Roles of long non coding RNAs in human embryonic stem cell pluripotency and neural differentiation 5

Since RMST interacts with hnRNPA2B1 and SOX2, they can be envisaged to engage in a RMST complex, in which these three components together would regulate neurogenesis.. To establish that

Figure 8.13: Western blot confirms that hnRNPA2B1 and SOX2 specifically

interact with RMST To validate findings from mass spectrometry, a Western blot

using the respective antibodies was performed following biotinylated RNA pulldown

Figure 8.14: RNA immunoprecipitation (RIP) established in vivo binding of RMST with hnRNPA2 and SOX2 Fold enrichment relative to the isotype IgG RIP

is presented, and the respective p-values are indicated (A) hnRNPA2-FLAG was

ectopically expressed in ReN-VM cells, and enrichment of RMST was observed in FLAG RIP, indicating physical association (B) Enrichment of RMST in SOX2 RIP also confirmed in vivo association of SOX2 and the lncRNA

Since RMST physically interacts with both hnRNPA2B1 and SOX2, the question of the assembly of the “RMST complex” arose It was found that hnRNPA2

and SOX2 assembled in an RNA-independent fashion In three independent immunoprecipitation (co-IP) experiments, SOX2 is co-immunoprecipitated with hnRNPA2-FLAG in RNase-treated cell lysate (Figure 8.15) In accordance with this observation, Fang et al (2011) found that SOX2 forms protein-protein interactions with several heterogenous nuclear ribonucleoproteins, including hnRNPA2B1 These

co-evidences suggest that the RMST complex is composed of the RNA-binding protein, hnRNPA2B1 binding to RMST, and SOX2 in turn associates with hnRNPA2B1

through its protein interaction domain (Figure 8.16)

Figure 8.15: Co-immunoprecipitation (Co-IP) of hnRNPA2-FLAG and SOX2 in the absence of RNA Cell lysate was treated with RNase, and the co-IP of hnRNPA2

and SOX2 indicated that the two proteins could associate by protein-protein interactions, rather than through a lncRNA scaffold

Figure 8.16: Proposed model of the RMST complex hnRNPA2B1 is a

RNA-binding protein and probably binds directly to RMST SOX2 and hnRNPA2B1 may

interact via protein-protein interactions The transcription factor SOX2 may then bind

to target chromatin regions to affect gene expression

8.2.5 RMST and SOX2 co-regulate a common pool of genes

During neurogenesis, RMST expression was upregulated Since RMST interacts with hnRNPA2B1 and SOX2, they can be envisaged to engage in a RMST complex, in

which these three components together would regulate neurogenesis Previous reports indicate a crucial role for SOX2 in neurogenesis Depletion of SOX2 in neural progenitors resulted in impaired neurogenesis in the central as well as peripheral nervous system (Cavallaro et al., 2008; Cimadamore et al., 2011; Ferri et al., 2004; Puligilla et al., 2010) Therefore, it was expected that the knockdown of SOX2 in the ReN-VM neural stem cells would result in decreased neurogenesis To establish that

the protein components of the RMST complex also played a role in neuronal

differentiation, siRNAs targeting hnRNPA2B1 and SOX2 were introduced into

ReN-VM cells, and the effect on neuronal differentiation was assayed 7 days post transfection by immunofluorescence (Figure 8.17)

Figure 8.17: Knockdown of components of the RMST complex prevents

neurogenesis ReN-VM neural stem cells were transfected with the indicated

siRNAs TUJ1 and MAP2, both markers for neurons, were assayed by immunofluorescence 7 days later In the non-target siRNA (si-NT) control, TUJ1+ and MAP2+ neurons were observed When expressions of RMST, hnRNPA2B1 and SOX2

were depleted, there were significantly fewer TUJ1+ cells, indicating a loss of neurogenesis The scale bar indicates 100 µm In the bottom panel, the efficiencies of knockdown are indicated

