Altered nuclear retention pathway in human ES cells

What is the fate of IRAlus in human ES cells? In differentiated cells, the fate of promiscuously edited RNA frequently involves nuclear retention in nuclear complexes containing p54^nrb (Zhang and Carmichael, 2001; Prasanth et al., 2005; Chen et al., 2008). We recently published that edited IRAlus in 3′-UTRs of genes can repress gene expression by sequestering mRNAs in the nucleus (Chen et al., 2008). We therefore asked whether the Lin28 transcript is retained in the nucleus in hESCs. Northern blotting was carried out with cytoplasmic and nuclear RNAs isolated from H9 and HeLa cells. Surprisingly, we observed efficient export of Lin28 in H9 cells (Fig. 1B). In contrast, when the Lin28 IRAlus were inserted into a reporter construct, strong retention was seen in HEK293 cells (Chen et al., 2008). This raised the possibility that hESCs export mRNAs with IRAlus in their 3′-UTRs more efficiently than other cells.

However, Lin28 is a stem cell-specific transcript and might have a unique regulation in hESCs. The mRNA for Nicolin 1 contains one pair of inverted Alu repeats in its 3′-UTR and shows strong nuclear retention in both HEK293 cells (Chen et al., 2008) and HeLa cells (data not shown). Since this mRNA is not transcribed in hESCs (Fig. 1E), we sought genes that are transcribed in both hESCs and differentiated cells. From the Gladstone Microarray Data Including Stem Cell Tissue available through the UCSC genome browser we identified several candidate genes containing IRAlus in their 3′-UTRs. Paics (phosphoribosylaminoimidazole carboxylase) is one of them (Fig. 1C, upper panel; Fig. S3) and some available EST sequences show promiscuous editing. Recent biochemical studies showed that PAICS is an important bifunctional enzyme in de novo purine biosynthesis and is especially crucial for rapidly dividing cancer cells which rely heavily on the purine de novo pathway for synthesis of adenine and guanine, while normal cells favor the salvage pathway (Li et al., 2007).

Northern blotting using a probe lying upstream of the inverted Alu repeats in its 3′-UTR showed similar transcription levels of Paics in both H9 cells and HeLa cells (Fig. 1C, lanes 1 and 4). Using equivalent amounts of RNAs from the cytoplasm and the nucleus, we observed distinct nuclear retention of the full-length Paics (3295 nt) in HeLa cells (Fig 1C, lanes 5 and 6, N/C ratio of 3.3), but not in H9 cells (Fig. 1C, lanes 2 and 3). Notably, the Paics probe also detected a shorter isoform of Paics (1562 nt) which corresponds to the polyadenylated cDNA BC019255. This shorter isoform is mostly cytoplasmic in both cell lines, suggesting that nuclear retention in HeLa cells is mediated by the inverted Alu repeats, as we have seen in other cases in differentiated cells (Chen et al., 2008). Consistent with these results, we have observed that multiple mRNAs with IRAlus in their 3′-UTRs (Lin28, Paics and Pccb) are efficiently exported to the cytoplasm in another hESC line, H14 (Fig. S4). Finally, in agreement with what we have generally observed for mRNAs containing IRAlus (Chen et al., 2008), Paics, Nicn1, and Lin28 mRNAs are associated with p54^nrb in vivo (Fig. 1E).

