Transketolase-like 1 ectopic expression is associated with DNA hypomethylation and induces the Warburg effect in melanoma cells

We quantified TKTL1 transcripts by qRT-PCR in a panel of normal human tissues and metastatic melanoma tumors. Figure 1a shows that high expression of TKTL1 mRNA was detected in testis but no other normal human tissues tested including skin and melanocytes. 15 of 38 (40 %) melanoma tumors assessed expressed TKTL1 to varying degrees. We examined TKTL1 by immunohistochemistry using a tissue microarray (TMA) comprising 81 tumors from patients with stage III and IV metastatic melanoma. Four representative tumors with high and diffuse intensity staining of tumor cells for TKTL1 are depicted in Fig. 1b. TKTL1 in tumors varied from homogenous to heterogeneous expression with clusters of TKTL1 positive cells interspersed with TKTL1 negative cells. TKTL1 expression in testis tissue was used as positive control and anti-IgG antibody was used as negative control (Fig. 1b). 31 of 81 (38 %) of metastatic melanoma tumors were scored positive for TKTL1 expression (Fig. 1c). High magnification image shows that TKTL1 localization is predominantly cytosolic and membrane staining is also seen in some melanoma tumor cells. Nuclear staining in some tumors is Melanin (brown) and not TKTL1 (red) (Additional file 1: Figure S1A). We extended our studies to a clinical outcome dataset that subdivided 57 stage IV melanoma samples into high-risk or low-risk melanoma groups based on transcriptome profiling (GSE22153) [38]. We found that tumors expressing high TKTL1 levels were statistically significantly associated with the high-risk group (log-rank p value 0.0277), risk being reduced survival and risk of relapse.

In a panel of 53 metastatic melanoma cell lines TKTL1 mRNA was detectable in 11 of 53 lines (20 %) by qRT-PCR (Additional file 1: Figure S1B). We selected two cell lines for further functional studies; LM-MEL-59, which showed the highest level of TKTL1 mRNA expression and LM-MEL-44 which lacked TKTL1 expression (Fig. 2a). Immuno-staining confirmed the expression of TKTL1 in LM-MEL-59 and its absence in LM-MEL-44 (Fig. 2b). Both of these cell lines lack activating mutations in the BRAF protein kinase. High magnification image confirms cytosolic TKTL1 localization (Additional file 1: Figure S1C).

Expression of TKTL1 has been reported to be controlled by methylation and demethylation of TKTL1 has been previously reported in lung and head and neck cancers [24, 39]. However, little is known about the methylation status of TKTL1 in melanoma. To investigate promoter cytosine-(phosphate)-guanine (CpG) dinucleotide DNA methylation, we quantified the level of TKTL1 promoter methylation using bisulfite modification and MS-qPCR. Melanocytes were found to be hypermethylated which correlated with the lack of TKTL1 expression in those cells (Fig. 2c), while metastatic melanoma cell lines tested showed reduced methylation levels when compared to melanocytes. LM-MEL-44 cell line with no TKTL1 expression showed hypermethylation of the TKTL1 promoter whereas a lesser degree of TKTL1 promoter methylation was detected in LM-MEL-59 with high TKTL1 mRNA expression. Promoter hypermethylation of TKTL1 could be reversed by a DNA demethylating agent 5-aza-2′-deoxycytidine (5aza) in LM-MEL-44 resulting in the expression of TKTL1 in treated LM-MEL-44 cells in contrast to DMSO treated cells (Fig. 2d).

We extended our studies to a publically available clinical melanoma patient dataset to determine whether deregulation of TKTL1 expression was linked to DNA hypomethylation. Using MethPrimer (www.​urogene.​org/​cgi-bin/​methprimer/​methprimer.​cgi), a CpG island prediction program, we identified two CpG sites in the TKTL1 promoter that were highly prone to methylation (data not shown). The primers we designed for MS-qPCR spanned these two CpG sites; cg09892236 (−100 bp of transcription start site (TSS)) and cg23106779 (−250 bp of TSS). We used an in silico approach to assess the methylation status of the promoter region on the TKTL1 gene in 385 metastatic melanoma tissues from the Skin Cutaneous Melanoma (SKCM) dataset of The Cancer Genome Atlas (TCGA) database (www.​cbioportal.​org) [40, 41] and estimated its association with TKTL1 mRNA expression. We first evaluated DNA methylation of the two CpG sites using the Illumina Infinium Human Methylation 450 k BeadChip array (450 k array). Methylation of the two CpG sites were statistically associated with each other with the melanoma dataset (correlation = 0.68, p-value = 2.2e-16). We next examined the exon and gene expression pattern of TKTL1 in these patient samples in the Illumina Hiseq RNA-Seq Platforms. Spearman correlation coefficients were calculated to test the correlations between DNA methylation and gene expression of TKTL1. We found that methylation status at both the CpG sites was statistically inversely correlated with TKTL1 gene expression in melanoma samples (Table 1 and Additional file 2: Figure S2A&B). Melanoma samples were further sub-divided into two groups, primary (n = 82) and metastatic (n = 303) in order to examine the role of DNA methylation in regulation of TKTL1 gene expression within primary and metastatic melanoma samples. We found positive correlation of TKTL1 expression with methylation in metastatic samples (Table 2). These results confirm that DNA hypomethylation is associated with aberrant activation of TKTL1 expression in metastatic melanoma.

