CTCF is an exquisitely sensitive sensor of stress signals in normal patient-derived human mammary epithelial cells (HMECs)

We analyzed the role of CTCF in both the acute and chronic stress response using primary cell cultures of normal HMECs. HMECs were propagated from disease-free reduction mammoplasty tissue from multiple healthy donors (age range 20–40 years old) chosen in an unbiased fashion. Epithelial-rich regions of tissue were dissected from surgical discard material and prepared as described [2,6,7], and as diagrammed (Fig 1A, left panel). With time in culture, a subpopulation of HMECs overcame the p16^INK4A-induced senescence barrier and continued to proliferate (Fig 1A, right panel). Previous studies have shown that these “variant” HMECs (vHMECs) exist in a stably chronic stressed state, as indicated by elevated p53 protein levels, activation of the p53 target gene p21, and downregulation of p16 protein [6,35,36]. We extended this analysis by examining CTCF protein levels during a time course of the HMEC to vHMEC transition from days 1–70 (Fig 1B). We observed that CTCF protein levels were downregulated in normal HMECs by day 15 in culture in “stressed” but otherwise normal HMECs (sHMECs) before the p16-induced senescence barrier and remained stably repressed in proliferating vHMECs. We confirmed the stress-responsive nature of CTCF by showing that downregulation of CTCF protein levels in HMECs is induced by diverse forms of stress, such as the genotoxic agent doxorubicin (DoxR) (Fig 1C) and the acute oxidative stressor hydrogen peroxide (H₂O₂) (Fig 1D). Hydrogen peroxide also elevates the detoxification enzyme and cancer stem cell marker aldehyde dehydrogenase 1 family member A3 (ALDH1A3) [37,38] (Fig 1D). Importantly, it is critical to use freshly prepared HMECs because they rapidly become stressed and show elevated ALDH1A3 and decreased CTCF levels simply by freeze-thawing and prolonged tissue culturing. In fact, elevated expression of ALDH1A3 is a sensitive indicator of whether control cells are already stressed and have downregulated CTCF. Stress-induced downregulation of CTCF was also observed in spontaneously immortalized, non-tumorigenic human mammary epithelial MCF10A cells (Fig 1D and 1E), although not in other human breast cancer cell lines that we tested (S1 Fig). Care must be taken by using fresh, low-passage aliquots of MCF10A cells as they undergo a stress response accompanied by the appearance of multiple CTCF isoforms upon prolonged cell culture. With both HMECs and MCF10A cells, the normal, unstressed state should be carefully maintained in order to observe maximal CTCF stress-responsiveness. Notably, the impact of stress on other chromatin and transcription regulators is variable depending upon the stressor. For example, levels of the histone variant H2A.Z are unaffected by oxidative stress in HMECs (Fig 1F). In fact, we find that CTCF protein abundance in primary and non-tumorigenic mammary epithelial cells is consistently downregulated by all forms of stress that we have tested unlike other regulators and is far more sensitive to low levels of stressors than well-known stress markers such as p53 or ALDH1A3. Therefore, CTCF is a very robust sensor of the stress response in normal and non-tumorigenic human mammary epithelial cells, especially during acute stress. In fact, the abundance and nuclear distribution of CTCF in our experiments is a very sensitive indicator of whether primary cells are “normal” or not, as this state can be quickly lost under certain experimental conditions. An important consideration is that distinct MW forms of CTCF can be distinguished by immunoblotting using antibodies raised against different CTCF epitopes. For our Western blots, we used an N-terminal CTCF polyclonal antibody (Active Motif) that predominately recognized a protein(s) of ~110–120 kDa. The Active Motif CTCF antibody has been used to co-immunoprecipitate both hyper- and hypo-poly(ADP-ribosyl)ated (PARylated) forms of CTCF, which are each recognized by Western blot analysis with an N-terminal CTCF monoclonal antibody (G8; Santa Cruz) [39]. Given the biochemical and functional diversity of CTCF complexes, we expect that the stress-sensitive species reported here may represent a specifically modified form of CTCF.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. CTCF protein levels are selectively downregulated in normal epithelial cells upon stress.

(A) HMEC isolation and cell culture diagram; population doublings and mRNA levels of tumor suppressors p16 and RASSF1A which are typically silenced in the classic HMEC/vHMEC primary culture model. Days in culture are indicated at the bottom. mRNA was measured by RT-qPCR and expressed as fold-change relative to day 1. (B) Protein extracts were prepared from patient-matched HMECs and vHMECs and analyzed by Western blot with the indicated antibodies. Representative Western blot analysis of CTCF, p16, p53, and actin levels in: HMECs (H); pre-selection, stressed HMECs (sHMECs); and post-selection vHMECs. (C) Western blot analysis of CTCF levels upon genotoxic stress by Dox (0.4μM, 24 hours) or (D) acute oxidative stress by H₂O₂ (100 mM, 30 min) in HMECs (day 1) and MCF10A cells. (E) MCF10A cells were treated with increasing concentrations of H₂O₂ over a period of 20 min to induce oxidative damage. Protein levels of CTCF, ALDH1A3, and actin (as a control) were assessed by Western blotting. (F) HMECs were treated with 0.1 mM H₂O₂ for 24 hours. Protein levels of CTCF, H2A.Z, EZH2, ALDH1A3, p53, Smad 3, RNA Pol II, p16, and actin were measured by Western blotting.

https://doi.org/10.1371/journal.pgen.1009277.g001

CTCF protein levels are downregulated by stress-induced changes in protein stability without significant alteration of CTCF-genome binding in HMECs

To investigate possible mechanisms by which CTCF is so effectively downregulated, we measured CTCF mRNA levels in HMECs during prolonged cell culture stress (vHMECs) or by acute exposure to H₂O₂. Comparing levels of gene expression between HMECs and vHMECs revealed that CTCF and p53 mRNA levels were only slightly changed, whereas other genes were significantly downregulated (p16, RASSF1A) or elevated (c-MYC, HDM2) (Fig 2A). Similarly, oxidative damage induced in HMECs or MCF10A cells resulted in very little change in CTCF mRNA levels (Fig 2B). These results suggest that the decrease of CTCF protein in stressed HMECs and vHMECs is regulated at the levels of protein synthesis and stability. Indeed, treatment of HMECs with the inhibitor MG132 significantly elevated levels of CTCF, indicating a ubiquitin-proteasomal regulation of CTCF protein stability upon exposure to free radicals generated by H₂O₂ or tBHQ, similar to control of p53 protein stability (Fig 2C) [40]. We conclude that upon acute stress, a fraction of CTCF, detectable with the specific antibodies we used, is rapidly downregulated by changes in protein stability.

