Nucleoporin 153 links nuclear pore complex to chromatin architecture by mediating CTCF and cohesin binding at cis-regulatory elements and TAD boundaries

The nuclear pore complex (NPC) components, nucleoporins (Nups), have been proposed to mediate spatial and temporal organization of chromatin during gene regulation. Nevertheless, we have little understanding on the molecular mechanisms that underlie Nup-mediated chromatin structure and transcription in mammals. Here, we show that Nucleoporin 153 (NUP153) interacts with the chromatin architectural proteins, CTCF and cohesin, and mediates their binding across cis-regulatory elements and TAD boundaries in mouse embryonic stem (ES) cells. NUP153 depletion results in altered CTCF and cohesin occupancy and differential gene expression. This function of NUP153 is most prevalent at the developmental genes that show bivalent chromatin state. To dissect the functional relevance of NUP153-mediated CTCF and cohesin binding during transcriptional activation or silencing, we utilized epidermal growth factor (EGF)-inducible immediate early genes (IEGs). We found that NUP153 binding at the cis-regulatory elements controls CTCF and cohesin binding and subsequent POL II pausing during the transcriptionally silent state. Furthermore, efficient and timely transcription initiation of IEGs relies on NUP153 and occurs around the nuclear periphery suggesting that NUP153 acts as an activator of IEG transcription. Collectively, these results uncover a key role for NUP153 in chromatin architecture and transcription by mediating CTCF and cohesin binding in mammalian cells. We propose that NUP153 links NPCs to chromatin architecture allowing developmental genes and IEGs that are poised to respond rapidly to developmental cues to be properly modulated.

Nucleoporin 153 (NUP153) interacts with the chromatin architectural proteins, CTCF and cohesin, and mediates their binding across cis-regulatory elements and TAD boundaries in mouse embryonic stem (ES) cells. NUP153 depletion results in altered CTCF and cohesin occupancy and differential gene expression. This function of NUP153 is most prevalent at the developmental genes that show bivalent chromatin state. To dissect the functional relevance of NUP153-mediated CTCF and cohesin binding during transcriptional activation or silencing, we utilized epidermal growth factor (EGF)-inducible immediate early genes (IEGs). We found that NUP153 binding at the cis-regulatory elements controls CTCF and cohesin binding and subsequent POL II pausing during the transcriptionally silent state. Furthermore, efficient and timely transcription initiation of IEGs relies on NUP153 and occurs around the nuclear periphery suggesting that NUP153 acts as an activator of IEG transcription. Collectively, these results uncover a key role for NUP153 in chromatin architecture and transcription by mediating CTCF and cohesin binding in mammalian cells. We propose that NUP153 links NPCs to chromatin architecture allowing developmental genes and IEGs that are poised to respond rapidly to developmental cues to be properly modulated.