As expected, knockdown of any of the components of the RMST complex

resulted in the loss of neurogenesis, indicated by decrease in the number of TUJ1+ and MAP2+ cells The similarity in loss-of-function phenotype, together with RNA-

protein interactions, suggested the possibility that RMST and SOX2 could be

regulating a common pool of genes To this end, a microarray experiment was carried

out ReN-VM cells were transfected with RMST siRNA RMST), SOX2 siRNA

(si-SOX2) or non-target siRNA (si-NT) RNA was extracted 48 hours after transfection

of siRNAs Microarray expression data was normalized to the si-NT dataset, and the extent of overlap between differentially expressed genes in the si-RMST and si-SOX2 datasets was examined

To summarize the microarray findings, 1171 genes were differentially detected (FDR < 0.05; absolute fold change > 1.5) in the si-RMST cells, while there were 626 differentially expressed genes in the si-SOX2 cells There was a 100% overlap between the 2 datasets in that the 626 genes differentially expressed in si-SOX2 were also differentially expressed in si-RMST Of the 1170 differentially expressed genes in si-RMST cells, 632 were upregulated while 539 were downregulated A similar trend was observed in the si-SOX2 cells, where 372 genes were upregulated, while 254 were downregulated (Table 8.1) The extent of overlap

was also large (more than 60%), indicating that RMST and SOX2 possibly regulate a

common pool of genes (Figure 8.18) In addition, a gene ontology (GO) analysis of

the 152 genes downregulated upon both RMST and SOX2 knockdown indicated an

enrichment of GO terms related to neurogenesis and neuronal function (Table 8.2) Taken together, the loss-of-function studies, coupled with whole genome gene

expression analyses, indicate that RMST and SOX2 regulate a common set of target

genes that are essential for neurogenesis

Table 8.1: Table summarizing microarray findings upon knockdown of RMST or

SOX2 in neural stem cells

Figure 8.18: RMST and SOX2 regulate a common pool of targets The extent of

overlap between differentially expressed genes in si-SOX2 and si-RMST datasets was examined by means of a microarray experiment Red circles indicate si-RMST, while blue circles indicate si-SOX2 datasets (A) Regardless of direction of fold change, the overlap between si-SOX2 and si-RMST was 100%, with 626 differentially expressed genes in si-SOX2 also misexpressed in si-RMST (B) Of these, 331 genes were upregulated in both si-RMST and si-SOX2 (C) 152 genes were downregulated in both si-RMST and si-SOX2

Table 8.2: Gene Ontology (GO) analysis of the 152 genes in the RMST and SOX2 overlap

Gene Ontology Biological Process

Gene Ontology Term p-value

2 Regulation of Notch signaling pathway GO:0008593 6.55E-06

3 Cellular response to stress GO:0033554 2.97E-05

4 Forebrain development GO:0030900 2.97E-05

6 Positive regulation of Notch signaling pathway GO:0045747 1.46E-04

7 Telencephalon development GO:0021537 1.48E-04

8 Rho protein signal transduction GO:0007266 1.67E-04

10 Oligodendrocyte differentiation GO:0048709 1.89E-04

The top 10 terms, and their respective p-values are shown

Gene clusters categorized into biological processes at levels 3-9 when analyzed with FatiGO

8.2.6 RMST does not regulate SOX2 expression

Since RMST and SOX2 share a large set of target genes, one question that arose was whether RMST regulates SOX2 expression To this end, SOX2 expression following RMST knockdown in ReN-VM cells was measured RNA was isolated 48 hours after transfection of siRNAs Quantification by qPCR showed that upon RMST knockdown, transcript levels of SOX2 and hnRNPA2B1 remained unchanged Similarly, knockdown of SOX2 had no effect on RMST or hnRNPA2B1 expression (Figure 8.19) This demonstrates that RMST did not regulate downstream target

expression by altering cellular levels of the transcription factor SOX2

Figure 8.19: Transcript levels of SOX2 remained unchanged following RMST

knockdown Knockdown of hnRNPA2B1 (bars shown in light blue) did not alter

expression levels of RMST and SOX2 significantly Similarly, knockdown of RMST

(bars shown in dark blue) did not result in a change in SOX2 and hnRNPA2B1 levels

