Unr defines a novel class of nucleoplasmic reticulum involved in mRNA translation

Csde1 (cold shock domain-containing E1, also known as and, hereafter referred to as Unr), was identified as a transcription unit located immediately upstream of N-ras in the genome of several mammalian species (Jeffers et al., 1990; Nicolaiew et al., 1991). Unr is a member of the family of proteins that contain an evolutionarily conserved nucleic-acid-binding domain termed cold-shock domain (CSD). This domain binds single-stranded DNA and RNA (Graumann and Marahiel, 1998), and CSD-containing proteins are involved in transcriptional and/or post-transcriptional control of gene expression (Mihailovich et al., 2010; Wolffe, 1994). The mammalian Unr proteins, composed of five CSDs, are highly similar by sharing >90% amino acid identity. Unr is a cytoplasmic RNA-binding protein that, in vitro, interacts preferentially with purine-rich motifs located in RNA loops (Jacquemin-Sablon et al., 1994; Triqueneaux et al., 1999). Unr has been characterized as a regulator of mRNA turnover (Grosset et al., 2000) and translation. During translation, Unr acts as a positive or negative regulator of specific transcripts; it either stimulates or represses the translation driven by internal ribosome entry sites (IRESs) (Boussadia et al., 2003; Dormoy-Raclet et al., 2005; Hunt et al., 1999; Mitchell et al., 2003) or represses cap-dependent translation (Abaza et al., 2006; Duncan et al., 2006; Patel et al., 2005). Recent studies have identified numerous direct Unr mRNA targets in Drosophila (Mihailovich et al., 2010) and in human melanoma (Wurth et al., 2016). In melanoma Unr regulates its specific target genes mainly at the level of translation elongation or termination (Wurth et al., 2016).

Genetic and biochemical studies have linked Unr to several cellular processes as well as to human diseases. Unr has been implicated in the control of cell death (Dormoy-Raclet et al., 2007), cell differention (Elatmani et al., 2011) and cell migration (Kobayashi et al., 2013). In humans, recent studies identified important roles for Unr in the promotion of melanoma cell invasion and metastasis (Wurth et al., 2016), as well as in disorders, such as autism or Diamond-Blackfan anemia (Sanders et al., 2012; Xia et al., 2014; Horos and von Lindern, 2012). Unr has also been characterized as being a critical regulator of embryonic development. In Drosophila, Unr inhibits dosage compensation in female flies, and Unr overexpression results in predominant male lethality at the larvae stage (Patalano et al., 2009).

Here, we report that Unr is crucial for mouse embryonic and placental development. We also identify an Unr-rich structure, forming a network of cytoplasmic invaginations into the nucleus, reminiscent of the nucleoplasmic reticulum (NR) (reviewed in Malhas et al., 2011). These so-called Unr-NRs, found in polyploid cells, constitute a novel type of NR and are found in polyploid cells, in which active translation takes place.

We have previously generated mice carrying an inactivated Unr allele as a result of the deletion of the Unr promoter and reported that the homozygous mutation of the Unr gene results in a null mutation (Boussadia et al., 2003). The lack of Unr protein led to embryonic lethality at mid-gestation, since the percentage of Unr^−/− mutant embryos was detected close to the expected Mendelian levels at 9.5 days of gestation (24%), but were less by day 10.5 (14%) and absent at 12.5 days of gestation and later (Fig. 1A). Unr^−/− embryos were indistinguishable from normal littermates at E7.5. Between E8.5 and E10.5, mutant embryos could be identified morphologically as they are smaller, present delayed growth and, at E9.5−E10.5, lack neural tube closure (Fig. 1B). Heart maturation was delayed in Unr^−/−embryos, which presented a defect of ventricular trabeculation and smaller than normal atrioventricular cushions at the atrioventricular canal (Fig. S1A). The heart was, however, not critically abnormal.

Malfunctions of extra-embryonic tissues are primary the cause of embryonic lethality at mid-gestation (Copp, 1995; Ihle, 2000; Rossant and Cross, 2001). Histological analyses of the yolk sac from Unr^−/− mice did not reveal obvious structural changes at E8.5 or later (Fig. S1B). In contrast, at E9.5−E11.5 obvious placental defects were detected in placentas of Unr KO mice, when compared to wild-type littermates. The murine placenta consists of maternal and embryonic parts, with the latter being composed of three distinct trophoblastic cell layers; namely, an outermost layer of trophoblast giant cells (TGCs), an intermediate spongiotrophoblast layer and the innermost labyrinthin layer. In placentas of Unr KO mice there was a marked atrophy of the spongiotrophoblast and labyrinthin layers, and an ∼60−75% decrease in the number of TGCs (Fig. 1C). These results show that Unr is essential for placental development between E8.5 and E11.5.