KEYWORDS
Nuclear pore complex; nucleoporin; chromatin structure; transcription; chromatin architecture; nuclear structure INTRODUCTION Establishment of cell lineage specification, maintenance of cellular states and cellular responses to developmental cues rely on gene regulation and spatial genome organization during development 1,2,3 . Emerging data point to highly coordinated activity between epigenetic mechanisms that involve nuclear architecture, chromatin structure and chromatin organization 4,5,6,7,8 . However, our understanding on how nuclear architectural proteins are causally linked to chromatin organization and impact gene regulation have been limited underscoring the importance of defining the molecular determinants.
Nuclear architecture is in part organized by the nuclear lamina composed of lamin proteins and the nuclear pore complex (NPC). Nucleoporin proteins (Nups) are the building blocks of the NPC, which forms a ~60-120 mega dalton (mDa) macromolecular channel at the nuclear envelope mediating nucleocytoplasmic trafficking of proteins and RNA molecules during key cellular processes such as cell signal transduction and cell growth (reviewed in 9 ). Beyond their role in nuclear transport, the NPC has been one of the nuclear structural sites of interest for its potential role in gene regulation by directly associating with genes (reviewed in 10 ). Studies in budding yeast and metazoans have shown that the NPC provides a scaffold for chromatin modifying complexes and transcription factors, and mediates chromatin organization. In metazoans, such compartmentalization supports nucleoporin-chromatin interactions that influence either transcriptional activation or silencing 11,12,13,14 . In yeast, the majority of the genes that position to the NPC are transcriptionally active. For example, inducible genes such as GAL, INO1, and HXK1 relocalize from the nucleoplasm to the NPC upon transcription activation -a process that has been proposed to be critical for establishment of transcription memory 15,16,17,18,19 . For several of these loci, NPC association facilitates chromatin looping between distal regulatory elements and promoters 20, 21 . This mechanism has been proposed to be critical for expression and transcriptional memory of developmentally regulated ecdysone responsive genes in Drosophila. Upon activation, ecdysone responsive genes exhibit NUP98-mediated enhancer-promoter chromatin looping at the NPC 22 . Notably, NUP98 has been shown to interact with several chromatin architectural proteins, including the CCCTC-binding factor, CTCF. These findings collectively suggest that Nups can facilitate chromatin structure in a direct manner by regulating transcription and in an indirect manner whereby Nup-mediated gene regulation relies on architectural proteins. Nevertheless, the functional relevance of Nup-architectural protein interactions in transcription regulation and chromatin structure is not well understood.
Chromatin architectural proteins, CTCF and the cohesin complex, facilitate interactions between cis-regulatory elements 23, 24 . These interactions influence the formation and maintenance of long-range chromatin loops that underlie higher order chromatin organization 25,26,27,28 . Long-range loops of preferential chromatin interactions, referred to as "topologically associating domains" (TADs) are stable, conserved across the species, and exhibit dynamicity during development 24,29 . Importantly, TADs segregate into transcriptionally distinct sub-compartments 30, 31 and exhibit spatial positioning 32 .
Current models argue that lamina-chromatin interactions may provide sequestration of specific loci inside the peripheral heterochromatin and promote formation of a silent nuclear compartment 33,34 . Despite the close interaction between the nuclear lamina and the NPC, we still know very little on how NPC-chromatin interactions influence transcription and chromatin organization at the nuclear periphery.
In mammals, Nups show variable expression across different cell types and their chromatin binding has been attributed to cell-type specific gene expression programs (reviewed in 10 ). NUP153 is among the chromatin-binding Nups which have been proposed to impact transcription programs that mediate pluripotency and self-renewal of stem cells in mammalian cells 35,36,37 . NUP153 chromatin binding sites have been detected at the promoters and across gene bodies 35,37 . Large proportion of NUP153 sites have also been detected at the intergenic sites containing enhancers that are linked to cell identity genes 36 . Furthermore, the epigenetic and transcriptional state of NUP153 bindings sites show variability whereby NUP153 can associate with both transcriptionally active and silent regions of the genome 35, 37 . Nevertheless, the molecular basis for how NUP153 association at the enhancers or promoters impact chromatin structure and transcription remain to be open questions.
Here, we directly tested the relationship between NUP153-chromatin interactions and gene regulation in pluripotent mouse ES cells. Towards elucidating NUP153mediated mechanisms that control transcriptional silencing vs activation, we further utilized immediate early genes (IEGs) as model loci. We report that NUP153 interacts with chromatin architectural proteins, cohesin and CTCF, and mediates their binding at enhancers, transcription start sites (TSS) and TAD boundaries in mouse ES cells.
NUP153 depletion results in differential gene expression that is most prevalent at bivalent genes 8 . To determine the mechanism by which NUP153 regulates CTCF and cohesin function during gene expression, we utilized IEGs, including Egr1, c-Fos and Jun loci, which we identified as NUP153 targets in mouse ES cells. We took advantage of the fact that transcription at the IEG loci can be efficiently and transiently induced using growth hormones such as the epidermal growth factor (EGF) in HeLa cells 38 . We found that NUP153 binding at the IEG cis-regulatory elements is critical for CTCF and cohesin binding and subsequent POL II pausing. We also found that this function of NUP153 is essential for efficient transcription initiation of IEGs. Notably, IEGs exhibit a NUP153dependent peripheral positioning during the basal state and reposition even closer to the periphery during transcriptional activation. Our findings reveal that IEG-NUP153 contacts are essential for IEG transcription via establishment of a chromatin structure that is permissive for POL II pausing at the basal state. Collectively, we propose that NUP153 is a key regulator of chromatin structure by mediating binding of the chromatin architectural proteins, CTCF and cohesin, at cis-regulatory elements and TAD boundaries in mammalian cells. Through this function, NUP153 links NPCs to chromatin architecture allowing developmental genes and IEGs that are poised to respond rapidly to developmental cues to be properly modulated.

