ABSTRACT
Using the programmable RNA-sequence binding domain of the Pumilio protein, we FLAG-tagged Xist (inactivated X chromosome specific transcript) in live cells. Affinity pulldown coupled to mass spectrometry was employed to identify a list of 138 candidate Xist-binding proteins, from which, the lupus autoantigen La (encoding gene Ssb) was validated as a protein functionally critical for X chromosome inactivation (XCI). Extensive XCI defects were detected in Ssb knockdown cells, including chromatin compaction, death of female ES cells during in vitro differentiation and chromosome-wide monoallelic gene expression pattern. Live-cell imaging of Xist RNA reveals the defining XCI defect: Xist cloud formation. La is a ubiquitous and versatile RNA-binding protein with RNA chaperone activity. Functional dissection of La shows that the RNA chaperon domain plays critical roles in XCI. In mutant cells, Xist transcripts are misfolded and unstable. These results show that La is involved in XCI as an RNA chaperone for Xist.
SIGNIFICANCE STATEMENT Xist RNA functions as a scaffold to recruit and assemble a protein machinery to epigenetically silence genes along one X chromosome in each female mammalian cell. Here, we devised a FLAG-out system to profile the Xist interactome, and identified the lupus autoantigen La as an Xist-binding protein. The RNA chaperone activity of La is essential for X chromosome inactivation (XCI). When La is downregulated, the folding of Xist RNA is altered, leading to shorter half-life of Xist RNA and compromised Xist cloud formation, which in turn results in XCI defects, including chromatin compaction, chromosome-wide monoallelic gene expression pattern, and death of female ES cells during in vitro differentiation.
INTRODUCTION
Xist RNA is a prototype long non-coding RNA (lncRNA) involved in X chromosome inactivation (XCI), a mammalian dosage compensation mechanism, in which one female X chromosome is transcriptionally silenced to balance the X-linked gene dosage between males and females (1). Upon the onset of XCI during early embryonic development, the coating of Xist RNA transcripts on the chromosome territory of the chosen inactive X chromosome (Xi) recruits epigenetic factors for heterochromatinization and establishes the chromosome-wide gene silencing (1). Intensive efforts have been spent in isolating Xist-binding proteins. Three attempts of comprehensive isolation of Xist-binding proteins were reported in 2015 (2-4). Chu et al. identified 81 candidate Xist-binding proteins (4). Using a more selective approach, McHugh et al identified a list of 10 candidate proteins (3). Meanwhile, through a more sensitive approach, Minajigi et al. generated a list of more than 700 proteins (2). 4 proteins were commonly identified by all three studies. These results generate an initial portrait of a fascinating and yet still poorly understood epigenetic machinery recruited/assembled by the Xist RNA. These results also illustrate the difficulty of balancing the assay’s specificity and sensitivity in profiling a complex lncRNA-protein interactome.
Here, we devised a different system for profiling the Xist interactome. We have previously shown that the Xist RNA can be efficiently tagged in live cells by the programmable RNA-sequence binding domain of the Pumilio protein (PUF: Pumilio-homology domain; PBS: PUF-binding site) (5). A transgenic cell line previously generated in the lab is a male mouse ES cell line carrying an X-linked single-copy inducible Xist transgene with PBSb sites fused to its 5’-ends (Fig. 1a) (5). We further engineered the cell line to have it stably express a PUFb-FLAG fusion protein (i-FLAG-Xist, Fig. 1a). Therefore, in this cell line, the Xist RNA is efficiently FLAG-tagged, which enables the usage of the highly specific and sensitive anti-FLAG antibody in the subsequent protein isolation work (Fig. 1b). Meanwhile, we engineered a negative control cell line, i-Empty, which stably expresses PUFb-FLAG but carries an empty inducible cassette (Fig. 1a). We name this method “FLAG-out”. In this study, using FLAG-out, we identified 138 candidate Xist-binding proteins, and further validated the lupus autoantigen La as an Xist-binding RNA chaperone functionally critical for XCI.
