Paxillin-dependent regulation of IGF2 and H19 gene cluster expression

Multicellular organisms are composed of many different cell types that become organized into distinct tissues and organs during development. Focal adhesions are among the key structures required for these cellular rearrangements, because they mediate the transfer of information about the cellular environment and are essential components of the response mechanism (Zamir and Geiger, 2001). A number of proteins are involved in focal adhesion regulation. Interestingly, focal adhesion proteins can be found not only at focal adhesion contacts but also inside the nucleus and they can shuttle out of it and back in (Hervy et al., 2006). This allows them to participate in a crosstalk between the nucleus and the cytoplasm by converting the extracellular signals into altered gene expression. Given that focal adhesion proteins play an important role in crucial cellular processes, including cell proliferation, survival and differentiation (reviewed in Hervy et al., 2006), their occurrence in the nucleus suggests that they are involved in a new mechanism for regulation of these processes. However, nuclear functions of adhesion proteins are just beginning to be understood.

The adaptor protein paxillin is one of the major components of focal adhesions. Its principal function is a coordination and propagation of downstream signaling from transmembrane integrin and growth factor receptors (reviewed in Brown and Turner, 2004). Paxillin has also been shown to shuttle between focal adhesions and the nucleus. The block of CRM1-dependent export pathway causes accumulation of paxillin inside the nucleus (Woods et al., 2002). The shuttling is regulated by phosphorylation and dephosphorylation because nuclear localization of paxillin requires mitogen-activated protein kinase (MAPK)-dependent phosphorylation (Sen et al., 2012), whereas its export from the nucleus requires dephosphorylation (Dong et al., 2009).

Paxillin has been recently linked to the regulation of H19 gene. Overexpression of paxillin downregulates the expression of H19 in mouse 3T3 cells and directly suppresses the mouse H19 promoter (Dong et al., 2009). This gene produces a 2.3-kb long, capped, spliced and polyadenylated non-coding RNA (Brannan et al., 1990; Milligan et al., 2002). The first exon of H19 RNA encodes two conserved microRNAs (miRNAs), miR-675-3p and miR-675-5p, that are proposed to be responsible for proliferation-repressive function of H19 (Mineno et al., 2006; Cai and Cullen, 2007; Keniry et al., 2012). The H19 and insulin-like growth factor (IGF2) gene form a reciprocally imprinted cluster (IGF2/H19) located on the human chromosome 11 (Reik and Walter, 2001). Imprinting restricts the expression of these genes to only a single allele. H19 expression is restricted to the maternal allele, whereas IGF2 is transcribed only from the paternal one (reviewed in Bartolomei and Ferguson-Smith, 2011). In addition, paternal expression of IGF2 and maternal expression of H19 are mechanistically coupled (Ratajczak, 2012). The current model of the imprinting mechanism includes an imprinting control region (ICR) positioned between the two genes, an enhancer located downstream of both of them, and long-range chromosomal interactions orchestrated by a cohesin complex and a CCCTC-binding factor (CTCF; reviewed in MacDonald, 2012). The zinc-finger insulator protein CTCF binds to the maternal unmethylated ICR and blocks the access of the enhancer to the IGF2 promoter (Bell and Felsenfeld, 2000; Hark et al., 2000). Paternal methylation of the H19 ICR inhibits CTCF binding, thus allowing the enhancer to activate the IGF2 promoter on the paternal chromosome (Murrell et al., 2004; Kurukuti et al., 2006). Maintaining this imprinting pattern is crucial for cell growth and development (reviewed in Ishida and Moore, 2013).

The transcription of the IGF2/H19 locus is further controlled by an evolutionarily conserved cohesin complex (Parelho et al., 2008; Wendt et al., 2008; Nativio et al., 2009) composed of four core subunits, SMC1A, SMC3, SCC1 (also known as RAD21) and SCC3 (also known as SA2 and STAG2) (Guacci et al., 1997; Michaelis et al., 1997; Losada et al., 1998). These proteins assemble in a ring-like structure (Haering et al., 2002), topologically entrapping DNA strands as a ring (Haering et al., 2002; Gruber et al., 2003). Cohesin (along with CTCF) regulates higher order chromatin conformation at the IGF2/H19 locus, forming distinct intrachromosomal loops (Nativio et al., 2009; reviewed in MacDonald, 2012). In addition, cohesin along with the protein complex known as mediator of RNA polymerase II (hereafter mediator) has been shown to mediate long-range looping between distal enhancers and the pluripotency-regulated genes (Kagey et al., 2010), which is important for maintenance of their expression (Kagey et al., 2010; Conaway and Conaway, 2011). However, the link between paxillin and IGF2/H19 transcription regulators has remained elusive. Our study expands on the current understanding of the role of paxillin in the expression of H19 and its functional antagonist IGF2, and demonstrates that paxillin regulates long-range chromosomal interactions formed between the promoters of the active IGF2/H19 alleles and the enhancer, and thus mediates the expression of the IGF2/H19 gene cluster. Finally, we show that the interaction of paxillin, cohesin and mediator plays a role in this regulation.