A proteomics screen identifies cohesin subunits as NUP153 interacting proteins
To understand the functional relevance of NUP153 in transcriptional regulation and chromatin structure, we utilized an unbiased proteomics screen using mouse NUP153 as bait in an affinity purification assay. We expressed FLAG-tagged mouse NUP153 (FLAG-mNUP153) in HEK293T cells and carried out immunoprecipitation (IP) followed by mass spectrometry (MS) ( Figure 1A, S1A-B). We identified several known NUP153 interacting proteins including TPR 39 , NXF1 40 , IPO5 41 , SENP1 42 , XPO-1, TNPO1, and RAN 43 . In addition, IP-MS revealed that NUP153 interacts with several chromatin interacting proteins including the cohesin complex components, SMC1A, SMC3 and RAD21 ( Figure 1A and data not shown).
NUP153 has been mapped to enhancers and promoters in mammalian cells and has been implicated in transcription regulation 35,36,37 . Nevertheless, whether NUP153 influences higher-order chromatin structure and how NUP153 impacts gene expression are not well understood. We, thus focused on the cohesin complex as it mediates higherorder chromatin organization, and regulates gene expression by facilitating and stabilizing enhancer-promoter interactions together with CTCF 44, 45 . Cohesin binding sites show ~70-80% overlap with CTCF chromatin interaction sites 46 . It has also been recently shown that cohesin positioning to CTCF binding sites is influenced by both transcription and CTCF 46,47 . To investigate functional communication between NUP153, CTCF and cohesin, we performed FLAG-NUP153 IP followed by western blotting and determined NUP153 interaction with CTCF and cohesin subunits ( Figure 1B).
To define the nuclear fraction at which NUP153 spatially interacts with CTCF and cohesin, we performed biochemical chromatin fractionation assay 48 using HeLa cells ( Figure 1C). We successfully collected the core histone, Histone H3, in the nuclear fraction (P1). Treatment of the nuclear fraction (P1) with the micrococcal nuclease (MNase) resulted in elution of chromatin binding proteins into the soluble fraction (S3) in comparison to the insoluble nuclear fraction (P2) ( Figure 1C-D). We detected the NPC component, NUP62, in the nuclear insoluble fraction (P2) in the absence or presence of MNase suggesting that the P2 fraction contains the intact nuclear membrane including the nuclear envelope and the NPC. Insoluble fraction has been also shown to contain several proteins, such as CTCF, that associate with the nuclear matrix 49 . We detected NUP153 both in the nuclear insoluble (P2), and the soluble (S3) fractions that contain chromatinbinding proteins ( Figure 1D). This data provides biochemical evidence supporting earlier cell biological reports that NUP153 associates with the nuclear membrane and is found as a soluble protein within the nucleoplasm 50, 51, 52 . It also suggests that the NUP153chromatin interactions might be established either at the nuclear periphery or in the nucleoplasm. Interestingly, similar to NUP153, we detected a proportion of CTCF and cohesin in the insoluble nuclear fraction (P2) even in the presence of MNase ( Figure 1D).
These findings argue that NUP153 may interact with CTCF and cohesin at the nuclear periphery, nuclear matrix or within the nucleoplasm.  Figure 2A). We next examined NUP153 distribution across three genetic elements including the TSS, enhancers and TAD boundaries ( Figure 2B). NUP153 binding was detected at the TSS ( Figure S2A) and we identified 31.5% of TSS (7,721/24,513) to be NUP153positive ( Figure 2B). To investigate the transcriptional state of the NUP153 chromatin binding sites, we performed RNA-Seq in ES cells and determined transcriptionally active vs inactive TSS based on Fragments Per Kilobase of transcript per Million mapped reads (FPKM). TSS with FPKM>1 (n=8,768) were denoted as active and TSS with FPKM£1 (n=11,861) were denoted as inactive. By utilizing previously published Histone 3 Lysine 4 trimethylation (H3K4me3) and H3K27me3 ChIP-Seq data sets 57 , we validated transcriptional activity and silencing at these sites ( Figure S2B). We found that NUP153 occupied both transcriptionally active and inactive TSS with a bias towards the active genes ( Figure S2B).
Compared to NUP153-negative enhancers, NUP153-positive enhancers exhibited higher H3K4me1, H3K27Ac and CBP/P300 occupancy ( Figure S2D). Distribution of NUP153 at the TSS and enhancers suggested that NUP153 may have a functional role in gene regulation.
Considering interaction of NUP153 with CTCF and cohesin complex, the third genomic region we focused on was the TAD boundaries which are characterized to be enriched for CTCF and cohesin ( 61 (reviewed in 62 )). To identify TAD boundaries, we utilized the previously reported coordinates based on Hi-C data in mouse ES cells 61 . We found that 66.9% of TAD boundaries (3,984/5,957) contained NUP153 binding ( Figure 2B) suggesting that NUP153 may functionally cooperate with CTCF and/or SMC3 at the TAD boundaries during chromatin organization.

