eIF3a is over-expressed in urinary bladder cancer and influences its phenotype independent of translation initiation

Abstract

Purpose

The eukaryotic translation initiation factor (eIF) 3a, the largest subunit of the eIF3 complex, is a key functional entity in ribosome establishment and translation initiation. In the past, aberrant eIF3a expression has been linked to the pathology of various cancer types but, so far, its expression has not been investigated in transitional cell carcinomas. Here, we investigated the impact of eIF3 expression on urinary bladder cancer (UBC) cell characteristics and UBC patient survival.

Methods and results

eIF3a expression was reduced through inducible knockdown in the UBC-derived cell lines RT112, T24, 5637 and HT1197. As a consequence of eIF3a down-regulation, UBC cell proliferation, clonogenic potential and motility were found to be decreased and, concordantly, UBC tumour cell growth rates were found to be impaired in xenotransplanted mice. Polysomal profiling revealed that reduced eIF3a levels increased the abundance of 80S ribosomes, rather than impairing translation initiation. Microarray-based gene expression and ontology analyses revealed broad effects of eIF3a knockdown on the transcriptome. Analysis of eIF3a expression in primary formalin-fixed paraffin embedded UBC samples of 198 patients revealed that eIF3a up-regulation corresponds to tumour grade and that high eIF3a expression corresponds to longer overall survival rates of patients with low grade tumours.

Conclusions

From our results we conclude that eIF3a expression may have a profound effect on the UBC phenotype and, in addition, may serve as a prognostic marker for low grade UBCs.

Electronic supplementary material

The online version of this article (doi:10.1007/s13402-014-0181-9) contains supplementary material, which is available to authorized users.

Introduction

The eukaryotic translation initiation factor (eIF) 3a has been described in a number of different carcinomas as being aberrantly expressed and, as such, to be involved in the development and progression of these malignancies [1], including adenocarcinomas (AC) of the breast, lung, colon and stomach, and squamous cell carcinomas (SCC) of the cervix, oral cavity and oesophagus [1]. In both carcinoma types eIF3a was found to be up-regulated during early tumour development. Due to its higher expression in low grade, less invasive tumours compared to its lower expression in de-differentiated, metastatic tumours, eIF3a has also been denoted as an anti-cancer protein [1]. In contrast, a putative pro-cancer role of eIF3a has been proposed in especially AC cells, as exogenous expression of eIF3a antisense cDNA led to reversal of the malignant phenotype in breast and lung cancer cell lines [2]. Additionally, two eIF3a single nucleotide polymorphisms (SNPs; reference IDs: rs10787899 and rs3824830) have been identified as risk factors for the development of AC of the breast and pancreas [1]. Transitional cell carcinomas (TCC) represent another major cancer type and is the common form occurring in the urinary bladder. Therefore, we investigated eIF3a expression in this carcinoma type to assess its pro- or anti-cancer properties.

Urinary bladder cancer (UBC) is the ninth most common type of cancer worldwide. It is the seventh most common malignancy in males and is three times less frequent in females. Its incidence varies widely between geographical regions and is especially high in developed countries [3]. The mortality rate of UBC is 4 in 100,000 for men and 1.1 in 100,000 for women [3]. The treatment of UBC encompasses Bacillus of Calmette and Guerin (BCG) immune-therapy, tumour resection, radical surgery, urinary diversion, radiotherapy and different chemotherapeutic regimens [4]. Individual therapy decisions are taken based upon EAU guidelines (http://www.uroweb.org/guidelines/online-guidelines). Chemotherapy for UBC is based on a combined application of Gemcitabin and Cisplatin, but also on M-VAC (Methtrexate, Vinblastin, Doxorubicin and Cisplatin) or Vinflunin, as second line approaches [5].

Suitable biomarkers to assist proper diagnostic and therapy decisions are, however, still lacking for UBC. Potential biomarkers that have been reported for UBC development and progression include survivin [6], DEK [7] and KISS1 [8] but, as yet, none of them has reached clinical application. An in-depth analysis of gene expression profiles of UBCs is, therefore, considered indispensable. Today’s understanding of the molecular mechanisms underlying UBC is that two different pathways of tumour development and progression exist [9]. These two pathways are marked by the accumulation of several pathogenic alterations in slow, papillary, minimally invasive tumours and, on the other hand, in non-papillary, highly malignant, invasive urothelial carcinomas [10]. Papillary tumours make up 80 % of urothelial neoplasms and are characterized by papillary and exophytic growth patterns. Molecular alterations in these tumours frequently include mutations in the HRAS and FGFR3 genes. These tumours display a high recurrence rate of approximately 70 %. Only about 15 % of these tumours will, however, progress to invasive lesions. In highly malignant progressive cases, mutations in the p53 tumour suppressor gene are frequently found. Involvement of p53 is a known primary indicator of neoplasms following an invasive, fast growing course. Such tumours lack the papillary phenotype and frequently originate from a carcinoma in situ. Besides p53, the tumour suppressor RB1 is frequently altered in invasive UBC. Another characteristic of these tumours is the occurrence of gross genetic distortions, including translocations, chromosomal duplications and losses [10].