We examined the expression of Unr protein within wild-type placenta at day 10.5. Immunofluorescence analyses of placental sections revealed that Unr is distributed throughout the cytoplasm of the three trophoblastic sub-populations (Fig. 2A, left panel). The specificity of the Unr antibody used in this study was shown by the absence of Unr staining in immunofluorescence analyses of placentas of Unr KO mice (Fig. 2B). In addition, we found that Unr also localized to distinguishable foci within the nuclei of a subset of TGCs, but not of spongiotrophoblasts (Fig. 2A, see magnified images of boxed areas). This finding was unexpected because no nuclear expression of Unr has been reported so far, prompting us to further characterize these Unr foci through physiological, molecular and functional analyses.

First, we performed, by immunohistochemistry, a temporal analysis of the presence of Unr foci in TGCs. The results (Fig. 2C,D) show that Unr foci were undetectable in TGCs at E8.5. The proportion of Unr-foci-containing TGCs increased up to 45% at E10.5, thereafter declining again. This temporal analysis revealed that Unr foci are transient structures that are regulated during placental development.

We next wanted to know whether Unr foci represent a novel type of intra-nuclear structure or whether they correlate with known intra-nuclear domains. In view of the RNA-binding properties of Unr, we focused on nuclear structures known to be involved in RNA metabolism. We first observed, by immunohistochemistry staining, that Unr foci are large (average diameter 4,03 µM±1.07, n=100, Fig. 2A) and clearly distinct from nucleoli (shown in Fig. 2C).

We then thought to determine whether the Unr foci are sub-nuclear bodies without membranes, or whether they are nuclear bodies wrapped by the nuclear envelope (Mao et al., 2011). Co-labeling of TGCs with antibodies against Unr and lamin A/C, and with DAPI (Fig. 3A) revealed that Unr foci are enclosed by the nuclear lamina and devoid of DNA (absence of DAPI staining, Fig. 3A, right panel). Of note, Unr was highly concentrated within these foci as revealed by the comparative quantification of the Unr protein in nucleoplasm, cytoplasm and Unr foci (Fig. S2A). Because Unr foci are surrounded by nuclear lamina and devoid of DNA, they might correspond to either the NR or to the infoldings of the inner nuclear membrane (INM). The NR consists of double nuclear membranes with embedded nuclear pores that are continuous with the endoplasmic reticulum (ER) and enclose a cytoplasmic core (Fricker et al., 1997; Malhas et al., 2011). INM infoldings are invaginations of the INM alone. They lack nuclear pore complex components and do not contain a cytoplasmic core (Jokhi et al., 2013; Speese et al., 2012). By using a combination of confocal immunofluorescence and transmission electronic microscopy (TEM) for the imaging of TGC sections, we determined that Unr foci: (i) are lined by a double wall nuclear membrane containing nuclear pores − marked by antibodies against the nuclear pore complex (NP) and visible in TEM images (Fig. 3B), (ii) contain ER − marked by antibodies against the ER resident protein calnexin and visible in TEM images (Fig. 3C) and, (iii) are continuous with the cytoplasm (Fig. 3D). These results indicate that Unr foci are structurally similar to the NR. In agreement, we found that Unr foci contain cytoskeletal cytoplasmic elements (vimentin and α-tubulin, Fig. S2B) that are frequently found in NRs (Gehrig et al., 2008). We next investigated whether Unr foci, as the NR, form a network of invaginations that is connected to the nuclear envelope. We performed 3D reconstruction by using confocal serial sections through the nucleus of a TGC co-stained with antibodies against Unr and Lamin A/C. Visualizing Unr foci from xy and z vantage points revealed that the observed dots (xy plane) are cross sections of tubular extensions of the nuclear membrane, clearly observed in the xz and z images (Fig. 3D and Movie 1). All Unr foci that we visualized in 3D appeared to form a network of branched tubular and vesicular structures in continuity with the nuclear membrane.

These results show that Unr is not expressed in nuclear bodies, and we identified a pronounced structural similarity between Unr foci and the NR, a network of cytoplasmic invaginations in the nucleus. Accordingly, we named this structure the Unr-rich NR (Unr-NR), i.e. the nuclear foci that contain Unr.

The high concentration of Unr throughout the tubular network suggested a function of Unr-NRs in mRNA metabolism that had never been described for the NR. To determine whether Unr-NRs contained poly(A) RNA, we combined RNA in situ hybridization − by using a fluorescent FITC-labeled oligonucleotide (dT) probe − and immunofluorescence, by using antibodies against lamin A/C to outline the Unr-NR. An intense oligo (dT) signal was detected in 100% of the Unr-NRs examined, and the signal was sensitive to RNase treatment (Fig. 4A). To better characterize the spatial organization of the RNA within Unr foci, we performed triple-staining analyses combining RNA FISH and immunofluorescence, by using antibodies marking the NPC and the ER (Fig. S2C). The fluorescent signals corresponding to the poly(A) RNA (green) and the ER (red) did not colocalize within the Unr-NR but were closely apposed, revealing an intermingle organization of these components within the Unr-NR.