NUP153 mediates CTCF and cohesin binding at cis-regulatory elements and TAD
boundaries CTCF and cohesin binding across the intergenic sites have been linked to their role in chromatin insulation or enhancer function during gene expression 63,64 . To determine the functional relevance of NUP153 interaction with CTCF and cohesin during gene regulation, we first mapped CTCF and cohesin binding sites by ChIP-Seq and correlated the data with NUP153 DamID-Seq to define co-occupied sites. In accordance with earlier reports 64 , CTCF and SMC3 were enriched across TSS, enhancers ( Figure   S2A, S2C) and TAD boundaries ( Figure 2D). We found that on average CTCF and cohesin binding sites were at ~5 kb distance with respect to the nearest NUP153 binding sites ( Figure S2E). Based on this criterion, we detected a robust co-localization whereby 48.9% of the CTCF and 44.4% of the SMC3 binding sites were co-occupied by NUP153.
Out of the CTCF + /NUP153 + co-occupied sites, 29.9% associated with TSS, and 24.2% associated with enhancers. SMC3 + /NUP153 + co-occupied sites presented a similar profile in that 23.9% of these sites associated with TSS, and 27.1% associated with enhancers.
By examining sites that are occupied by all three factors (NUP153 + /CTCF + /SMC3 + ), we found association of the sites with 10.4% of the TSS and 13.9% of the enhancers. These results pointed to a potential crosstalk between NUP153 and architectural proteins during regulation of gene expression and/or chromatin architecture.
To investigate the functional relevance of NUP153 in CTCF and cohesin distribution genome-wide and define its impact on gene regulation, we generated NUP153 knockdown (KD) ES cells by transducing cells with two different mouse NUP153-specific shRNA lentivirus. We confirmed ~55-60% downregulation of NUP153 expression by realtime PCR ( Figure S2F). Both control and NUP153 deficient cells showed typical pluripotent ES cell characteristics with their morphology and the presence of alkaline phosphatase activity 65 , suggesting that NUP153 depletion did not interfere with the pluripotent state of ES cells ( Figure S2G). By utilizing an oligo (dT)50-mer probe and performing RNA Flourescent in situ hybridization (FISH) 66 , we further validated that the Poly(A) + RNA export function of the NPCs was intact in NUP153 KD ES cells ( Figure S2G). We next evaluated how NUP153 depletion influenced distribution of CTCF and cohesin sites across different genetic elements ( Figure S3A). In control ES cells, 22.8% of CTCF binding was detected at promoters, while 45% binding was detected across gene bodies and 32.2% of binding was detected at intergenic sites. In contrast, both NUP153 deficient ES cells exhibited a higher CTCF binding at promoters (30-34%) and lower gene body-(38-41%) and intergenic-(27-29%) specific CTCF binding. Similar distribution patterns were detected for SMC3 in control and NUP153 KD cells ( Figure S3A). These results suggested that NUP153 may impact CTCF and cohesin binding across the genome. To this end, we compared number of CTCF and cohesin peaks between control and NUP153 KD ES cells ( Figure 2C). We identified a significant loss of genome-wide CTCF (~60%) and SMC3 (~86%) binding in NUP153 deficient ES cells ( Figure 2C).
Given that cohesin binding relies on CTCF 46, 67 , we focused on CTCF binding sites and showed that NUP153 is enriched at the CTCF-positive TSS (n=2,164; p=0, hypergeometric test), enhancers (n=2,272; p=0, hypergeometric test) and TAD boundaries (n=2,238; p=8.66e-103, hypergeometric test) ( Figure S3B). We next asked whether NUP153 regulates CTCF and/or cohesin binding selectively at a given genetic element and found that NUP153 depletion resulted in reduction in CTCF and cohesin binding across all three genetic elements ( Figure 2D). We next asked how NUP153 binding influence CTCF distribution. To address this question, we calculated the mean CTCF binding in NUP153 deficient cells in comparison to control cells and grouped the CTCF binding sites into two. Group I, contained CTCF sites that showed greater mean CTCF binding in control cells over NUP153 KD cells, and Group II contained CTCF sites that showed equal or lesser mean CTCF binding in control cells over NUP153 KD cells ( Figure 2E, S3C-E). Group I TSS sites constituted ~10% (1,123/11,726) of the total CTCF binding sites and half of these sites (~5%, 558/11,726) were NUP153 positive. Notably, metagene profiles across TSS, enhancer and TAD boundaries showed higher NUP153 binding at Group I sites over Group II sites ( Figure 2E). This data suggested that the degree of NUP153 binding correlates with differential change in CTCF binding at each genetic element. Based on these findings, we concluded that NUP153 mediates CTCF and cohesin binding at TSS, enhancer and TAD-boundaries. This raises the possibility that NUP153 may be critical for enhancer-promoter functions or chromatin organization functions of CTCF and cohesin during gene expression.

NUP153-mediated transcription regulation across the genome and at bivalent genes
Given the enriched association of NUP153 with gene regulatory elements and its influence on CTCF and cohesin binding at TSS and enhancers, we investigated the extent of transcriptional changes in NUP153 deficient ES cells. To this end, we performed RNA-Seq for control and two NUP153 KD ES cell lines. NUP153 depletion resulted in differential expression of 711 genes (fold change ≥1.5 and FDR<0.05) genome-wide ( Figure 3A). We next investigated how NUP153-dependent changes in CTCF binding may impact transcription. We found that ~34.4% (245/711) of the differentially regulated genes associated with CTCF-positive TSS. Majority of this gene set (~61%) showed transcriptional upregulation ( Figure 3B). Examining the biological function of these genes by GO analysis, we found that they associate with important cellular processes such as the cell migration (e.g. Ptk2b, Tcaf2, Wnt11), cell adhesion (e.g. Alcam, App, Itga3, Itga8, PLCb1), and cell differentiation (e.g. Foxa3, Flnb, Zfp423, Tnk2). Given that the CTCFpositive Group I sites showed drastic change in CTCF binding in NUP153 KD cells and they were enriched for NUP153 over Group II sites, we evaluated the number of differentially regulated genes between two groups. Group I sites associated with 19.4% Our analyses on the bivalent genes revealed that several Hox genes, which are known to present bivalent state in ES cells 1,71 , are NUP153 targets. Hox loci exhibit a tightly controlled genomic organization that relies on TADs with enriched CTCF binding 61,72 . This organization has functional relevance during developmental expression of Hox genes 73,74 . As presented in the representative tracks shown for the HoxA and HoxC clusters, we found that NUP153 depletion resulted in altered CTCF and/or cohesin binding at specific Hox genes ( Figure 3D, arrows). Importantly, three of these CTCF-binding sites ( Figure 3D, asterisks) have been reported to be critical in facilitating formation of TADs and providing an insulator function during developmental regulation of Hox gene transcription in mouse 71,75 . Based on our findings, we propose that NUP153 may contribute to the higher-order chromatin organization by regulating CTCF and cohesin positioning at specific developmental genes and mediate their gene expression.