Since eIF3a over-expression is known to differently affect the development of ACs and SCCs, eIF3a expression may represent an additional parameter corroborating individual growth patterns of UBCs. eIF3a is the largest subunit of the eIF3 complex, that further comprises the subunits eIF3b to eIF3m. The first role of the eIF3 complex in translation initiation is to bind to the solvent side of free 40S ribosomal subunits and to prevent premature binding to 60S ribosomal subunits [11]. In the 40S-associated state, eIF3a, eIF3b and eIF3d are capable of binding mRNA, which enhances the interaction of 40S ribosomal subunits and eIF2-tRNA_i ^Met-ternary complexes required for mRNA activation [12]. After successful completion of the scanning process and AUG recognition, eIF3 dissociates from the 40S ribosomal subunit after which translation proceeds. eIF3a is also suggested to take part in the post-termination disassembly and recycling of ribosomal subunits [11, 13]. In nasopharyngeal, oral cavity and lung cancers, eIF3a expression has been found to influence the patient’s sensitivity to DNA-damaging chemotherapy, i.e., high eIF3a expression in the tumours of these patients has been found to correlate with a better overall survival upon platinum-based chemotherapy. This effect could be explained by a specific eIF3-dependent regulation of nuclear excision repair proteins in vitro [14–16]. eIF3a’s association with tumour development and progression in various cancers and its eIF3 complex-independent occurrence underscores our hypothesis that eIF3a is not only a translation initiation factor, but also an independent regulator of tumour growth and progression.

Here, we show that eIF3a expression is up-regulated in UBC cells and that in vitro knockdown counteracts their malignant phenotype. We also found that inducible knockdown of eIF3a in a xenotransplant mouse model reduces its tumour load. Moreover, we observed a correlation of eIF3a expression with overall survival in a cohort of 91 UBC patients.

Materials and methods

Cell cultures

The UBC cell lines RT112, T24, 5637 and HT1197 were selected for this study. HT1197 cells were cultured in Minimum Essential Medium (MEM) supplemented with 10 % FCS, 1 % non-essential amino acids, 1 % L-glutamine and 1 % penicillin/streptomycin. RT112, T24 and 5637 were cultured in Roswell Park Memorial Institute (RPMI)-1640 medium containing 10 % FCS, 1 % L-glutamine and 1 % penicillin/streptomycin. The HT1197 cells were kindly provided by Dr. H. Klocker (Department of Urology, Innsbruck Medical University) and the RT112, T24 and 5637 were a kind gift of Dr. R. Illes and Dr. S. Butler (Middlesex University London). All cell lines were cultured at 37 °C in a humidified atmosphere containing 5 % CO₂. Passaging was conducted every 2–4 days, depending on cell density. Cells were split before reaching 80 % confluence.

Western blotting

Protein was isolated from PBS-washed cell culture pellets, lysed in NP-40 Lysis buffer (0.05 M Tris–HCl, 0.15 M NaCl, 0.5 % NP-40, 0.001 M Pefabloc, 0.001 M DTT) and homogenised with a Potter tissue homogenizer (Kontes Glass Co, Duall 20). Crude protein concentrations were determined using a Bradford protein assay (Biorad Protein Assay Dye Reagent, 500–0006; BioRad Laboratories GmbH, Munich, Germany). 30 μg of the proteins were loaded onto SDS-PAGE gels, subjected to electrophoresis in Mini-vertical electrophoresis units (Amersham Biosciences) and blotted onto PVDF membranes (Immobilin-P Tranfer Membrane; Millipore) using a Semi Dry Blotting Unit (JH BioInovations). After blocking by 5 % non-fat milk (Applied Chemistry) in TBS-Tween (0.1 %) primary antibodies were diluted in TBS-T, 5 % BSA and incubated o/n. Secondary antibody solutions for anti-mouse and anti-rabbit were purchased from Amersham. Detection was performed using an ECL Plus Western Blotting Detection Reagent (GE Healthcare) and ECL Hyperfilms (GE Healthcare).

Real-time RT-PCR

RNA was isolated from cells using a Qiagen RNeasy Mini Kit according to the manufacturer’s instructions (Qiagen). RNA concentrations and qualities (optical density (OD) ratio 260/280) were measured using a NanoDrop 2000 spectrophotometer (Peqlab). cDNA was synthesized from total RNA using a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems) according to the recommended protocol. The PCR reactions were performed using TaqMan® Gene Expression Assays containing gene-specific primers and fluorescently labelled probes (LifeTechnologies). The genes analysed were: CyclinB1 (Hs01030099_m1) and MDM2 (Hs01066930_m1), and the GAPDH (Hs03929097_g1) gene was used as reference. In brief, for one reaction 4 μl 2× TaqMan Gene Expression Mastermix, 0.4 μl 20× TaqMan Gene Expression Assay and cDNA diluted in ddH2O were combined in a 10 μl reaction mixture, with a final cDNA concentration of 30 ng. All runs were performed and evaluated using the 7900HT Fast Real-Time PCR System (Applied Biosystems).