The epigenetic reactivation of TKTL1 occurs simultaneously with DNA hypomethylation of CTAgs in head and neck cancer [39]. To determine whether TKTL1 expression in melanoma was linked to the expression of CTAgs, we queried the TCGA database comprising of 261 melanoma patient samples. We found a significant tendency toward co-occurrence of many CTAgs with TKTL1 expression in melanoma tumors (Table 3).

To determine the role of TKTL1 in melanoma, we sought to manipulate TKTL1 levels in melanoma cell lines by siRNA-mediated knockdown and vector-mediated over-expression. Effective suppression of TKTL1 on mRNA and protein levels in LM-MEL-59 was achieved at 10 nM concentration of siRNA (Fig. 3a & b). To confirm that the effects shown by TKTL1 were not off-target effects, we performed microarray [42] and qRT-PCR for the isoforms of TKTL1 namely, TKT and TKTL2. We found that TKT levels were high in both LM-MEL-59 and LM-MEL-44, while TKTL2 levels were low. Additionally, we have used testis tissue as a positive control for the expression of these genes. Upon knockdown with TKTL1 siRNA, TKT levels were affected by one of the siRNAs but only modestly by the other (Additional file 3: Figure S3).

Transfection with a TKTL1 expression plasmid resulted in marked overexpression of TKTL1 in LM-MEL-44, demonstrated by qRT-PCR and immunoblotting (Fig. 3c & d). No additional bands were detected for TKTL1 as evident in Additional file 4: Figure S4A&B. High transfection efficiency was also confirmed with immunofluorescence staining (Additional file 4: Figure S4C). Elevated glucose uptake and lactate production have been detected during malignant transformation. These metabolic parameters of cancer cells are consistent with a pronounced glycolytic state [43, 44]. To investigate whether TKTL1 expression in melanoma was driving the Warburg effect, glucose consumption and lactate production were assayed after TKTL1 interference. Measurements were made on conditioned medium of LM-MEL-59 cells after treatment with TKTL1 or control siRNA for 72 h. Depletion of TKTL1 in these cells caused a substantial decrease in glucose consumption and associated reduction in lactate production (Fig. 3e & f) while ectopic expression of TKTL1 in LM-MEL-44 conferred a more glycolytic phenotype (Fig. 3 g & h). Collectively, these results showed that TKTL1 enhanced aerobic glycolysis in melanoma.

TKTL1 plays an important role in tumor proliferation in gastric, colon and hepatocellular carcinomas [19, 26, 45]. We investigated whether TKTL1 plays a similar role in melanoma. TKTL1 knockdown in LM-MEL-59 led to an increase in the percentage of G0-G1 phase cells and a decrease in the percentage of S phase cell population compared to control transfected cells (Fig. 4a, b & e) as demonstrated by cell cycle analysis. Using MTS assay to determine total cell number after TKTL1 knockdown, similar changes were detected. However, effects were only modest after 48 h (Additional file 5: Figure S5A).

Conversely, TKTL1 overexpression led to a decrease in the percentage of G0-G1 phase cells in LM-MEL-44 and an increase in the percentage of S phase cells (Fig. 4c, d & f). Table 4 depicts the cell cycle phase distributions under the different conditions. Similarly, MTS assay demonstrated slightly enhanced proliferation in TKTL1 overexpressing cells compared to empty vector treated cells, although no significance was achieved (Additional file 5: Figure S5B). Overall, these results indicate that TKTL1 levels may affect melanoma cell proliferation.

Next, we sought to investigate the effect of TKTL1 on cell invasion. TKTL1 depleted LM-MEL-59 cells were significantly less invasive, while TKTL1 overexpressing LM-MEL-44 cells were more invasive than controls in the matrigel-coated Boyden chamber assay (Fig. 5a & c). TKTL1 depleted cells showed 62 % loss and TKTL1 overexpressing cells showed 38 % increase in invasion (Fig. 5b & d). These results demonstrate a substantive effect of TKTL1 levels on melanoma cellular invasion.

