Psip1/p52 regulates distal Hoxa genes through activation of lncRNA Hottip

Long noncoding RNAs (lncRNAs) have been implicated in various biological functions including regulation of gene expression, X-inactivation, imprinting, cell proliferation and differentiation. However, the functionality of lncRNAs is not clearly understood and conflicting conclusions have often been reached when comparing different methods to investigate them. Moreover, little is known about the upstream regulation of lncRNAs. Here we show that a transcriptional co activator – PC4 and SF2 interacting protein (Psip1)/p52, which is involved in linking transcription to RNA processing, regulates the expression of the lncRNA Hottip. Using complementary approaches – knockdown, Cas9 mediated lncRNA deletion, analysis of lncRNA binding by Chromatin isolation by RNA purification (ChIRP) - we demonstrate that Hottip binds to the 5’ Hoxa genes located in cis, which leads to their upregulation. Moreover, the synthetic activation of Hottip is sufficient to induce the expression ofpolycomb repressed Hox genes in mouse embryonic stem cells (mESCs).

Although very few of these molecules have been functionally characterised, a small number have been shown to function by binding to various protein complexes to regulate gene expression [2][3][4][5][6][7] . Some lncRNAs have been reported to affect gene expression in trans 8,9 , whereas others, such as Kcnq1ot1, Xact, Xist and Tsix, function in cis (reviewed in 10 ) . Other lncRNAs likely function in the cytoplasm through binding to other regulatory RNAs, e.g. miRNA 11 .
It has also been difficult to distinguish whether lncRNA function is conferred by the process of transcription or by the RNA molecule itself. Concerns have been raised with respect to limitations and discrepancies in various methodologies used to study lncRNA function 12 and contrasting conclusions have often been reached when comparing knockdown and knockout studies [13][14][15] .
With the exception of relatively well characterized lncRNAs like Xist 16 , H19 17,18 and Kcnq1ot1 19,20 , many recently described lncRNAs lack genetic evidence to support their function in vivo. Indeed, recent efforts to phenotype mouse knockouts for 18 lncRNA genes identified only 5 with strong phenotypes 21  Mammalian Hox loci are important model systems for the investigation of lncRNA functions. Many noncoding RNAs within Hox clusters are expressed tissue specifically [23][24][25][26][27] , and have indeed been linked to the regulation of Hox mRNA genes 8,13,28,29 . At the Hoxa cluster, the lncHoxa1/Halr -also known as Haunt ~ 50 kb away from 3' end of Hoxa (Fig 1a) has been shown to repress Hoxa1 expression in cis 30 .
Importantly, a recent study demonstrated that Haunt lncRNA plays distinct role as a repressor while its DNA sequence functions as an enhancer for Hoxa genes 14 .
HOTAIRM, located between Hoxa1 and Hoxa2, is expressed antisense to coding Hoxa genes, and is implicated in retinoic acid induced activation of HOXA1 and HOXA4 during myeloid differentiation 31 . HOTTIP lncRNA is transcribed in an antisense direction from the 5' end of Hoxa13 (Fig 1a), and is reported to be important for targeting MLL through interaction with WDR5 to maintain 5' HOXA expression in distal tissues 3 .
PC4 and SF2 interacting protein (Psip1) also known as LEDGF/p75 has been shown to be important for regulation of Hox genes in vivo 32 . We demonstrated the role of the p75 isoform of Psip1/p75 in recruiting Mll to expressed Hox genes 33

Psip1 is required for expression of lncRNA Hottip
In mammals, active Hox genes are maintained by complexes containing the MLL (Mix lineage leukemia) histone H3K4 methyltransferases, and repression is maintained by Polycomb (PcG) complexes 35 . We recently demonstrated a role for Psip1/p75 in recruiting Mll to expressed genes from the 5' end of Hoxa 33 . Most strikingly, at the extreme 5' end of Hoxa, beyond Hoxa13, where the Hottip lncRNA 3 is located, Mll1 binding is completely lost in Psip1 -/-MEFs compared to wild type (WT), this is accompanied by a concurrent loss of H3K4me3 and Menin -a common component of Mll1 and Mll2 36 (Fig. 1a). Loss of Psip1 also results in a complete loss of Hottip expression and reduced expression of Hoxa13, which has been described as a target of Hottip (Fig.1b) 3 . In contrast, other previously described Hottip target genes -Hoxa9, a10, and a11 3 are up-regulated in Psip1 -/-MEFs despite the loss of Hottip expression 33 . Nascent run-on analysis shows that these effects occur at the level of transcription (Fig. 1c). Together with the binding of Psip1 to the 5' Hoxa cluster (Fig 1a), this suggests that Psip1 has a direct role in regulating expression of

