Gly-tRF enhances LCSC-like properties and promotes HCC cells migration by targeting NDFIP2

Fifteen samples of histologically confirmed tumours and matched adjacent non-tumour tissues obtained from HCC patients who underwent radical hepatectomy at Lanzhou University First Hospital. The study was approved by the hospital ethics committee, and according to the institutional review committee's procedures, all patients signed an informed consent form before the study. A tissue microarray containing 90 tumour tissues and matched adjacent non-tumour tissues was purchased from Shanghai Outdo Biotech Co., Ltd (Shanghai, China). All patients provided written informed consent and were followed up for 5–6 years with clear prognostic information. Immunohistochemical staining was performed as previously described [23]. For immunohistochemical images, two experienced pathologists independently performed immunohistochemical staining scores according to the staining intensity (0: no staining; 1: weak staining; 2: moderate staining; 3: strong staining) and the percentage of positive cells (0: 0%; 1: < 25%; 2: 26–50%; 3: > 50%). The staining intensity score and the staining percentage score were summed to calculate the final immunohistochemical score. We defined an immunohistochemical of score 0 to 4 as low expression and a score of 5 to 6 as high expression.

The human liver cancer cell lines HepG2, Huh7 and HCCLM3 were purchased from the China Center for Type Culture Collection (CCTCC, Wuhan, China) and were identified by short tandem repeat (STR). L02 hepatocytes were gift from Zhongshan Hospital of Fudan University (Shanghai, China). The embryonic kidney cell line HEK-293 T was a gift from Shanghai GeneChem Co., Ltd. (Shanghai, China). All cells were grown in Dulbecco’s modified Eagle’s medium (DMEM; pH = 7.2, Gibco Company, Grand Island, NY, USA) containing 10% (v/v) foetal bovine serum (FBS, HyClone, Logan, UT, USA). All cells were cultured in a humidified incubator (Thermo Fisher Scientific, Waltham, MA, USA) at 37 °C and 5% CO2. All cells were tested for mycoplasma contamination.

Total RNA was harvested from cells and tissues using RNAiso Plus (Takara Holdings Inc., Kyoto, Japan), and the protocol recommended by the manufacturer protocol was followed for isolation of total RNA. NanoDrop 2000 (Thermo Fisher Scientific) was used to measure the quality and quantity of the isolated RNA.

Heavy modifications contained in tRNAs, such as 3′-aminoacyl, 3′-cP, m1A, m1G, and m3C modifications will severely interfere with reverse transcription. Therefore, conventional PCR methods may not be able to reflect the true expression characteristics of tRNA-derived fragments [24]. In this work, an rtStar™ tRF&tiRNA Pretreatment Kit (Arraystar Inc., Rockville, MD, USA. Cat #AS-FS-005) was used to remove various modifications from Gly-tRF before 3′ and 5′ adaptor and cDNA synthesis. cDNA was synthesized using APExBIO First-strand cDNA Synsthesis Supermix (APExBIO Inc., Houston, TX, USA; Cat #K1073). All steps, such as 3′-terminal deacylation, 3′-cP removal and 5′-P addition, demethylation and reverse transcription, were carried out in accordance with the manufacturer's instructions. All reactions were performed in an Mx3000P QPCR system (Agilent Technologies Inc., Santa Clara, CA, USA) using TB Green Premix Ex Taq II (Takara Holdings Inc., Kyoto, Japan) for real-time PCR according to the manufacturer's instructions. The primers are listed in Table 1. U6 or GAPDH was used as the normalized endogenous control for expression. The relative expression levels of Gly-tRF were analysed by the 2^−ΔΔCt method.

HCCLM3 and Huh7 cells were seeded in 6-well plates (Corning Life Sciences, USA) at a density of 4 × 10⁵ cells per well. When the cells were 40–50% confluent, Gly-tRF negative control, Gly-tRF inhibitor, and Gly-tRF mimic lentiviruses were used to transduce cells according to the manufacturer's instructions (Shanghai GeneChem Co., Ltd, Shanghai, China). The sequences of all lentiviral products are shown in Table 1. After 72 h of transduction, cells were cultured in complete medium containing 2 µg/µL puromycin for 72 h. Cells stably transduced with Gly-tRF were used for subsequent experiments.