In tumors expressing high TKTL1, oxythiamine has been used to block proliferation effectively both in vitro and in vivo [46]. We treated both LM-MEL-59 and LM-MEL-44 with the inhibitor in varying dose of 10 and 100 μM (Additional file 6: Figure S6A&B). In both cell lines no reduction in proliferation was detected. To assess if Oxythiamine induced a G1 phase cell cycle arrest, LM-MEL-59 cells treated for 72 h with 100 μM of the inhibitor was subjected to flow cytometric cell cycle analysis. However, Oxythiamine treatment in our hands did not cause a G1 phase arrest in melanoma cells (Additional file 6: Figure S6C). Furthermore, Oxythiamine treatment did not alter invasive potential (Additional file 6: Figure S6D & E).

Melanoma cell lines were established from resected melanoma metastases by mechanical dissociation of tissue with subsequent overnight digestion in media containing collagenase IV and DNAse at 37 °C. Established cell lines were Mycoplasma-tested using the MycoAlert test (Lonza Rockland, Inc.) and STR-profiled for identity confirmation.

Following pathological examination post-surgery tumour fragments were formalin-fixed and paraffin-embedded (FFPE) by staff at the Austin Health Tissue Bank. All melanoma tissue was derived from metastases resected from regions distant to the primary tumor; no locoregional metastases or primary tumors were included. Tissue microarray consists of tumor tissue embedded in a block and allows analysis of multiple patient tissue samples on the same slide. The study cohort consisted of 81 patients with stage III and IV metastatic melanoma. All tissue donors provided written informed consent for tissue collection and research, which was covered by protocols approved by the Austin Health Human Research Ethics Committee, Melbourne, Australia (approval number H2012/04446). All cell lines were matched with their donors by HLA-typing. Cells were cultured in RPMI1640 supplemented with 10 % fetal calf serum (FCS) as described previously [42].

Total RNA was extracted using the RNEasy mini kit (Qiagen). We have performed on column DNAase digestion using RNase-Free DNase at room temperature (20–30 °C) for 15 min in accordance to Qiagen RNEasy Kit instructions. Spectrophotometric quantification using Nanodrop confirmed purity of RNA and absence of DNA in our samples. In addition to this to prevent any interference by residual DNA contamination, we have performed qRT-PCR with primers that are intron spanning such that genomic DNA will not be amplified.

RNA was reverse transcribed using the High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Following reverse transcription, qRT-PCR was performed using the SensiFAST SYBR Lo-ROX kit (Bioline), and a Vii-A7 thermocycler (Applied Biosystems). The primers for TKTL1 and β-actin (internal control) were synthesized by Sigma Aldrich and the sequences were as follows: TKTL1 primer (forward) 5’-CCA CCT GAT TAC AGA GTT GGT G-3’ and (reverse) 5’-CTC TGT TGT TCG CGT AGC C-3'; β-actin (forward) 5'-CCA ACC GCG AGA AGA TGA-3’ and (reverse) 5'-CCA GAG GCG TAC AGG GAT AG-3’, TKT primer (forward) 5’-TGT GTC CAG TGC AGT AGT GG-3’ and (reverse) 5’-ACA CTT CAT ACC CGC CCT AG-3’; TKTL2 primer (forward) 5’-ACG ACC GGT TCA TCC TCT C-3’ and (reverse) 5’-TCC ACC CAA GCA GCA TAG A-3’.

Immunohistochemistry was performed on tissue microarrays constructed from formalin-fixed paraffin-embedded (FFPE) metastatic melanomas and melanoma cell lines after citrate buffer (pH 6.0) antigen retrieval for 30 min using the Dako Envision⁺ kit (Dako). Incubation with 1× protein blocking buffer for one hour at room temperature was followed by overnight incubation at 4 °C with mouse monoclonal anti-TKTL1 antibody (clone JFC12T10; 1:200 dilution) previously described by Langbein et al. [17]. 60 min incubation with secondary anti-mouse antibody HRP (Dako) was performed. 3-amino-9-ethylcarbazole (AEC) was used as chromagen (Vector Laboratories). Slides were counterstained with hematoxylin, images were captured and immunohistochemical reactivity was evaluated by two independent investigators. Patient tumor cores were scored TKTL positive based on cytosolic staining in minimum of 10 % tumor cells. Human testis tissue was used as the positive control for TKTL1 and a negative control, for which the primary antibody was substituted with the same concentration of mouse IgG.