NUP153-mediated POL II recruitment during the IEG paused state is critical for timely IEG transcription
To determine the mechanism by which NUP153 regulates CTCF and cohesin function during gene expression, we utilized IEGs including Egr1, c-Fos and Jun loci which we identified to be NUP153 targets in mouse ES cells ( Figure S4). It is well established that transcription at the IEGs is regulated by a proximally paused-POL II release mechanism 76,77,78 . This mechanism results in POL II occupancy at the promoter-proximal regions (20-50 bp downstream of TSS), allowing for rapid and transient responsiveness of the IEGs to stimuli such as growth hormones 38 .
By examining transcription and chromatin structure across the IEG loci, we determined that the TSS and distal regulatory elements of IEGs were occupied by CTCF and cohesin ( Figure S4). During the preparation of this manuscript, it was shown that IEG locus, EGR1, forms CTCF-mediated higher order chromatin structure which impacts EGR1 transcription in HeLa cells 79 . Based on these characteristics, we envisioned that the IEG loci would provide a powerful in vivo model to examine mechanisms of NUP153dependent gene expression, and provide a mechanistic understanding for the interplay between NUP153 and architectural proteins.
Earlier studies have revealed that IEG transcription kinetics showed variability in ES cells and thus could not be stably measured in this cell system 80 . To test the function of NUP153 in transcription regulation directly, we thus utilized HeLa cells. In these cells, IEG transcription can be reduced to a silent state by serum starvation and transcription initiation can be reproducibly induced within 15 minutes upon EGF treatment 38 . We generated NUP153 KD HeLa cells by transducing cells with NUP153-specific shRNA lentivirus and detected ~60-80% reduction in NUP153 expression in comparison to control cells by western blotting ( Figure 4A). We validated that NUP153 knockdown did not alter the nucleocytoplasmic trafficking at the NPCs by quantitating dexamethasone (Dex) responsive GFP-tagged glucocorticoid receptor (GR) nuclear import and export 81 Table   S1 for list of primer sets used for real-time RT-PCR). Knockdown of NUP153 reproducibly led to a significant reduction in IEG mRNA and pre-mRNA levels upon 15 min of EGF treatment when compared to control cells ( Figure 4B, S6A). We also detected a significant increase in the EGR1 and c-FOS pre-mRNA levels upon 30 min EGF treatment ( Figure   4B and S6A). These results argue that the suppression of IEG transcription during the initiation step leads to a delay in transcription or triggers a passive induction of negative feedback of IEG transcription 38 . These transcriptional changes at the IEG were NUP153 specific, as expression of FLAG-NUP153 in NUP153 deficient HeLa cells led to recovery of transcription initiation ( Figure 4C). These data collectively indicated that NUP153 acts as an activator of IEG transcription initiation.
Given that IEG transcription is mediated by the POL II pause-release mechanism 76,77,78 , we reasoned that NUP153 may control POL II occupancy pre-and/or posttranscription induction. To investigate, we performed POL II ChIP and quantitatively measured POL II occupancy at the TSS and across gene bodies (GB) of JUN and EGR1 using specific primer sets ( Figure 4D, and see Table S1 for primers used for ChIP realtime PCR). At the paused state (minus EGF), NUP153 knockdown led to significant reduction in the paused POL II amounts at the IEG TSS. Upon 15 min of EGF induction, POL II occupancy at the TSS and across the gene body of IEGs was also significantly lower in NUP153 KD HeLa cells. In contrast, POL II binding across the IEGs was comparable between NUP153 KD and control cells at 30 min of EGF induction. These results were in line with the real-time PCR data showing that NUP153 is critical for timely IEG transcription initiation ( Figure 4B). We concluded that NUP153 regulates IEG transcription initiation by controlling POL II occupancy across the TSS of IEGs at the paused state.

CTCF and cohesin binding at cis-regulatory elements of IEGs relies on NUP153
In mouse ES cells, NUP153 depletion led to significant reduction in CTCF and cohesin binding at the TSS and enhancers coupled with differential changes in transcription ( Figure 2, 3). Here, we took advantage of the inducible IEG loci to test the functional relationship between NUP153, and CTCF and cohesin during the paused state and transcription in a time course dependent manner. We utilized HeLa cell specific ENCODE ChIP-Seq data sets 82 (see Methods for details on the ENCODE datasets) and examined EGR1 and JUN loci-specific POL II occupancy along with chromatin structure by mapping CTCF and cohesin subunit, RAD21, enhancer-specific marks (H3K27Ac, H3K4me1, and CBP/P300), and H3K4me3 which positively correlates with transcription activation ( Figure S6B). Based on these maps, we designed primer sets to determine NUP153 kinetics at the IEG and evaluated NUP153-dependent changes in CTCF, cohesin occupancy across the predicted distal regulatory elements (enhancers) and IEG genetic elements including the TSS, promoter, GB and transcription termination sites (TTS) ( Table   S1).
NUP153 ChIP-Seq revealed that at the paused (minus EGF) state, NUP153 associates with the EGR1 and JUN enhancers (site 7 for the JUN locus, sites 2 and 3 for the EGR1 locus), TSS and TTS ( Figure 5 (left panel)). Furthermore, we found that NUP153 binding spreads across the loci in a transcription dependent manner ( Figure 5 (right panel)). These data suggested that the dynamics of NUP153 binding is tightly coupled to the transcriptional state of IEGs. Notably, similar to NUP153, CTCF and cohesin were also enriched around EGR1 and JUN enhancer sites at the paused state and both proteins dynamically dissociated from these sites upon transcriptional activation with EGF.
Furthermore, enhancer-specific binding of both proteins was dependent on NUP153. This is because CTCF and cohesin occupancy on IEG enhancers were significantly reduced in NUP153 KD cells compared to the control cells at the paused state (minus EGF), but were comparable between NUP153 KD and control cells in the transcriptionally activate state (+EGF) ( Figure 5). A recent report provided evidence that EGR1 transcription relies on CTCF-mediated higher order chromatin 79 . In the light of our findings, it is likely that NUP153 influences IEG chromatin organization by mediating CTCF and cohesin binding during the paused state.