RESULTS
“FLAG-out” the Xist-binding proteins
After Xist induction, we fixed the cells by UV-crosslinking and performed FLAG affinity pull-down (Fig.1b). Mass-spec was used to analyze two protein pull-down samples, a dox-treated i-FLAG-Xist sample and a dox-treated i-Empty sample (Fig.1b). 69 proteins were identified only in i-FLAG-Xist but not in i-Empty. Furthermore, for the proteins identified in both samples, we ranked them according to their protein scores in each sample and calculated the ranking gains. hnRNPM, a known Xist-binding protein, showed a ranking gain of 11 in i-FLAG-Xist. Therefore, additional 69 proteins with ranking gains higher or equal to 11 were also selected as candidate proteins. In total, 138 candidate proteins were selected (Fig. 1c&d and Supplementary Table S1). Among the selected candidates, 9 proteins were also identified as Xist-binding proteins in previous studies (2-4) (Fig. 1c&d and Supplementary Table S1). Interestingly, the 138 candidate proteins can be clearly classified into 6 functional groups: [1] nuclear actin and related proteins (38 proteins); [2] chromatin and related proteins (13 proteins); [3] DNA and RNA binding proteins (20 proteins); [4] nuclear RNP, ribosomal and nucleolar proteins (30 proteins); [5] membrane proteins (8 proteins); [6] other proteins (29 proteins) (Fig. 1c&d and Supplementary information). The complete list of the candidate proteins is shown in Supplementary Table S1. Selected candidate proteins are shown in Fig. 1d.
Validation of the functional significance of the lupus autoantigen La in the induced XCI
From the candidate proteins, we shortlisted three proteins for individual validation. Myb-binding protein 1A (Mybbp1a, Q7TPV4) and TAR DNA-binding protein 43 (Tardbp, Q921F2) were selected because they are known transcription repressors (6, 7). The Lupus autoantigen La (P32067, encoding-gene name: Ssb) was selected because systemic lupus erythematosus (SLE) is an autoimmune disease characterized by a strikingly high female to male ratios of 9:1(8). Moreover, its autoimmune antigen La is a ubiquitous and versatile RNA-binding protein and a known RNA chaperone (9). All the three selected candidates have also been identified as Xist-binding proteins in previous studies (2, 4). Moreover, the knockout of these three genes all lead to early embryonic death. Tardbp knockout causes embryonic lethality at the blastocyst implantation stage (10). Mybbp1a and Ssb knockout affect blastocyst formation (11, 12). Early embryonic lethality is a mutant phenotype consistent with a critical role of the mutated gene in XCI (1).
As the gene knockout of all three candidate genes causes early embryonic lethality, we chose to knock down their expression using shRNAs to validate the candidate genes’ roles in the induced XCI. The selected cell line is a male ES cell line carrying an inducible X-linked Xist transgene. Inducible Xist expression in this cell line causes cell death due to the inactivation of the single X chromosome in male cells. Massive cell death usually occurs after 4-5 days of Doxycycline treatment, which can be used as a convenient assay to assess the functionalities of XCI (Fig. 2a). We established clonal ES cell lines stably expressing the shRNAs and confirmed the shRNA knockdown efficiencies (Fig. 2b). Knocking down Ssb (the gene encoding La) showed a negative effect on the cell growth rate (Fig. 2c&d), consistent with the role of La as a ubiquitous RNA-binding protein involved in house-keeping functions such as tRNA biogenesis. Nonetheless, the cell survival rates show that knocking down Ssb significantly rescued the cells from the toxicity of induced XCI (Fig. 2c-e). These results confirm that La is involved in induced XCI. To rule out the off-target effect of shRNAs, we used a second shRNA construct and obtained similar results (Fig. 2c-e). In the forgoing experiments, cells were cultured as undifferentiated ES cells during the induction. We performed the Doxycycline-induced cell death assay in differentiating ES cells and obtained consistent results (Supplementary Figure S1). We further compared the X-linked genes’ expression levels in control cells and in Ssb knockdown cells during the induced XCI. The expression level of 8 selected X-linked genes were measured by quantitative RT-PCR (Fig. 2f&g). Compared to the empty vector control, higher expression levels of X-linked genes were exclusively detected in Ssb knockdown cells showing the impaired XCI status in Ssb knockdown cells. These results confirm that La is involved in the induced XCI in the male transgenic cell line.
Meanwhile, our results do not confirm the functional roles of Mybbp1a and Tardbp in XCI (Fig. 2c&e). However, it is possible that the knockdown efficiencies of these two genes were not high enough to overcome the thresholds required to disturb XCI. Alternatively, there might be some redundant factors for Mybbp1a and Tardbp that compensate the loss of these two factors.