Since CTCF is a critical regulator of chromatin architecture, we investigated how such a decrease in protein levels of this stress-responsive form of CTCF may affect occupancy of the almost 48,000 CTCF genomic binding sites in HMECs during the stress response [41]. The reduction in cellular CTCF protein levels observed upon stress in normal HMECs suggested that the interaction and function of this regulator with a subset of its numerous genomic sites were likely to be impacted. For example, RNAi-mediated downregulation of CTCF in certain cancer cell lines results in destabilization of chromatin boundary elements and aberrant epigenetic silencing of specific tumor suppressor genes [42]. We therefore expected to see substantial changes in the interaction of CTCF with a subset of its numerous genomic sites in HMECs during the stress response. To this end, a comparative ChIP-seq analysis of genome-wide CTCF-chromatin interactions in HMECs at day 1 and day 10 was performed. Limited changes in global CTCF interactions were observed (0.04% changes at a threshold of adjusted p-value <0.05), despite a significant stress-induced decrease in abundance of CTCF cellular protein in the range of ~120 kDa. As shown in the barplot, the number of differential CTCF peaks in HMECs between day 1 and day 10 was only 17 among 48,000 total peaks (Fig 2D). Comparative signal tracks of CTCF peaks between multiple samples from HMECs at day 1 and day 10 align with each other and with ENCODE data (Fig 2E). Thus, CTCF may facilitate the stress response in normal epithelial cells by modulating its binding to low-affinity genomic sites or through other activities [43]. Interestingly, live cell single molecule imaging studies have shown that ~20–50% of CTCF is not DNA-bound but freely diffusing with rapidly exchanging genome binding that is dynamically regulated throughout the cell cycle [44–47]. Recent studies show that CTCF may be transiently sequestered in small nuclear “zones” through its RNA-binding region to more efficiently locate its genomic binding sites [48]. In normal HMECs, stress signals that rapidly and reversibly modulate CTCF protein stability and nuclear localization may also impact its dynamic, exchanging DNA-interactions to facilitate the cellular damage response.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. CTCF protein levels are downregulated by stress-induced changes in protein stability without significant alteration of CTCF-genome binding in HMECs.

(A) Comparative mRNA levels of the indicated genes in vHMECs (day 40) relative to HMECs (day 1) as analyzed by RT-qPCR. (B) Comparative mRNA levels of CTCF upon H₂O₂ treatment in HMECs and MCF10A cells. (C) Oxidative stress in HMECs was induced with either H₂O₂ treatment for 24 hours or with the indicated concentrations of tBHQ. HMECs were stressed in the presence or absence of the proteasome inhibitor MG132. Protein levels of CTCF, p53, and actin were measured by Western blotting. (D) Chromatin immunoprecipitation ChIP-seq analysis of genome-wide CTCF-DNA binding in HMECs at day 1 and day 10. Shown are number of unique and common CTCF peaks. (E) ChIP-seq sample signal tracks of CTCF peaks demonstrate no changes in binding sites upon stress for the selected area.

https://doi.org/10.1371/journal.pgen.1009277.g002

Proteomic analysis of CTCF interaction partners identifies nuclear speckle- associated proteins involved in RNA processing

To investigate the possible function of stress-responsive CTCF complexes in HMECs, we conducted a mass spectrometry analysis of CTCF protein interaction partners. Since we were unable to isolate sufficient amounts of endogenous CTCF from patient samples, we used spontaneously immortalized, non-transformed mammary epithelial MCF10A cells as a source of CTCF. Similar to HMECs, MCF10A cells show stress-dependent downregulation of CTCF protein and CTCF nuclear localization (Figs 1D, 1E and S2). We planned to compare CTCF complexes in both normal and stressed MCF10A cells. However, due to the significant decrease in CTCF abundance, we were unable to obtain enough CTCF protein from stressed cells for mass spectrometry analysis. Thus, we focused on analyzing CTCF interaction partners in normal MCF10A cells to reveal the variety of interaction partners and potential function of CTCF complexes. Immunoprecipitated CTCF complexes were analyzed by mass spectrometry (Q Exactive mass spectrometer (Thermo) and Integrated Proteomics Pipeline–IP2 (Integrated Proteomics Applications). Identified proteins were filtered using DTASelect [49] with a target-decoy database search strategy to control the false discovery rate to 1%. The list of CTCF interaction partners yielded 314 IDs with a Homo sapiens background (S1 Table). To extract biological features associated with the identified mass spectrometry hits, DAVID Bioinformatics Resources 6.7 (DAVID; NIAID, NIH, http://david.abcc.ncifcrf.gov) was applied to analyze functional annotation clustering. This revealed that CTCF interaction partners are enriched in proteins involved in RNA biology, including RNA binding, RNA splicing, and nuclear export (Fig 3A) as indicated by the top-scoring hits (Fig 3B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Analysis of functional annotation clustering for CTCF mass spectrometry hits.

(A) DAVID Bioinformatics Resources 6.7 was used to analyze functional annotation clustering results from the mass spectrometry hits for CTCF cellular interaction partners, highlighting enrichment of RNA splicing and RNA binding proteins. (B) Table of top-scoring relevant CTCF mass spectrometry hits.

https://doi.org/10.1371/journal.pgen.1009277.g003

Many previous studies have identified numerous RNA binding partners for CTCF and demonstrated their critical role in regulating CTCF-dependent processes [14,17–22]. We also detected RNA binding and RNA splicing proteins (including snRNPs, hnRNPs, serine-arginine proteins) as the most prevalent CTCF interaction partners which predominantly localize to inter-chromatin SC-35 nuclear speckles [50,51]. CTCF has been proposed to regulate alternative RNA splicing through several different mechanisms [24,52–54], some of which may be co-translational in close association with chromatin. Thus, a CTCF-dependent activity at SC-35 nuclear speckles may serve to fine-tune processing of RNA transcribed from genes within its vicinity [50]. CTCF is known primarily to interact with chromatin-associated proteins and these new partners may reveal new functional roles for this diverse regulator.

CTCF forms stress-responsive complexes at SC-35 nuclear speckles in normal HMECs

To explore the potential involvement of CTCF in RNA processing in nuclear speckles (Fig 3), we visualized its sub-nuclear distribution in HMECs using Airyscan confocal microscopy (Fig 4). Our observation of downregulation of CTCF protein levels upon cellular stress prompted us to simultaneously examine whether CTCF nuclear localization was also affected by stress signaling. Previous studies have shown that stress signals can promote growth arrest accompanied by changes in differential CTCF poly(ADP-ribosyl)ation (PARylation) and genomic binding. In normal tissues and breast cancer cells, PARylated CTCF predominates upon cell cycle arrest but transitions to a hypo-PARylated form in proliferating cells, which is also associated with breast cancer progression [55,56]. This important work shows that cellular stress and cell cycle stage transitions can evoke significant changes in CTCF post-translational modifications, which modulates site-specific genomic occupancy and gene expression. We performed imaging in three independent experiments (from three different patient tissue specimens) using unstressed HMECs, HMECs acutely exposed to H₂O₂, or HMECs stressed by prolonged cell culture (vHMECs). We first analyzed co-localization of CTCF with histone H3 and DAPI, as a DNA marker. Intriguingly, this revealed that CTCF was enriched in large nuclear bodies within the inter-chromatin space that is excluded from DAPI and H3 staining (confirmed by a Pearson’s Correlation Coefficient close to zero). By contrast, the minority of CTCF signal was detected outside the inter-nuclear space in a punctate pattern highly associated with histone H3, as expected for genome-associated CTCF (Pearson’s Correlation Coefficient ~0.7) (Fig 4A). However, we cannot rule out interactions between CTCF and H3 at the border due to the signal detection threshold. Consistent with our observation that oxidative stress significantly downregulates a certain pool of CTCF through changes in protein stability (Fig 2C), the large CTCF inter-chromatin bodies disappeared upon exposure to H₂O₂ stress, leaving only the punctate, H3-associated fraction of CTCF detectable (Fig 4A). However, the large CTCF-associated nuclear bodies could be partially restored by inhibiting proteasome-mediated degradation with MG132 and elevating CTCF cellular abundance in stressed HMECs (Fig 4A). Formation of large nuclear clusters by CTCF has been described previously in a pioneering study by Zirkel and colleagues. They showed that decreased expression of the high-mobility group protein, HMGB2, initiates cellular senescence by inducing pronounced spatial redistribution of CTCF into nuclear depots with concomitant genomic reorganization [57]. This and other studies underscore the importance of large-scale changes in CTCF nuclear localization in potentially driving fundamental cellular processes like senescence, the cell cycle, and the cellular stress response.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. CTCF forms stress-sensitive complexes at SC-35 positive nuclear speckles in normal patient-derived HMECs.