Mll1 occupancy over Hoxa cluster is altered upon p52 & Hottip knockdown
It has been suggested that Hottip has a role in maintaining an MLL complex through interaction with the WDR5 component 3 . Consistent with this, Mll1 occupancy was significantly reduced across 5' Hoxa genes upon knockdown of p52 or Hottip compared to control knockdown (Fig. 2d), Intriguingly, upon depletion of p52 ( Fig. 2a and b) and Hottip (Fig. 2a), Mll1 was redistributed to 3' Hoxa genes, which is consistent with the increase in expression of 3' Hoxa genes upon p52 or Hottip depletion (Fig. 2d). Loss of Mll1 upon p52 and Hottip knockdown is consistent with the reduced Hottip expression along with reduced Mll1, H3K4me3 and Menin across Hoxa13 and Hottip loci in Psip1 -/-MEFs (Fig. 1a).

Deletion of Hottip leads to reduced expression of 5' Hoxa genes
Most lncRNA depletion studies are done by si/sh RNA mediated knockdown, but the conclusions reached have often been different from those after genetic deletion of the loci encoding the lncRNAs [13][14][15] . We therefore used two pairs of guide RNAs with cas9 nickase (cas9n) to delete the gene body of Hottip (HottipΔ) in limb mesenchymal cells, leaving the promoter intact (Fig. 3a). RT-qPCR of Hoxa genes in homozygous HottipΔ cells showed a >50% reduction in expression of Hoxa13 and a11 (Fig. 3b). Similar to Psip1 and Hottip knock down studies (Fig. 2a), expression of 3' Hoxa genes, such Hoxa2, a6 and a7 increased compared to WT cells.

Hottip RNA is localized at 5' Hoxa genes
To find the direct targets of Hottip lncRNA in limb cells, chromatin isolation by RNA purification (ChIRP) 38 was performed with multiple antisense oligo pools for Hottip.

Induction of Hottip lncRNA is sufficient to activate 5' Hoxa genes
It was possible that reduced expression of 5' Hox genes in HottipΔ cells is due to loss of cis regulatory elements located within the deleted region, rather than due to loss of the Hottip RNA, to rule out this possibility we synthetically activated Hottip expression in mES cells, by recruiting a catalytically inactive Cas9 (dcas9) fused to a synthetic activator -VP64/VP160 dcas9-VP160 (VP16 x10) [39][40][41][42] . CRISPR dCas9 mediated transcriptional activation has been shown to be ineffective when guides are targeted >1kb from TSS 39 . Unlike human HOTTIP which is transcribed bidirectionally from the HOXA13 CpG island promoter (Fig. 4a) 3 , the Mouse Hottip promoter is ~2 kb away from the TSS of Hoxa13 which allowed us to recruit dcas9-VP160 to the promoter of Hottip and not to that of Hoxa13 (Fig. 4a).
Microarray and RT-qPCR analysis showed specific upregulation of 5' (a13, a11 and a10), but not 3', Hoxa genes upon dcas9-VP160 mediated Hottip activation ( Fig. 4b & 4c). In contrast, specific recruitment of dCas9-VP160 to the Hoxa13 promoter led to up regulation of only Hoxa13 -expression of other Hoxa genes was unaltered (Fig. 4b). Recruitment of dCas9-VP160 to the Hottip promoter in HottipΔ limb cells led to only a modest upregulation of Hoxa13 compared to WT cells (Fig.   4c), pointing to the importance of Hottip RNA in the regulation of Hoxa genes.