Knockdown construct generation and lentiviral transfection

For the shRNA-mediated knockdown of eIF3a the following oligonucleotide sequences were used: 5′-gatcccccccttggcaggtattgagtaattcaagagattactcaatacctgaagggtttttggaaa-3′ (targeting the 3′UTR) and 5′-gatccccccactagagagttcgctgattcaagagatcagcgaactctctagtggtttttggaaa-3′ (targeting the 5′UTR). These oligos were cloned into the gateway compatible pENTR.THT.III vector and the pHR-Dest-tetR-GFP vector, respectively. Correctly cloned expression vectors, containing a GFP sequence and the independently expressed shRNA directed against eIF3a under the control of the tetracycline-sensitive RNAseP H1 promotor (THT), were used for lentivirus production in 293T packaging cells and the infection of target cells. These target cells (HT1197, RT112, T24 and 5637) were grown to approximately 50 % confluence and incubated with viral particles obtained from the supernatants of the 293T cultures. The transfection efficiency was assessed by the percentage of GFP expressing cells, which was around 80 % in all cell lines after 48 h. The efficiency of eIF3a knockdown was tested by Western blotting. In order to increase the knockdown efficiency in HT1197 cells, which responded only marginally to the first transfection, a double infection was performed prior to the microarray analyses. Despite the ultimately higher knockdown in HT1197 as compared to other cell lines, they were not used in all experiments due to their exceptional individual growth pattern and proliferation behaviour, which impeded a proper comparison between all cell lines.

In vitro phenotype analyses

The UBC-derived cell lines RT112, T24 and 5637, and the corresponding cells bearing the shRNA constructs against eIF3a (RT112 tetR eIF3a sh 3.1 GFP; T24 tetR eIF3a sh 3.1 GFP; 5637 tetR eIF3a sh 3.1 GFP) were analysed for alterations in proliferation, migration, invasion of basal membranes and colony formation. As an additional control, also the construct-bearing cells without shRNA induction were tested. shRNA expression was induced by the addition of 1 μg/ml doxycycline to the culture medium, which was changed every 48 h. Growth curves were obtained by seeding 20,000 cells per well in a 6-well plate, either with or without the addition of doxycycline, and counting of cells after 24, 48, 72 and 96 h. Triplicates were performed in three independent experiments. Clonogenicity was evaluated by seeding 100 or 1000 UBC cells/well into 6-well plates. Doxycycline was refreshed and medium changed every two days. After 9 days the medium was removed, cells were washed once in PBS and fixed in methanol prior to staining with 0.5 % crystal violet for 15 min. After removal of the staining solution and washing in tap water, the plates were analysed and photographed. Triplicates were performed in three independent experiments. For the assessment of UBC cell invasiveness, the CytoSelect™ 96-Well Cell Invasion Assay (Cell Biolabs, San Diego, CA, USA) was used according to the manufacturer’s instructions. In brief, a cell suspension was placed on top of a pre-set basement membrane in medium containing FCS. After 12 h transmigration towards the lower chamber compartment containing serum-supplemented medium was monitored. In contrast to the manufacturer’s protocol cells were counted directly, not fluorometrically.

Xenotransplantation experiments

RT112/T24/5637 tetR eIF3a sh 3.1 GFP expressing cells were harvested (10⁷ cells in 0.1 ml PBS) and injected into the flanks of six nude mice for each cell line. Half of the animals were treated with doxycycline (6 mg/ml) via drinking water, which was supplied ad libitum and exchanged twice a week. The other half received sucrose as a control. Body weight and tumour volume were measured weekly. After the tumours reached a size of 1.0 cm³, the animals were sacrificed, and the tumours were excised and cryoconserved. Immunohistochemical staining against eIF3a was performed on tumour slides to confirm the reduction in eIF3a expression levels upon doxycycline treatment.

Caspase activation as a measure of apoptosis induction

In order to measure cell death and to determine a putative correlation of eIF3a expression with caspase activation and apoptosis induction, the MultiTox-Fluor Assay was multiplexed with the Caspase-Glo 3/7 Assay according to the manufacturer’s recommendations (Promega, Madison, WI, USA). In brief, cells were treated with doxycycline to induce shRNA-mediated knockdown of eIF3a, with puromycin (Sigma Aldrich, P9620) to inhibit mRNA translation initiation and elongation globally, or with rapamycin (Sigma Aldrich R8781), which is a specific inhibitor of the translational regulator kinase mTOR. Cytotoxicity and apoptosis induction were measured after 6, 12, 24 and 48 h. The fluorescent signals of R110 (dead cells: ex 504/12 nm, em 542/27 nm) and AFC (life cells: ex 406/15 nm, em 465/35 nm) were recorded prior to addition of the Caspase-Glo 3/7 reagent, which was measured via the Luminescence channel. All measurements were performed on a Beckman Coulter Paradigm platform, using FI-DL Dual Label Detection (Beckman Coulter A68244) and Luminescence Cartridges (Beckman Coulter A56144).