(A) HMECs were exposed to H₂O₂ treatment in the presence or absence of the proteasome inhibitor MG132 and stained against anti-CTCF and anti-histone H3 antibodies as well as DAPI to visualize the nuclear DNA. Scale bar = 3 μm, (HMEC untreated: n = 20; HMEC + 50 uM H₂O₂ 24 hours: n = 20; HMEC + 50 uM H₂O₂ 24 hours + MG132: n = 27). (B) HMECs were exposed to cell culture stress, or H₂O₂ treatment, fixed and stained against antibodies to CTCF and nuclear speckle marker SC-35 and DAPI, (HMEC untreated: n = 20; HMEC prolonged TC stress: n = 20; HMEC + 50 uM H₂O₂ 24 hours: n = 23). Samples were visualized with Airyscan microscope and the Pearson’s coefficients for co-localization were determined using Imaris software; 4B showed a significant difference (****, P<0.0001). (C) Histograms demonstrate the presence of CTCF-associated clusters (red) in unstressed HMECs and the loss of CTCF-cluster association in H₂O₂-treated HMECs and in stably stressed vHMECs. By contrast, the distribution of SC-35 in nuclear speckles (“clusters”) (green) remains relatively stable across the cell samples. (D) Example of co-localization of CTCF with SC-35 inside and outside of nuclear speckles. (E) Western blot shows SC-35 co-immunoprecipitation with CTCF (CTCF IP untreated), while no SC-35 was co-immunoprecipitated with CTCF under stressed conditions (CTCF IP H₂O₂-treated). The control co-immunoprecipitation of hnRNP-K (as a protein known to reside in nuclear speckles) with SC-35 and a small amount of CTCF confirmed the SC-35 speckle identity.

https://doi.org/10.1371/journal.pgen.1009277.g004

To further investigate possible functions of stress-responsive CTCF inter-nuclear bodies, we analyzed their co-localization with other known nuclear markers. Nuclear bodies can concentrate factors like proteins, DNA, RNA, and/or lipids in reaction sites for specific biological processes, or segregate these factors away from unwanted sites [33,50,51]. Our previous [58] and current proteomic analyses identified a variety of RNA processing proteins as CTCF interaction partners (Fig 3). These included SC-35, a marker for nuclear speckles, which are inter-chromatin domains enriched in pre-mRNA splicing factors [50,51]. To examine whether stress-responsive CTCF protein co-localized with SC-35 in HMECs, we imaged CTCF both inside and outside of the SC-35-associated nuclear speckles in normal and stressed HMECs. Airyscan microscopy confirmed that the species of CTCF detected by the antibodies we used predominantly co-localized to the SC-35-positive nuclear speckles in normal HMECs (highly significant Pearson’s Correlation Coefficient). Upon chronic (vHMECs) or acute (H₂O₂) stress, CTCF clusters disappeared from the SC-35 nuclear speckles. However, SC-35 association with nuclear speckles remained stable in stressed cells despite the loss of CTCF, indicating that CTCF is not required to maintain the structure of these nuclear bodies (Fig 4B and 4C). A co-localization channel is illustrated for SC-35-positive cluster analysis (Fig 4D). SC-35 is associated within the same complexes as CTCF in HMECs and its interaction is regulated by stress signals, as shown by co-immunoprecipitation experiments (Fig 4E).

Several nuclear bodies have RNA-associated functions, such as mRNA retention, RNA splicing, or RNA export [33,50,51]. To determine whether the CTCF-SC-35 associated nuclear speckles contain RNA compartments, we combined fluorescently-labeled oligo d(T) FISH probes with immune-CTCF staining. Indeed, co-localization of CTCF nuclear speckles with poly(A) RNA was observed. CTCF also co-localizes with SC-35 nuclear speckle-associated proteins [59], PABPN1, which binds to nascent poly(A) tails, and hnRNP-K, an RNA processing protein, but not to NONO, a paraspeckle-specific marker (S2 Fig). Since CTCF has been shown to associate with nucleolar components, like Nucleolin [42], we examined and eventually excluded nucleoli as a co-localization site for the stress-responsive CTCF-associated speckles (S2 Fig). Nuclear speckles are a major nuclear body, yet their function is not clearly understood. Recent genomic mapping data revealed that a significant chromosomal compartment is in close proximity with nuclear speckles, suggesting that RNA processing in nuclear speckles may be tightly coupled to transcribing genes [50].

Taken together, these results implicate a role for CTCF in stress-responsive RNA biology at nuclear speckles in primary HMECs. The fact that SC-35 nuclear localization was unchanged in H₂O₂-treated cells suggests that certain forms of CTCF are specifically targeted for degradation by the stress response. We did not observe CTCF interaction with nuclear speckles in patient-matched fibroblasts derived from the same breast tissue (S1 Fig). Moreover, in the breast cancer cell lines that we examined CTCF is no longer responsive to stress-induced protein downregulation nor enriched at SC-35 nuclear speckles (S1 Fig). To validate our results and exclude the possibility that post-translational modifications (PTM) might obscure CTCF-antibody recognition sites, we used antibodies against two different epitopes of CTCF outside of known PTM sites (N-terminus, Active Motif; C-terminus B-5, Santa Cruz) and observed similar patterns of sub-cellular localization (S3 Fig). This highlights the unique function and biological relevance of CTCF during the stress response in primary human epithelial cells and suggests that this specific role of CTCF is a target of deregulation.

CTCF-RNA interactions in normal and stressed HMECs reveal functional roles in cell development and neuronal differentiation

The observed nuclear co-localization of CTCF with poly(A) RNA and splicing factors at SC-35 nuclear speckles in normal HMECs prompted us to identify stress-responsive changes in the landscape of CTCF-bound RNA targets. To this end, we performed RNA-protein immunoprecipitation (RIP) analyses in HMECs with either CTCF-specific or SC-35-specific (as a positive control for RNA at the nuclear speckles) antibodies and sequenced the RNA recovered from the immunocomplexes (RIP-seq) [60,61]. Input samples were also collected and sequenced for each cell type to determine input over specific RIP. To select for enriched RNAs that specifically associate with CTCF or SC-35 in HMECs but dissociate in vHMECs (correlated with stress-induced downregulation of CTCF at SC-35 nuclear speckles), we compared RNAs in samples obtained by an identical procedure from vHMEC extracts. Our RIP-seq analysis revealed that 360 different RNAs were bound to both SC-35 and CTCF in HMECs (Fig 5A, S2 and S3 Tables). Despite the profound loss of CTCF protein abundance and depletion from the SC-35 nuclear speckles, we identified only 48 genes that no longer interacted with CTCF in vHMECs but still remained bound to SC-35. These data demonstrate that the majority of RNAs that associate with SC-35 and CTCF remain unchanged, despite stress-induced downregulation of a particular CTCF pool and its depletion from nuclear speckles in vHMECs. We cannot rule out that low levels of CTCF in vHMECs remain at the nuclear speckles, sufficient to detect RNA interactions by RIP-seq, or that other forms of CTCF not detectable by the antibodies we used interact with SC-35 outside of the nuclear speckles. Nevertheless, a role for CTCF in specific alternative splicing events, RNA transport or nuclear export in either HMECs or vHMECs remains a possibility.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Stress-sensitive CTCF-RNA interactions that are shared with SC35.

