In the present study, we show that the FET proteins are ubiquitously expressed throughout human tissues and only a few cell types lack FET expression. The proteins display cell type-specific localization with FUS and TAF15 having highly similar expression patterns. FUS and EWS have previously been found in the nucleus as well as in the cytoplasm and shown to shuttle between these locations [14, 15]. We here report that TAF15 is also present in both of these compartments in numerous human cell types. This implies that TAF15 may participate in nuclear-cytoplasmic shuttling in much the same way as the other FET family members do. In general, FUS and TAF15 showed cytoplasmic localization in most cell types while EWS was more rarely seen in this compartment. Expression data obtained for EWS in the present study differed from that recently made available on the internet , where the EWS expression was judged to be restricted to the nucleus throughout human tissues. Differences in staining procedure and analysis could explain discrepancies between our studies. In this study, cytoplasmic EWS was mainly detected in secretory cell types, suggesting that EWS is involved in the expression of secreted proteins. In salivary gland, the EWS protein was expressed in both the cytoplasm and nucleus of ductal and serous cells. In contrast, EWS was restricted to a nuclear localization in mucous cells. This and other examples of differential expression of the FET proteins in closely related cell types indicate specific roles in regulation of specialized functions. In cultured cells of different tissue origins, EWS expression was limited to the nucleus whereas FUS and TAF15 were observed in both compartments. This pattern reflected the most common type of expression seen in tissues and may be explained by homogeneous culture conditions. The strong nuclear preference of EWS in most cell types in contrast to FUS and TAF15 may be connected to its more frequent mutation in human cancers were the oncoprotein is restricted to nuclear functions as an aberrant transcription factor. The localization of normal EWS in various subcellular compartments has previously been shown to be affected by methylation  but it is unknown whether localization of FUS and TAF15 is regulated in a similar manner.
A majority of cells overexpressing GFP-tagged FUS and TAF15 showed cytoplasmic aggregates that colocalized with the stress granule marker TIA-1. These results implied that the FET proteins might target SGs as part of cellular stress response. SGs are phase-dense particles that are composed of stalled translation pre-initiation complexes, mRNAs and RNA-binding proteins and appear in the cytoplasm of cells exposed to environmental stress . To further confirm SG localization of FET proteins, we exposed stable FET transfectants and HeLa cells to oxidative stress and heat shock. Exogenous as well as endogenous FET proteins localized to SGs in these experiments. However, exogenous EWS was detected in SGs only in occasional cells and endogenous EWS showed weak SG staining compared to FUS and TAF15. This could be explained by higher amounts of cytoplasmic EWS during and post mitosis. FUS was recently reported in stress granules in a small subset of thapsigargin treated cells and also when expressed as an RFP-tagged protein . We here further show that the entire FET family is targeted to SGs upon environmental stress. FUSA and GFP expressing cells showed similar degrees of signal in SGs which seemed to correlate with the amount of GFP tagged protein present in the cell. It is therefore possible that a certain amount of ectopically expressed protein associates with stress granules as a consequence of overexpression and non-specific protein aggregation. The SG signals from the full-length FET members were judged to far exceed those of the FUS mutant and the GFP protein alone. It is therefore likely that the RNA-binding domain of FUS is needed for stress granule targeting. Several other RNA-binding proteins have previously been found in stress granules, e.g. TIA-1, HuR, hnRNPA1, YB-1 and FMRP [19–21], and many of these proteins have been shown to regulate translation of specific mRNAs [22–25]. For the FMRP protein, this regulation was proposed to be mediated by miRNA-coupled translational repression . FMRP, hnRNPA1 and YB-1 have previously been found together with FUS and EWS in messenger ribonucleoprotein complexes [26, 27], suggesting related functions of these proteins. In addition, the FET proteins are part of nuclear miRNA processing complexes [8, 9] and in our work detected in cytoplasmic SGs known to contain miRNAs . Hence, it is possible that the FET proteins interact with miRNAs and shuttle in protein-RNA complexes containing these non-coding RNAs. FUS has previously been associated with polysomes  and implicated in regulation of localized protein synthesis in dendritic spines . The protein is also proposed to be a component of processing bodies, cellular structures with a direct role in mRNA degradation and with implications in RNAi-mediated post-transcriptional gene silencing . However, we could not detect any structures resembling p-bodies and the FET family was restricted to stress granules in our hands. Based on these data we speculate that the FET proteins have functions in regulation of post-transcriptional gene expression during both normal and stress-induced situations.