Microarray-based gene expression profiling and data analysis

Gene expression profiling analyses were performed by the Expression Profiling Unit at the Innsbruck Medical University. For total RNA isolation the RNAeasy Mini Kit (Quiagen, Valencia, CA) was used according to the manufacturer’s protocol. RNA quantity and purity were determined by optical density measurements (OD260/280) and RNA integrity by using the 2100 Bioanalyser (Agilent Technologies, Palo Alto, CA). Only high quality RNA was further processed. For Affymetrix GeneChip analysis 500 ng total RNA extracted from eIF3a knockdown and control HT1197 cells (three independent samples each) was processed to generate biotinylated hybridization targets using the One Cycle cDNA Synthesis and One Cycle Target Labelling Kits from Affymetrix (Affymetrix, Santa Clara, CA, USA) according to the manufacturer’s protocols. In brief, total RNA was reverse-transcribed into cDNA using an anchored oligo-dTT7-Primer, converted into double-stranded cDNA and purified with the Affymetrix Sample Clean-up Kit. Subsequently, cRNA was generated by T7 polymerase-mediated in vitro transcription introducing a modified nucleotide for subsequent biotinylation. Following RNA purification, 20 μg of cRNA was fragmented at 95 °C using the Affymetrix fragmentation buffer, mixed with hybridization buffer containing hybridization controls and hybridized to Affymetrix HGU133 plus 2.0 GeneChips. The microarrays were stained and washed in an Affymetrix fluidic station 450. Fluorescence signals were recorded by an Affymetrix scanner 3000 and image analysis performed using the GCOS software. Raw and pre-processed data were deposited in the Gene Expression Omnibus depository (accession number GSE45636). All further analyses were performed in R (http://www.r-project.org; version 2.15.1) using packages from the Bioconductor project [17]. Functions from the affyPLM package were used for quality assessments of the microarrays and only good quality microarrays were further analysed. GeneChip raw expression values of six microarrays were normalized and summarized using the GCRMA method [18]. The moderated t-test [19] was employed to assess the significance of differential expression patterns between the three eIF3a knockdown and three control samples. The resulting p-values were further adjusted for multiple hypotheses testing using the method from Benjamini and Hochberg [20] for a strong control of the false discovery rate. Probe sets with an adjusted p-value smaller than 0.01 and a M-value bigger than 1 or smaller than −1 (i.e., more than two-fold differentially expressed) were considered significant. For Gene Ontology analyses the Bioconductor’s GOstats [21] package was employed and all genes detected on the microarray were used as background gene set.

Sucrose density gradient centrifugation

Cells were seeded in 10 cm dishes (three per lysate) and incubated at 37 °C with 100 μg/ml cycloheximide for 15 min prior to harvest. All further steps were performed on ice. Cells were washed twice in cold PBS containing 100 μg/ml cycloheximide, scraped into 1 ml of lysis buffer (20 mM HEPES, pH 7.4; 15 mM MgCl2; 200 mM KCl; 1 % Triton X-100; 2 mM DTT; 100 μg/ml cycloheximide) and passed through a 27 gauge needle four times. Lysates were centrifuged at 14.000 g for 7 min at 4 °C. Supernatants were flash frozen in liquid nitrogen and stored at −80 °C until used. Sucrose solutions were prepared in gradient buffer (50 mM Tris–HCl, pH 7.4; 12 mM MgCl2, 50 mM NH4Cl; 1 mM DTT) at concentrations of 5 %, 15 %, 25 %, 35 %, 45 % for the gradient and at 60 % concentration, which was used to pump the “lighter” gradient through the analyser. 2.2 ml of each solution was pipetted into 13 ml centrifuge tubes (Beckman Coulter, Cat.No. 331372) and frozen at −80 °C before the next layer was added. The full gradient was kept at −80 °C for at least 24 h and placed at 4 °C to homogenize o/n prior to centrifugation. Five OD values of lysate were placed on top of the gradients and subsequently ultracentrifuged in an SW41 rotor (Beckman Coulter) for 2.5 h at 26,000 g and 4 °C. Polysomal profiles were analysed via an ISCO density gradient analyser unit, which analyses and simultaneously blots ribosomal distributions measured by an UA-6 detector with 254 and 280 nm filters (Teledyne ISCO, Lincoln, Nebraska, USA).

Primary patient samples and tissue microarrays

UBC biopsies, as well as UBC tumour excision specimens, were fixed in formalin and processed using routine pathological workup methods. The material used was derived from the archives of the Pathology Laboratory Dr. Obrist & Dr. Brunhuber OG, associated to the General Hospital Zams (Austria). UBC specimens of 107 patients were available from the period 2008–2010 (Table 1), in addition to tissue microarrays (TMA) containing 91 UBC patient samples from the University Hospital Innsbruck (Table 2). Five micrometres thick sections of this TMA were taken and mounted on adhesive-coated glass slides for immunohistochemical staining and analysis.

Table 1.

Patient cohort details of 107 archival FFPE samples of UBC cases (n = 107 patients)

Total number of patients			Average age at diagnosis [years]		eIF3a high [n]		eIF3a low [n]
	107	100 %		71.7	74	69 %	33	31 %
Male			Male
	73	68 %		71.2	46	63 %	27	37 %
Female			Female
	34	32 %		72.7	28	82 %	6	18 %
G1
	32	29 %			22	69 %	10	31 %
G2
	45	42 %			32	71 %	13	29 %
G3
	30	28 %			20	67 %	10	33 %
pTa
	43	40 %			28	65 %	15	35 %
pT1
	36	34 %			31	86 %	5	14 %
pT2
	14	13 %			8	57 %	6	43 %
pT3
	7	7 %			2	29 %	5	71 %
pT4
	7	7 %			5	71 %	2	29 %

Table 2.