(A) Venn diagram showing the number of unique RNAs interacting with CTCF and/or SC-35 in HMECs compared to vHMECs, based on RIP-seq data. (B) List of top CTCF interactions with RNA in HMECs that were lost in vHMECs, but still remained bound to SC-35.

https://doi.org/10.1371/journal.pgen.1009277.g005

We were also intrigued by the gene ontology (GO) term analysis for RNAs that are significantly bound by CTCF and SC-35 but dissociate upon stress. Surprisingly, this analysis revealed that most protein-coding RNAs that initially interacted with both CTCF and SC-35 in HMECs but lost their CTCF interaction in vHMECs, are involved in cell development/differentiation or neuronal function (Fig 5B). For example, genes involved in brain segmentation and neuronal outgrowth (CDH4) [62], axon guidance (ABLIM2, SEMA4D) [63,64], and cell lineage development (GLI2, TWIST2, UNC45B) [65–67].

Our nervous system is routinely exposed to different sources of oxidative stress, including free radicals from the redox potential of neurotransmitters, high levels of lipids as an oxidation substrate, and an elevated requirement for oxygen, which generates free radicals as by-products [68]. An inefficient response to oxidative stress in the developing brain or in differentiated neurons may compromise the nervous system leaving it vulnerable to neurodegenerative disorders [69,70]. Furthermore, a controlled oxidative stress response can influence the proper development of our nervous system, while the intracellular redox regulation and sensitivity to such stress varies between the embryonal stage, neural progenitor cells, and differentiated neurons [71,72]. This prompted us to investigate potential roles for CTCF in both the regulation of neural development and the response to oxidative stress at SC-35 nuclear speckles.

Stress-responsive CTCF complexes at SC-35 nuclear speckles form during a specific stage of human neuronal differentiation

To gain insight into possible functions of CTCF during neural differentiation, we analyzed its stress-sensitivity and nuclear distribution pattern in human pluripotent stem cell (hPSC)-derived neural progenitor cells (NPCs) and early neurons. Cell identities were assessed using previously established criteria [73]. Using immunofluorescence, the nuclear distribution of endogenous CTCF during neuronal differentiation and targeting to nuclear speckles were determined by co-immunostaining with antibodies against CTCF and SC-35. Airyscan imaging revealed that CTCF was diffusely distributed in the nuclei of both undifferentiated hPSCs and neurons (Fig 6A). Strikingly, in NPC nuclei CTCF associated with large clusters and a strong Pearson’s correlation coefficient was observed for co-localization of CTCF and SC-35, similar to the pattern in unstressed HMECs. CTCF and SC-35 co-localization was significantly reduced in pluripotent and differentiated cells, suggesting that the functional interaction between CTCF and SC-35 nuclear speckles is specific to a particular stage of neural differentiation. We also examined the stress-sensitivity of CTCF nuclear localization at different stages of neural differentiation by visualizing normal and oxidative-stressed cells using Airyscan imaging. Consistent with our previous results in HMECs, stress signals induced profound changes in CTCF nuclear localization only when associated with nuclear speckles in NPCs whereas CTCF distribution in hPSCs and neurons was unaffected. As previously reported, CTCF protein levels detected by the specific antibodies we used markedly decrease as pluripotent hPSCs undergo differentiation into a variety of lineages [74]. We found that the abundance of this species of CTCF protein is high in hPSCs (like HMECs) but decreases in NPCs and is very low in differentiated neurons. Interestingly, stress-induced downregulation of this CTCF species occurs exclusively in NPCs, consistent with our imaging results, showing stress-sensitive CTCF co-localization at SC-35 nuclear speckles (Fig 6B). In pluripotent stem cells and in neurons, the protein abundance of this form of CTCF and its nuclear distribution are stable to stress.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Stress-sensitive CTCF complexes at SC-35-associated nuclear speckles are exclusively found at the neural progenitor cells (NPCs) stage of neuro development.

(A) hPSCs, NPCs, and early neurons were exposed to H₂O₂ treatment and stained with anti-CTCF, anti-SC-35 and DAPI. Samples were visualized with an Airyscan microscope; (NPC: n = 21; NPC + H₂O₂: n = 21). (B) hPSCs, NPCs, and early neurons were exposed to H₂O₂ treatment and CTCF protein levels were analyzed by Western blotting. Developmental markers for different neuronal stages are also shown. (C) Histograms demonstrate loss of CTCF cluster-formation (red) from NPCs to H₂O₂-treated NPCs, whereas the distribution of SC-35-associated clusters (green) remains relatively stable.

https://doi.org/10.1371/journal.pgen.1009277.g006

Taken together, these data suggest that CTCF has the potential to serve a unique functional role which is highly responsive to cellular stress by associating with SC-35 nuclear speckles at a specific stage of neural development. In this capacity, CTCF complexes containing SC-35 and specific RNAs may maintain neural progenitor states through adaptable switches in RNA processing or fine-tuned gene regulation within speckles but rapidly lose this function during the stress response. Several recent studies have revealed important functions for CTCF in nerve regeneration and neuronal differentiation. One study showed that conditional loss of CTCF impaired the regenerative capability of sensory neurons by altering epigenetic programming and the transcriptional response to injury [75]. Intriguingly, CTCF protein abundance was shown to regulate a differentiation switch between myelinogenesis and Schwann cell maturation to promote myelin regeneration and Schwann cell myelination after nerve injury [76]. Underpinning the rapid, functional plasticity of CTCF observed during differentiation and environmental responses, is the finding that CTCF-genomic binding dynamics play a critical role in modulating gene expression frequency through stochastic gene choice to drive differentiation of neuronal progenitors to mature neurons [77].

Ethics statement

Studies using human tissue were approved by the Salk Institute Institutional Review Board (Protocol #10–0005). No human subjects were specifically recruited for our study and all patient specimens were de-identified.

