Nup93 and CTCF modulate spatiotemporal dynamics and function of the HOXA gene locus during differentiation

The nuclear pore complex (NPC) consists of 30 nucleoporins, which form a channel between the nucleus and the cytoplasm for the transport of RNA and macromolecules (D'Angelo and Hetzer, 2008). In addition to nuclear transport, nucleoporins also regulate gene expression by associating with chromatin (D'Angelo, 2018; Kuhn and Capelson, 2019; Sun et al., 2019). Based on their relative residence times within the NPC, nucleoporins are classified as ‘on-pore’ or ‘off-pore’. Nucleoporin occupancy modulates development and differentiation (D'Angelo, 2018; Sun et al., 2019). For instance, the on-pore nucleoporin Seh1 (encoded by SEH1L) regulates oligodendrocyte lineage progression and myelin production in rats (Liu et al., 2019). Seh1 is upregulated during in vitro differentiation of rat oligodendrocyte progenitor cells. Loss of Seh1 in mice downregulates the genes involved in myelination such as Mbp, Cnp, Plp1, Myrf and Sox10. Seh1 also interacts with the transcription factor Olig2, which is required for oligodendrocyte differentiation, and the chromatin remodeler Brd7 to nucleate a transcription complex at the nuclear envelope (Liu et al., 2019). A null allele of the nucleoporin Nup133 (a member of the Nup107-160 complex) disrupts neural differentiation of mouse embryonic stem cells (ESCs) (Lupu et al., 2008). Furthermore, Nup133 is upregulated in the neuroepithelium and paraxial mesoderm/somites during the development of mouse embryos (Lupu et al., 2008). The depletion of the chromatin-associated nucleoporin Nup153 results in premature differentiation and silencing of development-specific genes in mouse ESCs, mediated by the PRC1 complex proteins (Jacinto et al., 2015). Nup153 and Sox2 co-associate with promoters of pluripotency-specific genes in rat neural progenitor cells and regulate their differentiation (Toda et al., 2017). Additionally, co-immunoprecipitation assays reveal that Nup153 interacts with the chromatin architectural proteins CTCF and cohesin in mouse ESCs, implicating Nup153 in modulating enhancer–promoter interactions (Kadota et al., 2020). Furthermore, the mobile nucleoporin Nup98 interacts with CTCF and controls promoter–enhancer contacts in Drosophila cells (Pascual-Garcia et al., 2017). The off-pore nucleoporin Nup98 [a phenylalanine-glycine (FG) repeat-containing nucleoporin] associates with genes involved in development and differentiation (Liang et al., 2013). Nucleoporins associate with chromatin directly or indirectly or via transcription factors and modified histones, facilitating cellular differentiation (Capelson et al., 2010; Hou and Corces, 2010; Ibarra et al., 2016; Kalverda and Fornerod, 2010; Sood and Brickner, 2014; Vaquerizas et al., 2010).

Nup93 is an on-pore nucleoporin that functions as a chromatin anchor during differentiation (Gozalo et al., 2020; Iglesias et al., 2020). Nup93 is enriched on human chromosomes 5, 7 and 16, as shown by chromatin immunoprecipitation (ChIP)-tiling arrays (Brown et al., 2008). Dam-ID analysis shows that Nup93 binds to enhancers that regulate the expression of cell identity genes (Ibarra and Hetzer, 2015; Ibarra et al., 2016). In Drosophila S2 cells, Nup93 tethers the Polycomb-group factors at the nuclear envelope, thus repressing gene expression (Gozalo et al., 2020). We have previously shown that Nup93 occupancy represses HOXA gene expression in differentiated cells (Labade et al., 2016).

The human HOXA genes are a cluster of 13 genes (HOXA1–HOXA13) that are essential for development and differentiation (Mallo and Alonso, 2013; Montavon and Soshnikova, 2014). Disruption of the temporal expression profile of HOXA genes is associated with developmental defects in vertebrates (Aubin et al., 1997; Mallo and Alonso, 2013). Also, aberrant expression of HOXA genes is associated with breast and lung cancers (Bitu et al., 2012; Novak et al., 2006). Therefore, the precise regulation of the temporal activation of HOXA gene expression is essential for normal development and differentiation (Mallo and Alonso, 2013; Montavon and Soshnikova, 2014).

CTCF – a chromatin architectural protein – is a crucial regulator of HOXA organization and function (Xu et al., 2014). CTCF associates with multiple conserved CTCF-binding sites (CBSs) on the 5′ HOXA genes, regulating their organization and expression (Rousseau et al., 2014; Xu et al., 2014). Interestingly, Nup93 associates with the 3′ HOXA genes (HOXA1, HOXA3 and HOXA5) (Labade et al., 2016). However, the relationship between nucleoporin Nup93 and CTCF in regulating chromatin organization remains unclear.

Here, we uncover a novel and complementary role of Nup93 and CTCF in modulating the organization and function of the HOXA gene locus in the NT2/D1 embryonal human carcinoma cell line, which differentiates into a neuroectodermal lineage upon retinoic acid (RA) treatment. ChIP-seq analysis of Nup93 in differentiated cells revealed an overlap between Nup93- and CTCF-binding sites. We found that Nup93 and CTCF differentially modulate the expression levels of the different HOXA genes in undifferentiated cells. Furthermore, three-dimensional fluorescence in situ hybridization (3D-FISH) analysis showed that Nup93, but not CTCF, tethers the HOXA locus to the nuclear periphery. Interestingly, ChIP-quantitative PCR (qPCR) analysis of Nup93 and CTCF revealed a dynamic association of the HOXA gene locus with the nuclear periphery during differentiation. These findings provide key insights into how a nuclear pore complex protein (Nup93) and the chromatin organizer (CTCF) co-modulate chromatin dynamics during differentiation.