The FUS protein has earlier been detected in spreading initiation centers, focal adhesion-like complexes that assembles upon early cell spreading . We here show that TAF15 is also present in these structures. However, EWS was undetectable in SICs, possibly is due to a low abundance of cytoplasmic EWS protein under normal conditions. FUS has previously been reported in NMDA receptor-adhesion protein signaling complexes  and all three FET proteins have been found to interact with v-Src , a protein known to indirectly induce adhesion turnover and actin remodeling . de Hoog et al. have also found that perturbation of RNA-binding proteins (in particular FUS) affects cell spreading . These data suggest that at least two of the FET proteins are involved in focal adhesion-related processes. We noted that stable transfectants expressing FUS variants were somewhat larger than other stably transfected cells but seemingly not flatter when visually inspected in the z-axis by confocal microscopy. The reason for this larger phenotype of FUS-expressing cells and a putative connection with cell spreading is currently not understood.
In addition to in vivo cell type-specific localization, the FET proteins showed heterogeneous expression levels within the same cell type in multiple organs. These results suggested that the expression of the FET proteins may be regulated in individual cells by external factors provided by neighboring cells and the microenvironment. Observations from cell cultures with uniform growth conditions and cell populations, showing only small variances in FET expression between individual cells, supported this assumption. However, we saw no change in FET expression following experimental serum starvation or stimulation of fibrosarcoma tumor cells, implying that externally provided serum factors are not directly affecting FET expression in these cells. An alternative hypothesis is that the observed heterogeneity depends on differentiation, attributing a developmental role for the proteins. We investigated this hypothesis experimentally by measuring FET expression in SH-SY5Y neuroblastoma cells induced to differentiate by retinoic acid. In these experiments, FET expression was markedly decreased in cells receiving RA treatment compared to untreated cells. To further investigate expression of the FET family during differentiation, we measured FET gene expression in spontaneously differentiating human embryonic stem cells. We conclude from these data that the three FET genes are regulated upon early differentiation in a lineage-specific manner.
Altered FUS and EWS expression during differentiation has been reported by several studies [34–37]. Furthermore, FUS and EWS have been shown to be required for B lymphocyte development and for spermatogenesis in mice [38–40]. One study showed that an alternative EWS isoform is augmented during neuronal development . In the present study, we could however not distinguish between EWS isoforms. Nevertheless, our data together with previous reports point to functions for the FET family in specialized cells, rather than the housekeeping functions inferred from earlier promoter studies [42–44]. This conclusion is further supported by the lack of FET expression in terminally differentiated melanocytes and cardiac muscle cells. Bertolotti et al. previously showed that the FET proteins associate with both common and distinct TFIID complexes and RNA polymerase II subunits [3, 45]. These results implied overlapping as well as unique functions within the FET family. Our data showing differences in regulation and expression patterns of the individual FET proteins in specific cell types supports this interpretation and suggests that these unique functions are manifested at a cell type-specific level.
A previous study showed that the FUS homolog pigpen is regulated during the transition between proliferating and quiescent endothelial cells , providing an alternative explanation for the heterogenic FET expression seen in tissues and under experimental conditions. Therefore, we investigated a possible correlation between proliferation and FET expression by comparing FET expression in proliferating and growth arrested cells. However, no relationship between active proliferation and FET protein expression was seen in these experiments. In support of our data is an earlier study showing that FUS expression is uncorrelated to proliferative status . Thus, we conclude differentiation rather than proliferation as an expression determinant for the FET family proteins.
FET proteins have been shown to be part of the splicing machinery  and oncogenic variants of the FET proteins are reported to promote aberrant splicing [48, 49]. Interestingly, altered subcellular localization of the FUS and EWS associated hnRNPA1 protein results in alternative splicing. In addition, an overall change in subcellular distribution of splicing factors has been proposed to influence pre-mRNA processing . We thus speculate that the heterogeneous tissue and cell type-specific expression patterns shown by the FET proteins and their involvement in RNA processing link these proteins to cell type-specific splicing. The miRNA profile of a given cell could in a similar manner be affected by FET protein expression as many miRNAs are spliced out of introns of protein-coding genes . The tumor type-specific FET fusion oncogenes present in multiple human cancers have documented strong transforming properties and the tumors display few other cytogenetic aberrations [52, 53]. As alternative splicing and miRNA maturation are emerging as central for both development and disease [54, 55], abnormal FET oncoproteins disturbing these vital processes could thereby instigate significant biological changes resulting in cancer. Analogously, altered adhesion and stress response are common traits of many human cancers and these properties could be targeted by oncogenic variants of FET proteins.