Discussion
Psip isoforms regulate the expression of Hox genes through two distinct mechanisms. We have previously shown that the longer p75 isoform binds directly to Mll through its MLL or integrase binding domain (IBD) and recruits Mll to active Hox clusters 33 . Here we have demonstrated that the shorter isoform Psip1/p52 -that lacks the C-terminal Mll1 binding domain of p75 -controls 5' Hoxa genes by activating the expression of Hottip lncRNA.
With the recent increase in the number of lncRNAs identified using highthroughput sequencing technologies, the list of these transcripts with unknown upstream transcriptional regulation and downstream functional mechanism is growing. Till date only few proteins have been shown to transcriptionally regulate lncRNAs -Oct4 and Nanog have been shown to directly regulate expression of several lncRNAs in mESCs 43  Although transcription at noncoding enhancer elements has been associated with enhancer activity 46,47 . Exosome mediated degradation of enhancer transcripts suggests that in most of the cases the process of transcription itself could be sufficient but not functional RNAs 48,49 . Enhancer like function of lncRNAs has been demonstrated in only few cases including HOTTIP 3 .
Together, we demonstrate the specific transcriptional regulation of Hottip by a transcriptional coactivator Psip1/p52, which has been shown to have higher in vitro coactivator activity than Psip1/p75 isoform 50 Table 1) were transduced as described by the manufacturer (Sigma Aldrich).
RTqPCR: cDNA prepared from superscript II reverse transcriptase was used to perform expression qPCR. All qPCRs were performed with three biological replicates in a LightCyler 480 (LC480, Roche), and the data was normalized to Gapdh. Details of the oligos are given in the Supplementary Table 2 Chromatin Isolation by RNA Purification (ChIRP) Anti-sense oligo probes tiling the mouse Hottip RNA were designed using the web tool from Stellaris FISH Probe Designer (https://www.biosearchtech.com/support/education/stellaris-rna-fish) Biosearch Technologies, CA, USA). Biotinylated oligos were synthesized by Sigma-Aldrich.
ChIRP was performed in limb mesenchymal cells as described previously 38 . RNA was isolated from 20% of the ChIRPed beads and used for RT-qPCR for Hottip, 7sk and Gapdh specific primers and rest was used to perform qPCR for Hoxa genes.
Primer details are given in Supplementary Table 2.

RNA In situ
RNA in situ hybridization for mHottip in 10.5 days mouse embryos were performed as previously described 52 . Details of oligos used to PCR amplify Hottip cDNA including T7 (sense) and T3 (Antisense) promoter sequences are given in Supplementary Table 3.
ChIP, antibodies and data analysis: ChIP was performed as described previously 33 , using antibodies for Mll1 (Active Motif 61295, 61296), ChIP DNA was hybridized to a custom Hox array and data was normalized as described previously 33 .

CRISPR mediated deletion of Hottip
Guide RNAs were designed using Zhang laboratory web tool (http://crispr.mit.edu). Table 4) were designed to target the murine Hottip genomic locus beyond the transcription start site (TSS) and before the transcription end site (Figure 2e). gRNAs were cloned into the D10A nickase mutant version of cas9 (cas9n) containing pSpCas9n(BB)-2A-GFP (PX461) 53 . A pool of four gRNA containing plasmids were transfected into limb cells using FuGENE HD transfection reagent and FACS sorted 48 hours after transfection for GFP + cells.

Paired guide RNAs (gRNAs) (Supplementary
Homozygous deletion of Hottip was confirmed by PCR and Sanger sequencing, primers used are given in Supplementary Table 2.

Cas9-mediated activation of Hottip and Hoxa13
Guide RNAs targeting the promoters of Hottip and Hoxa13 (Supplementary Table 5) were designed as above and cloned into pSLQ1371 49,54 . gRNA containing plasmids encoding mCherry and puromycin resistance were co-transfected with dCas9-  Table 5). Cells were harvested 48hrs after transfection for both RNA expression analyses by RT-qPCR and expression microarrays. Plasmids containing non-targeting guide RNAs and dCas9 alone (dcas9Δ) served as controls.