Co-regulatory function of NUP153 and CTCF during IEG paused state
Based on these results, we hypothesized that NUP153-mediated CTCF and cohesin binding to the IEG enhancer sites might be necessary for POL II occupancy at the proximal-promoter sites during the IEG paused state. Given that cohesin distribution depends on CTCF binding and POL II elongation 47, 83, 84 , we focused on the functional relationship between NUP153 and CTCF. We generated CTCF knockdown HeLa cells by using shRNA against CTCF ( Figure 6A) and measured impact of CTCF downregulation on transcription and POL II occupancy at the IEG loci in a time course dependent manner.
Similar to the phenotype of NUP153 KD HeLa cells ( Figure 4B), depletion of CTCF also resulted in significant reduction in IEG transcription initiation ( Figure 6B). We also detected significant reduction in POL II occupancy at the promoter and TSS during the IEG paused state (minus EGF) ( Figure 6C). Importantly, upon targeting both NUP153 and CTCF by shRNA (NUP153/CTCF KD), we did not detect an additive effect in downregulation of IEG transcription ( Figure 6D) suggesting that NUP153 and CTCF mediate IEG transcription through the same regulatory mechanism.

NUP153-dependent spatial positioning of IEGs during transcription regulation
Here, we investigated the spatial organization at the c-FOS locus and its dependency on NUP153 during the paused and transcriptionally active state in a time course dependent manner. To this end, we examined the sub-nuclear position of c-FOS DNA with respect to the nuclear periphery in control and NUP153 KD HeLa cells. We used a c-FOS gene containing Bacterial Artificial Chromosome (BAC) clone (RP11-293M10) as a DNA probe and performed DNA FISH in combination with immunofluorescence using an anti-Lamin B1 antibody to label the nuclear periphery ( Figure 7A). We determined that the HeLa cells contained three c-FOS alleles. We utilized the microscopy images to measure the distance between each c-FOS locus and the nuclear periphery by Fiji software (see Methods for details on the calculation of normalized distance (ND) based on cell area). Analysis of cumulative frequency graphs has revealed that c-FOS locus is closely positioned (ND£0.12) to the nuclear periphery in ~30% of the control cells at the paused state (minus EGF) and that the loci moved even closer to the periphery (ND£0.10) upon transcription induction (+EGF). In contrast, NUP153 deficiency led to positioning of the locus almost similar to the paused state in control cells whereby the locus remained distal to the periphery independent of the transcriptional state ( Figure 7B and S7). These results argue that NUP153-dependent NPC positioning of IEG is critical during transcription regulation. These data also support our biochemical and genome-wide analyses showing that NUP153-mediates spatial positioning of CTCF and cohesin to the NPC during IEG transcription.

DISCUSSION
In this study, we aimed to provide a mechanistic understanding on how NUP153 mediates chromatin structure and influences transcription. Through an unbiased proteomics approach, we identified NUP153 association with chromatin architectural proteins, CTCF and cohesin, and revealed that NUP153 is a critical regulator of chromatin structure and transcription by affecting CTCF and cohesin binding across enhancers and promoters in mammalian cells. Specifically, we determined that NUP153 depletion altered CTCF and cohesin binding in mouse ES cells at the cis-regulatory elements and TAD boundaries. Examination of the number of CTCF or SMC3 differential binding sites between the two different NUP153 KD mouse ES cell lines showed that they vary ( Figure   2B). Nevertheless, CTCF and cohesin distribution patterns across different genetic elements showed similar pattern in different NUP153 KD cells ( Figure S3A). These data argue that NUP153 facilitates CTCF and cohesin binding to their putative binding sites rather than recruiting them.
NUP153 is one of the mobile Nups that associates with the NPC basket or is found within the nucleoplasm 50, 52 . Our data suggest that the co-regulatory function of NUP153 and architectural proteins likely occurs around the NPC. Even though we cannot exclude the fact that nucleoplasmic or nuclear matrix association is not possible, two of our results support NPC specific association. First, we have provided evidence that NUP153, CTCF and cohesin are detected in the nuclear insoluble fraction that contains the nuclear envelope and the nuclear matrix ( Figure 1D). Second, we have shown that the IEG loci, which we show to rely on cooperation between NUP153 and architectural proteins for efficient transcription, present a close spatial positioning to the periphery, which is distorted upon NUP153 depletion (Figure 7, S7). There is already evidence in yeast showing that the inducible genes such as the HXK1 facilitate chromatin looping between distal regulatory elements and promoters at the NPC 20, 21 . A similar mechanism has been also proposed to be critical for NUP98-dependent regulation of ecdysone responsive genes during Drosophila development 22 . Thus, use of mouse models or in vitro cell differentiation models will be necessary to further determine the regulatory role of NUP153 in chromatin organization and examine dependency of these processes to the subnuclear distribution of NUP153. These studies would provide critical insights on the significance of NUP153 in cell type-specific genome function and transcription regulation.
Regulation of bivalent gene expression has been proposed to be necessary for maintenance of ES cell pluripotency 1,8 . Bivalent state is established through the simultaneous catalytic activity of MLL and Polycomb Repressive Complex 2 (PRC2).