HCC cells were prepared as single-cell suspensions for staining. All antibodies used for staining were purchased from Miltenyi Biotec (Bergisch Gladbach, Germany), and included a phycoerythrin (PE)-conjugated anti-CD133 antibody (Cat #130-110-962), a PE-Vio770-conjugated anti-CD13 antibody (Cat #130-120-727), an allophycocyanin (APC)-conjugated anti-EpCAM antibody (Cat #130-111-000), an APC-Vio770-conjugated anti-CD44 antibody (Cat #130-113-339) and anti-REA control antibody (Cat # 130-113-438, 130-113-440, 130-113-434, 130-113-445). The percentages of CD133⁺, CD13⁺, EpCAM⁺, and CD44⁺ within the HCC cell population were determined according to the manufacturer’s instructions. In brief, 1 × 10⁶ cells were centrifuged and resuspended in 98 µL of buffer, 2 µL antibody was added, and the cells were incubated for 10 min in the dark at 4 °C. The cells were washed with 1 mL of buffer, centrifuged at 300 g for 10 min, and resuspended in 400 µL of buffer for detection. Data were acquired with a BD LSRFortessa. All samples were analysed in triplicate.

Cells (2000 cells per well) were planted in a 6-well ultra-low adhesion plate (Corning, USA). After 8 days of incubation, spheres were counted and photographed (15 random fields/well) under a stereomicroscope (Olympus, Tokyo, Japan). The diameter of the spheres was measured with Image-Pro Plus 6.0 software (Media Cybernetics Inc., Rockville, MD, USA), and colonies with a diameter greater than 20 µm were considered positive for sphere formation.

Cell migration experiments were performed in a 24-well Transwell plate (8.0 µm pore size, Corning Life Sciences, Costar, USA). Stably transduced cells were starved for 6 h in serum-free medium, trypsinized and adjusted to 2 × 10⁵ cells/mL after counting. Then, 600 µL of complete medium containing 30% (v/v) serum was added to the lower chamber, 200 µL of the cell suspension was added to the upper chamber, and the cells were cultured for 48 h. The cells in the upper chamber were removed, and the cells remaining on the membrane were fixed with 4% paraformaldehyde (Solarbio, Beijing, China). After staining with 0.5% crystal violet (Solarbio, Beijing, China), the cells were observed under a microscope and imaged. All experiments were repeated three times.

Stably transduced cells were trypsinized and seeded in a 6-well plate. When the cells were 90% confluent, a 200 µL sterile pipette tip was used to uniformly make vertical scratches in the wells of the 6-well plate. Cells were removed by washing 3 times with PBS and multiple random fields were selected to observe cell migration at 0 h, 24 h, and 48 h. The area and width of the scratches were quantified with Image-Pro Plus 6.0.

Western blotting was performed as previously described [25]. The primary antibodies used for Western blotting were as follows: anti-NDFIP2(1:1000, Bioss Antibodies Inc., Beijing, China; Cat # bs-19059R), anti-pan AKT (1:1000, Abcam, Cambridge, UK; Cat # ab8805), anti- phospho-AKT1 (1:1000, Abcam, Cat # ab66138), anti-N-cadherin (1:1000, Abcam, Cat # ab18203), anti-E-cadherin (1:10,000, Abcam, Cat # ab40772). An anti-β-actin antibody (1:2000, Sigma, USA) was used as an internal control to ensure equal amounts of protein loading.

Stably transduced cells were grown overnight on glass coverslips. The cells were fixed with 4% paraformaldehyde for 10 min at room temperature. The cells were washed 3 times with -cold PBS. The cells were then incubated with PBS (containing 0.3% Triton X-100) for 10 min. The cells were washed 3 times with PBS for 5 min each. The cells were blocked 3% BSA for 30 min at room temperature. The cells were then incubated with an anti-NDFIP2 antibody (1:200, Bioss Antibodies Inc., Beijing, China; Cat # bs-19059R) overnight at 4 °C. The cells were washed 3 times with PBS-T, and were then incubated with Cy3-conjugated goat anti-rabbit (1:400, Servicebio, Wuhan, China; Cat # GB21303) at room temperature in the dark for 60 min. The cells were then incubated with 4′,6-diamidino-2-phenylindole (DAPI, Servicebio) for 5 min. After washing with PBS, images were acquired using a fluorescence microscope (Nikon Eclipse C1; Nikon Corporation). Fluorescence quantitative analysis was performed using Image-Pro Plus 6.0.