The central role of nucleoporins is nuclear transport. The stable nucleoporin Nup93 also associates with chromatin and modulates gene expression (Breuer and Ohkura, 2015; Brown et al., 2008; Ibarra et al., 2016; Iglesias et al., 2020). Using ChIP-qPCR, we previously showed that Nup93 is enriched on the HOXA gene locus (Labade et al., 2016). Here, we determined the genome-wide occupancy of Nup93 in the differentiated colorectal cancer cells (DLD-1). ChIP-seq analysis revealed that Nup93 associates with sites widely across the genome (∼404 high-confidence sites) (Fig. S1A–C). Remarkably, ∼8% of these sites are enriched on gene promoters (Fig. 1A; Fig. S1D). Gene Ontology (GO) analysis showed Nup93 enrichment on genes involved in developmental processes, such as the canonical Wnt-signaling pathway, osteoblast development, organ morphogenesis, and development of the central nervous system (Fig. 1B). Furthermore, Nup93-bound sequences are enriched for consensus DNA motifs for transcription factors, such as EGR3, ZNF394 and ZNF502, which are also involved in regulating development and differentiation (Fig. 1C) (Eldredge et al., 2008; Quach et al., 2013). Interestingly, ChIP-seq analyses showed an overlap between Nup93- and CTCF-binding sites (Fig. 1D). Furthermore, ReMap analyses of Nup93-binding sequences revealed an overlap of Nup93-binding sites with transcription factors that include CTCF, MYC and AR (Fig. S1E). Of note, both Nup93 and CTCF showed an enrichment on the HOXA1 promoter (Fig. S1F). In summary, Nup93- and CTCF-binding sites show a genome-wide overlap in differentiated cells, suggesting a potential co-regulatory role in controlling the expression of genes involved in development and differentiation.

We previously showed that Nup93 is enriched on the 3′ HOXA genes whereas CTCF associates with the 5′ region of the HOXA gene cluster in differentiated cells (Ferraiuolo et al., 2010; Labade et al., 2016; Rousseau et al., 2014; Wang et al., 2015). Here, we examined the role of Nup93 and CTCF in the regulation of HOXA gene expression upon RA-mediated differentiation in pluripotent embryonal carcinoma (NT2/D1) cells. We observed distinctive changes in cell morphology from polygonal to elongated – a characteristic of neuronal cells (Fig. S2A; day 6 of RA treatment). Furthermore, pluripotency genes OCT4, SOX2 and NANOG showed decreased expression levels from day 2 of RA addition (Fig. S2B, immunofluorescence staining of OCT4, and Fig. S2C). In contrast, the expression levels of the neuronal marker PAX6 was significantly upregulated, consistent with the neuronal lineage adopted by these cells upon RA-mediated differentiation (Fig. S2D). Notably, both protein and transcript levels of Nup93, Nup188 and Nup205 were not significantly altered during differentiation (Fig. 2A,B; Fig. S2E). Consistent with the pattern of transcriptional induction of HOXA genes, expression levels of the 3′ HOXA genes HOXA1 and HOXA5 showed a marked increase, followed by the 5′ HOXA genes HOXA9 and HOXA13, respectively (Fig. 2C) (Xu et al., 2014). In addition, we performed RNA-FISH to examine the induction of HOXA expression during differentiation. Single-cell imaging revealed about three or four foci of HOXA transcripts after ∼48 h of RA treatment, which corroborates the transcriptional activation of HOXA (Fig. 2D; Fig. S2F). Taken together, these results underscore that HOXA genes are transcriptionally activated in a temporal manner upon RA-induced differentiation of NT2/D1 cells.

Nup93 and CTCF are enriched on the 3′ and 5′ regions of the HOXA gene cluster, respectively (Labade et al., 2016; Xu et al., 2014). We asked whether Nup93 or CTCF depletion impacts HOXA expression. Remarkably, siRNA-mediated knockdown of Nup93 significantly upregulated expression of the 3′ HOXA genes HOXA1 and HOXA5, whereas that of the 5′ HOXA genes HOXA9 and HOXA13 was downregulated (Fig. 3A). In contrast, CTCF knockdown caused a significant downregulation of the 3′ HOXA genes HOXA1 and HOXA5 and an upregulation of the 5′ HOXA genes – HOXA9 and HOXA13 (Fig. 3A). Although Nup93 is enriched on the promoter of 3′ HOXA genes (but not on 5′HOXA genes), Nup93 depletion represses 5′ HOXA gene expression (Fig. 3A). Conversely, while CTCF occupies 5′ HOXA genes, nevertheless, loss of CTCF downregulates 3′ HOXA gene expression. Interestingly, co-depletion of Nup93 and CTCF rescued HOXA expression to basal levels, suggesting a functional complementarity between Nup93 and CTCF in the regulation of HOXA gene expression (Fig. 3A). Additionally, the loss of either Nup93, CTCF or their co-depletion did not significantly alter the expression levels of the pluripotency genes OCT4, SOX2 and NANOG [Fig. 3B,C (protein levels of OCT4); Fig. S3A].

Although nucleoporins regulate chromosomal ploidy and mitosis, cell cycle profiles were unaltered upon Nup93 loss (Fig. S3B) (Nakano et al., 2011). Furthermore, nuclear export of poly-A RNA was only marginally affected upon Nup93 depletion in NT2/D1 cells, as detected by an increased nuclear accumulation of polyA-RNA. In sharp contrast, depletion of the off-pore Nup98 caused a significant increase in the nuclear accumulation of polyA-RNA (Fig. S3C,D). This suggests a comparatively marginal effect of Nup93 loss on nuclear export.