This spatial organization, i.e. poly(A) RNA concentrated in the vicinity of nuclear pores and the ER, suggests that Unr-NRs are sites of active translation of newly exported mRNAs. As mRNAs exit the nucleus, they undergo a pioneer round of translation, initiated by the cap-binding complex (CBC) comprising NCBP1 and NCBP2 (Maquat et al., 2010). The CBC is then replaced by the main translation initiation factor 4E (eIF4E), which directs steady-state rounds of mRNA translation. Poly(A)-binding protein nuclear 1 (PABPN1) is then replaced by poly(A)-binding protein cytoplasmic 1 (PABPC1) at the poly(A) tail (Hosoda et al., 2006; Lemay et al., 2010). In TGCs, we examined the sub-cellular localization of NCBP2, eIF4E, PABPN1, PABPC1 and of the elongation factor eEF2 through coimmunostaining experiments by using Lamin A/C antibodies to delineate Unr-NRs. The specificity of these colocalization experiments was demonstrated in Unr-NRs by the absence of survival of motor neuron 1 and 2 (SMN1 and 2, respectively) RNA-binding proteins, not linked to mRNA translation (Fig. 4G). We found that all these factors, except NCBP2, localized and were concentrated within Unr-NRs (Fig. 4B-F). Because NCBP2 is replaced by eIF4E, the pioneer round of translation was likely to be completed for most of the transcripts. However, the presence of both PABPN1 and PABPC1 is surprising, and might be related to the reported cytosolic function of PABPN1 (Lemay et al., 2010).

Next, we used TEM to determine whether the ER present in Unr-NRs is studded with ribosomes. High-magnification TEM micrographs showed that free ribosomes as well as ER-bound ribosomes are clearly visible within these nuclear bodies (Fig. 4H).

Together, these results, showing that poly(A) RNA, translation factor as well as ribosomes are concentrated within Unr-NRs, support a role of these structures in mRNA translation. Since a role of NRs in mRNA metabolism has never been reported, our next efforts were aimed at determining (i) whether Unr-NRs are present in other tissues and cell lines and (ii) whether Unr-NRs are sites of active mRNA translation.

We next used Unr immunohistochemistry to analyse Unr expression in a murine tissue microarray with 27 cores and in paraffin-embedded embryos (examples are shown in Fig. S1C-F). This screening revealed that none of the examined tissues contained Unr-NRs except liver, which we further studied by using immunofluorescence. The Unr-NRs present in hepatocytes were highly similar to those of TGCs, i.e. consisting of nuclear membrane invaginations in which Unr, poly(A) RNA and translation factors were concentrated (Fig. 5A,B). A tubular structure reaching deep in the nucleus was also visualized, (Fig. 5C and Movie 2); however, with a less-branched organization as that seen in TGCs. As a result, the number of Unr-NRs per nucleus appeared to be lower in hepatocytes than in TGCs. That Unr-NRs are restricted to liver − in addition to the placenta − sustains the idea that Unr-NR formation is linked to polyploidy. In agreement, quantitative analysis of nuclear areas and DAPI intensities (Fig. 5D) showed that hepatocyte nuclei containing Unr-NRs were approximately two-fold larger and had an approximately two-fold higher DNA content than hepatocytes devoid of Unr-NRs.

NRs have been described in vitro in a variety of cultured cells (Malhas et al., 2011) but their Unr content has not been investigated. It is, therefore, conceivable that Unr-NRs and NRs are identical structures, and NRs might have a not-yet-explored role in mRNA metabolism. To address this question, we used confocal microscopy to examine MDA-MB-231 cells, reported to present the highest ratio of NRs per cell (Johnson et al., 2003). Although we detected double-wall nuclear membrane invaginations devoid of DNA (Fig. S3A), and containing microtubules (Fig. S3B), these structures neither concentrate Unr nor eIF4E (Fig. S3A, C). Based on their expression profile, Unr-NRs, therefore, do not correspond to the classic NRs.