Validation of the functionality of La in endogenous XCI in female cells
To validate the functionality of La in endogenous XCI, we performed shRNA knockdown of Ssb in female ES cells (Fig. 3a). We confirmed the knockdown efficiency on the selected clonal cell lines stably expressing shRNAs against Ssb (Supplementary Figure S2). XCI can be triggered in female mouse ES cells by in vitro differentiation. Massive cell death usually occurs during this process in female mouse ES cells with XCI defects (1). Consistent with this notion, we observed significant cell death in the Ssb knockdown cells during in vitro differentiation (Fig. 3b&c). Importantly, on the Ssb knockdown background, significantly more cell death was observed in female cells than male cells (Fig. 3b&c). Although Ssb knockdown showed negative effect on the cell growth rate of male ES cells (Fig. 2c&d), the morphology of male embryoid bodies (EBs) on day 4 of in vitro differentiation were comparable to the wild type EBs (Fig. 3b). The EBs could attach to the surface of the tissue culture flask and showed significant amount of expansion two days later (Fig. 3b). This is in clear contrast to the female EBs. With Ssb knockdown, the female EBs were small and unhealthy on day 4 of in vitro differentiation (Fig. 3b). The EBs barely attached to the surface of the tissue culture flask and showed no or very limited amount of expansion two days later (Fig. 3b). These results support a functional role of La in XCI.
As La is an RNA-binding protein involved in various house-keeping functions, the massive cell death, which occurred in Ssb knockdown female ES cells during in vitro differentiation, may not be directly and solely attributed to XCI defects. Therefore, we performed ATAC-seq (Assay for Transposase-Accessible Chromatin with high-throughput sequencing) to examine the chromatin compaction along X chromosome. We also performed padlock SNP capture to directly assess the XCI status by allelotyping the X-linked gene expression chromosome-wide.
ATAC-seq was performed using male ES cell lines carrying an X-linked inducible Xist transgene so that the sequencing reads mapped onto X chromosome only reflect the chromatin status of one X chromosome, the inactive X. In ATAC-seq, the density of transposon insertion reveals the status of chromatin compaction. We analyzed the transposon insertion sites genome-wide and defined “Cut Count” as the number of transposon insertion sites identified within a 1-Mb region (Fig. 3d). The chromatin compaction status during the induced XCI was then quantitatively measured by the ratio of the Cut Counts between the uninduced undifferentiated cells and the differentiated cells treated with dox (Fig. 3d). Induced XCI clearly caused chromatin compaction in wild type cells, meanwhile the induced expression of a mutant Xist transgene with the critical A-repeat region deleted (i-Δ A-Xist) failed to generate chromatin compaction (Fig. 3d, Supplementary Figure S3). These results serve as controls to validate the experimental system. On the Ssb knockdown background, induced XCI also caused chromatin compaction, but to a lesser degree than that in wild type cells (Fig. 3d, Supplementary Figure S3). This observation was clearly made after 14 hours dox treatment. After 48 hours dox treatment, the chromatin compaction status was comparable between the wild type sample and the Ssb knockdown sample, although a slightly lesser degree of chromatin compaction was still visible in the knockdown sample (Fig. 3d, Supplementary Figure S3). These results directly connect La to the heterochromatinization of Xi, supporting a functional role of La in XCI. Meanwhile, the results also show that the Ssb knockdown cells encountered difficulties during the early onset of XCI. We further discuss this issue in the later sections.
To directly assess the XCI status of X-linked genes, we comprehensively allelotyped X-linked genes. The shRNA knockdown was performed on a female ES cell line 3F1 (13), which carries Xs from two genetic backgrounds, the 129 mouse strain (X129) and the Mus musculus castaneus (CAST/Ei) mouse strain (XCast) (Fig. 3a). The “preemptive choice” mutant phenotype of 3F1 cells causes non-random inactivation of the X129 allele (13) (Fig. 3a). Therefore, the XCI status of an X-linked gene can be evaluated by RNA allelotyping of X129 and XCast, which provide ample choice of single nucleotide polymorphisms (SNPs). Padlock SNP capture, a high-throughput and high-resolution RNA allelotyping method (14), was performed to profile the XCI status of X linked genes chromosome-wide. The padlock probe library was designed to target 2,969 SNPs covering 1,110 (∼55%) of the X-linked genes (14). 457 X-linked genes were successfully allelotyped in the experiment. Padlock SNP capture detected bi-allelic expression of genes along the entire X chromosome in Ssb knockdown cells, demonstrating obvious XCI defects (Fig. 3e). This result provides the direct evidence confirming the critical functionality of La in XCI.