In order to address the role of Nup93 and CTCF in regulating HOXA genes during differentiation, we independently knocked down Nup93 and CTCF followed by RA treatment (Fig. 3D). Surprisingly, RA treatment upon Nup93 depletion further enhanced the expression levels of the 3′ HOXA genes (HOXA1 and HOXA5; Control, siLacZ plus RA, Fig. 3E and Fig. S3E). However, expression levels of the 5′ HOXA gene HOXA9 was attenuated (Fig. 3E; Fig. S3E). We surmise that the loss of Nup93 primes the 3′ HOXA genes for transcriptional activation, further enhancing HOXA1 and HOXA5 expression upon RA treatment (Fig. 3E; Fig. S3E). However, CTCF depletion attenuates 3′ HOXA gene expression (HOXA1 and HOXA5) in response to RA treatment (Fig. 3F; Fig. S3F).

Since RA is a potent and rapid inducer of HOXA, we asked whether HOXA expression is induced and regulated at early time points of differentiation (Mills et al., 1996). We independently knocked down Nup93 or CTCF and examined its effect on HOXA1 expression within ∼4 h of RA treatment (Fig. 3G,H; Fig. S3G,H). While Nup93 depletion showed a relatively early induction of HOXA1 (from ∼2 h), CTCF knockdown delayed the induction of HOXA1 expression relative to RA treatment (Fig. 3G,H; Fig. S3G,H). In conclusion, Nup93 and CTCF depletion modulate HOXA expression in a differential manner even at early time points, unraveling a functional complementarity between Nup93 and CTCF in the regulation of HOXA gene expression.

The spatial localization of the HOXA gene locus in the interphase nucleus correlates with its expression status (Morey et al., 2007). We performed 3D-FISH followed by confocal imaging and analyses to determine whether Nup93 or CTCF depletion impacts the spatial organization of the HOXA gene locus (Fig. 4A). Interestingly, loss of Nup93 showed a significant shift of the HOXA gene locus away from the nuclear periphery, whereas loss of CTCF did not alter the localization of the HOXA gene locus with respect to the nuclear border (Fig. 4B). Taken together, these results suggest a novel role for Nup93 in tethering and repressing the HOXA gene locus at the nuclear periphery.

We next studied the association of Nup93 and CTCF with the HOXA gene locus during differentiation. As a first step, we asked whether the Nup93 sub-complex (Nup188–Nup93–Nup205) is (1) stably maintained and (2) interacts with CTCF during differentiation. Co-immunoprecipitation of Nup188 (an interactor of Nup93) showed that the interaction between Nup93 and Nup188 is stably maintained during differentiation (Fig. 5A). Consistent with previous studies, Nup188 does not interact with Nup205 (Fig. 5A) (Braun et al., 2016; Labade et al., 2016; Miller and Forbes, 2000; Theerthagiri et al., 2010). Furthermore, we found that CTCF does not interact with the Nup93 sub-complex during differentiation (Labade et al., 2016). These results suggest that the Nup93 sub-complex is stable during differentiation with no direct protein–protein interaction with CTCF.

Nup93 and CTCF have distinct binding sites on the HOXA gene locus (Fig. 5B) (Labade et al., 2016). Nup93-binding sites are predominantly enriched on the 3′ HOXA genes (HOXA1, HOXA3 and HOXA5), while CBSs are enriched on the 5′ HOXA genes (Fig. 5B). We determined the relative enrichment of CTCF and Nup93 on the CBSs and the 3′ HOXA genes (HOXA1 and HOXA5) during differentiation (Fig. 5C–H). Interestingly, the occupancy of CTCF decreased across the CBSs by day 8 of differentiation (day 8, Fig. 5C–F; Fig. S4A–D). Of note, RA inhibits CTCF binding on HOXA, thereby promoting HOXA expression during differentiation (Oh et al., 2018). Remarkably, CTCF showed a striking increase in its occupancy on CBSs by day 21 of RA treatment (day 21, Fig. 5C–F). In contrast, Nup93 hardly showed an enrichment on the CBSs (except CBS2 on day 21) suggesting reduced occupancy of Nup93 on 5′ HOXA genes (Fig. 5C–F). Nup93 is enriched on the promoters of the 3′HOXA genes HOXA1 and HOXA5 prior to differentiation (day 0), but shows reduced occupancy during differentiation (day 4 and day 8) (Fig. 5G,H; Fig. S4E,F). Interestingly, Nup93 was re-enriched on the HOXA1 promoter upon differentiation (day 21, Fig. 5G,H). Collectively, these results show an association of Nup93 with the HOXA gene locus and consequently HOXA repression, both prior to differentiation (day 0, undifferentiated) and post differentiation (day 21, differentiated).

We examined the binding specificity of Nup93 and CTCF by testing their occupancy on regions outside the HOXA1 promoter (∼2 kb upstream and downstream of the HOXA1 promoter region) (Fig. 5B; Figs S5A,B and S6A,B). Nup93 and CTCF were hardly enriched either upstream or downstream of the HOXA1 promoter (∼2 kb) (Figs S5A,B and S6A,B). However, PanH3 (a-core histone H3) which served as a control, was enriched on both HOXA1 and HOXA5 promoters and the CBSs, thus validating the efficiency of pulldown in our ChIP-qPCR experiments (Figs S5C–J and S6C–J). In summary, these results show that Nup93 and CTCF co-modulate HOXA function during differentiation.

We previously showed that the activation of HOXA gene expression correlates with its disengagement from the nuclear envelope upon Nup93 depletion (Labade et al., 2016). We next determined the spatial localization of the HOXA gene locus during differentiation (Fig. 6A,B). 3D-FISH analyses revealed that the HOXA gene locus is proximal to the nuclear periphery (Fig. 6B,C) in undifferentiated cells but shifts away from the periphery during early differentiation. Notably, the HOXA gene locus shifts closer to the nuclear periphery upon differentiation (Fig. 6B,C). Taken together, the transcriptional activation and repression of the HOXA gene locus is associated with its disengagement and re-engagement with the nuclear periphery during differentiation.

Nucleoporins regulate chromatin organization and function across various cell types. Nup93 regulates spatiotemporal organization and function of the HOXA genes, which are essential for development and differentiation. Although nucleoporins are implicated in gene regulation, their mechanistic and functional relationship with genome organizers such as CTCF or cohesin is unclear.