Cell culture and stress induction

Human mammary epithelial tissues were reliably procured from the UCSD Tissue Biorepository and the National Disease Research Interchange (NDRI). HMECs were propagated from reduction mammoplasty tissue from multiple, cancer-free female donors (age range 20–40 years old) chosen in an unbiased fashion. Upon notification from the surgeons and approval from the pathology test, the fresh breast tissue was either shipped on ice or picked up in person within 24 hours of surgery. Breast tissues were immediately documented for appearance, catalogued, and dissected in a laminar flow hood. In sterile 150 mm petri dishes (Thermo Fisher), crude surgery sites, fat areas and fibrotic tissue were discarded, while mammary gland-rich areas were isolated. The latter was further cut into smaller cubical-like sections, using disposable scalpels (Feather) and prepared as described [2,6,7]. In brief, the dissected material was placed into 50 ml Falcon tubes and digested overnight at 37°C in DMEM/F-12 (Invitrogen) supplemented with 10 μg/ml insulin, 10% FBS (Gibco), 200 U/ml crude collagenase (Sigma) and 100 U/ml hyaluronidase on a tube rotator (Fisher Scientific). The digested tissues were centrifuged at 600 x g for 5 minutes at room temperature. First, the upper fat-lipid section was pipetted off into another 50 ml tube for disposal, followed by the supernatant containing mostly media. The resuspended pellet containing the organoid pool was concentrated on top of a 40 μm strainer (Greiner), followed by a quality check under a light microscope. If organoids were not visibly cleaned of connective tissue, another round of digestion was performed. 100 mm tissue culture dishes were pre-rinsed with MEGM media (Lonza), followed by pipetting the organoids drop-wise in a small volume (2–5 ml) of MEGM media and in a spaced manner–leaving at least 2 cm space between organoids–to improve their attachment. Organoids were left for 10 minutes at room temperature to attach. After initial attachment, the 100 mm tissue culture dish was slowly and carefully filled up to 20 ml MEGM media without disrupting the organoid attachment to the tissue culture dish. Organoids were cultured on 100 mm tissue culture dishes in MEGM media (Lonza) for 10 days, until HMEC mitotic outgrowth was achieved. HMECs and vHMECs were maintained in MEGM media (Lonza). MCF-10A cells (ATCC) were cultured in DMEM/ F-12 supplemented with 100 ng/ml cholera toxin, 20 ng/ml epidermal growth factor (EGF), 0.01 mg/ml insulin, 500 ng/ml hydrocortisone, and 5% horse serum. All growth factors were purchased from Sigma. MDA-MB-231 and MCF-7 cells were obtained from ATCC and cultured following ATCC recommendations. hPSCs (EC#11) were derived from human umbilical vein cells (Lonza) as previously described [103] and provided by the Salk Institute Stem Cell Core for subsequent culturing and differentiation into neural progenitor cells (NPCs), and early neurons following previously established methods [73]. In brief, 6 well plates were coated with Matrigel (BD Bioscience) in DMEM-F12 overnight. hPSCs were thawed and seeded in presence of Rock inhibitor in mTeSR1 media (Stem Cell Technologies). Subsequently, media was changed daily in absence of Rock inhibitor. Colonies were monitored for areas of de-differentiation, which were marked and aspired. Once colonies were large and nearly touching in a ~70% confluent plate, colonies were detached with collagenase (Invitrogen) at 37°C until they were floating, and then collected and washed in DMEM/F12 medium. Cells were resuspended in Neural Induction Medium (NIM: containing DMEM/F12 as base, 1 x N2 Supplement (Invitrogen), 1 x B27 supplement (Invitrogen) and transferred to a 10 cm ultra-low-attachment plate. After one week of culturing, cells were transferred to a 10 cm polyornithine/laminin-coated plate in NIM supplemented with 1 μg/ml laminin. After one week, rosettes were manually isolated and plated on polyornithine/laminin-coated wells in Neural Progenitor Medium (NPM: containing DMEM/F12 as base, 1 x N2 supplements, 1 x B27 supplements, 100 ng/ml FGF8 (Peprotech), and 200 ng/ml SHH (R&D). For neural differentiation, cells were dissociated with Accutase (Innovative Cell Technologies) with vigorous pipetting and then plated on polyornithine/laminin plates at 10⁴ cells/cm² in NPM. The next day, the medium was switched to Neural Differentiation Medium (NDM: containing DMEM/F12 as base, 1 x N2 Supplement (Invitrogen), 1 x B27 supplement (Invitrogen), 20 ng/ml BDNF (Peprotech), 20 ng/ml GDNF (Peprotech), 200 nM ascorbic acid (Sigma), 1 μg/ml laminin (Invitrogen). Cells were maintained in NDM for at least 4–6 weeks. Oxidative stress in cultured cells was induced as indicated by the addition H₂O₂ (Thermo-Fisher) to the culture media. Genotoxic stress was induced as indicated by the addition of 400 nM Doxorubicin to the culture media for 24 hours. Importantly, it is critical to use freshly prepared HMECs and low-passage aliquots of MCF10A cells to preserve “normalcy”. Each cell type rapidly undergoes a stress response accompanied by elevated ALDH1A3 and decreased CTCF levels or the appearance of multiple CTCF isoforms simply by freeze-thawing and prolonged tissue culturing.

Antibodies

All antibodies are listed in S4 Table.

Chromatin Immunoprecipitation (ChIP)

ChIPs were performed essentially as described [104] with minor modifications. Unless stated otherwise, all chemicals were purchased from Sigma. Briefly, normal HMECs and stressed HMECs were cultured on 150 mm tissue culture plates until cells were 60–70% confluent. After 1 x PBS wash, cells were crosslinked for 15 minutes at room temperature by addition of formaldehyde (Invitrogen) to a 1% final concentration in 1 x PBS. Crosslinking was stopped upon dropwise addition of glycine to a final concentration of 125 mM for 5 minutes. After washing cells twice with 1X PBS, cells were trypsinized and counted. 2 x 10⁷ cell/ml were collected in RIPA buffer containing protease inhibitors (150 mM NaCl, 1% NP-40, 0.5% Sodium Deoxycholate, 0.1% SDS, 50 mM Tris-HCl pH 8.0, 5 mM EDTA, 0.2 mM NaF, 0.2 mM sodium orthovanadate, 5 μM trichostatin A, and 5 mM sodium butyrate) and incubated on ice for 10 minutes. Lysates were subsequently sonicated on ice four times for 15 sec each with a small probe at medium setting to generate DNA fragments that are smaller than ~500 bp. Samples were centrifuged at 4°C, 13,000 rpm for 10 minutes. The collected supernatant was measured for DNA quality control on a nanodrop spectrophotometer (NanoDrop Technologies Inc.) and protein content was determined with a BCA kit (Pierce). 500 mg of lysate was diluted with 10 volumes of dilution buffer (1% Triton X-100, 150 mM NaCl, 2 mM EDTA pH8, 20 mM Tris pH8, 1X protease cocktail Roche) and pre-cleared by rotating for 1 hour at 4°C using 40 ml of a 50% slurry of 1:1 protein A- and protein G-Sepharose beads (GE Healthcare). Samples were centrifuged for 10 sec at 1500 rpm at 4°C to pellet the beads and remove the supernatant. For immunoprecipitation, 5 μg anti-CTCF (Millipore; 07–729) was added to precleared lysates along with 40 μl of a 50% slurry of 1:1 protein A/G beads that were preblocked with 1 mg/ml BSA and 0.3 mg/ml salmon sperm DNA followed by incubation at 4°C overnight. Immuno-complexes were washed with IP wash buffer (100 mM Tris-HCl pH 8.5, 500 mM LiCl, 1% NP-40, 1% Sodium Deoxycholate) and rotated for 5 minutes at 4°C. Samples were centrifuged as described above and supernatant was removed. Washes were repeated 2 more times with 1X TE buffer. Elution was performed in buffer containing 70 mM Tris-HCl pH 8.0, 1 mM EDTA and 1.5% SDS by incubation at 65°C for 10 or 30 minutes. Crosslinks were reversed from immuno-complexes by addition of 200 mM NaCl and incubation at 65°C for 6 hours or overnight. DNA was purified by incubation with proteinase K and phenol-chloroform extraction and a fraction was used as template in real-time PCR reactions. Primers were designed with Primer Express 2.0. PCR products range in size between 50 and 75 bp (see RT-PCR section for sequences). PCR reactions contained 1× SYBR Green Mix (Applied Biosystems), 1/100 fraction of the ChIP-enriched DNA, and 100 nM primers. Standard curves from 1–200 ng of sonicated genomic DNA was run alongside ChIP samples for each individual primer; input samples were treated identically and were used to subtract/normalize the values from ChIP samples.