Knockdown of Ssb impairs Xist cloud formation
To further study the functional roles of La in the endogenous XCI, we investigated the Xist cloud formation in Ssb knockdown cells. In undifferentiated female ES cells, Xist expression is detected as a pinpoint signal associated with its gene locus in cis. Upon differentiation, Xist expression is up-regulated along the chosen Xi and Xist RNA transcripts spread out to cover the chromosome territory in cis. Thus, in differentiated female cells, Xist expression is detected as a cloud signal (the Xist cloud) enveloping the Xi chromosome territory. We performed Xist RNA FISH in differentiating ESCs (Fig. 4a, Supplementary Figure S4). After 6 days of in vitro differentiation, Xist clouds could be detected in more than 80% of the wild type cells. However, in Ssb knockdown cells, Xist clouds were detected in a significant lower percentage of cells (Fig. 4a, Supplementary Figure S4). Given that significant cell death occurs in Ssb knockdown cells during in vitro differentiation and RNA FISH cannot be applied to dead cells, the percentage of cells showing faulty Xist cloud formation should be even higher in Ssb knockdown cells. Taken together, these results show that La is involved in Xist cloud formation.
To avoid the shortcomings of RNA FISH not applicable to dead cells, and to observe the process of Xist cloud formation during the early onset of XCI, we performed live-cell imaging of Xist RNA. We previously established a live-cell imaging system of the Xist RNA in a transgenic male mouse ES cell line carrying an X-linked single-copy inducible Xist transgene with PBSb sites fused to its 5’-ends (Fig. 1a). The cell line also stably expresses a PUFb-GFP fusion protein. Therefore, Xist RNA can be efficiently GFP-tagged in live cells. We chose an inducible transgene to study the Xist RNA in live cells, because the system provides a synchronized onset of XCI in the cell population. In female cells, during the in vitro differentiation, the onset of the endogenous XCI occurs in an unsynchronized manner. For a given cell, the onset of XCI may occur at any time during a window of a few days. However, high-quality live-cell imaging can only be performed within a 2-hour time window due to technical issues, such as photo-bleaching and phototoxicity. In our experimental system, dox treatment was started after the cells were in vitro differentiated for 16 hours. Live-cell imaging was performed 1 hour after dox treatment and lasted for 2 hours with a 2-minute time interval. Because the selected time window belongs to the early onset stage of XCI, the issue of cell death caused by XCI defects is circumvented. In wild type differentiating cells, the behavior of Xist cloud formation is synchronized and consistent (5). The Xist RNA signals first appear as small pinpoint signals that then gradually grew into ∼2 µm large Xist RNA clouds within 60-90 min (Fig. 4b, Movies S1-3). In Ssb knockdown cells, the Xist cloud formation clearly encountered difficulties (Fig. 4c-e, Movies S4-7). In some cells, an Xist RNA signal first appeared as a small pinpoint signal, but the pinpoint signal failed to stabilize. Instead of growing into a large cloud signal, the pinpoint signal was on and off during the first 60-90 min after the signal first appeared (Fig. 4c, Movies S4). In some other cells, a pinpoint Xist RNA signal grew into a faint cloud signal, and the cloud signal quickly diffused and vanished (Fig. 4d, Movies S5). We analyzed 64 Xist signals in wild type cells and 52 Xist signals in Ssb knockdown cells. The total fluorescent intensity of each Xist signal was measured every 2 minutes during the first 40 minutes after the signal first appeared (Fig. 4e). The results clearly revealed the faults of Xist cloud formation on the Ssb knockdown background. In summary, the observed behaviors of Xist cloud formation in Ssb knockdown cells are heterogenous but consistently show the difficulties encountered in forming the Xist cloud. In mutant cells, the Xist RNA transcripts failed to spread out and coat the X chromosome territory. It is likely that the RNA transcripts are quickly diffused and/or degraded. As Xist cloud formation is the initial step which triggers XCI, the defective Xist cloud formation should be the primary reason behind other XCI defects observed in Ssb knockdown cells. These results confirm that the defining XCI defect in Ssb knockdown cells is faulty Xist cloud formation.