Here, we studied the role of the nucleoporin Nup93 subcomplex in regulating the spatiotemporal organization and function of the HOXA gene cluster during RA-induced neuronal differentiation. We show that Nup93 and CTCF differentially regulate 3′ and 5′ HOXA expression of the HOXA gene locus. Furthermore, the occupancy of Nup93 on the 3′ HOXA genes decreased during differentiation, which correlates with the disengagement of the HOXA gene locus from the nuclear envelope. Remarkably, post differentiation, the HOXA gene locus re-associates with the nuclear envelope, consistent with a restored occupancy of Nup93 and CTCF on HOXA.

Nup93 is an on-pore nucleoporin (Rabut et al., 2004). ChIP-seq shows an enrichment of Nup93 on genes associated with development and differentiation (Fig. 1; Fig. S1). Nucleoporins such as Nup153, Nup50, Nup210, Seh1 and Nup98 also regulate expression of genes involved in development (Buchwalter et al., 2014; Jacinto et al., 2015; Kalverda and Fornerod, 2010; Liang et al., 2013; Liu et al., 2019). Furthermore, Nup93 is enriched on genes implicated in Hippo and Wnt-signaling pathways (Fig. 1B). Nup153, a basket nucleoporin, unlike Nup93, dynamically associates with the NPC, consistent with its relatively higher turnover rates (Daigle et al., 2001). Nup153 directly interacts with CTCF and the cohesin subunits SMC1A and SMC3, and modulates their binding on early EGF-inducible bivalent genes (Kadota et al., 2020), suggesting a partnership between nucleoporins and CTCF in regulating gene expression. Interestingly, RNA polymerase II binding decreases upon co-depletion of Nup153 and CTCF, resulting in attenuated expression of immediate early genes (IEGs) (Kadota et al., 2020). Furthermore, Nup2 and Nup98 function as insulators in Saccharomyces cerevisiae and Drosophila (Ishii et al., 2002; Kalverda and Fornerod, 2010). Nup98, although an off-pore nucleoporin, facilitates enhancer–promoter contacts along with CTCF in Drosophila S2 cells (Pascual-Garcia et al., 2017). On-pore nucleoporins such as Nup107 and Nup93 differentially associate with chromatin (Gozalo et al., 2020). While Nup107 associates with active chromatin, Nup93 occupancy overlaps with the inactive H3K27me3 mark, which is enriched on the repressed HOXA locus in differentiated cells (Labade et al., 2016). Furthermore, Nup93 also associates with and represses polycomb-target sites in the genome (Gozalo et al., 2020). We surmise that Nup93 functions as a stable scaffold that facilitates tethering of the chromatin loop potentially formed by CTCF, which effectively controls temporal HOXA organization and function required during differentiation.

CTCF is a chromatin architectural protein that also functions as an insulator (Huang et al., 2021; Kim et al., 2007; Narendra et al., 2015). The insulator function of CTCF depends on the number of CBSs, its binding strength, and orientation of the target DNA sequence. In addition, the position and number of CBSs are cell type specific (de Wit et al., 2015; Heger et al., 2012; Huang et al., 2021; Narendra et al., 2015). In NT2/D1 cells, there are five distinct CBSs on the 5′ end of the HOXA gene locus (HOXA6–HOXA13) (Fig. 5B). CTCF occupancy at the 5′ end induces chromatin looping (Oh et al., 2018; Rousseau et al., 2014; Wang et al., 2015; Xu et al., 2014). Also, the 5′ region (HOXA6–HOXA13) of the HOXA gene locus is enriched for the heterochromatin mark H3K27me3 (Xu et al., 2014). Interestingly, CTCF demarcates the boundary between the potential euchromatic (HOXA1–HOXA5) and heterochromatic regions (HOXA6–HOXA13) of the HOXA gene locus (Narendra et al., 2015; Oh et al., 2018; Wang et al., 2015; Xu et al., 2014). Although CTCF does not bind to the 3′ end of the HOXA gene locus (and prevents heterochromatinization), the 3′ end of HOXA (HOXA1–HOXA5) remain poised in undifferentiated cells potentially tethered by Nup93. However, during differentiation, the HOXA locus transitions from a poised state to an active state accompanied by a temporal wave of transcriptional activation from HOXA1 to HOXA13. We found that in undifferentiated NT2/D1 cells, Nup93 occupies the 3′ region of the HOXA gene locus while CTCF occupies the 5′ end. These findings suggest that Nup93 is involved in repressing the 3′ end of the HOXA gene locus in the absence of CTCF, whereas the 5′ end of the HOXA gene locus remains repressed due to the occupancy of CTCF (Ghasemi et al., 2021; Oh et al., 2018; Wang et al., 2015; Xu et al., 2014). Interestingly, upon RA treatment, the poised HOXA locus transitions into an active state due to the probable untethering of its 3′ end from the nuclear periphery, as evidenced by a decreased occupancy of Nup93 (Figs 3A and 5G,H). On the other hand, CTCF shows a gradual decrease in its occupancy on the CBSs at the 5′ end of HOXA, facilitating a temporal expression starting from the 3′ to the 5′ end of HOXA gene locus (Fig. 5C,D).

RA is a potent activator of retinoic acid receptors (RARs), and RAR occupies specific DNA sequence elements [retinoic acid response elements (RAREs)] within the HOXA gene locus (Oh et al., 2018; Wang et al., 2015). Binding of RARs to RAREs induces the eviction of CTCF from its binding sites, depending on its binding strength, leading to expression of HOXA during differentiation in a temporal manner (Oh et al., 2018). Interestingly, both CTCF and Nup93 re-occupy the HOXA locus on their respective binding sites post differentiation, thereby enabling its re-looping and re-tethering to the nuclear periphery, respectively (Figs 5B–H and 6A–D). Nup93 provides a stable tether at the NPC, further preventing an untimely activation of the HOXA gene locus in differentiated cells (Fig. 6D). In summary, the nucleoporin Nup93 and the chromatin organizer CTCF engage in a remarkable partnership to regulate and fine-tune the spatiotemporal organization and function of the HOXA gene locus during differentiation.

