Tissue microarrays (TMA) were stained with antibodies against FUS, EWS and TAF15 and the percentage of positively staining cells within 35 organs were estimated (Table 1). The FET proteins showed almost ubiquitous expression with FUS and TAF15 having highly correlated expression patterns (Table 2). However, FET proteins were not detected in melanocytes and cardiac muscle cells and neither FUS nor TAF15 were detected in cardiac endothelium. Most cell types showing FET expression had nuclear localization of the proteins but FUS and TAF15 was absent from the nuclei of hepatocytes. Moreover, FUS and TAF15 often showed cytoplasmic localization while EWS was more rarely found in this compartment (Table 1). Cytoplasmic EWS was mainly detected in secretory cell types. In salivary gland, EWS showed a striking divergence in cytoplasmic expression between different cell types. EWS was restricted to the nuclei of mucous cells while being expressed in both the nucleus and cytoplasm of ducts and serous cells (Figure 1).

In cultured human cells, the FET proteins displayed smooth nuclear localization with infrequent nuclear speckles (Figure 2a and Additional file 1). None of the proteins localized to nucleoli. FUS and TAF15 showed a diffuse distribution in the cytoplasm but were in some cases localized to small cytoplasmic granules. EWS was not detected in the cytoplasm. These patterns were similar for all cell lines tested. Stably expressed GFP-tagged FET proteins demonstrated similar expression patterns as the corresponding endogenous proteins (Figure 2b) while overexpressed FET proteins showed protein aggregation (see below). FET protein expression in stable transfectants was confirmed by western blot analysis (Figure 2c). Cell measurements performed on stable transfectants showed that cells with ectopic FUS or FUSA expression were somewhat larger than other cells (Additional File 2).

Cells transiently transfected with FET-GFP expression vectors showed occasional nuclear and more commonly cytoplasmic FET-GFP protein aggregation (Figure 3a). The cytoplasmic aggregates resembled stress granules (SGs) 11. When such cells were stained with antibodies against the SG marker TIA-1 12, colocalization was seen between this marker and the ectopically expressed GFP-tagged FET proteins (Figure 3a). However, there were clear differences between the individual FET proteins. EWS-GFP rarely localized to SGs (in less than 1% of the cells), while FUS-GFP and TAF15-GFP localized to stress granules in the majority of transiently transfected cells. To further confirm that the FET proteins localize to SGs following stress, stable FET transfectants (not showing stress granules under normal growth conditions (Figure 2b)) and HeLa cells were exposed to oxidative stress by sodium arsenite treatment (Figure 3b and 3c). FET-GFP proteins as well as endogenous FET proteins localized to stress granules following these experiments. EWS-GFP was only detected in a minority of cells and in cases were two smaller cells were located in close proximity to each other. Endogenous EWS on the other hand was found in SGs in all HeLa cells. The FET proteins showed similar SG localization after both heat-shock and arsenite treatment (not shown). A stably expressed FUS mutant lacking the RNA-binding domain showed background signal in SGs similar to the GFP control (Figure 3d).

FUS and TAF15, but not EWS, were detected in large cytoplasmic granules forming close to the plasma membrane in a subset of adhering F470 and HT-1080 cells. These cells were rounded-up and seemingly not completely attached to the underlying surface. When such cells were further co-stained with the focal adhesion/spreading initiation center (SIC) markers Vinculin, FAK and RACK1 13, colocalization was apparent between these markers and both FUS and TAF15 (Figure 4).

When comparing individual cells of the same type within tissues, we found that in addition to differences in localizations there were quantitative differences in the amount of FET protein expression. Epithelial cells of the esophagus displayed the most pronounced variations ranging from strong to complete lack of expression within the same cell type (Figure 5a). Similar observations were made for many other cell types and tissues (not shown). In addition, markedly attenuated FET expression was seen in secretory endometrium compared with proliferative endometrium (Figure 5b). Immunostained cultured cells showed more homogeneous FET expression (Figure 2a).

To investigate a possible connection between FET expression and differentiation, we analyzed FET expression in SH-SY5Y neuroblastoma cells treated with all-trans retinoic acid (RA). Cells were monitored by light microscopy and harvested after 3, 6 and 9 days of treatment. An increased neurite formation was visual in RA treated cells compared to untreated cells. Western blot analysis of protein extracts from these cells demonstrated a marked decrease in FET expression upon prolonged retinoic acid treatment (Figure 6). No change in localization of the FET proteins was detected by immunofluorescence analysis after RA treatment of SH-SY5Y cells (not shown).