pcDNA3.1 was used as the vector to construct the NDFIP2 overexpression plasmid (pcDNA3.1 + NDFIP2 OE), and pGL6 (Beyotime, Shanghai, China) was used as the vector to construct the NDFIP2 3' UTR wild-type (NDFIP2 wt) and NDFIP2 3' UTR mutant (NDFIP2 mut) luciferase reporter plasmids. All constructed plasmids were verified by sequencing (TSINGKE, Beijing, China). The Renilla luciferase reporter plasmid pRL-TK and the pGL6 promoter empty vector were co-transfected with NDFIP2 wt or NDFIP2 mut into the HEK-293T cells in each well using Exfect Transfection Reagent (Vazyme Biotechnology Co., Ltd, Nanjing, China) following the manufacturer's instructions. In brief, 50 ng of pRL-TK and 400 ng of NDFIP2 wt, NDFIP2 mut or pGL6 were added to Opti-MEM, mixed with 1 µL liposomes and incubated for 10 min at room temperature. Forty-eight hours after transfection, the cells were lysed, and a Dual-Luciferase® reporter analysis system (Promega, Madison, WI, USA) was used to perform dual-luciferase reporter assays. A GLOMAX 20/20 luminometer (Promega, Madison, WI, USA) was used to detect luciferase activity. All samples were analysed in triplicate. The same method was used to transfect HCCLM3 cells with pcDNA3.1 + NDFIP2 OE and pcDNA3.1 empty vector (pcDNA3.1 + vector). In brief, 3 µg of pcDNA3.1 + NDFIP2 OE or pcDNA3.1 + vector and 9 µL of liposomes were added to the cells. After 48 h of culture, follow-up experiments were performed.

Gene expression profiles and clinical data were downloaded from the TCGA XENA database (https://​xena.​ucsc.​edu/​) to identify differentially expressed genes (DEGs) between HCC tissues and matched non-tumour tissues. Gene Ontology (GO) enrichment analysis was performed with the for DEGs.

The differential expression of Gly-tRF in L02 hepatocytes and HCC cell lines (HCCLM3, Huh7, HepG2) by were determined qRT-PCR. The expression of Gly-tRF in HCC cell lines was higher than that in L02 cells (Fig. 1A). In the 15 HCC specimens, the expression of Gly-tRF in tumour tissues was elevated compared with that in adjacent non-tumour tissues (Fig. 1B). Based on the online database OncotRF (http://​bioinformatics.​zju.​edu.​cn/​) [26], the median expression level of Gly-tRF in tumour tissues of HCC patients was higher than that in normal tissues (755.06RPM vs 655.40RPM) (Fig. 1C). The abnormal expression of Gly-tRF in HCC suggested that it might be involved in HCC progression. To evaluate the effect of Gly-tRF on HCC cell functions, Gly-tRF negative control, Gly-tRF inhibitor, and Gly-tRF mimic lentiviruses were transduced into HCCLM3 and Huh7 cells. We used qRT-PCR to confirm the Gly-tRF inhibitor blocked the expression of Gly-tRF in HCCLM3 and Huh7 cells (Fig. 1D, Fig. 1E). In addition, whether Gly-tRF mimic transduction boosts Gly-tRF expression was assessed (Fig. 1D, Fig. 1E).

The existence of liver cancer stem cells (LCSCs) is generally considered to be the primary cause of HCC metastasis, malignant growth and treatment failure [27]. We investigated whether Gly-tRF has any effects on LCSCs. LCSC surface markers (CD133/CD13/EpCAM/CD44) is often considered to represent the LCSC population [28]. We compared the percentages of CD133⁺, CD13⁺, EpCAM⁺ and CD44⁺ in HCCLM3 and Huh7 cells transduced with Gly-tRF mimic using flow cytometry. Interestingly, our results showed that HCCLM3 cells stably transduced with the Gly-tRF mimic had higher percentages of CD13⁺, EpCAM⁺ and CD44⁺ (Fig. 2A). However, the percentage of CD133⁺ cells showed no significant difference attributable to discrete detection data (Fig. 2A). Similarly, Gly-tRF elevated the percentages of CD133⁺, CD13⁺, EpCAM⁺ and CD44⁺ in Huh7 cells (Fig. 2C). This result suggests that Gly-tRF expression may be involved in the regulation of LCSC.

Sphere formation assay is a widely used model for studying CSCs [29]. Next, the effect of Gly-tRF on the sphere formation ability was further investigated. Our study reveals that cells transduced with the Gly-tRF mimic had a higher LCSC sphere formation ability (Fig. 2B, D). These findings indicate that in human HCC, Gly-tRF plays a role in the maintenance of LCSC-like properties.

Convincing evidence shows that the existence of CSCs is responsible for EMT in cancer cells [30]. In addition, a large amount of credible evidence shows the driving role of CSCs in tumour migration [31]. Therefore, we continue to evaluate the effect of Gly-tRF on the migration of HCC cells.

Through Transwell assays, we observed that the number of migrated cells increased after Gly-tRF mimic transduction and decreased after Gly-tRF inhibitor transduction (Fig. 3A, B, E, F) in HCCLM3 and Huh7 cells.

Wound healing assay was used to confirm the effect of Gly-tRF on HCC cell migration. Our results showed that Gly-tRF mimic transduction accelerated wound healing, while Gly-tRF inhibitor transduction slowed wound healing at 48 h, (Fig. 3C, D, G, H). These data suggest that Gly-tRF inhibits migration in HCC.