Since in vivo the presence of Unr-NRs was restricted to polyploid cell types, we reasoned that, in vitro also, hyperploidy might be essential for Unr-NRs formation. To this end, we selected two human cancer cell lines, BeWo and Hep3B − derived from trophoblasts and hepatocytes, respectively − to test the capacity of cytotoxic treatments known to induce polyploidy to trigger Unr-NRs formation. The drugs we used included etoposide − a DNA-damaging drug, and Taxol − a spindle toxin (Litwiniec et al., 2013; Marth et al., 1995). Whereas Unr-NRs were detectable in untreated cultures, their incidence markedly increased following drug treatment in both cell cultures (Figs 5E and 6A). The proportion of Unr-NR-containing cells reached ∼23% in etoposide-treated Hep3B cells and ∼70% in Taxol-treated BeWo cells (Figs 5F and 6B, left panels). Quantitative analysis of nuclear areas revealed that, simultaneously, etoposide and Taxol induced a ≥two-fold enlargement of nuclei (Figs 5F and 6B, middle panels). Moreover, in this cell population, nuclei of Unr-NR-positive cells were twice as big (Figs 5F and 6B, right panels).

We then extended our investigations to two further cancerous human cells lines of other origin, MDA-MB231 and HCT116, derived from breast and colon cancer, respectively. We found that Unr-NRs, barely detectable in untreated cell cultures, were efficiently induced by etoposide in MDA-MB231 cells (Fig. S3D,E). As in BeWo and Hep3B cells, Unr-NRs appeared in MDA-MB231 cells with enlarged nuclei (Fig. S3E). Fig. S3F presents a clear Unr-rich nuclear invagination, observed in MDA-MB231 cells. We did not detect Unr-NRs in HCT116 cells, whatever the dose of etoposide or Taxol used.

In summary, we infer from these results that, in vitro, Unr-NR formation is related to both stress and polyploidy. Moreover, cancer cells treated with anticancer drugs seem prone to form Unr-NRs (three Unr-NR-positive cell lines of four cell lines tested).

The development of in vitro cell culture models made it possible to decipher the translational status of mRNAs localized to Unr-NRs. Since Taxol treatment of BeWo cells turned out to be the most effective in UNR-NRs induction, we selected this condition for further analyses. First, we confirmed that the in vitro formed Unr-NRs were highly similar to their in vivo counterpart. Indeed, in Taxol-treated BeWo cells, Unr, eIF4E, and poly(A) RNA were concentrated in nuclear membrane invaginations (Fig. 6A,C).

To directly visualize localized translation in live cells, we used the ribopuromycylation method (RPM), which works on the basis that puromycin (PMY) is incorporated into nascent chains, whose association with ribosomes is maintained by the presence of the chain elongation inhibitor emetine (David et al., 2012). In control BeWo cells, RPM coupled with PMY staining produced an intense signal, distributed throughout the cytoplasm, and co-localized with Unr (Fig. 6D,E, left panels). Blocking translation by anisomycin before the puromycin pulse abrogated the PMY signal (Fig. 6D, right panels), demonstrating the specific labeling of ribosome-associated nascent chains by the PMY antibodies. In Taxol-treated BeWo cells, the distribution of the Unr and PMY signals clearly changed, Unr being mostly localized to Unr-NRs, whereas the PMY signal was observed both in the cytoplasm and in Unr-NRs (Fig. 6E; Fig. S4B,C). PMY and the ER marker calnexin showed a similar distribution, present in the cytoplasm of control cells, and in both the cytoplasm and Unr-NRs of Taxol-treated cells (Fig. S5C).

These results demonstrated that an active mRNA translation process is taking place within Unr-NRs (>80% of the Unr-NRs were PMY positive). However, we cannot exclude that Unr-NRs also represent a storehouse for non-translating mRNAs. Indeed, a number of studies has provided evidence that cytoplasmic RNA granules are structures in which mRNAs are ‘masked’, i.e. in a translational repressed state (Anderson and Kedersha, 2009; Eulalio et al., 2007). To test whether Unr-NRs harbor non-translating mRNAs, we used antibodies against TIA-1 and Rck/p54 (also known as DDX6), markers shared by P granules, P-bodies and stress granules (Buchan and Parker, 2009). We observed that, in TGCs, most of the Unr-NRs appeared positive for TIA-1 and Rck/p54 (Fig. S2D), suggesting that Unr-NRs store a pool of non-translating mRNAs. But does active and repressed mRNA translation take place within Unr-NRs? To investigate this hypothesis, we performed RPM in live BeWo cells, coupled with PMY, Unr and TIA-1 triple staining. As expected, TIA-1 was undetectable in unstressed control cells (Fig. S4A). In Taxol-treated cells, TIA-1 was expressed in most Unr-NRs, and the PMY and TIA-1 signals colocalized within Unr-NRs (Fig. S4B).

Altogether, the aforementioned data suggest a dual role of Unr-NRs. On one side, they are sites of active mRNA translation and, on the other side, they are likely to maintain a pool of mRNAs in a translationally repressed state.