Genomic DNA was extracted using phenol chloroform (Qiagen), and bisulfite modification of genomic DNA was performed according to the manufacturer’s instructions (Qiagen). Methylation specific and methylation independent primers (Sigma Aldrich) were used to quantify the relative levels of methylated and unmethylated products within PCR amplified samples. The primers are as follows: TKTL1 methylated primer (forward) 5'-GAC GTC TAA AAA ACG AAT AAC GC-3' and (reverse) 5'-AAA GAA CAT TTT GTA TTC GCG C -3'; TKTL1 unmethylated primer (forward) 5'-CAA CAT CTA AAA AAC AAA TAA CAC C-3' and (reverse) 5'-GAA AGA AGA TTT TGT ATT TGT GTG G-3'; β-actin primer (forward) 5'-TGG TGA TGG AGG AGG TTT AGT AAG T -3' and (reverse) 5'-AAC CAA TAA AAC CTA CTC CTC CCT TAA -3'.

We acquired data form The Cancer Genome Atlas (TCGA) data portal (www.​cbioportal.​org). We started our analysis with the Level 3 data download (normalized data) of Illumina Infinium Human Methylation 450 k BeadChip array (450 k array) (http://​www.​illumina.​com/​products/​methylation_​450_​beadchip_​kits.​html), and Illumina Hiseq RNA-Seq Platforms with RNA-seq gene and RNA-seq exon expression datasets (http://​www.​illumina.​com/​applications/​sequencing/​rna.​html) of melanoma from the TCGA website (https://​tcga-data.​nci.​nih.​gov/​tcga/​dataAccessMatrix​.​htm?​mode=​ApplyFilter&​showMatrix=​true&​diseaseType=​SKCM&​tumorNormal=​TN&​tumorNormal=​T&​tumorNormal=​NT). Methylation measurement of each CpG probe was called β and was calculated as the methylated signal divided by the sum of the methylated and the unmethylated signal. Two methylation probes in accordance with Illumina annotation and UCSC genome browser that overlapped with our targeted region by MS-QPCR were selected. As there were some CpG sites between the two probes that was not covered by 450 k platform, we only investigated the correlation between these two probes in the 450 k platform by Spearman correlation coefficients statistical test (p < 0.05) to estimate methylation level of our targeted region. Subsequently, correlations between the two probes were separately assessed with TKTL1 gene expression and exon expression. All analysis was performed by R statistical programming language (3.1.1).

Melanoma cell line LM-MEL-44 was treated with 0.1 μM of 5aza (Sigma) for 4 days. The medium and the 5aza were refreshed on alternate days. Cells were confluent at time of harvest.

The plasmid encoding TKTL1 (clone ID 4825931) was obtained from Life Technologies and the TKTL1 gene was subcloned into pcDNA3.1(+) vector (Promega). Cells were transfected using Lipofectamine 2000 (Life Technologies). Empty vector control used was pcDNA3.1(+). For transient siRNA transfection, cells at 30 % confluence were transfected using a control siRNA and two different Silencer select siRNAs targeting TKTL1 (s224894 and s15774) at 10nM final concentration (Ambion) with Lipofectamine RNAiMAX according to the manufacturer’s protocol (Invitrogen).

LM-MEL-44 transfected with TKTL1 overexpression plasmid construct was incubated for 72 h and then fixed with 4 % paraformaldehyde, stained with anti-TKTL1 rabbit polyclonal antibody (ab87187, Abcam) which was applied at 3 μg/mL concentration overnight at 4 °C and with Alexa flour 555 (anti-rabbit) conjugated secondary antibody for 45 min at room temperature (Molecular probes, USA). Cells were counter stained with DAPI for 10 min.

Cell-free supernatant of cultured melanoma cells was collected, and glucose and lactate concentrations were determined using colorimetric glucose and lactate assay kits (Biopharma). The assays were performed according to the manufacturer’s instructions and absorbance was measured at 340 nm.

MTS colorimetric assay was performed using standard protocols. Twenty thousand cells were seeded per well of a 96 well plate and treated as indicated. Relative cell numbers were measured using the CellTiter 96® AQueous One Solution Cell Proliferation Assay (Promega Corporation).

Invasion assays were performed with Boyden chamber insert (6.5 mm diameter inserts with 8.0 μm pores) coated with Matrigel (Becton, Dickinson and Company) as described previously [70]. Briefly, melanoma cells were seeded at 40,000 cells per insert. 600 μl of media containing 10 % FCS was added as chemo attractant to the lower compartment. The cells were incubated overnight at 37 °C and subsequently fixed with 4 % paraformaldehyde, stained with 0.1 % Crystal Violet and cells on the upper surface of the insert were removed. Invaded cells on the bottom side were counted from three random fields of view observed with a 10X objective lens. Average of invaded cells across three fields of view was calculated from independent experiments repeated three times.

Melanoma cells were treated with varying dose of oxythiamine (Sigma) as indicated for 24, 48 and 72 h. Following treatment cells were subjected to proliferation, invasion and cytofluorometric analyses.