This study provides an interesting perspective on the expression pattern and the spatiotemporal regulation of the genes involved in development and differentiation. Our findings support the emerging evidence showing that nucleoporins exert gene regulatory functions beyond their canonical roles in nuclear transport. We surmise that the NPC functions as a tether to modulate the temporal regulation of HOXA gene expression during differentiation by recruiting other chromatin modulators, such as the PRC1 or PRC2 complex. High-resolution approaches, such as ChIP-Mass spectrometry, BioID and proximity ligation assays, are likely to unravel the potential intermediary interactors that facilitate HOXA locus tethering. In addition, super-resolution microscopy-based technologies would be useful to resolve sub-micron level changes in chromatin organization. Furthermore, chromatin accessibility assays using scATAC-seq are likely to elucidate the role of nucleoporins in regulating chromatin architecture in the context of differentiation.

In conclusion, this study unravels a complementarity between the nucleoporin Nup93 and the chromatin architecture protein CTCF in the regulation of HOXA genes during differentiation.

The NTERA-2 cl. D1 (NT2/D1) cell line was a kind gift from the lab of Sanjeev Galande (IISER Pune, India) with permission from Peter Andrews (The University of Sheffield, UK). Cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) (Gibco, 11995) supplemented with 10% fetal bovine serum (Sigma, F2442), 100 U/ml penicillin, 100 µg/ml streptomycin (Gibco, 15070-063), and 2 mM L-glutamine (1× Gibco® GlutaMAX™ Supplement, 35050061) at 37°C with 5% CO₂. The cultures were free of Mycoplasma contamination, as monitored by DAPI staining the cells and medium periodically. The authenticity of NT2/D1 cells was validated by karyotyping.

Retinoic acid (RA) stock (10 mM, Sigma-Aldrich, R2625) was prepared in DMSO and stored at −80°C, protected from light. Once thawed, the aliquot was stored at −20°C and used within 1 week. For all experiments, cells were treated with 10 µM RA, and cell culture medium was replaced every 24 h with fresh RA. RA treatments were performed for the specific time mentioned in the respective figures and legends.

siRNA-mediated knockdowns were performed using siRNA oligonucleotides from Dharmacon, USA. NT2/D1 cells (∼0.2×10⁶) were seeded in each well of a six-well plate at 24 h prior to transfection. The cells were transfected with siRNA oligonucleotides at the final indicated concentrations (Nup93, 20 nM; Nup188, 20 nM; Nup205, 50 nM; CTCF, 50 nM) using RNAiMax transfection reagent (Invitrogen, 13778) in reduced serum Opti-MEM medium (Gibco, 31985). Cell culture medium containing transfection mix was replaced with complete DMEM at 24 h post-transfection. Cells were harvested after 48 h and 72 h of knockdown and processed for RNA extraction or western blotting. siLacZ (Dharmacon) was used as a control. The siRNA sequences are provided in Table S1.

Cells were trypsinized and washed with 1× PBS. Fixation was performed with chilled 70% ethanol which was added to cells with tapping. Cells were centrifuged at 200 g at 4°C for 10 min. RNase A (1 µg/µl) treatment was performed for 45 min at 37°C. Cells were filtered through a 0.45 µm filter. Propidium iodide (10 µg) was added to cells and cells were sorted using the FACS-Calibur flow cytometer. An analysis was performed using BD Cellquest Pro. Representative images were created using Flowing Software (Turku lab).

Immunoblotting was performed as described previously (Labade et al., 2016). Briefly, cells were lysed in RIPA buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.1% SDS, 0.1% sodium azide, 0.5% Na-deoxycholate, 1 mM EDTA, 1% NP-40 and 1× protease inhibitor cocktail) and centrifuged at 13,000 g for 10 min at 4°C. Total protein was estimated using a BCA kit (cat. no. 23225; Thermo Fisher Scientific). Protein samples (20 μg) were prepared in 1× Laemmli buffer (Tris-HCl pH 6.8, 2% SDS, 20% glycerol, 0.2% Bromophenol Blue and 0.025% β-mercaptoethanol) and denatured at 95°C for 5 min. Proteins were resolved by SDS-PAGE and transferred to activated PVDF membrane (Millipore, cat. no. IPVH00010), followed by blocking with 5% non-fat dried skim milk in 1× Tris-buffered saline (10 mM Tris-HCl pH 7.4, 150 mM NaCl) plus 0.1% Tween-20 (TBST) for 1 h at room temperature. Primary antibodies were diluted in 0.5% milk in 1× TBST buffer. All antibody dilutions are within the linear range of detection. Antibodies used were: rabbit anti-Nup93 (1:500, sc-292099, Lot-E0211, Santa Cruz Biotechnology, CA [Note: anti-Nup93 antibody (sc-292099, Lot E0211, Santa Cruz Biotechnology) has been discontinued. We recommend using an alternative antibody from Sigma Millipore cat. no. HPA017937], rabbit anti-Nup188 (1:1000, Abcam, ab86601, Lot-GR43443-4), rabbit anti-Nup205 antibody (1:500, HPA024574, Lot-R11937, ATLAS antibodies) and rabbit anti-CTCF antibody (1:500, 07-729, Lot-2375606, Millipore). Secondary antibodies were donkey anti-rabbit-IgG conjugated to horseradish peroxidase (HRP) (1:10,000, GE NA9340V) and sheep anti-mouse-IgG conjugated to HRP (1:10,000, NA9310V), and were diluted in 0.5% milk in 1× TBST. Blots were developed using ECL Prime (Amersham 89168-782) at incremental exposures of 10 s, and images were acquired using LAS4000 (GE). Densitometry analysis of western blots was performed using ImageJ software from three independent biological replicates. GAPDH was used as an internal control for normalization.