To address the role of Unr in the formation of Unr-NRs, we examined whether Unr-NRs are formed in TGCs null for Unr. To be able to visualize a nuclear tubular network in the absence of Unr, we used antibodies against the nuclear pore. Five placentas each of wild-type and Unr KO mice at E10.5 and three each at E11.5 were analyzed. By using either 2D or 3D confocal imaging, we found that the perinuclear membrane was stained normally in Unr KO TGCs, whereas intranuclear invaginations − that are visible as granules in cross sections − were completely absent (Fig. 7B). Moreover, the nuclei of Unr^−/− TGCs neither accumulate ER resident proteins nor translation factors in NR-like structures in (Fig. 7C). Unr^+/− TGCs did not differ from Unr^+/+ TGCs, regarding the proportion of Unr-NR-positive cells or their molecular composition. These results demonstrate that Unr is required for the formation of NR-like structures that concentrate polyadenylated RNA and translation factors in TGCs.

In this study, we used an Unr knockout model to determine the role of this RNA-binding protein during mouse development. The two main findings are that Unr (a) is required for mouse embryonic and placental development, and (b) defines a novel type of NR by having a role in mRNA translation.

Unr KO embryos die between E10.5 and E12.5 but the phenotypic defects they exhibit before death, i.e. absence of neural tube closure, delay in heart maturation and small size, are unlikely to be its cause (Copp, 1995). In contrast, major placental defects were evident. These are presumably the primary cause for growth retardation and death of Unr KO embryos, as a result of insufficient materno-fetal nutrient exchanges. Thus, Unr − as several other RNA-binding proteins (Katsanou et al., 2009; Lu et al., 2005; Shibayama et al., 2009; Stumpo et al., 2004) − has important roles during mouse development.

Unr expression studies revealed that, surprisingly, Unr is localized to bright foci within nuclei of a subset of TGCs. We determined that Unr foci were not isolated in the nucleus but were cross sections of a tubular network of nuclear membrane invaginations enclosing a cytoplasmic core. These structures are related to the NR (Fricker et al., 1997; Malhas et al., 2011) and, accordingly, were named Unr-NRs.

What is the role of Unr-NR? Several studies have provided evidence that NRs are involved in Ca²⁺ signaling within localized sub-nuclear regions (Echevarria et al., 2003; Lui et al., 1998). Because of the Unr RNA-binding properties, we thought that Unr-NRs might have a role in mRNA metabolism that had not yet been explored in NRs (Malhas et al., 2011). And, indeed, we found that Unr-NRs present in TGCs are structures that contain an impressive amount of poly(A) RNA and translation factors. We then considered two possibilities: that (i) either a function in mRNA metabolism had not yet been explored in classic NRs, and Unr-NRs are not functionally different from classic NRs or (ii) they are two types of NR. The first, described in many cell types, does not concentrate the translational machinery and has been shown to be involved in Ca²⁺ signaling; the second, Unr-NR, is restricted to specific cell types and has a function in mRNA metabolism. To answer this question, we searched for Unr-NRs in a variety of cell types and generated in vitro cell culture models to analyse the function(s) of Unr-NR.

In an effort to link Unr-NRs to cell physiology, we searched for these structures in murine tissues. A tissue microarray screen combined with immunohistochemistry analyses of some murine tissues revealed their scarcity since none of the examined tissues contained Unr-NRs, except liver, a tissue rich in polyploid cells (Chen et al., 2012; Pandit et al., 2012). The low incidence (3%) of Unr-NRs suggests that, in liver, the increased ploidy, modest as compared to that reached in TGCs (up to 1024 copies), is necessary but not sufficient to produce Unr-NRs. An adverse context, such as oxidative stress, might also contribute to their formation. It will be important to scrutinize other polyploid cell types not examined here, such as megakaryocytes and cardiomyocytes, to confirm the restriction of Unr-NRs to enlarged polyploid cells.

These findings helped us to generate cell culture models exhibiting Unr-NRs. Our hypothesis, linking Unr-NRs to polyploidy, was correct since Unr-NRs were inducible in the hepatoblastic Hep3B cell line treated with etoposide, and in the trophoblastic BeWo cell line treated with Taxol (Litwiniec et al., 2013; Marth et al., 1995). Unr-NRs were preferentially formed in cells with enlarged nuclei, their frequency in the surviving cell population reaching up to 23% in etoposide-treated Hep3B cells and 70% in Taxol-treated BeWo cells. Unr-NRs are not restricted to cell lines derived from placenta and liver because they are also efficiently induced by etoposide in breast-cancer-derived MDA-MB231 cells. Neither etoposide nor Taxol led to Unr-NRs formation in the colorectal-carcinoma-derived HCT116 cells. An hypothesis is that HCT116 cannot endoreplicate their genome because they express wild-type p53 (Liu and Bodmer, 2006), which is assumed to prevent cells from undergoing endoreduplication and polyploidy (Di Leonardo et al., 1997). BeWo, Hep3B and MDA-MB231 cells either express an inactive p53 protein, carry p53 mutations or lack p53 (Hau et al., 2006; Lin et al., 2000; Negrini et al., 1994).