Knockdown of Ssb compromises the enrichment of Polycomb marks on Xi
Even though Xist cloud formation is less efficient in the mutant cells, the knockdown cell population behave heterogeneously and Xist clouds are formed in a substantial fraction of surviving cells. We further investigated whether La regulates other events during the later stages of XCI. It is known that, in order to inactivate one large chromosome, Xist cloud recruits other silencing factors to heterochromatinize Xi with multiple layers of epigenetic modifications (1). Two histone modifications enriched along Xi are the Polycomb marks, H3K27me3 and H2AK119ub (15). We performed immuno-RNA FISH to study the role of La in Polycomb marks enrichment along Xi. After 6 days in vitro differentiation, most of the Xist clouds signals in RNA FISH overlapped with enrichments of Polycomb marks detected by the immunostains (Fig. 5a&b). However, the enrichment of Polycomb marks along Xi is significantly disrupted in Ssb knockdown cells. These results suggest that La is also involved in establishing the enrichment of Polycomb marks along Xi.
The RNA chaperone domain of La is critical in XCI
La is an RNA-binding protein with 5 defined functional domains including 3 RNA-binding domains (La motif, LAM; RNA recognition motif 1, RRM1; RNA recognition motif 2, RRM2), an RNA chaperone domain (RCD) and a nuclear localization signal (NLS) (Fig. 6a). We generated plasmid constructs carrying in-frame deletions of the selected domains of La (Fig. 6a) and stably transfected the Ssb knockdown cells with these “rescue” plasmid constructs (Fig. 6b). GFP fusion was included in the plasmid design for checking transfection efficiencies and the transgene expression levels in the selected clonal cell lines. We first rescued female ES cells with a full-length Ssb construct (Supplementary Figure S5). Indeed, the full-length Ssb construct showed significant rescue effects on cell survival, Xist cloud formation and Polycomb marks enrichment (Supplementary Figure S5). Meanwhile, all Δ constructs showed significant rescue effects of various degrees on the cell survival during in vitro differentiation, except for Δ RCD and Δ NLS, indicating the critical roles of these two functional domains (Fig. 6c). For Xist cloud formation, Δ NLS and Δ RCD showed no rescue effect, while other Δ constructs showed significant rescue effects of various degrees (Fig. 6d). These results show that all three RNA-binding domains are required for the functionality of La in XCI and highlight the critical functions of the RCD and NLS. As XCI is a nuclear event, it is obvious why the NLS of La is critical for its functionality in XCI. On the other hand, the critical role of RCD (RNA chaperone domain) of La in XCI is intriguing.
Xist RNA is misfolded and less stable on Ssb knockdown background
We believe La is an Xist-binding RNA chaperone important for folding of the RNA. To access the folding of Xist RNA, we performed SHAPE-MaP (selective 2’-hydroxyl acylation analyzed by primer extension and mutational profiling) (16). The RNA structure was probed by 1-methyl-7-nitroisatoic (1M7), which reacts with 2’-hydroxyl group and forms adducts along the RNA with a preference of unstructured regions. Chemical adducts along the treated RNA affect the fidelity of reverse transcription reaction and are detected as “mutation rates” by sequencing, which can be used to access the RNA structure. Using male ES cell lines carrying the inducible Xist transgene, we performed in-cell SHAPE experiments in wild type cells (duplicate samples), Ssb knockdown cells (duplicate samples) and DMSO control (duplicate samples) (Fig. 6e). Along the ∼18kb Xist RNA, we only analyzed the 3,800 nucleotides which were covered by more than 10,000 reads in each sample (“profiled nucleotides”, Fig. 6e). We selected 0.1% mutation rate as the threshold. A mutation rate greater than 0.1% is considered “positive”. 0.1% is close to the error rate of the current illumina sequencing technology and occurs to be a natural boundary separating the two control samples and the four 1M7-treated samples in our experiments. Among the profiled nucleotides, all the 291 incidents of positive mutation rate were detected in 1M7-treated samples (Fig. 6f). The distribution patterns of the nucleotides with positive mutation rates are consistent between the two wild type samples. These data validate the SHAPE experimental system. Interestingly, the distribution patterns of the wild type samples are clearly distinguishable from mutation samples; and the two mutant sample also different from each other (Fig. 6g-i). This observation can be best illustrated by the 467 profiled nucleotides along the 5000th-6000th nucleotide region of Xist (Fig. 6g). Principle component analysis of the global SHAPE reactivity profiles and the structure predictions based on SHAPE reactivities are also in consistent with this notion (Fig. 6h&i). These data show that Xist RNA is misfolded in mutant cells. Very likely, the misfolded RNA forms a random pool of transcripts which lacks a consensus structure and shows poor consistency from sample to sample.