For co-immunoprecipitation (co-IP), cells were lysed in a co-IP buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl and 0.5% NP-40) supplemented with a protease inhibitor cocktail (PIC; Roche 04693159001). Lysates were pre-cleared using protein-A Dynabeads (Invitrogen, 10002D) for 1 h at 4°C. After pre-clearing, total protein was quantified using a BCA assay. Anti-NUP188 antibody (2 μg), was incubated with 15 μl Protein-A Dynabeads in 500 μl 1× PBST for 1 h at 4°C. Antibody-coated Dynabeads were incubated overnight at 4°C with 500 μg of total protein in the lysis buffer on the end-to-end rotor at 9 rpm. Dynabeads were washed nine times with a co-IP lysis buffer. Samples for western blotting were prepared in Laemmli buffer.

Cells were washed with 1× PBS, and total RNA was extracted using the Trizol method (Rio et al., 2010). cDNA was synthesized from total RNA with the ImProm-II reverse transcription kit. Real-time quantitative PCR (RT-qPCR) was performed using the Bio-Rad real-time PCR instrument (CFX96 Touch) in 5 μl of reaction mixture containing KAPA SYBR Green mix and 2 μM each of the forward and reverse primers (intron-exon primers) (Table S1). The β-actin-encoding gene was used as an internal control. Fold change was calculated by double normalization of the threshold cycle (Ct) values to the internal control and untreated samples by the 2^−ΔΔCt method (Livak and Schmittgen, 2001). All primer sequences are listed in Table S1.

Cells were rinsed with 1× PBS and fixed with 4% paraformaldehyde (PFA, pH 7.4) for 10 min at room temperature, followed by washes in 1× PBS, permeabilization with 1× PBST for 1 h at room temperature. Cells were incubated with primary antibodies (mouse anti-OCT4, DHSB, cat. no. PCRP-POU5F1-1D2; 1:500) diluted in 1% BSA (in 1× PBST) for 2 h at room temperature in a humidified chamber. After three washes with 1× PBS, cells were incubated with secondary antibody (anti-mouse-IgG conjugated to antibody Alexa Fluor 568, Molecular Probes Cat: A11004) diluted 1:1000 in 1% BSA (in 1× PBST) for 1 h at room temperature in a humidified chamber. Cells were washed with 1× PBS and counterstained with DAPI (0.1 µg/ml) in 2× sodium saline citrate (SSC; 300 mM NaCl, 30 mM sodium citrate, pH 7.0) for 5 min at room temperature. Coverslips were mounted on glass slides using Slowfade Gold antifade. Image acquisition was performed on a Zeiss LSM710 confocal microscope with a 63× Plan-Apo1.4 NA oil immersion objective with 405 nm, 488 nm and 561 nm laser at 1 to 2.5 digital zoom.

Cells were grown to a confluence of ∼60% and arrested at metaphase with 0.1 µg/ml Colcemid (Roche 10 295 892 001) for 90 min. Cells were harvested by trypsinization and treated with a hypotonic solution (using 0.075 M KCl) for 30 min. Hypotonic treatment was terminated by fixing cells with 4–5 drops of fixative (methanol:acetic acid, 3:1), followed by centrifugation at 200 g for 10 min at 4°C. The cell pellet was washed three times with a fixative solution and resuspended in a fresh fixative solution. Cells were dropped from a height onto clean glass slides. Metaphases were stained with DAPI (0.05 µg/ml in 1× PBS, pH 7.0).

A bacterial artificial chromosome (BAC) DNA for the HOXA locus (RP11- 1192 1132K14 for HOXA1-A9 and RP11-163M21 for HOXA3-A13) was purchased from CHORI BACPAC Resources. BAC-DNA extraction was performed with the Hi-Pure Plasmid DNA Extraction kit (Invitrogen K210017). Nick translation of the BAC-DNA was performed with the Nick Translation Kit (Roche 11745 808 910). Nick-translated DNA was precipitated using ethanol and resuspended in hybridization mix (50% deionized formamide plus master mix containing 10% dextran sulfate, 0.1 mg salmon sperm DNA in 2× SSC solution, pH 7.4).

NT2/D1 cells (∼0.2×10⁶) were seeded on coverslips in a six-well plate. After 48 h of Nup93 (20 nM), CTCF (50 nM) or Nup93 (10 nM) plus CTCF (50 nM) knockdowns, the cells were washed with ice-cold 1× PBS and treated with cytoskeletal (CSK) digestion buffer [0.1 M NaCl, 0.3 M sucrose, 3 mM MgCl₂, 10 mM PIPES (pH 7.4), 0.5% Triton X-100] for 5 min followed by fixation with 4% PFA in 1× PBS (pH 7.4) for 10 min at room temperature. The cells were permeabilized in 0.5% Triton X-100 (prepared in 1× PBS) for 10 min and incubated in 20% glycerol (prepared in 1× PBS) for 60 min followed by four freeze-thaw cycles in liquid nitrogen. The cells were washed three times with 1× PBS and treated with 0.1 M HCl for 10 min, followed by three washes in 1× PBS for 5 min each. The cells were incubated in 50% formamide (FA)/2× saline sodium citrate (SSC) (pH 7.4) overnight at 4°C or until used for hybridization. Cells were hybridized with 3 µl of human whole chromosome 7 paint [Applied Spectral Imaging (ASI), Israel, or MetaSystems, USA] and a nick-translated BAC DNA probe for the HOXA gene locus (3 µl). Post hybridization, coverslips were washed in 50% FA/2× SSC (pH 7.4), three times for 5 min each time at 45°C, followed by three washes for 5 min each in 0.1× SSC at 60°C. Coverslips were then counterstained with DAPI for 2 min, washed in 2× SSC and mounted in Slowfade Gold antifade (Invitrogen S36937).