By using RPM, we demonstrated that Unr-NRs are sites of active mRNA translation. Nevertheless, a number of studies has provided evidence that mRNAs, aggregated as microscopically visible RNA granules, are translationally repressed (Anderson and Kedersha, 2009; Eulalio et al., 2007). Specific examples include germ cells, perinuclear P granules, P-bodies and stress granules that all share protein components including TIA-1 (Kedersha and Anderson, 2007). TIA-1 has been found in most (but not all) Unr-NRs, suggesting that a pool of untranslated mRNAs is also localized herein. As proposed for P granules (Sheth et al., 2010), newly exported mRNAs could be transiently stored in Unr-NRs because their diffusion rate into the cytoplasm is low.

That Unr-NRs are sites of active mRNA translation and storage is a somewhat unusual situation since most of RNA granules are dedicated to the storage of non-translating mRNAs. Two recent reports, however, have documented RNA granules either translationally active or as having a dual translational status, i.e. in which translated and untranslated mRNAs are both present (Buchan, 2014; Weil et al., 2012; Yasuda et al., 2013).

A diagram illustrating the structure and function of Unr-NRs is presented in Fig. 8. We propose that the formation of Unr-NRs is an adaptative response of the cell following metabolic or cytotoxic stress situations. TGCs encounter a metabolic stress condition because of their vastly amplified genome producing an amount of mRNA that exceeds their capacity of translation. We propose that, in order to cope with this stress, polyploid TGCs expand the nucleo-cytoplasmic interface and the ER compartment by forming the Unr-NR network, which then facilitates mRNA export and translation at the rough ER. They also maintain a pool of untranslated mRNAs localized within the core of the Unr-NR network. This model is consistent with recent reports having established that, under stress conditions, mRNA translation on free ribosomes is repressed, whereas mRNA translation on ER-bound ribosomes is sustained (Lerner and Nicchitta, 2006; Unsworth et al., 2010).

Many questions arise from the results reported here. Unr-NRs are regulated structures, appearing at a precise developmental stage of the placental development or in response to cytotoxic stresses. A current theme is that polyploidy confers resistance to environmental stresses not tolerated by diploid cells (review in Schoenfelder and Fox, 2015). It will be very interesting to see whether Unr-NRs facilitate survival of giant cancer cells that are subjected to chemotherapy.

Finally and importantly, we determined that Unr-NRs do not form in Unr-null TGCs in the absence of Unr. What is the role of Unr in Unr-NR formation? Although the role of Unr is a matter of speculation, we propose that Unr (a) promotes the nuclear and membrane expansion required for the acquisition of a NR, by stimulating lipid and protein synthesis (Gehrig et al., 2008); (b) recruits mRNAs fated to be translated on ER-bound ribosomes. This later possibility relies on Unr being located at the ER, a position that is conserved between Drosophila and mammals (Abaza et al., 2006; Jacquemin-Sablon et al., 1994).

In conclusion, the main finding of this study is the identification of an Unr-rich structure that we have named Unr-NR because of its structural similarity with the NR. The novelty of this structure, identified in vitro and in vivo, is that it comprises a the high amount of poly(A) RNA and translation factors and that its function in mRNA translation that has not been reported previously. Unr-NRs − like NRs − are not found in normal cells, but are restricted to polyploid cells (in vivo) and to stressed cells (in vitro). We propose that this novel type of NR provides an extended surface for mRNA translation at the rough ER, helping cells to ensure correct translational control. Our results, linking Unr-NRs to cancerous cells subjected to anticancer drugs, might be of importance for drug resistance.

Unr^+/− mice were obtained as previously described (Boussadia et al., 1997). Unr^+/− mice were maintained on a C57BL/6 genetic background. Mice were routinely genotyped by PCR of tail DNA. Embryos were genotyped by PCR of yolk sac DNA at E9.5, and by PCR at earlier stages. This study was performed in accordance with the European Community Standards on the Care and Use of the Animals and was approved by the Animal Care and Use Committee of the University of Bordeaux. Adult livers were obtained from C57-BL/6, Nog or NMRI mice.

Primers used for genotyping. The targeted allele was detected by using neo primers (neoFo 5′-CGTGTTCCGGCTGTCAGCGCAGG-3′; neoRo 5′-CAACGCTATGTCCTGATAGCGGTCC-3′) that give a product of 565 bp. The normal allele was detected by using Unr promoter primers (unrFw 5′ -AACGCAGAGATTGCTTCTGG-3′; unrBw 5′-CCACTTTAACAGTGAGGTCG-3′) that give a product of 290 bp.

