The initiator methionine tRNA drives cell migration and invasion leading to increased metastatic potential in melanoma

Abstract

The cell's repertoire of transfer RNAs (tRNAs) has been linked to cancer. Recently, the level of the initiator methionine tRNA (tRNA_i^Met) in stromal fibroblasts has been shown to influence extracellular matrix (ECM) secretion to drive tumour growth and angiogenesis. Here we show that increased tRNA_i^Met within cancer cells does not influence tumour growth, but drives cell migration and invasion via a mechanism that is independent from ECM synthesis and dependent on α5β1 integrin and levels of the translation initiation ternary complex. In vivo and ex vivo migration (but not proliferation) of melanoblasts is significantly enhanced in transgenic mice which express additional copies of the tRNA_i^Met gene. We show that increased tRNA_i^Met in melanoma drives migratory, invasive behaviour and metastatic potential without affecting cell proliferation and primary tumour growth, and that expression of RNA polymerase III-associated genes (which drive tRNA expression) are elevated in metastases by comparison with primary tumours. Thus, specific alterations to the cancer cell tRNA repertoire drive a migration/invasion programme that may lead to metastasis.

Keywords: Cell migration; Extracellular matrix; Integrin; Invasion; Melanoma; Metastasis; RNA polymerase III (Pol III); tRNA repertoire; tRNAiMet.

Fig. 1.

Elevated levels of tRNA_i^Met promote α5β1-dependent cell migration. (A) Immortalised mouse embryonic fibroblasts (iMEFs) were stably transfected with a vector encoding tRNA_i^Met or an empty vector control (Vect.) (2 independent pools of each). The migration speed of cells plated subconfluently, or of those moving into scratch-wounds was determined using time-lapse microscopy followed by cell tracking. Representative cell trajectories are displayed in the left panels. Data are represented as box and whisker plots (whiskers: 10-90 percentile), n=3 independent experiments; ****P<0.0001; ***P<0.001; Mann–Whitney test. Scale bar: 100 μm. (B) iMEFs expressing tRNA_i^Met or an empty vector control (Vect.) were transfected with siRNAs targeting α5 integrin (siRNA-α5; either a SMARTPool or two individual siRNA oligonucleotides as indicated), a non-targeting control (siRNA-nt) (right panel) or were left untransfected (left and centre panels). Cell were plated onto plastic surfaces in the presence of blocking antibodies against α5 integrin (mAb16; left panel), the RGD-containing integrin-binding site in fibronectin (16G3; centre panel) or the appropriate isotype-matched control antibody (IgG). Cell migration speed was then determined as for A. Data are represented as box and whisker plots (whiskers: 10-90 percentile), n=3 independent experiments; ****P<0.0001; ns, not significant; Mann–Whitney test. (C) Conditioned medium was collected from iMEFs stably transfected with a vector encoding tRNA_i^Met or an empty vector control. Conditioned medium was then incubated with iMEFs and the migration speed of the cells determined as for A. Data are represented as box and whisker plots (whiskers: 10-90 percentile), n=3 independent experiments; ns, not significant; Mann–Whitney test. (D) iMEFs expressing tRNA_i^Met were transfected with siRNAs targeting collagen II (siRNA-Col2) or a non-targeting control (siRNA-nt), or iMEFs that had the collagen II gene disrupted using CRISPR (CRISPR-Col2) and their appropriate CRISPR control (CRISPR-Con) [see Clarke et al. (2016)] and their migration speed was determined as for A. Data are represented as box and whisker plots (whiskers: 10-90 percentile), n=3 independent experiments; ns, not significant; Mann–Whitney test.

Fig. 2.

tRNA_i^Met-driven cell migration is dependent on translation initiation ternary complex formation. iMEFs stably expressing tRNA_i^Met or vector control (Vector) were treated with salubrinal (75 μM) or vehicle control (DMSO) for 2 h (A). Alternatively these cells were transfected with a vector encoding the phosphatase, GADD34 or mock control (B). iMEF migration speed was then determined as for Fig. 1A. The trackplots indicate representative migration trajectories of these cells over a 17 h period. The western blots indicate the influence of salubrinal and GADD34 overexpression on levels of phosphorylated eIF2α. Data are represented as box and whisker plots (whiskers: 10-90 percentile); n=3 independent experiments; ****P<0.0001; ns, not significant; Mann–Whitney test. Scale bar: 100 μm.

Fig. 3.