We further investigated FET family gene expression during spontaneous differentiation of cultured human embryonic stem cells (hESCs). Samples were collected from morphologically distinct inner and peripheral parts of the hESC colonies (Figure 7a). Quantitative RT-PCR was used to assay FET gene expression as well as markers for differentiation and proliferation (Figure 7b). FET gene expression was over time downregulated in the peripheral parts of the hESC colonies. However, the expression of individual FET genes diverged partially from each other. FUS and TAF15 expression was considerably reduced in the peripheral parts while EWS showed only weak attenuation here. In addition, TAF15 was slightly downregulated in the inner parts, while FUS and EWS expression was maintained in these cells. An overall downregulation of POU5F1 (OCT4) indicated that all hESCs had initiated differentiation (POU5F1 expression is restricted to undifferentiated cells). Over time, peripheral colonies spontaneously differentiated towards a mesodermal cell fate defined by an upregulation of VIM expression. The inner parts showed a very weak upregulation of SOX2, an early ectodermal marker as well as downregulation of the proliferation marker CCNA2. FUS, EWS and TAF15 expression at the protein level was confirmed in hESCs by immunofluorescence (Figure 7c).

As FET gene expression correlated with the proliferation marker CCNA2 in hESCs we investigated whether actively proliferating cells have increased FET expression compared to quiescent cells. Proliferating, semi-confluent F470 primary human fibroblasts were compared with confluent contact-inhibited cells. Also, experiments studying FET expression following serum-stimulation/starvation were performed in HT-1080 cells. These experiments did no reveal any significant differences in FET expression as analyzed by western blot (Additional file 3).

Human normal organ tissue arrays (Super Bio Chips) containing 59 core biopsies per slide were stained with primary antibodies against human FUS 56, EWS and TAF15 (Additional file 4), according to the protocol supplied by the manufacturer. Briefly, slides were deparaffinized in xylene and rehydrated by successive incubations in ethanol followed by immersion in water. Antigens were retrieved by microwave treatment in citrate buffer pH 6.0 and endogenous peroxidase activity was quenched using hydrogen peroxide. Slides were incubated with FET antibodies in Tris-buffered saline pH 7.4 with 2% bovine serum albumin (BSA, Sigma-Aldrich) and 0.05% Tween 20 (Sigma-Aldrich) overnight at 4°C. Primary antibodies were detected using biotin conjugated secondary antibodies (Multi Link, DakoCytomation), which were further incubated with Streptavidin/HRP (P0397, DakoCytomation) and then treated with metal enhanced DAB solution (Pierce). Tissue arrays were counterstained with Mayer's hematoxylin (Histolab), dehydrated in ethanol and xylene, and mounted in Pertex (Histolab). Slides were analyzed with an Olympus BX51 light microscope and selected tissues were photographed on a Nikon Eclipse E1000M light microscope fitted with a ProgRes 3012 digital camera (Kontron Elektronik). FET expression patterns for individual cell types in each tissue were scored based on the amount of positively staining cells (Table 1). Nuclear and cytoplasmic FET expression was estimated by the following criteria: +++ (100-76% positively staining cells), ++ (75-26% positively staining cells), + (25-1% positively staining cells) and - (negative).

HT-1080 fibrosarcoma cells 57 and F470 primary human fibroblasts 58 were maintained in RPMI1640 medium (Sigma-Aldrich). HeLa cells (a kind gift from Dr. Tommy Nilsson, Department of Medical and Clinical Genetics, Goteborg University) were cultured in Dulbecco's Modified Eagles Medium High Glucose (E15-011, PAA) and U1242MG glioblastoma cells 59 were grown in Basal Medium Eagle with Earle's salts (41010, Gibco). The neuroblastoma cell line SH-SY5Y 60 was maintained in Minimal Essential Medium with Earle's salts. Growth medium was further supplied with 2 mM L-Glutamine, 10% fetal bovine serum (FBS), penicillin (50 U/ml) and streptomycin (50 μg/ml) and cells were cultured at 37°C in 5% CO₂. Cells were seeded in Lab-Tek flaskettes (Nalge Nunc International) one day prior to immunofluorescence analysis and transfection (see below). Undifferentiated human embryonic stem cell lines SA 121 (Cellartis AB, Gothenburg, Sweden), HUES1 61 and HUES3 61 were maintained on mitotically inactivated mouse embryonic fibroblasts (in the Semb laboratory). SA121 was mechanically passaged every 4–7 days and half of medium was changed every second day as previously reported 62. HUES1 and HUES3 were enzymatically passaged and cultivated as described 61. In vitro experiments using hESCs were performed according to Swedish ethical guidelines. Stress experiments were performed by treating selected cell types with 0.5 mM sodium arsenite (Sigma-Aldrich) or heat-shock at 44°C for 1 h as described 12.

Cells were fixed in 3.7% formaldehyde in phosphate-buffed saline (PBS) pH 7.2 at 37°C and stained with primary antibodies for FUS, EWS, TAF15 (Additional file 4) in PBS supplied with 2% BSA and 0.2% Triton x-100 (Merck). Spreading initiation centers were assayed in F470 cells three hours after seeding and in HT1080 cells 16 hours after seeding and stained with antisera directed at Vinculin, FAK and RACK1 (Additional file 4). Stress granules were detected with a TIA-1 antibody (Additional file 4). Primary antibodies were detected using goat Cy3 conjugated secondary antibodies (Fluorolink, Amersham Biosciences) or combinations of donkey/goat/rabbit AlexaFluor 488/568 conjugated secondary antibodies (Molecular Probes). Slides were mounted using Prolong Gold antifade with DAPI (Molecular Probes) and allowed to cure overnight. Cellular fluorescence was imaged using a Zeiss LSM510 META confocal microscope system or a Zeiss Axioplan 2.