BeWo (human trophoblast-derived choriocarcinoma) cells were maintained in Ham's F-12 K medium (Gibco) supplemented with 2 mM L-glutamine, 20% fetal bovine serum and 100 U/ml penicillin−streptomycin (Invitrogen). Hep3B (human hepatocellular cell carcinoma), MDA-MB231 (human breast cancer) and HCT116 (human colorectal carcinoma) cells were maintained in DMEM containing 4.5 g/l glucose, supplemented with 10% fetal bovine serum and 100 U/ml penicillin−streptomycin. For Unr-NR induction, BeWo, Hep3B and MDA-MB231 cells were plated in 24-well dishes (4×10⁴ cells/well); 24 h later, BeWo cells were treated with 0.5 µM Taxol (paclitaxel; Sigma) or 0.2% DMSO (as vehicle) for 48 h. The medium with Taxol or 0.2% DMSO was changed every 24 h. Hep3B and MDA-MB231 cells were treated with 1.5 µM etoposide for 24 h (Hep3B) or 48 h (MDA-MB231), and cultured for an additional 48 h in drug-free medium (Hep3B).

Embryos, yolk sacs, placentas were fixed at 4°C in 4% paraformaldehyde in PBS for 2 h − adult livers were fixed overnight, rinsed in PBS, dehydrated in graded ethanol, cleared in toluene and embedded in paraffin. Sections (4 µm) were stained with hematoxylin and eosin (H&E) following standard protocols, or used for immunohistochemistry. Antigens were retrieved by 30 min incubation at 98°C in target-retrieving solution (10 mM Tris-HCl pH 9, 1 mM EDTA) followed by 30 min cooling at room temperature. Sections were incubated for 2 h at room temperature with primary antibody (anti-Unr; 1:100, HPA018846, Sigma), and bound antibodies were visualized by using the anti-rabbit or anti-mouse RTU Vectastain Kit (Universal Elite ABC Kit, PK-7200, Vector) and NovaRED Substrate Kit (SK-4800, Vector). Nuclei were detected by light counterstaining with Mayer's hemalun.

After deparaffinization and antigen retrieval, sections were incubated in a blocking solution (1% BSA and 10% SVF in PBS) for 30 min and with primary antibodies for 1 h at room temperature. Sections were then incubated with Alexa Fluor-conjugated secondary antibodies for 1 h before mounting in DAPI Fluoromount-G (SouthernBiotech).

Cells grown on coverslips were fixed with 4% paraformaldehyde in PBS for 10 min, permeabilized with 0.2% Triton X-100 for 10 min before incubation with antibodies as described above.

The following primary antibodies were used in this study: anti-Unr (1:100, HPA018846, Sigma), anti Lamin A/C (1:500, provided by Harald Wodrich, UMR-CNRS5234, Bordeaux, France), anti-nuclear pore complex (1:100, clone Mab414, ab24609, Abcam), anti-Calnexin (1:100, ab93355, Abcam), anti- KDEL (1:100, clone 10C3, Enzo Life Sciences), anti-vimentin (1:40, clone V9, Dako), anti-alpha tubulin (1: 100, clone DM1A, Sigma-Aldrich), anti-eIF4E (1:100, clone 87, BD Transduction Laboratories), anti-NCBP2 (1:100, ab124632, Abcam), anti-PABPN1 (1:100, AJ1580b, Abgent), anti-PABPC1 (1:100, clone 10E10, Sigma-Aldrich), anti-eEF2 (1:100, ab40812, Abcam), anti-SMN (1: 100, 610647, BD Transduction Laboratories), anti-PMY (1:100, Millipore, MabE343), anti-TIA-1 and anti-Rck/p54 were gifts from Hervé Moine and Catherine Tomasetto (Université Louis Pasteur, Illkirch, France) (1:500 each). Secondary antibodies used were: Alexa Fluor 488-, Alexa Fluor 594- and Alexa Fluor 647-coupled anti-mouse and anti-rabbit, anti-goat and anti-guinea-pig (1:400, Life Technology). DNA was visualized with DAPI (0.5 µg/ml).

RNA fluorescence in situ hybridization (FISH) on paraffin-embedded tissue sections was performed according to de Planell-Saguer et al. (2010). After deparaffinization and antigen retrieval, sections were incubated with prehybridization solution (4× SSC, 10% formamide) for 20 min at room temperature. Where prehybridization RNase treatment was applied, sections were treated with 0.1 µg/µl RNase A in PBS for 2 h at room temperature. Hybridization was performed for 1 h at room temperature, in hybridization solution (4×SSC, 10% formamide, 0.5 mg/ml single-stranded DNA, 0.5 mg/ml tRNA, 10% dextran) containing FITC-conjugated oligo (dT)₄₀ probe (Eurogenetec) at a final concentration of 0.5 ng/µl. After washes, sections were either mounted and examined, or subjected to immunofluorescence before mounting.