The probe for RNA FISH for the HOXA locus was prepared from the BAC clone (RP11-1132K14) by nick translation. The probe was resuspended in 10 μl of deionized FA and mixed with an equal volume of 2× hybridization mix (10% dextran sulfate and 0.1 mg salmon sperm DNA in 2× SSC solution, pH 7.4) containing 2 mM vanadyl ribonucleoside complex (VRC), and incubated on ice for 30 min. Cells were washed three times in ice-cold 1× PBS with 2 mM VRC (5 min each) and treated with CSK buffer with 2 mM VRC on ice for 5 min followed by fixation using 4 % PFA with 2 mM VRC (7 min at room temperature). The cells were incubated with 70% ethanol at −20°C for 60 mins (or stored in 70% ethanol at −20°C until further use) followed by washes in an ethanol series (70%, 90% and 100% ethanol) and air-dried. Cells were hybridized with HOXA RNA probe by incubating at 37°C overnight, followed by washes with 50% FA/2× SSC (with 2 mM VRC) and 2× SSC (with 2 mM VRC, pH 7.2–7.4) at 42°C (three washes each of 5 min each). The cells were mounted using Antifade, followed by imaging on a confocal microscope.

Cells (∼0.2×10⁶) were seeded on coverslips in a six-well plate. After 48 h of Nup93 knockdown, the cells were fixed with 4% PFA in 1× PBS (pH 7.4) for 15 min at room temperature. Cells were re-fixed and permeabilized with chilled methanol for 5 minutes, followed by incubation in 2× SSC at room temperature for 10 min. Cells were hybridized with 100 µl of hybridization mix (40% formamide, 10% dextran sulfate, 0.1 mg salmon sperm DNA and 5 ng/ml of FAM oligo dT prepared in 2× SSC solution) at 37°C for 3 h. Coverslips were washed twice with 2× SSC, followed by washes with 0.1× SSC. Cells were stained with DAPI and mounted in an antifade solution. Images were acquired using confocal microscopy using a 63× objective (NA 1.4), using 488 nm and 405 nm lasers, with the zoom set to 1.0. The mean fluorescence intensity of the FISH signal was determined for each cell and nucleus (demarcated by DAPI) and expressed as a ratio of the nuclear to cytoplasmic fluorescence intensity. Nuclear/cytoplasmic (N/C) fluorescence intensity ratios were calculated and plotted using GraphPad Prism software. Statistical analysis was performed using the Mann-Whitney U test.

Image acquisition was performed on a Zeiss LSM710 confocal microscope with a 63× Plan-Apo1.4 NA oil immersion objective with 405 nm, 488 nm, and 561 nm laser at 1 to 2.5 digital zoom. Acquisition of Z-stacked images (voxel size of 0.105 μm × 0.105 μm × 0.34 μm) was at 512 × 512 pixels per frame using 8-bit pixel depth for each channel and a pinhole size of 0.7 μm [1.0 Airy units (AU)]. The line averaging was set to 4.0, and images were collected sequentially in a three-channel mode. Distances of gene loci from the nuclear periphery were measured (in μm) in 3D using the boundary of the DAPI signal as a marker of the nuclear periphery (Shachar et al., 2015). For quantification, confocal images were loaded into the object analysis tool of the Huygens Professional software (Version 18.10.0p2 64b, build Nov 20, 2018). This tool performs surface rendering according to the threshold segmentation of different groups of voxels that are separated from the background into a 3D object. 3D reconstruction was performed using surface rendering for the nucleus (blue channel), HOXA gene locus (red channel) and Chromosome Territory 7 (CT7; green channel). Comparable threshold and seeding levels were used for surface rendering for all images. The nucleus was selected as an anchor, and the shortest distance of each gene locus (Object) from the nuclear periphery was measured using the ‘analyze object tool’ in Huygens professional software (Version 18.10.0p2 64b, build Nov 20, 2018).

ChIP was performed as described previously (Labade et al., 2016). In brief, cells (∼1.0 ×10⁷) were cross-linked using 1% formaldehyde for 10 min at room temperature. Cross-linking was quenched using 150 mM glycine, and cells were lysed in 1 ml swelling buffer [25 mM HEPES pH 8.0, 1.5 mM MgCl₂, 10 mM KCl, 0.1% NP-40 and 1× protease inhibitor cocktail (PIC)], and nuclei were recovered by centrifugation at 500 g. Fixed nuclei were re-suspended in 1 ml sonication buffer (50 mM HEPES pH 8.0, 140 mM NaCl, 1 mM EDTA, 1% Triton-X-100, 0.1% sodium deoxycholate, and 0.1% SDS) supplemented with PIC, and sonicated using a Bioruptor Tween sonicator (Diagenode) to generate fragment sizes of ∼100-500 bp. The supernatant was pre-cleared using Protein A-Dynabeads (Invitrogen) for 1 h at 4°C. The amount of DNA was estimated using Nanodrop 2000. Anti-Nup93 antibody (sc-202099, Santa Cruz Biotechnology) and anti-CTCF antibodies were used at a concentration of ∼2 µg per ∼100 µg of chromatin sample, and were diluted to ∼1 ml in sonication buffer and incubated overnight at 4°C. IP complexes were captured using Protein-A-Dynabeads (pre-blocked with 0.5 % BSA in 1× PBS) and washed three times (at 4°C, 11 rpm on end to end rotor) each with sonication buffer, wash buffer-1 (50 mM HEPES pH 8.0, 500 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate and 0.1% SDS), wash buffer-2 (20 mM Tris-Hcl pH 8.0, 1 mM EDTA, 0.5% NP-40, 250 mM LiCl, 0.5% sodium deoxycholate and 1× PIC) and TE buffer. Immunoprecipitated chromatin was eluted twice in 200 µl of elution buffer (50 mM Tris-HCl pH 8.0, 1 mM EDTA, 1% SDS and 50 mM NaHCO₃) at 65°C for 10 min. The Input and IP fractions were treated with 20 µg RNase A for 1 h at 42°C followed by 40 µg Proteinase K for 1 h at 65°C. Reverse crosslinking was performed at 65°C overnight. DNA was extracted with phenol:chloroform:isoamyl alcohol (25:24:1), and ethanol precipitated using 3M sodium acetate (pH 5.2) and 2 µg of glycogen. DNA samples were washed with 70% ethanol and re-suspended in 10 µl of nuclease-free water.