The full-length coding regions of FUS, EWSR1 and TAF15 were cloned into EGFP-N1 expression vectors (Clontech) as described 58. A FUS mutant (here named FUSA) expressing amino acids 1–175 was generated in a similar manner. Cloning primer sequences are available as supplementary data (Additional file 5). FET-EGFP expression vectors were transiently transfected into cells using the FuGENE 6 transfection reagent (Roche), according to the instructions supplied by the manufacturer. On the following day, transfected cells were fixed, mounted and imaged as before. HT-1080 cells stably expressing FUS, FUSA, EWS and TAF15 tagged with EGFP as well as control cells expressing EGFP were obtained after two weeks of prolonged culture of transfected cells under selection of 800 μg/ml geneticin (G418, Invitrogen). These cells were subsequently subjected to single-cell dilution cloning. Stable transfectants were maintained in RPMI1640 medium supplied with 500 μg/ml geneticin. Stable transfectants were fixed and imaged as before. Total protein was obtained by lysis in radioimmunoprecipitation (RIPA) buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton x-100, 1% Sodium deoxycholate, 0.1% SDS) containing 1× Complete Mini Protease Inhibitor (Roche). For cell measurements, stable transfectants were fixed as earlier and stained with 0.2% Evans Blue (Merck) in PBS (whole-cell staining) before red fluorescence was recorded with an Axio Imager Z1 microscope (Zeiss). Images were thresholded for background and average cell areas were obtained by using the "Analyze Particles" function in the public domain ImageJ software.

Differentiation of neuroblastoma cells was performed as previously described 63. Briefly, SHSY-5Y cells were treated with 1 μM all-trans retinoic acid for 3, 6 or 9 consecutive days to induce neuronal differentiation. Treatment with vehicle (DMSO) was used as control. Protein extracts were obtained by lysis in buffer containing 1% Nonidet-P40, 10% glycerol, 20 mM Tris buffer pH 8, and 197 mM NaCl. In addition, FET protein expression was assayed in actively proliferating and quiescent F470 cells as well as in serum stimulated and starved HT-1080 cells. For F470 cells, 80% confluent, actively proliferating cells were lysed and compared with contact-inhibited cells that had grown to full confluence over seven days. For HT-1080 cells, serum-free medium was added to 50% confluent cells grown in 6-well plates for 24 hours after which half of the wells received full-serum medium (10% FBS) and the rest serum-free medium for 20 hours. Total protein was extracted with RIPA lysis buffer as before.

Protein concentrations in cell extracts were determined using a bicinchoninic acid (BCA) protein assay kit (Pierce) and diluted for equal loading on gels. Samples were mixed with 4× LDS sample buffer (Invitrogen), 10% 0.5 M dithiothreitol and run on NuPage 4–12% Bis-Tris gels (Invitrogen). Proteins were blotted onto PVDF membranes (Immobilon) and probed with FET antibodies (Additional file 4). GFP protein was detected using an anti-GFP antibody and beta actin expression was used as loading control (Additional file 4). Bands were visualized by horseradish peroxidase-conjugated secondary antibodies by chemiluminscent detection (SuperSignal West Dura Extended Duration Substrate, Pierce) or alternatively membranes were developed with BCIP/NBT tablets (Sigma-Aldrich). Chemiluminscent membranes were imaged using a FluorChem imaging system (Alpha Innotech Corporation) and bands were quantified using ImageJ.

Inner and peripheral rings of hESC colonies were separated mechanically and total RNA was extracted with GenElute Mammalian Total RNA kit (Sigma-Aldrich). RNA concentrations were measured with the NanoDrop ND-1000 spectrophotometer. Reverse transcription was performed with SuperScript III (Invitrogen) according to the instructions of the manufacturer using a mixture of 2.5 μM oligo(dT) and 2.5 μM random hexamers (both Invitrogen) as primers. Real-time PCR measurements were performed on an ABI PRISM 7900HT Sequence Detection System (Applied Biosystems). Twenty microliter reactions contained 10 mM Tris (pH 8.3), 50 mM KCl, 3 mM MgCl₂, 0.3 mM dNTP, 1 U JumpStart Taq polymerase (all Sigma-Aldrich), 0.5 × SYBR Green I (Invitrogen) and 400 nM of each primer (MWG-Biotech). Primer sequences are available as supplementary data (Additional file 5). 1× Reference Dye (Sigma-Aldrich) was used as passive reference dye. Formation of correctly sized PCR products was confirmed by agarose gel electrophoresis (2%) for all assays and melting curve analysis for all samples. Gene expression data was normalized against ACTB expression after reference genes evaluation 64.

Spearman's rank correlation coefficients were calculated for nuclear and cytoplasmic expression of FET proteins in tissues using the SPSS 15.0 software. Coefficients were considered significant at the 0.01 level using two-tailed tests.