For RNA FISH on cell cultures, cells were fixed in 4% paraformaldehyde for 10 min at room temperature, washed in PBS and permeabilized for ≥24 h in 70% ethanol at 4°C. Prior to hybridization, coverslips were incubated in 2×SSC, 15% formamide, at 65°C for 10 min. Hybridization and washes were carried out as described above, except that hybridization was performed for 16 h at 37°C.

To visualize newly synthetized proteins within cells, we used the ribopuromycylation (RPM) method as described by David et al. (2012). BeWo cells grown on coverslips were incubated for 15 min at 37°C in complete H12 medium supplemented with 208 µM emetin (EMD, Sigma). In protein synthesis inhibitor control experiments, cells were pretreated with 40 µM anisomycin (Sigma) for 30 min at 37°C before incubation with EMD. Cells were then treated with 355 µM cycloheximide (Sigma) for 2 min on ice in permeabilization buffer [50 mM Tris-HCl pH 7.5, 5 mM MgCl₂, 25 mM KCl, 0.015% digitonin, EDTA-free protease inhibitor, 10 U/ml RNaseOut (Invitrogen)]. Cells were then washed and incubated on ice in polysome buffer (50 mM Tris-HCl pH 7.5, 5 mM MgCl₂, 25 mM KCl, 0.2 M sucrose, EDTA-free protease inhibitor, 10 U/ml RNaseOut) supplemented with 91 µM puromycin (PMY, Sigma) for 10 min. After rapid washing in in polysome buffer, cells were fixed in 4% formaldehyde for 15 min at room temperature. After fixation, cells were washed twice with PBS and immunostained with anti-PMY antibody as described above.

Slides were scanned by using a digital slide scanner (Pannoramic Scan; 3D HISTECH Ltd, Budapest, Hungary) with a Zeiss objective (Plan Apochromat 40×; numerical aperture 0.95; ZEISS, Oberkochen, Germany) and a high-resolution color camera (CIS VCC-FC60FR19CL, 4MP, CIS Corporation, Japan). The images were read by using the Pannoramic Viewer software (3D HISTECH Ltd, Budapest, Hungary).

Cells were imaged as previously described (Juin et al., 2014). Cells were imaged with a SP5 confocal microscope (Leica, Leica microsystems GmbH, Wetzlar, German) using a ×63/numerical aperture (NA) 1.4 Plan Neofluor objective lens. To prevent contamination between fluorochromes, each channel was imaged sequentially by using the multitrack recording module before merging. z-stack pictures were obtained using LAS AF, Leica software. Three-dimensional reconstructions were obtained from z-cut pictures, by using Imaris software (Bitplane, Zurich, Switzerland).

We used a macro with ImageJ software that allowed measurement of Unr-NRs number and diameter. Nuclear DNA content was quantified by using total DAPI fluorescence intensity per nucleus using ImageJ. A minimum of 100 cells were counted for each field.

Data are reported as mean±s.d. Significance was measured using two-tailed Student's test (P<0.05).

For transmission electron microscopy (TEM), placentas were washed with ice-cold PBS, fixed in 2.5% glutaraldehyde in 0.1 M phosphate buffer pH 7.4 for 2 h at 4°C, and post-fixed in 1% OsO₄ in phosphate buffer for 2 h at room temperature. Tissue blocks (1−2 mm³) were then processed following standard procedures. After fixation, tissue blocks corresponding to the embryonic part of the placenta were dehydrated, embedded in Epoxy resin and sectioned through a series of graded ethanol and propylene oxide. Blocks were sectioned at a thickness of 60 nm by using an Leica Ultracut UCT ultramicrotome. Sections were collected onto copper grids, contrasted with 3% uranyl acetate and lead citrate, and examined by using a transmission electronic microscope (HITACHI H7650).

This study was performed in accordance with the European Community Standards on the Care and Use of the Animals and was approved by the Animal Care and Use Committee of the University of Bordeaux, France.

We thank Christophe Grosset, Eric Chevet, Hervé Moine, Juan Iovana, David Bernard, Marc Landry, Michel Moenner, Jean-Philippe Merlio and Jean Rosenbaum for helpful discussions. We thank Hélène Boeuf and Violaine Moreau for critical comments on the manuscript. We thank the experimental histopathology platform, US 005 UMS 3427-TBM CORE, a service unit of the CNRS-INSERM and Bordeaux University. We thank Fabrice Cordelieres from the Bordeaux Imaging Center (BIC) for help in fluorescence quantification, and Lucie Geay and Etienne Gontier from the Pole d'Imagerie Electronique of the BIC for transmission electronic microscopy. We are grateful to Dr. Florence Bernex from the Réseau d'Histologie Expérimentale de Montpellier, for providing us with the murine tissue microarray used in this work.