Real time PCR (RT-PCR)

For gene expression analysis, adherent cells were washed 1 X PBS, followed by addition of trypsin for 5 minutes in a 37°C cell culture incubator, with frequent checking under the microscope for cell rounding and lifting. Cells were pipetted off the plate with 10% FBS in DMEM and centrifuged to harvest the cell pellet in buffer 1 of the RNeasy kit (Qiagen) (following the manufacturer’s instructions). Next, samples were subjected to DNase (Invitrogen) treatment according to the manufacturer’s instructions. The extracted RNA concentration was measured with a nanodrop spectrophotometer (NanoDrop Technologies Inc.) and adjusted with ddH₂O to a 0.5 mg RNA concentration. cDNA was synthesized from this 0.5 mg of RNA using Superscript III Reverse transcriptase (Invitrogen) according to the manufacturer’s protocol. In brief, PCR tubes containing 0.5 mg RNA were mixed with 1 μl Hexaprimer and 1 μl 10 mM dNTPs and were incubated at 65°C for 5 minutes. The following RT-PCR reaction was carried out as follows: 25°C for 10 minutes, 50°C for 50 minutes, 85°C for 5 minutes. cDNAs were analyzed by RT-PCR using SYBR Green and specific primers. All the selected primers had an annealing temperature of 60°C: CTCF (FWD CCCACGAGAAGCCATTCAAG, Rev GCTGGCATAACTGCACAAAC), p16 (FWD CATAGATGCCGCGGAAGGT, Rev CCCGAGGTTTCTCAGAGCCT), RASSF1A (FWD TATAGCCTGGGCAAGTCCTG, Rev GTACAGGGCGATCCACACTT), PUMA (FWD AGAGGGAGGAGTCTGGGAGTG, Rev GCAGCGCATATACAGTATCTTACAGG), p53 (FWD GAGCTGAATGAGGCCTTGGA, Rev CTGAGTCAGGCC CTTCTGTCTT), c-Myc (FWD GCCATTACCGGTTCTCCATA, Rev CAGGCGGTTCCTTAAAACAA), HDM2 (FWD GGCGATTGGAGGGTAGACCT, Rev CACATTTGCCTGGATCAGCA), and GAPDH (FWD TGCACCACCAACTGCTTAGC, Rev GGCATGGACTGTGGTCATGAG). Primer-specific PCR mix was obtained by diluting Power SYBRGreen MasterMix (Applied Biosystems) 1:1 and primers to a final concentration of 100 nM. Two μl of a 1:200 cDNA dilution was added to each well. For each experiment, samples were prepared in triplicates on MicroAmp plates (Applied Biosystems), while plates were sealed, gently mixed, and centrifuged at 2000 rpm for 1 minutes. Samples were run for 40 reaction cycles in Applied Biosystem 7900 HT Fast RT-PCR System. The curves obtained were quantified with the software SDS 2.4 (Applied Biosystems).

Western blotting and co-immunoprecipitation

For Western blot analysis standard methodologies were used. Cells were trypsinized for 5 minutes at 37°C, washed in 1 x PBS, centrifuged at 3000 rpm for 3 min at 4°C. The pellet was collected in RIPA buffer and snap-frozen in liquid nitrogen for storage at -80°C. Samples were thawed on ice for 45–60 minutes prior to centrifugation at 10,000 rpm for 10 minutes at 4°C. The supernatant was collected, and the pellet was discarded. Protein content was determined using a BCA kit (Pierce). 20 μg protein/sample in a total volume of 25 μl RIPA buffer + 5 μl 5 x SDS loading dye were separated by electrophoresis using 4–12% Bis-Tris Plus Gels (Invitrogen). Transfer was performed on a gel transfer device (Invitrogen) with similar results using the Nitrocellulose or PVDF iBlot Transfer Stacks (Invitrogen). The following transfer program was used: 20V for 1 minutes, 23V for 3 minutes, 25V 4 minutes. Membranes were blocked in 5% skim milk in 20 mM Tris (pH 7.4)-buffered saline (150 mM NaCl) and 0.1% Tween 20 (TBST) for at least 2 hours at room temperature. Membranes were incubated using specific antibodies overnight at 4°C: rabbit anti-CTCF (Active Motif; AB_2614975; 1:1000 dilution); mouse anti-p16 (Sigma; NA29; 1:1000 dilution); rabbit anti-H2A.Z (Active Motif; 39113; 1:1000 dilution); rabbit anti-actin (Sigma; A2066; 1:5000 dilution); mouse anti-p53 (Sigma; DO-1; 1:1000 dilution); rabbit anti-ALDH1A3 (Novus; NBP2-15339; 1:2000 dilution); rabbit anti-Smad2/3 (Cell Signaling; 3102; 1:1000 dilution); rabbit anti-EZH2 (Active Motif; AB_2614956; 1:1000 dilution); rabbit anti-RNA Pol II (Millipore; clone CTD4H8; 1:1000 dilution); mouse anti-SC-35 (Novus; NB100-1774; 1:2000 dilution); mouse anti-hnRNPK (Santa Cruz: D6; 1:500 dilution); mouse anti-Nestin (Abcam; ab22035; 1:500 dilution); mouse anti-beta III Tubulin (tuj1) (Abcam; ab78078; 1:1000 dilution); rabbit anti-Nanog (Abcam; ab80892; 1:1000 dilution); and rabbit anti-Sox2 (Abcam; ab97959; 1:1000 dilution). Membranes were then washed in 1 x TBST for 1 hour at room temperature, with a buffer change every 10–15 minutes. Finally, secondary antibody in the form of either HRP-conjugated anti-rabbit (Cat. No sc-2004), or anti-mouse (Cat. No sc-2005) antibodies (1:10000 dilution) (Santa Cruz) were added to membranes for 1 hour at room temperature, followed by 1 hour washes in 1 x TBST at room temperature, changing buffer every 10–15 minutes. Membranes were incubated with ECL system (Pierce) for 30 seconds– 5 minutes at room temperature, following the manufacturers protocol and visualized with a gel documentation system (Biorad). For co-immunoprecipitation, cells were washed in 1 x PBS, trypsinized at 37°C for up to 5 minutes, collected in 10% FBS in DMEM media, counted and centrifuged at 3000 rpm for 10 min at room temperature. Cells were lysed using twice the volume of whole cell extract lysis buffer (20 mM Tris HCl pH 7.5, 420 mM NaCl, 2 mM MgCl₂, 1 mM EDTA, 10% glycerol, 0.5% NP-40, 0.5% Triton-X-100) for 45–60 minutes on ice. Samples were centrifuged at 13,000 rpm for 10 minutes at 4°C. The supernatant was recovered, and the protein content was determined with the BCA kit. 2 mg lysate were diluted 6X with IP buffer (20 mM Tris HCl pH 7.5, 50 mM NaCl, 10 mM MgCl₂, 2 mM EDTA, 0.5% Triton-X-100) and pre-cleared using 50 ml of protein-G Sepharose beads, rotating the samples for 2 hours at 4°C. To remove the pre-clearance beads, samples were centrifuged at 1,500 rpm for 10 seconds at 4°C and the pre-cleared lysates in the supernatant were transferred into a new tube on ice. To pre-cleared lysates, 2–5 mg of antibody was added and incubated for 1 hour at 4°C. Subsequently, a 50% slurry of protein-G beads was added and further incubated for 2–4 hours. Immunocomplexes were washed with IP buffer containing 0.5% Triton-X-100 thrice and once with IP buffer containing 0.1% Triton-X-100. In between washes, samples were rotated at 4°C, then centrifuged at 1,500 rpm for 10 seconds at 4°C. 2X SDS loading buffer was added to the centrifuged pellet containing beads and immune-precipitated samples, then boiled for 10 minutes at 95°C in a heating block before SDS-PAGE and Western blotting using the antibodies indicated above.