Human ES cells lack paraspeckles

In mammalian cells, p54^nrb and its partners, such as PSF and PSP1α, likely participate in the regulation of the nuclear retention of inosine containing RNAs (Zhang and Carmichael, 2001; Prasanth et al., 2005; Chen et al., 2008). p54^nrb is a multifunctional protein and has been implicated in a variety of nuclear processes (Basu et al., 1997; Emili et al., 2002; Ishitani et al., 2003; Kameoka et al., 2004; Kaneko et al., 2007; Straub et al., 1998; Yang et al., 1993; Yang et al., 1997; Zhang et al., 1993; Zhang and Carmichael, 2001). It was the first RNA-binding protein described which shows high affinity for edited transcripts (Zhang and Carmichael, 2001). p54^nrb contains two tandem RNA recognition motif (RRM)-type RNA binding domains and a putative helix-turn-helix motif followed by a highly charged region with a predicted coiled-coil structure. These four regions together comprise the DBHS (Drosophila behavior and human splicing) domain, which is highly conserved among p54^nrb, PSF, and PSP1α (Fox et al., 2002; Yang et al., 1993). PSF and PSP1α can each form heterodimers with p54^nrb (Akhmedov and Lopez, 2000; Fox et al., 2005; Myojin et al., 2004; Zhang et al., 1993). PSF is also a multifunctional protein. It binds both RNA and DNA (Zhang et al., 1993) and can act in splicing (Gozani et al., 1994; Lindsey et al., 1995; Patton et al., 1993) and transcription (Mathur et al., 2001; Urban et al., 2000). Like p54^nrb, PSF also binds tightly to edited substrates (Zhang and Carmichael, 2001). PSP1α is somewhat less abundant, but shares about 50% sequence similarity with p54^nrb and PSF in the conserved DBHS domain, and has been recently found to be a marker for nuclear paraspeckles (Fox et al., 2002). Paraspeckles are cell cycle-regulated subnuclear domains of currently unknown function that are dependent on RNA for their structural integrity (Fox et al., 2005). The edited mouse CTN-RNA was shown to at least partialy localize to paraspeckles, suggesting paraspeckles could be sites of nuclear retention of at least a subset of edited dsRNAs (Prasanth et al., 2005). Our recent results on the nuclear retention of IRAlus-containing mRNAs are also consistent with paraspeckle association (Chen et al., 2008).

As shown in Fig. 1D, p54^nrb, PSF and PSP1α are expressed at a level in H9 cells similar to that seen in HEK293 cells and HeLa cells. The same result was seen in hESC H1 cells (Fig. S1B) and H14 cells (data not shown). Indeed, in HeLa cells p54^nrb, PSF and PSP1α all colocalize in paraspeckles (Fig. 2, panels a–h), although a fraction of all of them are found outside of these structures. Importantly, even though hESCs abundantly express these proteins, typical paraspeckles are not seen (H9 cells, Fig. 2, panels i–l; H14 cells, Fig. 2, panels q–t and Figs. S5 and S9). PSP1α and p54^nrb show an almost uniform distribution and colocalization throughout the nucleoplasm, but are excluded from nucleoli. In contrast, hESCs show no apparent defect in nuclear SC35 splicing speckles (Fig. 2, panels u–x). Consistent with what has been observed in other cells (Akhmedov and Lopez, 2000; Fox et al., 2005; Myojin et al., 2004; Zhang et al., 1993), p54^nrb, PSF and PSP1α interact with one another in H9 cells, even in the absence of RNA (Fig. 3A and B ).

Open in a separate window

Figure 2

Human ES cells lack paraspeckles. HeLa cells were stained with anti-p54^nrb and anti-PSP1α antibodies (a–d) or anti-PSFand anti-PSP1α antibodies (e–h). DAPI was used to indicate DNA. Note that p54^nrb, PSF, and PSP1α colocalize with each other and indicate paraspeckles. p54^nrb and PSP1α colocalize throughout the nucleoplasm in H9 cells, but show no apparent nuclear paraspeckles (i–l). The same batch of H9 cells was also stained with anti-SSEA4 (stem cell marker) and anti-Troma-1 (trophoblast differentiation marker) (m–p). p54^nrb and PSP1α also colocalize throughout the nucleoplasm in H14 cells (q–t), but show no apparent nuclear paraspeckles. H14 cells were also stained with anti-SC35 and anti-PSP1α (u–x). The white scale bar in all images denotes 10 μm.

Open in a separate window

Figure 3

A. RNA-independent interaction of p54^nrb with PSF and PSP1α in H9 cells. Total extracts of H9 cells with or without RNaseA treatment were immunoprecipitated with p54^nrb antibody or mock antibody, and then immunoblotted with anti-p54^nrb anti-PSF and anti-PSP1α antibodies. B. The same immunoprecipitation experiments were performed with anti-PSF antibody in H9 cells. Anti-actin was used as a negative control. C. RNA in situ hybridization was performed with green-dUTP-labeled antisense hNEAT1 probe in HeLa cells, and representative images are shown. p54^nrband PSP1α are in red. hNEAT1 colocalizes with p54^nrb (upper panel) and PSP1α (lower panel). D. hNEAT1 associates with p54^nrband PSF in HeLa cells. IP with anti-p54^nrb, anti-PSF, or mock antibody was carried out in extracts from H9 cells or HeLa cells. RT-PCR of hNEAT1 from the IP showed amplification in HeLa cells by anti-p54^nrb IP and anti-PSF IP, but not by IP with mock antibody. Oct3/4 is a marker of H9 cells.