Elevated expression of tRNA_i^Met drives melanoblast migration in the developing embryo. (A) Wild-type (wt) and 2+tRNA_i^Met mice were crossed with a line expressing β-galactosidase under a melanoblast-specific promoter (DCT-LacZ). Embryos were removed at E13.5 and stained for β-galactosidase expression to visualise melanoblasts (A), and the total number (B; left panel) and proportion of β-galactosidase-positive cells within regions 4–6 (B; right panel) of the forelimbs was scored (B, left and right panels). Data are represented as box and whisker plots (whiskers: 10-90 percentile); *P<0.05; ns, not significant; Mann–Whitney test. (C) Wild-type and 2+tRNA_i^Met mice were crossed with animals that were null for p16INK4A and that expressed a mutant allele of NRas under the melanoblast-specific tyrosinase promoter (Tyr-Nras^Q61K; INK4a^−/−). Melanocyte cell lines derived from the early pup skin of these Tyr::Nras^Q61K/°; INK4a^−/−; wild-type (wt) and Tyr::Nras^Q61K/°; INK4a^−/−; 2+tRNA_i^Met mice were plated onto plastic surfaces and their migration determined using time-lapse microscopy as for Fig. 1A. Data are represented as box and whisker plots (whiskers: 10-90 percentile) (right panel) and as mean±s.e.m. (left panel) as indicated; *P<0.05; ***P<0.001; Mann–Whitney test.

Fig. 4.

Elevated levels of tRNA_i^Met promote migration and invasion of melanoma cells. (A) Untransfected WM852 melanoma cells (WM852) or those stably transfected with a vector encoding tRNA_i^Met (two independent pools), tRNA_e^Met, or an empty vector control (Vect.) were plated onto plastic dishes and their migration speed determined as for Fig. 1A. (B) Control (Vect.) and tRNA_i^Met-expressing WM852 cells (pool 2 from A) were transfected with siRNAs targeting α5 integrin (siRNA-α5; either a SMARTPool or two individual siRNA oligonucleotides as indicated) or a non-targeting control (siRNA-nt) (right panel), or were left untransfected (left and centre panels). Migration of these cells was then determined as for Fig. 1A in the presence and absence of an α5β1 integrin blocking antibody (mAb16), an antibody which blocks integrin-fibronectin association (16G3) or the appropriate isotype-matched control antibodies (IgG). (C) WM266.4 melanoma cells were stably transfected with a vector encoding tRNA_i^Met (two independent pools), tRNA_e^Met, tRNA^Thr or empty vector (Vect.) (two independent pools). Cells were allowed to migrate into Matrigel plugs towards a gradient of EGF and serum for 72 h, and then visualised by Calcein-AM followed by confocal microscopy. Optical sections were taken every 15 μm and consecutive images are displayed as a series running from left to right (C; left panels). Cell invasion beyond 45 μm was quantified (C; right panel). Data in A-C are represented as box and whisker plots (whiskers: 10-90 percentile); ****P<0.0001; ***P<0.001; ns, not significant; Mann–Whitney test. All data are from at least 3 independent experiments with multiple internal replicates. (D) WM266.4 cells stably transfecting with a control vector (Vect.) or tRNA_i^Met (two independent pools of each) were plated onto plastic dishes and their rate of proliferation over a 96 h period was determined by cell counting. Values are mean±s.e.m., n=3.

Fig. 5.

tRNA_i^Met drives melanoma metastasis, but not primary tumour growth. (A) WM266.4 cells stably expressing a vector encoding tRNA_i^Met (two independent pools) or an empty vector (Vector) (two independent pools) and were injected subcutaneously into the flank of CD1 nude mice. Subcutaneous tumour growth was measured by callipers three times a week and tumour volume was calculated from these. Values are mean±s.e.m. (B) tRNA_i^Met and empty vector expressing WM266.4 cells were injected subcutaneously into CD1 nude mice. The resulting tumours (5 from each condition) were lysed and their fibronectin and β1 integrin content determined by western blotting. HSP70 is used as a loading control. (C,D) tRNA_i^Met and empty vector expressing WM266.4 cells (two independent pools of each as for A) were injected via the tail vein into CD1 nude mice. The lungs of these animals were assessed for the presence of tumours by visual inspection (C), by determination of lung weight (n=7+7 pool 1 and 7+9 pool 2, pLHCX and tRNA_i^Met respectively) (D, left panel), and by qPCR to quantify the proportion of human genomic DNA (from the WM266.4 cells) with respect to mouse genomic DNA (from the host animal) (D, right panel). n=7+7 pool 1 and 7+11 pool 2, vector and tRNA_i^Met respectively. Values are expressed as box and whisker plots (whiskers: 5-95 percentile). The ratios of human to mouse DNA are expressed on a Log¹⁰ scale. **P<0.01; P<0.05; van Elteren Test (stratified Mann–Whitney). N.B. The injection of cells from pool 1 and pool 2 were conducted in experiments that were independent from one another.

Fig. 6.

Pol III and its products are associated with invasion and metastasis. Heat maps from Oncomine data sets showing pol III (POLR3), TFIIIC (GTF3C) and TFIIIB (Brf1, BDP1) subunit mRNA expression in primary tumours versus patient-matched metastases in melanoma (Xu and Riker datasets; A,B) or prostate cancer (Grasso dataset; C). The heatmap scale bar is indicated at the top, and the average degree of upregulation of transcripts in metastasis and statistical significance values of these are tabulated on the right-hand side. In Oncomine, all data are normalised as follows: negative values were not included and all data were log²-transformed, median-centred per array, and the standard deviation was normalised to one per array. Depending on the type of microarray (one or two colours) data are presented either as log² median-centered intensities (Xu and Riker), or log² median-centered ratios (Grasso). The scale bar is log².