Mass spectrometry

CTCF complexes were immunoprecipitated from three 80% confluent MCF-10A 15 cm dishes with either 2 mg/ml of mouse anti-CTCF (B-5; Santa Cruz) or an IgG control. The resulting samples in the form of centrifuged pellets, containing beads and immune-precipitated samples in 2X SDS loading buffer, were given to the mass spec facility where samples were prepared and analyzed according to established protocols. In brief, samples were precipitated by methanol/chloroform. Dried pellets were dissolved in 8M urea/100 mM TEAB, pH 8.5. Proteins were reduced with 5 mM Tris(2-carboxyethyl)phosphine hydrochloride (TCEP, Sigma-Aldrich) and alkylated with 10 mM chloroacetamide (Sigma-Aldrich). Proteins were digested overnight at 37°C in 2M urea/100 mM TEAB, pH 8.5, with trypsin (Promega). Digestion was quenched with formic acid, 5% final concentration. The digested samples were analyzed on a Q-Exactive mass spectrometer (Thermo). The digest was injected directly onto a 30 cm, 75 μm ID column packed with BEH 1.7 μm C18 resin (Waters). Samples were separated at a flow rate of 300 nl/minute on an nLC 1000 (Thermo). Buffers A and B were 0.1% formic acid in water and 0.1% formic acid in 90% acetonitrile, respectively. A gradient of 1–30% B over 95 minutes, an increase to 40% B over 20 minutes, an increase to 90% B over another 10 minutes and then held at 90% B for 5 minutes of washing prior to returning to 1% B was used for 140 minutes total run time. The column was re-equilibrated with 20 μl of buffer A before injection of sample. Peptides were eluted directly from the tip of the column and nanosprayed directly into the mass spectrometer by application of 2.5 kV voltage at the back of the column. The Q Exactive was operated in a data-dependent mode. Full MS scans were collected in the Orbitrap at 70K resolution with a mass range of 400 to 1800 m/z. The 10 most abundant ions per cycle were selected for MS/MS and dynamic exclusion was used with an exclusion duration of 15 seconds. Hits are listed in S1 Table, sorted by spectral count.

Immunofluorescence

Cells were grown on coverslips (Zeiss; Cat. No 10474379) in 6 well tissue culture plates under normal culture conditions at least overnight, or until treatments were finished. At the end point, media was removed, cells were gently washed with 1 x PBS and fixed with 4% paraformaldehyde in PBS for 20 minutes at room temperature, followed by 5 washes every 10 minutes in PBS at room temperature. Cells were then permeabilized using 0.5% Triton-X-100 in PBS for 10 minutes. Cells were washed twice with 2 x PBS for 10 minutes each. Permeabilized cells were blocked with 4% BSA in PBS for 1 hour at room temperature and incubated with specific antibodies overnight at 4°C in a humidified chamber. In this chamber, 150 μl of antibody solution drops were pipetted on parafilm and inverted coverslips were placed on top, cautiously preventing air bubbles. Handling of the coverslips was best performed with standard surgical pointed forceps. Antibodies were: mouse anti-histone H3 (Millipore; clone 6.6.2; 1:400 dilution); rabbit anti-CTCF (Active Motif; AB_2614975; 1:200 dilution); mouse anti-CTCF (Santa Cruz; B-5; 1:100 dilution); mouse anti-SC-35 (Novus; NB100-1774; 1:200 dilution); rabbit anti-PAPBN1 (Invitrogen; JM11-28; 1:200 dilution); mouse anti-hnRNPK (Santa Cruz, Clone D6; 1:100 dilution); rabbit anti-NONO (generous gift from the Panda lab, Salk Institute); or mouse anti-Nucleolin (Santa Cruz; sc-13057; 1:400 dilution). Coverslips were placed with forceps in 6 well tissue culture plates and washed 5 times for 10 minutes each with PBS. Inverted cover slips were incubated with 150 μl of secondary antibody drops conjugated with Alexa 488 or Alexa 546 fluorophores (Life Technologies; A21202; A21206; 1:400 dilutions) on parafilm, in a humidified chamber and wrapped in aluminum foil for 1 hour at room temperature. Cover slips were placed back with forceps in 6 well tissue culture plates, followed by 5 washing steps in 1 x PBS for 10 minutes at room temperature, each. 150 μl of 1μg/mL of DAPI in PBS drops (Sigma) were placed on parafilm and incubated with inverted coverslips for 5 minutes at room temperature, followed by 3 quick dipping washes in ddH₂O in a 500 ml beaker. Excess water was removed by carefully holding one edge of the cover slip on Kimwipes. Coverslips were mounted on a glass slide using Vectashield (Vector laboratories). The edges of the coverslip were sealed with clear nail polish (Sally Hansen top coat). Slides were stored in a cardboard slide tray overnight at room temperature and at 4°C for long-term storage. All images were acquired using Airyscan mode on a ZEISS 880 LSM Airyscan confocal microscope. Samples were imaged using a 63x 1.4NA objective with a pixel size of 49nm and a z-step size of 159nm. Laser power settings were 2.74 mW (561nm), 1.05 mW (488nm), and 1.11 mW (405nm), with the detector gain set to 600. For all samples, the laser power, detector gain, pixel size, and pixel dwell time (0.66 ms/pixel) were identical. All images were processed with the same “Airyscan parameter” of 7.3.