Recent work in pluripotent ES cells suggests that MLL2 deficiency results in increased
Polycomb binding coupled with loss of chromatin accessibility across the promoters and alterations in long-range chromatin interactions 8 . These results suggest that maintenance of bivalent state might be causally linked to higher order chromatin organization during regulation of developmental gene expression. In this study, we provided evidence that NUP153 is enriched over enhancers, and TAD boundaries and is critical for CTCF and cohesin binding at these sites ( Figure 2D)  and architectural proteins suggests that mammalian NUP153 may have a role in multistep organization and/or insulation of site-specific higher-order chromatin around the NPCs.
NUP153 interaction with architectural proteins at the enhancers and/or TAD boundaries may promote formation of a chromatin compartment at which transcription factors downstream of specific signal transduction pathways associate with target loci-a compartment that can provide spatial and temporal organization of gene expression in response to cellular cues. Based on our findings, Hox loci and IEGs are among the NUP153 targets which may be subjected to such regulation. Collectively, our findings fill a gap in the field by providing evidence that NUP153 acts as a nuclear architectural component that associates with CTCF and cohesin mediating their binding across TSS, distal regulatory elements and TAD boundaries. Such function can be attributed to its potential role either in higher-order chromatin organization and/or transcription regulation.
We have determined that transcription of human IEG loci and a small subset (~5%) of the mouse ES cell genes rely on NUP153-mediated CTCF and/or cohesin binding at TSS.
Our results were in accordance with earlier findings showing that only ~10% of all TSS bound CTCF associated with promoter activity 23 . Thus, future studies focusing on the role of NUP153 in chromatin structure and chromatin organization are critical.
Several genome-wide studies have shown that paused POL II distribution shows positive correlation with CTCF and cohesin binding across metazoan genomes 95,96,97 .
CTCF is thought to induce POL II pausing by creating "roadblocks" on the DNA template obstructing transcription elongation 83 . In this study, we provide new evidence that NUP153 cooperates with CTCF in regulation of POL II occupancy at the IEG loci during paused state. Specifically, CTCF depletion results in altered POL II recruitment at the IEG loci during the paused state-a phenotype that mimics NUP153 depletion ( Figure 6C).
Knockdown of both CTCF and NUP153 did not result in an additive effect on POL II occupancy ( Figure 6D) arguing that NUP153 and CTCF mediate IEG transcription through the same regulatory mechanism. During the preparation of this manuscript, two recent reports have supported our findings by showing that CTCF-mediated chromatin organization impacts IEG transcription 79,98 . Based on our data, we propose that NUP153 interacts with CTCF and mediates its binding at the cis-regulatory elements which subsequently leads to cohesin recruitment and chromatin looping between gene regulatory elements and/or TADs at the NPC. This state is essential for the establishment of a poised chromatin environment at which efficient transcription initiation can be rapidly induced through a POL II pause-release mechanism in response to stimuli ( Figure 7C).
NUP153-dependent localization to the NPC might thus provide an advantageous spatial position to genes that are poised to respond rapidly to developmental cues during ES cell pluripotency and/or differentiation. Time course dependent analyses of NUP153 distribution along the IEGs also allowed us to examine NUP153 dynamics during transcription. We determined that NUP153 exhibits a wide distribution across the promoter and the gene bodies of IEGs during transcriptional activation ( Figure 5) suggesting that there might be a tight functional correlation between NUP153 and POL II activity during transcription. Chromatin sites that are engaged with stalled or active POL II might therefore allow for the differential NUP153 binding and can provide its selectivity towards transcriptionally silent or active chromatin domains, respectively.
We found that CTCF and cohesin binding sites were on average ~5 kb distance from the nearest NUP153 binding sites ( Figure S2E). Despite this fact, we detected a robust reduction in CTCF and cohesin binding at the cis-regulatory elements, suggesting that NUP153 may influence CTCF and cohesin binding directly or indirectly. One possible mechanism through which NUP153 can influence binding of architectural proteins is through the scaffold feature of NPCs. In various organisms, NPC acts as a scaffold at which specific genes associate with chromatin regulatory proteins or transcription factors (reviewed in 10 ). For instance, in mammalian cells, activated MYC associates with NUP153 and another NPC basket protein, TPR, at the nuclear periphery and triggers formation of a transcriptionally permissive environment that includes SAGA complex and MYC-target genes that mediate cell proliferation and migration 93 . Second possible mechanism might be through establishment of a NUP153-mediated chromatin structure that provides optimal chromatin environment at putative CTCF and cohesin binding sites. CTCF-binding

Immunoprecipitation (IP) assay
For IP assay, HEK293T cells that were transfected with FLAG-GFP or FLAG-NUP153 were considered dynamic mass modifications. All searched spectra were imported into Scaffold (v4.4, Proteome Software) and scoring thresholds were set to achieve a peptide false discovery rate of 1% using the PeptideProphet algorithm.

Chromatin fractionation assay
The chromatin fractionation assay was performed as described previously  Table S1. Relative gene expression was calculated by the relative standard curve method. GAPDH expression was used to normalize data.

