated with FOXK1 (Fig. 4i). HUMT-KO or YBX1-KD could significantly decrease the FOXK1 expression level detected by qRT-PCR (Fig. 4j). Moreover, western blot confirmed that FOXK1 was upregulated following YBX1 or HUMT overexpression, and these effects could be partly reversed by HUMT or YBX1 silencing (Fig. 4k). In addition, we found that HUMT-KO and YBX1-KD significantly suppressed cell proliferation of MDA-MB-231 and BT549 (Fig. 4l). Further bioinformatic analysis from the Breast Cancer Gene-Expression Miner (bc-GenExMiner v4.3) showed that FOXK1 predicted a poorer metastasis-free survival in patients with basal-like breast cancer, which was consistent with our results (Fig. 4m). Together, these data demonstrated that HUMT, YBX1, and FOXK1 could form a transcription complex, and HUMT could regulate the FOXK1 expression by recruiting YBX1.

Previous studies have identified a crucial role of FOXK1 in cancers [33‐35]. To explore whether HUMT exerted its function through FOXK1-mediated cell proliferation and vascular endothelial growth factor (VEGFC) signaling, which was a previously reported key lymphangiogenesis pathway [36, 37], we analyzed the effect of HUMT-KO by western blot. Our results showed that the protein level of FOXK1, p-Akt, p-mTOR, and VEGFC was markedly downregulated in response to HUMT-KO, but it could be partially reversed by FOXK1 overexpression in MDA-MB-231 cells (Fig. 5a).

To further validate the role of FOXK1 in HUMT-mediated tumor progression, we transfected MDA-MB-231 and BT549 cells stably overexpressing HUMT or empty vector with si-FOXK1 or scramble vector. An in vitro experiment showed that knockdown of FOXK1 could reverse the effect of HUMT overexpression. The effect of promoting migration and invasion by HUMT overexpression was abolished by the FOXK1 knockdown (Fig. 5b–d). The tube formation assay showed that the effect of promoting lymph node metastasis by HUMT overexpression could be reversed by the FOXK1 knockdown (Fig. 5e, f). Moreover, the migration activity of HLECs was also affected in the Transwell assay (Fig. 5g). Western blot of the rescue experiments showed that FOXK1 knockdown could partially reverse the increased expression of key signalings mediated by HUMT in MDA-MB-231 cells (Fig. 5h).

Taken together, our findings indicated that lncRNA HUMT promotes cancer cell proliferation, lymph node metastasis, and lymphangiogenesis by HUMT/YBX1/FOXK1-mediated Akt/mTOR and VEGFC signaling.

We further conducted experiments in vivo to validate the oncogenic characteristics of HUMT and FOXK1. The xenograft model showed that HUMT-KO could significantly suppress tumor growth compared with the control group, and it could be reversed by FOXK1 overexpression (Fig. 6a, b). Lymph node metastasis was widely considered as the first step for breast cancer metastasis, and cancer cells further extended to distant sites, especially lung metastasis [38]. To confirm the function of HUMT in triple-negative breast cancer metastasis in vivo, the lymph node metastasis and lung metastasis model of 231LNM3 were established. The mice in HUMT-KO exhibited a lower volume of lymph nodes compared with controls, and it could be reversed by FOXK1 overexpression (Fig 6c, d). The lymph nodes of nude mice were removed for IHC stain, and metastatic cancer cells were significantly less enriched in the HUMT-KO group than controls (Fig. 6e). Besides, the ratio of metastatic lymph nodes was significantly decreased in the HUMT-KO group (Fig. 6f). HUMT-KO could also suppress lung metastasis in nude mice (Fig. 6g, h). We further confirmed in the tumor that HUMT-KO significantly downregulated FOXK1 expression by qRT-PCR (Fig. 6i).

As patient-derived xenograft (PDX) models reflect the effect of microenvironment on tumor growth and treatment response for the origin patients, we further explored the potential role of HUMT as a therapeutic target by intratumoral ASO injection. The tumor volumes and tumor weights of the HUMT-KD group were significantly lower compared with the control group (Fig. 6j, k). These results revealed that HUMT-KO significantly inhibited tumor proliferation in vivo.

Taken together, HUMT promotes tumor growth and metastasis in vivo.

We used the KM plotter (http://​kmplot.​com/​analysis/​) to evaluate the prognostic value of HUMT in different cancers. HUMT was correlated with poorer OS and relapse-free survival (RFS) in specific cancers, not just in basal-like breast cancer (Additional file 7: Fig. S7). Moreover, the clinical relevance of HUMT and FOXK1 was assessed in patients with TNBC from SYSUCC. The qRT-PCR analysis confirmed the significant correlations between HUMT and YBX1, FOXK1, which was consistent with the results above (Fig. 7a, b). ROC analysis of OS and DFS revealed that a combined panel of HUMT expression, T stage, and N stage showed an additional predictive value (Fig. 7c, d). Western blot showed a higher level of FOXK1 protein in triple-negative breast cancer patients with metastatic lymph nodes (Fig. 7e), further confirming a crucial role in lymph node metastasis.

In summary, we put forward a proposed mechanism for HUMT-mediated triple-negative breast cancer progression, including cell proliferation and lymph node metastasis (Fig. 7f). Interestingly, the bioinformatic analysis indicated that HUMT upregulation was correlated with a lower fraction of activated NK cells calculated by CIBERSORT in TNBC from TCGA and GSE58812 (Additional file 8: Fig. S8a). The NK cells were stained, and stromal NK cells were quantified. Patients with high FOXK1 expression in the TCGA-TNBC cohort also showed a higher level of TGF-β, which was considered as an immunosuppressive factor in cancer (Additional file 8: Fig. S8b). We found that patients with high-HUMT expression exhibited a lower rate of positive stromal NK cells (Additional file 8: Fig. S8c, d). Moreover, HUMT might be correlated with cancer cell stemness as fewer spheroids were observed in the HUMT-KO group compared with the controls, indicating that the repression of HUMT could inhibit the stemness feature of MDA-MB-231 and BT549 (Additional file 9: Fig. S9).

Considering that FOXK1 was reported to be associated with M2-type macrophages attraction and thus promoted tumor growth [39], the HUMT/YBX1/FOXK1 axis might also be functioning in immune escape. But the mechanism of these results still needs further exploration in the future.