Patient cohort details of 91 TMA embedded FFPE samples of UBC cases (n = 91 patients)

Total number of patients			Average age at diagnosis [years]		eIF3a high [n]		eIF3a low [n]
	91	100 %		69.8	64	70 %	27	30 %
Male			Male
	67	74 %		69.1	48	72 %	19	28 %
Female			Female
	24	26 %		71.5	16	67 %	8	33 %
G1
	25	27 %			19	54 %	6	24 %
G2
	30	33 %			22	73 %	8	27 %
G3
	36	40 %			23	64 %	13	36 %
pTa
	26	29 %			21	81 %	5	19 %
pT1
	36	40 %			23	64 %	13	36 %
pT2
	16	18 %			7	44 %	9	56 %
pT3
	9	10 %			9	100 %	0	0 %
pT4
	4	4 %			4	100 %	0	0 %

Histology and immunohistochemistry

Haematoxylin and eosin (H&E) stainings were performed using a DAKO Autostainer [Dako Denmark A/S, Glostrup, Denmark] on 2.5 μm thick tissue sections for light microscopic evaluation and visualization. A monoclonal rabbit antibody (eIF3A (D51F4) XP® Rabbit mAb #3411, Cell Signaling Technology, Danvers, MA, USA) was used to assess eIF3a expression by immunohistochemistry. Staining was performed using a Dako Autostainer Universal Staining System as previously described [22]. Endogenous peroxidase was blocked by incubation of the TMA slides with 3 % hydrogen peroxide for 5 min. The primary anti-eIF3a antibody was applied at a dilution of 1:100 for 60 min, followed by incubation with a peroxidase-labelled secondary antibody for 30 min and substrate-chromogen 3.3′-diaminobenzidine tetrahydrochloride for 8 min. Counterstaining was performed by aqueous haematoxylin for 45 s. The expression of eIF3a was assessed by two independent, experienced and board-certified pathologists (P.O., J.H.). The staining density, indicating the actual percentage of tumour cells with a positive staining signal, was rated from 0 to 100 %, whereas the intensity score (IS) was rated “0”, “1”, “2” or “3”, indicating weak to intense, light to dark brown signals, respectively. In a second approach, density scores were adapted to a numeric scale, where 0 % = “0”; <30 % = “1”; <60 % = “2”; <80 % = “3”; ≤100 % = “4”, defined as staining proportion score (PS). PS and IS were multiplied to obtain the final eIF3a total immunostaining score (TIS), ranging from 0 to 12. A cut-off of TIS = 2.5 was set to discriminate eIF3a over-expressing samples from low eIF3a expressing samples.

Statistical analysis

Basic statistical evaluations, including average calculations, standard deviation calculations, and t-tests were performed using MS Excel. All calculations and the statistical analyses associated with the survival data of UBC patients were performed using the statistical software program SPSS for Windows (SPSS, Inc. Chicago, IL). Differences between groups were tested for statistical significance applying the χ² test. Kaplan-Meyer statistics were used to describe and calculate survival curves. Patients who were lost during follow-up were censored in the follow-up time parameter. For this method, p values were evaluated by the log-rank test for censored survival data. For all analyses a p value <0.05 was defined as statistically significant.

Results

Knockdown of eIF3a impairs proliferation, colony formation and transmigration of UBC cells

In order to investigate the effect of eIF3a expression on UBC cells, four different cell lines (RT112, T24, 5637, HT1197) were selected for performing knockdown experiments using an inducible eIF3a specific shRNA. The different cell lines were derived from low to high grade UBC tumours, as described previously [23–25]. They all expressed similar levels of eIF3a, which did not significantly depend on the proliferation rate as revealed by comparison of logarithmically growing and confluency-arrested cultures (Fig. 1a). The cells were transfected with an inducible shRNA and the subsequent knockdown efficiencies were assessed in stable transfectants three days after induction by addition of doxycycline to the culture medium. An inhibition close to 50 % was achieved in all cell lines, with the exception of HT1197 in which more than 90 % knockdown was observed (Fig. 1b). Doxycycline treatment strongly reduced the proliferation of all UBC cell lines which expressed the eIF3a specific shRNA, but had no effect on non-transfected control cells, indicating that reduced expression of eIF3a impairs cell growth (Fig. 2a). Furthermore, clonogenicity as assessed by crystal violet staining of colonies 9 days after seeding was reduced in all cell lines tested (Fig. 2b). The effect of eIF3a knockdown on UBC cell motility was investigated by determining the efficiency of cells to transmigrate through filters coated with an extracellular matrix. Suitable cell lines used for this latter assay were RT112 and 5637. Both cell lines exhibited a reduced capability to transmigrate upon eIF3a knockdown (Fig. 2c). Thus, knockdown of eIF3a affects the proliferation, colony formation from single cells and migratory abilities of UBC-derived cells in vitro.

Fig. 1.

a RT112, T24, 5637 and HT1197 UBC cell lines express high levels of eIF3a, which is not influenced by contact inhibition or seeding density b Knockdown of eIF3a by doxycycline-inducible shRNA reduces eIF3a expression levels to approximately 50 % in RT112, T24 and 5637 cells. HT1197 shows a high knockdown efficiency of approximately 90 %. Bar Chart depicts average knockdown efficiencies of seven independent Western blots analyzed by ImageJ Software; a representative Western Blot is shown

Fig. 2.