in the cells

8.3 Discussion

8.3.1 RMST forms part of a complex that is required for neurogenesis

In this chapter, I established that RMST interacts with hnRNPA2B1, as well as SOX2,

and loss of any of these three components of the RMST complex resulted in loss of neurogenesis In accordance with literature, Sox2 induces neuronal formation, and the loss of Sox2 had been reported to cause neurodegeneration and impaired neurogenesis

in the mouse brain (Ferri et al., 2004; Puligilla et al., 2010) SOX2 has also been reported to associate with hnRNPA2B1 via protein-protein interactions (Fang et al., 2011), and this interaction was also observed in this study via the co-IP experiment hnRNPA2 and hnRNPB1 (referred to as hnRNPA2B1) are spliced isoforms, and are RNA-binding proteins that regulate alternative splicing (Chen et al., 2010; Tauler et

al., 2010) Therefore, apart from serving as the adaptor molecule linking RMST to

SOX2, it is also probable that hnRNPA2B1 may regulate the splicing of RMST into its

three isoforms However, in this current study, the functions of individual isoforms are not established

8.3.2 RMST may change the binding patterns of SOX2 to chromatin

From the microarray experiment described in Figure 8.17, it was evident that RMST

and SOX2 share many common targets, supporting the notion that the two are part of

the RMST complex that regulates genes crucial for neurogenesis There are at least two possibilities in which RMST could function in the complex First, association of RMST with SOX2 could alter binding of the transcription factor to chromatin Preliminary evidence suggests that depletion of RMST in ReN-VM cells reduced

SOX2 occupancy at some target genes (Figure 8.20), indicating that this is a plausible

mechanism in which RMST may modulate neurogenesis

Figure 8.20: Depletion of RMST resulted in decreased SOX2 occupancy at target genes Cells were transfected with either a non-target siRNA or RMST siRNA, and

SOX2 ChIP was performed 48 hours later Fold enrichment at SOX2 targets (pDLX1, pJAG1 and pGLI2) were decreased upon RMST depletion pNECDIN was a negative

control (Engelen et al., 2011)

The second possibility is that RMST may recruit SOX2 to specific gene loci to regulate transcription Several lncRNAs, including HOTAIR, KCNQ1OT1 and Evf2

(Bond et al., 2009; Chu et al., 2011; Mohammad et al., 2010), have been reported to recruit transcription factors or chromatin modifiers to specific regions in the genome

to regulate gene expression

8.3.3 RMST may bind to proteins other than hnRNPA2B1 and SOX2

Depletion of the lncRNA RMST resulted in differential expression of more than 1000

putative target genes In contrast, depletion of SOX2 resulted in only 626

misexpressed genes (Figure 8.17A) This suggested that RMST could associate with

other nuclear proteins, apart from hnRNPA2B1 and SOX2 However, from the RIP

experiments in Chapter VII, it is indicated that RMST are neither bound by SUZ12, a

component of the PRC2 repressive complex, nor the transcription factor REST It is

likely that RMST may bind to a repressive chromatin modifier, such as DNMTs, to

silence non-neuronal genes during neurogenesis More experimental validation would

however, be required to provide full mechanistic insight of the role of RMST in

neurogenesis

8.4 Conclusion

This chapter focuses on elucidating the mechanism of RMST involvement in neuronal differentiation RMST is a conserved lncRNA that is repressed by REST During neurogenesis, when expression of REST is lost, RMST is upregulated Although RMST showed region-specific expression in the developing mouse brain, its role in

neurogenesis has not been established prior to this work Though association studies, I

identified two protein partners of RMST, namely RNA-binding protein hnRNPA2B1

and transcription factor SOX2 In addition, the data presented in this chapter indicates