We further investigated whether the misfolded Xist transcripts are subjected to degradation. To access the stability of Xist RNA on Ssb knockdown background, we studied the disappearance of the Xist cloud signals after doxycycline removal (the “sunset” process). Cells were cultured in differentiating conditions and treated with doxycycline overnight before doxycycline removal. The rate of the sunset process reflects the stability of the Xist RNA. We observed a significantly faster sunset rate on Ssb knockdown background (Fig. 7a&b), especially within the first two hours after dox removal. To further access the stability of Xist RNA, we performed quantitative RT-PCR to quantify the Xist RNA during sunset. The cell lines used were male ES cell lines carrying the inducible Xist transgene (Fig. 7c&d). Xist transcription was blocked by doxycycline removal and actinomycin D treatment. The data was best fitted using a first-order exponential decay model (Fig. 7c). Half-life of Xist RNA was then calculated (Fig. 7d). Consistent with our live-cell imaging data, quantitative RT-PCR results show that the half-life of Xist is significantly shorter on Ssb knockdown background. Interestingly, it seems the decay reaction kinetics of Xist in Ssb knockdown cells can be separated into two phases (Fig. 7c). The first hour is the first phase of the reaction, which is a short phase and only contains two data points. The rest of the data points form the second phase of the decay reaction and can be better fitted in one reaction curve separately from the 0 time point. The half-life of Xist RNA estimated from the data of the second phase is shorter and with more statistical significance (Fig. 7d). We provide our interpretation on this in the discussion section.
Taken together, SHAPE results show that the Xist RNA transcripts are misfolded on Ssb knockdown background, and the shorter half-life of Xist RNA suggests that the misfolded Xist RNA transcripts are subjected to faster degradation.
DISCUSSION
Here we show that FLAG-out is a useful method for profiling the interactome of lncRNAs. The advantage of FLAG-out is that once the target RNA is FLAG-tagged, the high specificity and high sensitivity of anti-FLAG antibody can be applied for the subsequent protein isolation. It should be noticed that selection of the PBS insertion site along the target RNA affects the outcome of FLAG-out. In this study, the PBSb sites are fused at the 5’ end of Xist, a position close to the most critical functional domain of Xist. In our previous study, we have shown that the 5’ PBSb fusion does not affect the Xist functionality (5).
La is a known RNA chaperone able to utilize ATP to unwind RNA-RNA and RNA-DNA duplex (9). Base on the results of this study, we propose that La is an Xist-binding RNA chaperone involved in Xist RNA folding and spreading (Fig. 7e). In the mutant cells, the misfolded Xist RNA transcripts are unstable and unable to spread and form the Xist clouds. These defects eventually lead to failure of Xist cloud formation and collapse of XCI.
La is an autoimmune antigen found in the serum of patients with autoimmune diseases. More than 80% of the patients with autoimmune disorders are female (8). It is a possibility that XCI is related to the striking sex ratio distortion of autoimmune disorders. Multiple hypotheses have been developed. As numerous X-linked genes are related to immune functions, dysregulated expression of these genes may cause malfunction of the immune system. Therefore, skewed XCI pattern (loss of mosaicism), imbalanced X-linked gene dosage (X chromosome reactivation in early T cell lineage and haploinsufficiency in X chromosome monosomy) have all been hypothesized as possible causes of the sex ratio distortion in autoimmune disorders (8, 17). In Ssb knockdown cells, dysregulated gene expression occurs on many X-linked genes including the genes functionally related to immune functions (Fig. 2f&g). These results suggest that Ssb may be a potential genetic factor of autoimmune diseases. However, identifying the lupus autoantigen La, as a protein involved in XCI does not help to further explain the sex ratio distortion. Rather, we believe it is a consequence of autoimmune disorders. Nuclear RNPs are immunogenic (18). The inactive X chromosome coated with Xist RNA and its associated proteins is a large piece of nuclear RNP in female cells, which can trigger autoimmune responses when exposed to the immune system under abnormal situations.