a Growth rates analyzed over 96 h following eIF3a knockdown, starting with 20,000 cells (500 for HT1197) seeded in 6-well plates. “+” next to the cell line description indicates doxycycline treatment, “−” cell lines were left untreated. The upper panel reveals a decrease in proliferation for all cell lines after eIF3a knockdown. The lower panel shows the proliferation curve of control cell lines b Clonogenicity is dramatically reduced after eIF3a knockdown in RT112 and 5637 cells. One hundred and 1000 cells were seeded and treated with (+) or without (−) doxycycline for 9 days. Colony formation is visualized by crystal violet staining c Invasiveness of cells assessed by their capability to transmigrate through a basement membrane layer towards a lower transwell compartment containing medium with FCS 10 h after doxycycline incubation (CellBiolabs CytoSelect™ 96-Well Cell Invasion Assay). Twenty-five thousand cells per well were seeded. The decrease in transmigrating cells after eIF3a knockdown is significant in RT112 (p = 0.03) and 5637 (p = 0.03) cells

Knockdown of eIF3a impairs tumour growth in a xenotransplant mouse model

To test whether reduced eIF3a expression also affects tumour cell growth in vivo, UBC-derived cell lines expressing inducible shRNA were xenotransplanted into nude mice and the effect of adding doxycycline to drinking water on the kinetics of the tumour growth was determined. As expected, doxycycline treatment led to a reduction of eIF3a expression in the tumours. Similar to the in vitro situation, the tumour cell growth rates tended to be reduced after eIF3a knockdown (Fig. 3), although this effect did not reach statistical significance in this experimental set-up.

Fig. 3.

A xenotransplant mouse model showing that all analyzed cells are tumourgenic in nude mice. The tumour burden is reduced by decreasing eIF3a levels after doxycylcine administration. Mouse tumour histology of eIF3a knockdown cell lines with and without induction of knockdown by means of doxycycline administration (6 mg/ml) in drinking water is shown at 20× magnification

Effects of eIF3a knockdown are distinct from translation inhibition by puromycin and rapamycin

To test whether the phenotypic changes observed after eIF3a expression reduction are due to its general function in protein translation, the consequences of the reduction were compared to the effects of the established translation inhibitors puromycin and rapamycin. Puromycin is a structural analog of aminoacyl-tRNA, which stably integrates into growing polypeptide chains, thereby causing premature termination of translation and release of unfinished polypeptidyl-puromycin derivatives [26, 27]. Rapamycin is a specific inhibitor of the mTOR kinase, which triggers the release of eIF4e via phosphorylation of 4eBPs and, thereby, initiates translation. mTOR has been shown to activate eIF4b, a regulatory unit in the initiation of mRNA translation, via S6K [28, 29]. UBC-derived cell lines containing the inducible eIF3a shRNA were treated with either doxycycline, puromycin or rapamycin and tested for its effects on cell viability and concomitant caspase 3 activity. Puromycin and rapamycin were used at two concentrations. Our results show that, in contrast to eIF3a knockdown, treatment with puromycin and rapamycin reduced cell viability. Moreover, only puromycin caused distinct caspase activation, which is implicated in the induction of apoptosis (Fig. 4). Sensitivity to drug treatment was higher in the T24 and 5637 cells as compared to the RT112 cells. The effects of puromycin and rapamycin to inhibit translation, and in case of rapamycin even without activation of caspase to induce apoptosis, did not resemble those of eIF3a knockdown after doxycycline treatment, in which case no cell death was observed.

Fig. 4.

The effect of eIF3a knockdown on cell viability and apoptosis induction measured by caspase 3 activation in UBC cells is compared to the established translational inhibitors puromycin and rapamycin; LC live control, L Dox live doxycycline 2 μg/ml, L Pu1 live puromycin 0.1 μM, L Pu2 live puromycin 1 μM, L Ra1 live rapamycin 0.5 μg/ml, L Ra2 live rapamycin 5 μg/ml. Rapamycin and puromycin treatment lead to reduced live cell counts; only puromycin treatment additionally causes caspase activation

Knockdown of eIF3a does not inhibit the formation of 80S ribosomes

The effects of eIF3a knockdown on translation initiation were investigated using polysomal profiles. Successful translation initiation is characterised by the formation of functional 80S ribosomes and the presence of polyribosomal RNAs which can only form when the ribosome is moving along mRNA, enabling multiple ribosomes to attach to the same mRNA. Without doxycycline treatment, all cell lines exhibited similar levels of 60S and 80S ribosomes and low numbers of polysomes, typical for cells with a relatively low translational activity. After eIF3a knockdown, an increased number of 80S ribosomes and more polyribosomes were observed (Fig. 5), which may be explained by the known role of eIF3a in blocking the interaction of 60S and 40S ribosomal subunits. Hence, eIF3a knockdown may lead to a reduced 60S-40S blockage and an increased subunit assembly, respectively. Another likely explanation for the observed increase in 80S ribosomes could be a decrease in ribosome recycling as a consequence of eIF3a expression reduction [30].

Fig. 5.