Nup93 ChIP-DNA (two biological replicates) and Input DNA were outsourced for high-throughput sequencing to Genotypic Technology, Bangalore, India. ChIP-Seq libraries for sequencing were constructed according to the NEXTflex™ ChIP-Seq library protocol outlined in the NEXTflex™ ChIP-Seq Kit (5143-01). Briefly, DNA was subjected to a series of enzymatic reactions that repairs frayed ends, phosphorylates the fragments, adds a single nucleotide A overhang and ligates adaptors (NEXTflex ChIP Barcodes-48 kit). The libraries were enriched using PCR (5 cycles), fragments were size-selected using a 2.0% low-melting-point agarose gel and purified using MinElute Gel Extraction Kit (QIAGEN). The libraries were further enriched using PCR (14 cycles), post PCR cleanup was performed using Agencourt AMPURE XP beads (Beckman Coulter #A63881). The prepared libraries were quantified using a Qubit fluorometer and validated for quality by running an aliquot on a High Sensitivity Bioanalyzer Chip (Agilent).

Quality control for all Illumina sequence reads was performed using a FastQC tool (Version 0.72) (Afgan et al., 2018). All quality-controlled reads were mapped against the reference human genome (hg19) using Bowtie. Reads that aligned uniquely to a single genomic location were used for further analysis. Correlation between two biological replicates of ChIP-seq was determined using the ‘multiBamSummary’ and ‘plotCorelation’ tool on the GALAXY server. Aligned BAM files were visualized using the genome browser within the EaSeq platform as well as on the UCSC genome browser. Since both replicates were significantly correlated with one another, only one replicate was processed for further analysis.

The MACS peak (version 1.4.2 20120305) calling algorithm was used to find peaks representing likely binding sites for Nup93. Parameters used for peak calling are as follows: # effective genome size = 2.70e⁺⁰⁹; # band width = 300; # model fold = 10,20; # p-value cutoff = 1.00e⁻⁰⁵; # Large dataset will be scaled towards smaller dataset; # Range for calculating regional lambda is 1000 bps and 10000 bps. Significant peaks identified by MACS were further filtered and ranked based on the P-value. Wiggle files generated using MACS were used for visualization on the genome browser. Statistically significant peaks obtained from MACS peak caller were used for peak annotation. Peaks were annotated using the CEAS (cis-regulatory element annotation system) (Shin et al., 2009). CEAS annotates each peak summit with the genomic annotations in the following four categories: (1) promoters, (2) bidirectional promoters, (3) downstream of a gene, and (4) gene bodies (3′ UTRs, 5′ UTRs, coding exons, and introns). Promoter regions were defined as −1 kb, −2 kb, and −3 kb upstream of the transcription start site (TSS). Downstream regions were defined as +1 kb, +2 kb, and +3 kb downstream of the transcription end site (TES). Each peak was annotated with the closest gene found within a 500 kb window from the center of the summit. We also performed independent gene annotation using the EaSeq platform (Lerdrup et al., 2016).

Gene ontology analysis of Nup93 associated genes was performed using DAVID (<https://david.ncifcrf.gov/home.jsp>). We also performed the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis of Nup93 associated genes (<http://www.genome.jp/kegg/>). For motif analysis, Nup93-associated FASTA sequences were submitted as an input to MEME ChIP, and motif analysis was performed with default parameters against the HOCOMOCO Human (v11 FULL) database (Machanick and Bailey, 2011; Bailey et al., 2009). Out of 20 motifs identified by MEME-ChIP, the top six motifs were selected for further analysis. We performed TOMTOM (Tomtom compares one or more motifs against a database of known motifs) analysis on the six motifs to identify similar motifs from the HOCOMOCO Human (v11 FULL) database (Bailey et al., 2009). To count the number of high-probability CTCF motifs within Nup93-binding sequences, we used MEME motif finding software [FIMO version 5.0.5, Release date: Mon Mar 18 20:12:19 2019 -0700].

We performed a transcription factor enrichment analysis of Nup93-associated peaks using ReMap (Regulatory Map of TF Binding Sites; Chèneby et al., 2018). A total of 404 Nup93 binding peaks (.bed) were submitted to ReMap analysis (with 10% Minimum overlap) against the ReMap catalog of transcription factor binding peaks. The ChIP-atlas database (<http://chip-atlas.org/>) was used to look at the enrichment of transcription factors and histone marks, specifically in DLD-1 cells. We downloaded ‘BAM’ and ‘bigwig’ files for CTCF (DRX013180), H3K4me1 (DRX013183), H3K4me3 (DRX013175), H3K27ac (DRX013177), H3K27me3 (DRX013182) and H3K36me3 (DRX013178) from DLD-1 cells. Correlation plots and heatmaps were generated using the EaSeq platform (Lerdrup et al., 2016).

For qRT-PCR experiments, statistical significance was calculated using an unpaired Student’s t-test with one-tailed distribution and assuming unequal variance in Microsoft Excel. For DNA FISH and Poly-A FISH experiments, significance was calculated using the Mann-Whitney U test between control and treatment samples, assuming unequal variance using GraphPad Prism 6 (Version 6.01).

We are grateful to IISER-Pune for facilities and support for microscopy and imaging. We thank the laboratories at IISER-Pune for their generous contribution of reagents. We thank Rini Shah and Sanjeev Galande for their guidance, insights and timely suggestions. We thank all members of Chromosome Biology Lab (CBL) IISER-Pune for their critical comments on the manuscript.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.259307.