Poly A RNA in situ immunofluorescence

All buffers and washes were prepared with RNase-free solutions. Cells were grown on coverslips (Zeiss; Cat. No 10474379) in 6 well tissue culture plates at least overnight, or until treatments were finished. At the end point, media was removed, cells were gently washed with 1 x PBS and fixed with 4% paraformaldehyde in PBS for 10 minutes at room temperature, followed by fixation with ice cold methanol for 10 minutes on ice, and a rehydration step in 70% EtOH for 10 minutes on ice. Cells were incubated in 1M Tris pH 8.0 in PBS for 5 minutes at room temperature, followed by two washing steps in 1 x PBS. A 150 μl solution of biotinylated oligo d(T) probes (1:1000, Promega) in hybridization buffer (5% BSA, 10% Dextran sulfate, 25% formamide de-ionized, 1 mg/ml yeast tRNA, in 2 x SSC) was placed on parafilm in a humidified chamber. Inverted coverslips were placed on the probe drops with the help of standard surgical pointed forceps. Samples were incubated for 2 hours at 37°C in a humidified chamber, followed by placing the cover slips back in a 6 well tissue culture plate and washing steps in 4 x SSC in PBS, and 2 x SSC in PBS for 10 minutes each at room temperature. Primary antibody staining with mouse anti-CTCF (Santa Cruz; B-5; 1:100 dilution) in 2 x SSC with 0.1% Triton-X-100 in PBS was carried out as inverted cover slips on 150 μl staining drops on parafilm overnight at 4°C in a humidified chamber. Cover slips were placed back in a 6 well tissue culture plate, and washed in 2 x SSC in PBS for 10 minutes at room temperature and then incubated for 1 hour at room temperature with 150 μl staining drops of Alexa Fluor 568-labeled streptavidin (Molecular Probes; S11226; 1:400 dilution) to detect the biotin-labeled poly d(T) probe; and Alexa Fluor 488 anti-mouse (Molecular Probes; A21202; 1:400 dilution) to detect CTCF, in a humidified chamber at room temperature. Cover slips were placed back in a 6 well tissue culture plate and washed five times with 2 x SSC in PBS for 10 minutes each at room temperature. 150 μl of 1 μg/ml of DAPI in PBS drops (Sigma) were placed on parafilm and incubated with inverted coverslips for 5 minutes at room temperature, followed by 3 quick dipping washes in ddH₂O in a 500 ml beaker. Excess water was removed by carefully holding one edge of the cover slip on Kimwipes. Coverslips were mounted on a glass slide using Vectashield (Vector laboratories). The edges of the coverslip were sealed with clear nail polish (Sally Hansen top coat). Slides were stored in a cardboard slide tray overnight at room temperature and at 4°C for long-term storage. All images were acquired using Airyscan mode on a ZEISS 880 LSM Airyscan confocal microscope. Samples were imaged using a 63x 1.4NA objective with a pixel size of 49nm and a step size of 159nm. Laser power settings were 2.74 μW (561nm), 1.05 μW (488nm), and 1.11 μW (405nm), with the detector gain set to 600. For all samples, the laser power, detector gain, pixel size, and pixel dwell time (0.66 μs/pixel) were identical. All images were processed with the same “Airyscan parameter” of 7.3.

Image quantification

Quantification was performed with Imaris software (Bitplane). Images were preprocessed with background subtraction function. “Clusters” for anti-SC-35 fluorescence signal were defined by surfaces generated with automatic threshold settings in Imaris, which makes use of K-means clustering algorithms to classify each pixel based on their intensity [105]. “Clusters” for CTCF fluorescence signal were defined similarly but with manual thresholding due to the large intensity variation. Pearson’s correlation of CTCF and SC-35 or CTCF and H3 were then measured inside and outside of the clusters. The ratio of CTCF inside:outside of clusters was similarly obtained. All raw and analyzed data together with the detailed creation parameters are hosted in Zenodo: https://doi.org/10.5281/zenodo.4081952.

Statistical analysis

Statistical analyses were performed using Prism (GraphPad Software). The data distribution of all control and experimental groups were first examined. In analyses where all groups passed the D’agostino & Pearson normality test, we used ANOVA and Tukey’s multiple comparisons tests. Otherwise, the non-parametric Kruskal-Wallis and Dunn’s multiple comparisons tests were used. Paired t test was used for two-condition comparison (Fig 4). Normally distributed data were shown as mean ± standard error of the mean (SEM), while median [interquartile range] was reported for the rest. Outliers from intensity inside:outside of clusters were removed using robust regression and outlier removal (ROUT) method with a false discovery rate of 1% permitted [106].

CTCF and SC-35 based RNA-immunoprecipitation sequencing (RIP-seq)

Briefly, normal HMECs and stressed HMECs were cultured in 150 mm tissue culture dishes under normal tissue culture conditions at 37°C. Cells were washed in 1 x PBS, followed by a crosslinking step for 15 minutes at room temperature by addition of 0.1% formaldehyde in 1 x PBS. Crosslinking was stopped upon dropwise addition of glycine to a final concentration of 125 mM for 5 minutes at room temperature. After washing with 1X PBS, cells were scraped and collected in RIPA (150 mM NaCl, 1% NP-40, 0.5% Sodium Deoxycholate, 0.1% SDS, 50 mM Tris-HCl pH 8.0, 5 mM EDTA) containing protease inhibitors into a 15 ml collection tube on ice. Lysates were subsequently sonicated four times for 15 sec with a small probe at medium setting on ice. Protein concentration was measured with a BCA kit (Pierce) and 500 mg of lysate was pre-cleared by rotating samples for 1 hour at 4°C using 40 μl of a 50% slurry of 1:1 protein A- and protein G-Sepharose beads (GE Healthcare). Samples were centrifuged for 15 seconds at 1500 rpm at 4°C and the supernatant with the precleared lysate was transferred into a fresh collection tube. For immunoprecipitation, 2 mg/ml anti-CTCF (Active Motif; AB_2614975), or mouse anti-SC-35 (Novus; NB100-1774) was added to precleared lysates together with 40 ml of a 50% slurry of 1:1 protein A/G beads and incubated at 4°C overnight. Samples were centrifuged for 15 seconds at 1500 rpm at 4°C and the pellet containing beads with immuno-complexes were washed with IP wash buffer (100 mM Tris-HCl pH 8.5, 500 mM LiCl, 1% NP-40, 1% Sodium Deoxycholate) for 1 h every 15 minutes at 4°C. After immunoprecipitation, CTCF- or SC-35-associated RNAs were isolated from the immunoprecipitate according to the manufacture's protocol (Wako). After elution from the beads, RNA was prepared using a miRNeasy kit (Qiagen). To remove possible genomic DNA contamination, RNase-free DNase was used during the RNA purification steps. RNA concentration was determined using a Nanodrop spectrophotometer (NanoDrop Technologies Inc.) and its integrity was ascertained by migration on a 2% agarose gel and analyzed by displaying 28S and 18S rRNA. The RNAs were analyzed using small RNA deep sequencing (summarized in S2 Table).

Bioinformatic analysis

Libraries were sequenced by generated 100 bp paired-end reads on HiSEq. 2000 (Illumina). Sequenced reads were quality-tested using FASTQC [107] and aligned to the hg19human genome using the STAR [108] version 2.5.1b. Mapping was carried out using default parameters (up to 10 mismatches per read, and up to 9 multi-mapping locations per read). The genome index was constructed using the gene annotation supplied with the hg19 Illumina iGenomes collection and sjdbOverhang value of 100. Raw gene expression was quantified across all gene bodies, using the top-expressed isoform as proxy for gene expression, and differential gene expression was carried out using the DESeq2 [109] package version 1.14 using replicates to compute within-group dispersion. Differentially bound transcripts were defined as having a false discovery rate (FDR) <0.05 and a log2 fold change >1 when comparing two experimental conditions. Transcripts showing sc35-specific binding and CTCF-specific binding were defined as those significantly upregulated in the sc35 RIP or CTCF RIP compared to the control RNA-seq, respectively. Transcripts showing both significant binding of CTCF and SC-35 in HMECs but no significant binding in the vHMECs were annotated with HOMER [110] and tested for term overrepresentation using WebGestalt [111]. ChIP-seq analysis was carried out using HOMER findPeaks and mergePeaks subroutines using default parameters (four-fold enrichment over input control, four-fold enrichment over local tag count, Poisson p-value < 0.001, and style factor). Peaks common to replicate 1 and replicate 2 at day 1 or day 10 were kept and merged into a common peak file. Raw read counts were then assigned to peaks using HOMER annotatePeaks using all replicates from both days. Differential CTCF peaks were found using raw counts of merged peaks from both day 1 and day 10 HMECs with DESeq2 using lenient parameters of FDR < 0.05 and log2fold > 1. Normalized read densities were visualized using the UCSC genome browser [112]. Data was submitted to GEO (GSE139886). A summary of the read statistics for ChIP-seq and RIP-seq is shown in S3 Table.