Increased formation of functional 80S ribosomes recorded after knockdown of eIF3a, which resulted in detection of established, polysome bound RNAs at higher frequency. The polysomal profile of 5637 tetR eIF3a sh 3.1 GFP cells with induced knockdown is shown in red, and the profile of untreated, eIF3a high 5637 tetR eIF3a sh 3.1 GFP cells is shown in blue

Knockdown of eIF3a does affect the transcriptome

Next, we investigated whether deregulation of translation through eIF3a knockdown could lead to changes in the transcriptome, e.g. through an impact on proteins of the transcriptional machinery. For that purpose, the mRNA profiles of shRNA containing HT1197 cells with and without doxycycline treatment were compared after 72 h hours on microarrays (Affymetrix HGU133 plus 2.0 GeneChips). By using these microarrays, 54,000 different probe sets covering more than 47,000 different transcripts and 38,500 genes were investigated. 6 % of these probe sets highlighted genes with more than two-fold changes in expression. This large number of regulated genes was subsequently subjected to further gene ontology analysis. By doing so, we found that all of the 245 examined ontology terms were significantly altered after eIF3a knockdown. This indicates that the changes in the transcriptome in response to eIF3a knockdown observed are rather unspecific and do not preferentially affect subgroups of genes with a common ontology. Interestingly, the effect of eIF3a knockdown on the transcription of other eIF3 subunits was only marginal, i.e., no other eIF3 subunit was found to be differentially expressed in a highly significant manner (pBH_eIF3a = 0.015; pBH_eIF3b = 0.120; pBH_eIF3c = 0.865; pBH_eIF3d = 0.374; pBH_eIF3e = 0.709; pBH_eIF3f = 0.384; pBH_eIF3g = 0.158; pBH_eIF3h = 0.077; pBH_eIF3i = 0.024; pBH_eIF3j = 0.612; pBH_eIF3k = 0.310; pBH_eIF3l = 0.345; pBH_eIF3m = 0.357; Supplemental Table 1). This independence of eIF3a knockdown from the expression levels of other eIF subunits was also verified at the protein level. Here, the eIF3f levels remained essentially unaltered in both the untreated and the doxycycline or rapamycin treated cells. Of note, the levels of eIF3c, eIF3i and eIF4e did not change significantly after eIF3a knockdown by doxycycline treatment, but tended to decrease upon treatment with rapamycin, indicating their involvement in canonical translational signalling (Fig. 6b).

Fig. 6.

Changes in mRNA and protein levels of selected proliferation markers and eIF subunits

Different cellular responses upon eIF3a knockdown or rapamycin treatment

For a closer examination of the effects of eIF3a knockdown on cellular proliferation, potential marker genes were identified through gene ontology analysis of the microarray data, including all significantly altered genes associated with the gene ontology term ‘cell proliferation’ (Supplemental Figure 1). Based on this analysis, CyclinB1 and MDM2 were selected and their expression was assessed at the transcriptional level in shRNA containing HT1197, T24, RT112 and 5637 cells after 24 and 48 h of treatment with doxycycline or rapamycin. We found that CyclinB1 was down-regulated in all cell lines, with more profound reductions after rapamycin treatment than after eIF3a knockdown. MDM2 was not down-regulated in any of the cell lines tested. In contrast, increased mRNA levels were observed after rapamycin treatment in HT1197, T24 and RT112 cells. Via Western blot analyses, the mRNA results could be reproduced (Fig. 6). Hence, a clear difference in response of UBC-derived cells after eIF3a knockdown or translation inhibition by rapamycin treatment was found.

eIF3a is up-regulated in primary UBC samples

The correlation of eIF3a down-regulation in UBC cells in vitro with reduced cellular growth and motility rates, and its implication as tumour marker in various forms of cancer, suggest a potential similar function in primary UBCs in vivo. However, no expression data from UBC patient samples have been gathered so far. Therefore, we set out to analyse the expression of eIF3a in 107 archival formalin fixed paraffin embedded (FFPE) samples by immunohistochemistry. The intracellular distribution of eIF3a staining was found to be mainly peri-nuclear. An increased eIF3a expression was observed in samples exhibiting a higher tumour grade and dedifferentiation. The staining scores significantly increased from G0 to G1 tumours (p = 0.046), from low grade to high grade tumours (p = 0.001), and from G3 to G4 tumours (p = 0.000). No correlation with tumour stage or invasiveness was detected (Fig. 7a). In addition to the FFPE samples, four TMAs comprising 91 UBC samples were subjected to eIF3a immunohistochemistry and scored alike. The same expression trend as observed in the initial 107 UBC samples was also observed in the independent set of 91 UBC samples. For the TMA samples, also the patient survival data were analysed and correlated to eIF3a expression, as shown in Fig. 7b. Kaplan-Meyer curves were calculated using SPSS, censoring patients that were lost during follow-up. The numbers of patients included were n[G1] = 13, n[G2] = 29 and n[G3] = 48. The overall survival rates ranged from 1 to 142 months, with a median of 60 months. The benefit in overall survival for patients with a high eIF3a expression, visible as a trend in the summarized Kaplan-Meyer curve, was significant when considering the low grade tumours only (p = 0.027). Here, a benefit of 20.2 months in overall survival for eIF3a over-expressing patients was observed. Therefore, especially in low grade tumours, eIF3a could serve as a valuable prognostic biomarker.