that RMST and SOX2 regulate a large set of target genes that are essential for

neurogenesis, and supports the hypothesis that the dynamics of SOX2/chromatin

binding may be altered by RMST

Chapter IX – Conclusion and Perspectives

9.1 Overall conclusions

It is now evident that large numbers of lncRNAs exist in the mammalian transcriptome (Carninci et al., 2005; Guttman et al., 2009) and they are emerging players in embryogenesis and cell fate decisions Recent publications show the involvement of lncRNAs in cell cycle progression of progenitors in the retina (Meola

et al., 2012), apoptosis in erythroid differentiation (Hu et al., 2011), muscle differentiation (Cesana et al., 2011), retinal development (Rapicavoli et al., 2011) and

in maintenance of pluripotency (Dinger et al., 2008; Guttman et al., 2011; Sheik Mohamed et al., 2010) The roles of the above-mentioned lncRNAs were established

in the mouse model system, and very few human lncRNAs have been studied with such extent In this thesis, I focused on examining lncRNAs involved in the pluripotency maintenance of human embryonic stem cells (hESCs), and human neuronal development, using hESCs as a cellular model for neural differentiation

9.1.1 Identification of functional human lncRNAs

Using a microarray approach, lncRNAs highly expressed in hESCs were identified as

“pluripotent lncRNAs”, with a plausible function in pluripotency maintenance; while lncRNAs highly expressed in hESC-derived neurons were classified as “neuronal lncRNAs”, possibly playing a role in neurogenesis Using an RNAi approach, functional lncRNAs were identified Knockdown of these functional lncRNAs resulted in an observed phenotypic change: loss of pluripotency in hESCs or loss of neurogenesis in neural progenitors

9.1.2 LncRNAs specific to hESCs maintain the pluripotent state

LncRNAs whose expressions were regulated by pluripotent transcription factors OCT4 and NANOG showed specific expression in hESCs and iPSCs These pluripotent lncRNAs could be targeted for knockdown by siRNAs, which resulted in their exit from the pluripotent state, accompanied by upregulation of differentiation markers Insight into molecular mechanisms revealed that some of these pluripotent lncRNAs associate with SUZ12, a component of the PRC2 repressive complex, and pluripotent transcription factor SOX2 Taken together, this indicates the integration of lncRNAs into the pluripotency gene regulatory network

9.1.3 Neuronal lncRNAs support neurogenesis by associating with transcription factors

Neuronal lncRNAs, or lncRNAs highly expressed in mature neurons compared to pluripotent stem cells and neural progenitors, were shown to be indispensable in neurogenesis by loss-of-function studies Association studies, by means of RNA immunoprecipitation, indicated that some neuronal lncRNAs modulate neuronal differentiation by binding to polycomb complexes, as well as transcription factors such as REST

The molecular functions of RMST, a specific neuronal lncRNA were also

elucidated in this thesis Data indicated that RMST is abundant in neurons, and it interacts with the SOX2 via the RNA-binding protein hnRNPA2B1 Expression analyzes following perturbations of cellular SOX2 and RMST levels demonstrated

that SOX2 and RMST shared many downstream targets, which were enriched for

biological processes relating to neuronal development Preliminary data also showed

that depletion of RMST resulted in decreased SOX2 occupancy at its target gene promoters Together, these results suggest that RMST could regulate neurogenesis by

modulating the activity of SOX2

9.2 Limitations and future work

9.2.1 Discovery of novel lncRNAs

Differentially expressed lncRNAs were detected by means of a microarray This technique makes it possible to examine the expression of thousands of genes simultaneously and rapidly However, the main limitation of the microarray is the inability to detect novel transcripts The lncRNA microarray used in this study was based on known, annotated human lncRNA genes (Jia et al., 2010), and potentially functional but previously unidentified lncRNA would have been missed out in this study Therefore, whole transcriptome sequencing or RNA-sequencing (RNA-seq) is

a very attractive alternative to microarrays, because it is not limited to detecting transcripts that correspond to existing transcriptomic information, and it can also quantitatively determine splice isoforms (Wang et al., 2009)