IEG transcription induction in HeLa cells
HeLa cells (1x10 6  were subjected to EGF treatment as described above.

Chromatin immunoprecipitation (ChIP) assay
ChIP experiments were performed as previously described 112 (Table S1) by real-time PCR as described above.

Immunostaining and DNA FISH
For sequential Lamin B1 immunostaining and c-FOS DNA FISH, HeLa cells (5.5x10 3 ) were grown on 12-well glass slides (Invitrogen) overnight at 37°C, and IEG transcription was induced as described above. Immunostaining was performed as previously described 108 . Briefly, fixed cells were subjected to immunostaining using anti-Lamin B1 (Abcam,

Poly(A) + RNA FISH and alkaline phosphatase staining
Poly(A) + RNA FISH was performed by using 5' Cy3-labled oligo-dT 50mer (Sigma-Aldrich) as previously described 66 . Briefly, hybridization was performed using 0.5 µg oligo-dT probe per sample overnight at 37°C in a humidified chamber. Following hybridization, cells were washed twice for 15 min at 42°C with 2XSSC, and once for 15 min at 42°C in 0.5XSSC. Slides were mounted using mounting medium containing DAPI (Vector labs) and cells were imaged by fluorescence microscopy. Alkaline phosphatase staining was performed using Red Alkaline Phosphatase Substrate kit (Vector labs, SK-5100) according to the manufacturer's protocol. Bright field images were taken using Leica EC3 color camera attached to Leica DM5500B microscope.

Nuclear transport assay
Hela cells were co-transfected with Rev-Glucocorticoid Receptor-GFP (RGG) expression vector (Gift from K. Ullman (University of Utah)) 113 and control (scrambled) or hNUP153specific shRNA vectors using Xfect reagent. Import and export assays were performed as previously described 81 . Briefly, for import assay, transfected HeLa cells were grown overnight on 12-well glass slides at 37°C and treated with 250 nM dexamethasone (Dex) (Sigma-Aldrich, D4902) to induce RGG nuclear import for the indicated times. For the export assay, 120 min Dex-treated cells were washed with 1xPBS (pH:7.2) and cultured in fresh culture medium for the indicated times. At the end of each time point, cells were fixed using 4% paraformaldehyde and mounted using DAPI containing mounting medium (Vector labs) Images were obtained with a Leica DM5500B microscope, a Leica DFC365 FX CCD camera, and examined to calculate percentage of cells with nuclear GFP-RGG signal.

RNA-Seq
Total Sequencing was done at 50bp single-end and generated about 110x10 6 reads per sample.
Sequence data was demultiplexed and Fastq files generated using Illumina's Bcl2Fastq v2 conversion software.

RNA-Seq data analyses
RNA-Seq reads were trimmed by Trim Galore (0.4.1, with -q 15) and then mapped with TopHat (v 2.1.1, with parameters --b2-very-sensitive --no-coverage-search and supplying the UCSC mm10 known gene annotation). The ERCC spike-in sequences were mapped separately. Gene-level read counts were obtained using the featureCounts (v1.6.1) by the reads with MAPQ greater than 30. Bioconductor package RUVseq (v 1.16.0) was used to normalize the read counts and edgeR (v 3.24.0) was employed for differential expression analysis. Fold change greater than 1.5 and false discovery rate (FDR) less than 0.05 was used to filter the significant differentially expressed genes.

ChIP-Seq
Definition of regulatory regions: Several analyses in the manuscript rely on ChIP-Seq analyses across different regulatory regions namely enhancers and promoters. Below we describe how these regulatory regions were defined.

ChIP-Seq data analyses
ChIP-Seq reads were trimmed by Trim Galore (0.4.1, with -q 15) and then mapped with bowtie2 (2.2.5, with parameters --very-sensitive) to mouse genome (UCSC mm10). The were utilized to examine chromatin structure across the JUN and EGR1 genes ( Figure   S6B) using Human hg19 as a reference genome.

Statistical analysis
Quantitation of data was performed using the following statistical tests. Significance of the difference between control and knockdown cells for variables was analyzed with parametric Student's t-test. The nonparametric Kolmogorov-Smirnov (KS)-test was applied to calculate significance between the control and knockdown cells during the analyses of c-FOS locus spatial positioning with respect to nuclear periphery in a time course dependent manner.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Gene expression profiles,  S.K. and E.Y. wrote the paper with input from all authors.

COMPETING INTERESTS
The authors declare no competing interests.

MATERIALS & CORRESPONDENCE
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Eda Yildirim (eda.yildirim@duke.edu).     Table S1) as denoted in the schematics showing EGR1 and JUN genes. Data shown are percent (%) of input at each genetic element. Values are mean ± standard deviation. Student's t-test was applied to calculate significance. *p<0.05; **p<0.01. Experiments were repeated more than 3 times. Nt, nucleotide.  Table S1 for primer sequences). Data shown are percent (%) of input at each genetic element. Values are mean ± standard deviation. Student's t-test was applied to calculate significance. *p<0.05; **p<0.01. Experiments were repeated more than 3 times. Nt, nucleotide.                   for details on GEO information for each data set). ChIP-Seq read numbers are indicated at the right y-axis per data set. Human hg19 reference genome was used to analyze data sets. Asterisk (*) denotes sites that contain putative CTCF binding.