Fig. 7.

a Staining pattern of eIF3a in UBC FFPE samples, of G1 (A), G2 (B), G3 (C), G4 (D). eIF3a staining is mainly detected in the peri-nuclear location with a tendency to expand towards the cytoplasm in higher grade tumours. Staining score of eIF3a in UBC patient material from 107 archival FFPE tumour tissue samples. eIF3a expression is increasing significantly with higher tumour grade, but is not significantly changed at different tumour stages. To represent the G0 population, eIF3a staining was evaluated on five tumour-free archival FFPE urinary bladder specimens. Staining scores were significantly increased from G0 to G1 tumours (p = 0.046), from low grade to high grade tumours (p = 0.001) and from G3 to G4 tumours (p = 0.000). Statistics were calculated using Student’s t-test b Kaplan-Meyer curves of 91 UBC patients analyzed on a tissue microarray that was immunohistochemically stained using an anti-eIF3a antibody. Kaplan-Meyer curves are also shown separately for grade 1, grade 2 and grade 3 tumours. Numbers of patients included are n[G1] = 13, n[G2] = 29 and n[G3] = 48. Differences in survival were calculated using the log-rank test for censored data in SPSS. A significant difference (p = 0.027) is seen in grade 1, but not in grade 2 (p = 0.061) and grade 3 (p = 0.618) tumours when comparing high eIF3a expressing to low eIF3a expressing tumour specimen

Discussion

In this study, we aimed to assess the effect of eIF3a expression on the pathology of urinary bladder cancer (UBC). eIF3a is the largest subunit of the eIF3 complex, which is responsible for scaffolding multiple interactions required during translation initiation and the assembly of the 43S pre-initiation complex [31]. eIF3 has previously been reported to be necessary and sufficient for several non-canonical translation initiation processes, and to be essential for translation initiation in tumour cells. Beyond its physiological role in translation initiation, eIF3a has also been suggested as a tumour marker in various cancer entities including those of the cervix, colon, oesophagus, lung, mamma, oral cavity and stomach [16, 32–37]. In all these entities, eIF3a was found to be up-regulated in the tumour tissues and thus, to be associated with malignancy. Grouping the entities in the categories of adenocarcinoma (AC) and squamous cell carcinoma (SCC) revealed that there are more major differences between entities when looking at the tissue of origin, than when looking at the degree of dedifferentiation. The UBC tissues included in this study were derived from transitional cancer cells (TCC). TCC, therefore, represents an as yet un-described variant of tumour in the context of differential eIF3a expression. Besides, TCC of the bladder represents an interesting model system, as UBCs frequently develop in subcategories bearing features of SCC. The diagnostics of UBC is currently based on clinical staging and histopathological analyses of biopsy material, or urine cytology. In addition, FISH analyses have been applied to the identification of genetic alterations. The multi-target, multi-colour UroVysion (UroVysion/Abbott Laboratories) test is a FDA-approved FISH test that allows the determination of genetic alterations in chromosomes 3, 7, 9 and 17 [38]. Established UBC biomarkers are, however, rare although several putative candidates have been described in the literature, including DEK, [7], nuclear matrix protein 22 [39], CIP2A [40], calreticulin [41], UHRF1 [42], KISS1 [8] and survivin [6]. Based on our current results, this list of biomarkers could be extended by eIF3a. We found that eIF3a up-regulation in UBC confers a prognostic benefit in low grade tumours of the urinary bladder, which is also seen in SCCs of the cervix, oesophagus and oral cavity. Previously, we and others found a significant increase in overall patient survival when eIF3a levels were increased in the respective tumour tissues [33, 34, 16]. On the contrary, in ACs of the colon eIF3a over-expression has been reported to correlate with a poor prognosis and a shorter life expectancy [37].

We also found that UBC cell proliferation, invasion and tumourigenicity decreased upon eIF3a knockdown, whereas no concomitant cell death could be detected. In addition, we found that the polysomal profiles changed, suggesting that eIF3a regulates signalling cascades besides translation. If and how eIF3a could achieve this broad regulation via targeted translation initiation of specific genes (such as e.g. p27, RRM2, tyrosinated α-tubulin) remains to be elucidated.

When comparing our results to those from other published SCC or AC studies, differences between AC and TCC become evident. In gastric AC [36], for example, proliferation was not significantly reduced in high-eIF3a tumours, whereas the apoptotic indices were dramatically increased. In SCC of the nasopharynx such susceptibility was not checked, which hampers a comparison [14]. In summary, from a molecular point of view eIF3a over-expression in TCC seems to resemble the situation found in SCC, whereas in AC different molecular mechanisms appear to be at work.

Trying to fit our findings into the dual-track model of the molecular pathogenesis of UBC, and to determine the anti-cancer and oncogenic features of eIF3a expression, a tentative explanation may be provided. The papillary pathway, which is characterised by FGFR3 and HRAS mutations, resembles a SCC-like pattern, based both on its molecular features and the fact that high eIF3a expression is associated with a good prognosis. In line with this notion, the non-papillary pathway follows a rather AC-like pattern of development and progression, being characterised by p53 and RB1 involvement. Further, in these UBC cases eIF3a over-expression does not improve the prognosis of the patient, but as known for other AC scenarios, has only a marginal negative impact on patient survival.

Future work will determine whether eIF subunits, specifically eIF3a, might gain impact in UBC diagnostics or even therapy via rapamycin, rapalogs or other targeted drugs against the translational cascade [43]. eIF3a represents an interesting drug target, not only due to its broad applicability, but also due to the low cytotoxicity observed after its knockdown. This suggests that inhibition or reduction of eIF3a levels in tumours could slow down the pace of tumour progression rather than driving cells into non-responsive cell death or, at worst, necrosis.

Electronic supplementary material

Below is the link to the electronic supplementary material.