9.2.2 Epigenetic regulation of pluripotency

In the proposed model of lncRNA involvement in pluripotency (Figure 6.12), pluripotent lncRNAs function as a modular scaffold to recruit PRC2 to silence

differentiation genes Many lncRNAs, including HOTAIR, have been suggested to

behave as molecular tethers on which different protein complexes assemble (Tsai et al., 2010), and this is an attractive model to consider However, experimental validation is required to prove this hypothesis First, genome-wide analysis of histone

marks (such as H3K27me3 catalyzed by PRC2 complex) after lncRNA knockdown can be determined by ChIP-seq Next, it is also expected that SUZ12 (or the PRC2 complex) and pluripotent lncRNAs would regulate a common pool of genes

In addition, to demonstrate that lncRNAs could function as modular scaffolds, the different protein-binding domains of the RNA could be elucidated by means of a series of lncRNA truncation experiments

9.2.3 RMST modulation of SOX2 activity

In Chapter VIII, I established a role of RMST in neurogenesis RNA/protein interaction studies demonstrated that SOX2 is a partner of RMST Preliminary evidence suggests that RMST could modulate SOX2 binding to chromatin, highlighting the possibility that RMST may recruit SOX2 and the transcriptional

machinery to specific genomic loci for the transcription of neurogenic genes This

hypothesis can be validated by mapping the genome-wide binding sites of RMST

using a novel method – chromatin isolation by RNA purification (ChIRP; Chu et al., 2011) Coupling ChIRP to next generation sequencing (ChIRP-seq), the genome-wide

binding profile of RMST can be determined Overlapping this data with SOX2 ChIP would yield valuable information on the co-regulation of genes by SOX2 and RMST Identification of an “RMST motif” will also shed light on how the lncRNA recruits

SOX2, or other protein partners to regulate gene expression and neurogenesis

9.2.4 SOX2 and lncRNA association

SOX2 is a transcription factor that is important in both the maintenance of pluripotency, and specification of neural cell fate This study indicates that SOX2 not

only binds chromatin, but also lncRNAs to maintain pluripotency or support neurogenesis (Figure 6.11C; Figure 8.13B) It appears that SOX2 has the propensity

to bind to RNAs, and SOX2 RNA immunoprecipitation, coupled to high throughput sequencing (RIP-seq) will identify other lncRNAs bound to SOX2 that might also be functional

9.2.4 Long non-coding RNAs or short peptides?

In this thesis, the Coding Potential Calculator (CPC) was used to bioinformatically predict the coding potential of lncRNAs Recent advances in sequencing technologies allowed for this prediction to be done experimentally Using ribosome profiling, a technique based on the deep-sequencing of ribosome-protected RNA fragments, Ingolia et al (2011) established that many lncRNAs can be exported to the cytoplasm and effectively engaged by the protein translational machinery Interestingly, almost half of the lincRNA candidates which were earlier shown to be required for the maintenance of pluripotency in mESCs (Guttman et al., 2011) were also found to be engaging elongating ribosomes Some lincRNAs were even demonstrated to be actively translated Therefore, this points to the possibility that lncRNAs could actually code for small peptides, which could contribute to cellular function and differentiation

9.3 Concluding remarks

In this thesis, I have established a highly efficient and robust neuronal differentiation system, which is the basis for the transcriptome profiling microarray experiments Functional lncRNAs that regulate pluripotency and neurogenesis were identified from

the microarrays, and from loss-of-function studies Similar to observations by Khalil

et al (2009), I found that these functional lncRNAs tend to associate with the PRC2 complex component SUZ12, as well as transcription factors I probed into the

molecular functions of RMST, a lncRNA that was indispensable in neuronal differentiation Association of RMST with SOX2 was established, and they regulate a common pool of neurogenic genes, indicating that RMST regulates neurogenesis by

SOX2 association

This work implies that biological processes such as pluripotency and cell fate determination are not just controlled by transcriptional networks, fine-tuned by microRNAs, but also modulated by lncRNAs, adding another layer of complexity to the control of gene expression that specifies cellular fate