Noncoding hNEAT1 RNA localizes to paraspeckles and is not expressed in hESCs

RNA immunoprecipitation experiments revealed that mRNAs for Paics and Lin28 associate with p54^nrb in H9 cells (Fig. 1E, lane 2), as in differentiated cells (lane 5). However, these mRNAs are efficiently exported to the cytoplasm in H9 cells (Fig. 1B and 1C) and H14 cells (Fig. S4), indicating that binding to p54^nrb is not sufficient for nuclear retention. Rather, retention correlates with the presence of paraspeckles, and these are missing in hESCs (Fig. 2 panels i–t, Fig. S5).

What underlies the failure of paraspeckles to assemble in hESCs? While it is known that RNA plays a role in paraspeckle formation (Fox et al., 2005), structured mRNAs such as those with IRAlus can not themselves play a crucial role because they are expressed in hESCs. Hutchinson et al. (2007) identified NEAT1 (nuclear enriched abundant transcript 1) as a large (3.7 kb) polyadenylated RNA transcript displaying striking nuclear enrichment in both human and mouse cells. In rats, NEAT1 expression can be further induced by virus infection (Saha et al., 2006).

Using in situ hybridization we demonstrated that hNEAT1 localizes to paraspeckles in HeLa cells. As shown in Fig. 3C, hNEAT1 RNA is restricted to a small number of large, distinct nuclear speckles in HeLa cell nuclei. Importantly, hNEAT1 accumulation colocalizes very well with both p54^nrb and PSP1α (Fig. 3C). Further confocal z-section studies showed such colocalization occurs throughout the nucleus (Fig. S6 and Fig. S7). These results are in complete agreement with recent work in both mouse (Sasaki et al., 2009; Sunwoo et al., 2009) and human (Clemson et al., 2009) cells showing that NEAT1 RNA (MENepsilon/beta RNA in mouse) plays an essential role in the assembly and architecture of paraspeckles. To further examine the relationship between hNEAT1 RNA and other paraspeckle components, we carried out RNA immunoprecipitation experiments in HeLa cells. In further agreement with Clemson et al. (2009), we observed association of hNEAT1 RNA with p54^nrb and PSF in HeLa cells (Fig. 3D, lanes 5 and 11).

Surprisingly, hNEAT1 RNA was undetectable in all hESCs we examined (H9 cells, Fig. 3D, Fig. 4A, B; H14 cells, Fig. 4C and H1 cells, Fig. S8). This absence of hNEAT1 was further confirmed using different PCR primers across hNEAT1 sequences (Fig. 4C and Fig. S8) and also using H9 cells that were cultured under different experimental conditions (Fig. 4B). Curiously, the noncoding NEAT2 RNA (MALAT-1), which lies immediately downstream of NEAT1 on chromosome 11 (Hutchinson et al., 2007), is expressed at high levels in hESCs (data not shown), suggesting independent regulation of expression of these ncRNAs.

Open in a separate window

Figure 4

hNEAT1 and paraspeckles appear upon hESC differentiation. A. Northern blot showing absence of hNEAT1 RNA in H9 cells. B. RT-PCR showing absence of hNEAT1 RNA in H9 cells (lane 1, H9 cells cultured in mTeSR; lane 4, H9 cells cultured in CM medium), but its presence in differentiated cells (Lane 3). C. RT-PCR showing hNEAT1 absence in undifferentiated H14 cells, but expression in differentiated H14 cells. D. H9 cells cultured in CM medium were induced with BMP4 for 5 days, and then fixed for immunostaining. Day 5 differentiated H9 cells show the trophoblast marker, Troma-1 and the loss of stem cell marker, SSEA4 (lower panel). p54^nrb and PSP1α colocalize with each other and form multiple large accumulations (white arrows), indicating paraspeckles. E. Day 4 differentiated H14 cells show the p54^nrb and PSP1α accumulations (white arrows) in the nuclei, indicating paraspeckles. F. hNEAT1 (in situ hybridization, green) is expressed in trophoblasts derived from H9 cells and colocalizes with p54^nrb.