Understanding the structure of lncRNAs is critical for understanding lncRNAs’ functionality. Initial efforts to elucidate the Xist RNA structure have been reported, but significant inconsistency exists among the results of the initial studies (19-22). Nonetheless, several interesting and important issues are revealed. First, the in-cell structure of the RNA is significantly different from the ex-vivo structure of the RNA purified with mild conditions. Second, subregions of Xist may form dynamic structures in living cells. SHAPE analysis suggests that the dynamic structures function as loading pads for protein recruitment (20). In PARIS, a method in which the RNA structures are fixed in cells and the interacting regions of RNAs (duplex sequences) are directly sequenced, significant amount of conflicting duplex sequences is detected, which shows the Xist structure is highly dynamic in living cells (21). In our study, identifying La as an Xist-binding RNA chaperon provides an important piece to the puzzle. Interestingly, La possesses three RRMs, which may function in a synergistic manner in regulating Xist structure in living cells.
As a protein involved in tRNA maturation, a significant fraction of La is located in the nucleolus (23, 24). The nucleolus may function as a general factory for assembling nuclear RNP particles, not just for rRNAs. Xist and its associated proteins forms a unique RNP complex in female cells. The fact that La is an Xist-binding RNA chaperone may be one reason behind the frequent association of the Xi with the nucleolus (25, 26).
It is an interesting observation that the decay reaction of Xist in Ssb knockdown cells seems to consist of two phases. The first hour of the reaction is a transient phase in which the decay reaction shows a slow kinetics. The rest of the data points form the second phase of the reaction which shows a significantly faster kinetics. To interpret the data, we assume that the decay reaction of Xist is catalyzed by both endonucleases and exonucleases. Endonuclease creates nicks along the RNA, which further serve as the starting points for exonucleases. Without the RNA chaperone activity of La, misfolded Xist RNA may be depleted for protein-binding and the RNA may be randomly folded in order to “protect” the hydrophobic bases from the hydrophilic environment. Therefore, compared to the protein-binding biologically active form, the misfolded structures may be more compact, which prevents the initial endonuclease reaction. Once the endonuclease reaction generates enough nicks to trigger the action of exonucleases, the degradation reaction quickly speeds up. However, we do not have enough data to study the transient first phase in detail, therefore, this interpretation may be speculative at this point. The details of the decay reaction and, more importantly, the structural details of Xist await future studies.
MATERIALS AND METHODS
Cell lines and culture
Mouse ES cells were cultured in medium containing 1000 unit/ml LIF. Feeder cells used were Drug Resistant 4 Mouse Embryonic Fibroblasts (DR4-MEF) (Applied StemCell; ASF-1002). Feeder-free ES cells were cultured on culture dishes pre-treated with 0.2 % gelatin (Sigma-Aldrich; G2500).
For in vitro differentiation, cells were cultured in differentiation medium containing 50 μg/ml L-ascorbic acid (Sigma). For the first 4 days of in vitro differentiation, embryoid bodies (EBs) were cultured in suspension. On day 4, EBs were transferred to a gelatinized T75 tissue culture flask. On day 5, EBs failed to attach to the tissue culture flask were washed away. On day 6, surviving cells were harvested for subsequent experiments.
Detailed information is provided in Supplementary Materials and Methods.
DECLARATION OF INTERESTS
The authors declare no competing interests.
ACKNOWLEDGEMENTS
L-F.Z. was supported by Singapore Ministry of Education Academic Research Fund (MOE2015-T2-1-093) and by the Singapore National Research Foundation under its Cooperative Basic Research Grant administered by the Singapore Ministry of Health’s National Medical Research Council (NMRC/CBRG/0092/2015). L.C. was supported by the National Key R&D Program of China (Grant No. 2018YFC1313003 and 2018YFA0107002), the National Natural Science Foundation of China (Grant No. 31622038, 31671497 and 31871485), the Natural Science Foundation of Tianjin (Grant No. 18JCJQJC48400), the 111 Project Grant (B08011), and the Fundamental Research Funds for the Central Universities.