EMT allows cancer cells to infiltrate into the surrounding tissues and metastasizes [32], cancer cell migration is often accompanied by EMT. Next, we used Western blotting to detect the abundance of key markers of EMT. The results showed that in the cells transduced with the Gly-tRF inhibitor, the abundance of N-cadherin decreased, while the abundance of E-cadherin increased. It was also observed that Gly-tRF mimic transduction correspondingly reversed this change (Fig. 3I, J). These results demonstrate that Gly-tRF as a cancer-promoter, plays critical roles in HCC by promoting cell migration and EMT.

Studies have shown that tRNA-derived fragments inhibit the biological functions by binding to the mRNA 3′ UTR of the target gene and exerting effect similar to that of miRNAs [33, 34]. The online database miRDB allows the use of user-provided tRNA-derived fragment sequences for custom target prediction [35]. To investigate the regulatory effect of Gly-tRF on downstream genes, we used miRDB to predict several Gly-tRF target genes (Table. 2).

Then, we evaluated the effect of Gly-tRF mimic transduction on the expression of selected predicted target genes by qRT-PCR. Nedd4 family interacting protein 2(NDFIP2) is an activator of Nedd4 family E3 ubiquitin ligases [36], and its mRNA level decreased after Gly-tRF mimic transduction (Fig. 4A). Interestingly, the NDFIP2 level was found to be significantly reduced in 371 HCC tissue, compared with 50 normal liver tissues by matching the expression levels of NDFIP2 in the TCGA database (https://​portal.​gdc.​cancer.​gov/​), and low NDFIP2 expression was found to imply a poor prognosis of HCC based on the Kaplan–Meier Plotter database (http://​kmplot.​com/​) (Fig. 4B, C). Immunohistochemistry was used to verify the expression of NDFIP2 in HCC, and the results showed that 58/84 (69%) of the samples had lower NDFIP2 expression in tumour tissues than in non-tumor tissues (Fig. 4D). These evidences support our preliminary prediction that Gly-tRF may target to regulate NDFIP2 in HCC.

Dual luciferase reporter assays were employed to determine whether Gly-tRF binds to the seed sequence in the NDFIP2 3′ UTR. There were potential targeted binding sites of Gly-tRF on NDFIP2 3′ UTR based on the miRDB database (Fig. 4E). Our results showed that luciferase activity was significantly reduced when HEK-293T cells were co-transduced with Gly-tRF mimic and NDFIP2 wt reporter plasmid (Fig. 4F). However, pGL6 empty vector and NDFIP2 mut reporter plasmid did not inhibit luciferase activity (Fig. 4F).

We detected the effect of Gly-tRF mimic transduction on the protein expression level of NDFIP2 by immunofluorescence and Western blotting analyses. When cells were transduced with the Gly-tRF mimic, the protein expression level of NDFIP2 was decreased (Fig. 5A–C). Our experiments demonstrate that Gly-tRF specifically targets NDFIP2 in HCC.

Next, we further investigated whether overexpression of NDFIP2 abolishes or reverses the promotive effect of Gly-tRF on the migration of HCC cells and LCSC-like properties. We constructed the NDFIP2 overexpression plasmid, and when the NDFIP2 overexpression plasmid was transfected into HCCLM3 cells, the expression levels of NDFIP2 mRNA and protein were verified to be increased through qRT-PCR and Western blotting analyses (Additional file 2: Fig. S2A, B). Our results showed that overexpression of NDFIP2 reliably reversed the increase in LCSC sphere formation ability induced by Gly-tRF in HCCLM3 cells (Fig. 6A, B). Co-transfection of the NDFIP2 overexpression plasmid with the Gly-tRF mimic abolished Gly-tRF-induced cell migration and EMT-related marker abundance in HCCLM3 cells (Fig. 6C–G). Collectively, these functional reversal experiments further confirmed that Gly-tRF acts as a cancer-promotor by targeting NDFIP2.

Next, we aimed to identify the signalling pathways through which Gly-tRF/NDFIP2 functions. Through GO molecular function (GO-MF) analysis, we found that Gly-tRF/NDFIP2 is mainly involved in ubiquitin protein ligase activity and phosphatidylinositol 3-kinase (PI3K) activity (Fig. 7A).

Previous studies have shown that depletion of NDFIP2 can inhibit AKT activation, and promote HeLa cell proliferation [37]. Western blotting was used to investigate whether the AKT signalling pathway is involved in the function of NDFIP2. We found that in Gly-tRF mimic transduced HCCLM3 cells, the abundance of phosphorylated AKT was increased and was reversed by overexpression of NDFIP2 (Fig. 7B).

In summary, the above data indicate that Gly-tRF negatively regulates the expression of NDFIP2, thereby activating the AKT signalling pathway to promote HCC (Fig. 7C).