hNEAT1 is expressed after differentiation

To examine whether hNEAT1 expression correlates with paraspeckle formation during hESC differentiation, we differentiated H9 cells to trophoblasts using BMP4 treatment (Xu et al., 2002; Xu et al., 2005). The majority of H9 cells became trophoblasts after 4–5 days of BMP4 treatment as shown by the presence of the trophoblast marker, Troma-1, and the loss of the stem cell marker, SSEA-4 (Fig. 4D, lower panels). ADAR1 and the paraspeckle-related proteins, p54^nrb, PSF and PSP1α were also detected in trophoblasts (Fig. 1D, lane 4). Importantly, we observed the return of paraspeckles in trophoblasts derived from H9 cells (Figs. 4D) and H14 cells (Fig. 4E and Fig. S9). PSP1α and p54^nrb accumulated in numerous dots per trophoblast nucleus (Figs. 4D, 4E and S9), similar to the pattern observed in HeLa cells (Fig. 2, a–d) and other cell types (data not shown). In addition, PSF and PSP1α also showed colocalization throughout the nucleus (data not shown). Importantly, the expression of hNEAT1 RNA was detected during trophoblast differentiation (Fig. 4B, C and F). Taken together, these observations led us to speculate that the lack of paraspeckles in hESCs might mechanistically underlie the weak nuclear retention for mRNAs containing inverted Alu repeats.

Paraspeckle associated proteins are multifunctional

A role in nuclear retention most likely represents only one of a multitude of functions of the p54^nrb protein. While p54^nrb, PSF and PSP1α accumulate in paraspeckles, these proteins are also found elsewhere throughout the nucleoplasm and thus may have functions apart from retention. These proteins can heterodimerize with each other and, since each contains two tandem RNA recognition motifs (RRMs) of unknown binding specificity, it is quite possible that there are numerous important RNA targets in addition to edited RNAs. In addition, both p54^nrb and PSF are also known to be DNA-binding proteins and there have been reports of important roles for these factors in the regulation of both pre-mRNA splicing and transcription (Basu et al., 1997; Emili et al., 2002; Gozani et al., 1994; Ishitani et al., 2003; Kameoka et al., 2004; Kaneko et al., 2007; Lindsey et al., 1995; Mathur et al., 2001; Patton et al., 1993; Straub et al., 1998; Urban et al., 2000; Yang et al., 1993; Yang et al., 1997; Zhang et al., 1993; Zhang and Carmichael, 2001). Thus, while paraspeckle-associated proteins are multifunctional, NEAT1 may have only one function. Paraspeckles may represent subnuclear compartments that utilize a noncoding RNA to recruit and organize a group of multifunctional nuclear proteins for a specialized function, mRNA nuclear retention. We still do not know how hNEAT1 association with these factors leads to formation of paraspeckles and results in the nuclear retention of a subset of mRNAs and edited transcripts. Since p54^nrb, PSF and PSP1α form heterodimers, it is possible that if one polypeptide were to contact hNEAT1 RNA while the other were to contact an mRNA target, then retention might result from tethering to hNEAT1 RNA, which might itself be anchored in the nucleus via interactions that have not yet been elucidated. In addition, it is possible that paraspeckles may be devoid of export factors. If this were the case, RNAs directed to paraspeckles would be unable to access the export machinery. Cleavage of retained RNAs to remove the anchoring sequence(s), as described in the work of Prasanth et al. (2005), would release the mRNAs from paraspeckles, making them available for export.

RNA nuclear retention analysis

Nuclear and cytoplasmic RNA isolation in HeLa cells was performed as described (Chen et al., 2008), and in hESCs was carried out with modifications detailed in a supplemental file. In all analyses, cell-equivalent amounts of cytoplasmic and RNA samples were used.

For northern blotting, Dig labeled Lin28, Paics, Adar1 and U6 probes were made by the DIG-High Prime DNA Labeling and Detection Starter Kit (Roche); Dig labeled antisense tRNA^lys was made using T7 RNA polymerase with the DIG Northern Starter Kit (Roche), and the actin probe used was provided with the kit. Total, cytoplasmic and nuclear RNA were isolated from 10⁷ cells. Nuclear/cytoplasmic ratios were normalized to actin mRNA.

RNA in situ hybridization and immunofluorescence microscopy

To detect hNEAT1 RNA, cells were rinsed briefly in PBS and then fixed in 3.6% formaldehyde plus 10% acetic acid in PBS (pH 7.4) for 15 min at RT. Cells were permeabilized in PBS containing 0.2–0.5% Triton X-100, and 5 mM VRC (Invitrogen) on ice for 5 min. Cells were washed in PBS 3×10 min and rinsed once in 2×SSC prior to hybridization. Hybridization was carried out using nick-translated cDNA probes (nick-translation kit; VYSIS Inc.) in a moist chamber at 37°C for 12–16 hr as described (Spector et al., 1998). A partial hNEAT1 clone was a gift from Dr. K. Prasanth. For colocalization studies, after RNA-FISH cells were again fixed for 5 min in 2% formaldehyde, and IF and imaging were performed as described (Chen et al, 2008). Antibodies used in IF are listed in a supplemental file. Images were taken with a Zeiss LSM 510 microscope.

Immunoprecipitation (IP) and RNA-protein-complex IP

HeLa cells and H9 cells on 10 cm dishes were rinsed twice with ice-cold PBS before harvesting in 10 mL ice-cold PBS by scraping. Cell pellets were resuspended in 1 mL IP buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 0.05% Igepal, 1 mM PMSF, proteinase inhibitor cocktail), and subjected to 2 rounds of gentle sonication and centrifuged to obtain cell extracts. For RNase treatment, 200 μg/mL RNaseA was added to cell extracts and incubated at 4°C for 2 hr. For IP, the RNaseA treated and non-treated cell extracts were incubated with 40 μL Protein A/G PLUS-Agarose beads (Santa Cruz Biotechnology) and 2.5 μg mouse anti-p54^nrb antibody or anti-PSF antibody at 4°C for 2 hr. The beads were washed five times with IP buffer, resuspended in 2x SDS sample buffer and boiled for 5 min before loading on SDS PAGE. RNA-Protein-Complex IP were carried out in HeLa and H9 cells as previously described (Chen et al, 2008).

Antisense-oligonucleotide and dsRNA treatment

To knockdown the expression of hNEAT1 RNA, phosphorothioate-modified oligodeoxynucleotides (Seq 1: GGTTCTCGGAAAACTGGTGA and Seq 2: GTAACAGAATTAGTTCTTACCA) were synthesized at University Core DNA Services, University of Calgary. In addition, hNEAT1 siRNA sense: 5′-/5Phos/rGrUrGrArGrA rArGrU rUrGrC rUrUrA rGrArA rArCrU rUrUC C-3′, and antisense: 5′-rGrGrA rArArG rUrUrU rCrUrA rArGrC rArArC rUrUrC rUrCrArCrUrU-3′ (Clemson et al., 2009) were synthesized at Integrated DNA Technologies. Pooled DNA oligonucleotides or annealed dsRNAs were introduced to HeLa cells using Lipofectamine RNAiMax (Invitrogen). Optimal Lipofectamine RNAiMax/oligonucleotide ratios were empirically determined.

RT-PCR

After treatment of RNA samples with DNase I (DNA-free^™ kit; Ambion), cDNA was transcribed with SuperScript II (Invitrogen) using oligo (dT) or random hexamers. Primers are listed in supplementary material.

We thank K. Prasanth for a probe for hNEAT1, K. Morris for help throughout the work and Li Yang for useful advice throughout. We also thank S. Garren, A. Günzl, Y. Huang, K. Morris and D. Moschenross for helpful comments on the manuscript. Human embryonic cell lines H1, H9 and H14 were obtained from the WiCell Research Institute and the CT Stem Cell Core Facility. This work was supported by grant CA04382 from the National Cancer Institute and an award from the State of Connecticut Stem Cell Initiative.