Overexpression of paxillin in mouse cells has been shown to block H19 expression (Dong et al., 2009). To explore the role of human paxillin in the expression of H19, we depleted paxillin from HepG2 cells by lentiviral transduction of short hairpin RNA (shRNA) directed against human paxillin. As expected, paxillin shRNA (shPXN) reduced the level of paxillin transcripts (Fig. 1C) and the protein level (Fig. 1C) to ∼20% of controls. Importantly, in accordance with published data (Dong et al., 2009), the depletion of paxillin resulted in upregulation of H19 transcription by approximately twofold (Fig. 1A) compared to control cells. Three different clones of shPXN were tested with similar results. The clone with the highest knockdown efficacy was selected for further experiments.

Genes H19 and IGF2 form a gene cluster on human chromosome 11 and share DNA as well as protein regulatory elements (Wendt et al., 2008; Yao et al., 2010). Therefore, we also tested the effect of paxillin depletion on IGF2 expression in HepG2 cells. We detected a slight decrease in IGF2 expression, but the difference was insignificant (P>0.05, Fig. 1A).

H19 and IGF2 genes have the opposite effect on cell proliferation; IGF2 encodes a growth-promoting peptide hormone, whereas the H19 non-coding RNA gives rise to proliferation-repressing miRNAs (Cai and Cullen, 2007; Gabory et al., 2010; Dey et al., 2014). We thus wondered how paxillin depletion in HepG2 cells would affect cell growth, and whether differences in cell growth would reflect the expression changes of proliferation-regulatory genes. Proliferating cells were identified by incubation with the synthetic thymidine analog BrdU, which is incorporated into newly synthesized DNA during S phase. A BrdU assay showed that the knockdown of paxillin and subsequent H19 upregulation resulted in fewer cells incorporating BrdU within DNA (Fig. 1B), that is, undergoing replication. Thus, consistent with the growth-suppressing function of H19, when paxillin level is low the proliferation of HepG2 cells is accordingly slower.

To investigate whether the change in expression of H19 following the paxillin knockdown is a consequence of deregulated imprinting, we examined allele-specific expression of H19 and IGF2. To distinguish between the alleles, we employed single nucleotide polymorphism (SNP) assays in SAOS2 cells, where the the IGF2/H19 polymorphisms in genomic DNA are heterozygous (Fig. 2A,B; also see Discussion). RNA isolated from control and paxillin-knockdown cells was reverse transcribed into cDNA, the region covering SNP rs680 for IGF2 and rs2839704 for H19 was then amplified by PCR, and digested with ApaI or RsaI for IGF2 SNP or H19 SNP, respectively (Fig. 2C). The analysis clearly showed that H19 was still expressed only from the maternal allele both in control and paxillin-knockdown SAOS2 cells, thus excluding the hypothesis that the lack of paxillin affects imprinting of H19. IGF2 showed biallelic expression in both control and paxillin-knockdown cell lines. Thus, IGF2 imprinting is deregulated regardless of paxillin expression and independently of H19.

The monoallelic expression of H19 is further controlled by methylation of the ICR lying upstream of H19. To test whether paxillin plays a role in ICR methylation, we tracked changes in the ICR methylation pattern after paxillin-knockdown in SAOS2 cells. Isolated genomic DNA was treated with bisulfite, and the region corresponding to the ICR, covering 13 CpG islands (as specified in supplementary material Fig. S1A), was amplified by PCR, cloned into a vector and sequenced. The selected region contained a polymorphic site rs2071094 making it possible to differentiate between alleles (parental origin is unknown). Therefore, we then analyzed the methylation status of each allele. Consistent with the observation that paxillin had no effect on the allelic expression of IGF2/H19 (Fig. 2C), ICR methylation was properly maintained in the control sample as well as in the paxillin-knockdown cells, and no gain of methylation on the second allele was detected (supplementary material Fig. S1B). Thus, we conclude that the ICR imprinting barrier of the IGF2/H19 cluster is unaffected by paxillin knockdown.

To explore the effect of paxillin on IGF2 and H19 promoter activity, we constructed luciferase reporter vectors. IGF2 and H19 promoters (Fig. 3A) from HepG2 genomic DNA were inserted upstream of the Firefly luciferase gene (pH19P-luc, pIGF2P3-luc; Fig. 3B). Whereas H19 is transcribed from a single promoter, the expression of IGF2 is controlled by four different promoters (P1, P2, P3, P4; Sussenbach, 1989; Holthuizen et al., 1990; van Dijk et al., 1991). As we found that 95% of IGF2 RNA was transcribed from P3 in HepG2 cells (data not shown), we used the P3 promoter in further experiments. Then, we examined luciferase activity in paxillin-knockdown as well as in control HepG2 cells. Surprisingly, no change in luciferase activity was detected after paxillin knockdown (Fig. 3C), suggesting that promoter activity was unaffected by reduced paxillin level.

The expression of IGF2/H19 genes is mediated by a distant enhancer located downstream of the H19-coding region (Fig. 3A). Both genes, IGF2 and H19, compete for this enhancer (Leighton et al., 1995a; Kaffer et al., 2000). We adopted the construction of the reporter plasmids for testing of enhancer-stimulated expression described by Ishihara et al. (Ishihara et al., 2006). We added a 5-kb-long enhancer region (+6033 to +10,972 relative to H19 transcription start site) into the luciferase reporter vectors downstream of the luciferase gene, mimicking its position relative to the IGF2/H19 genes (pH19P-luc-E, pIGF2P3-luc-E; Fig. 3B), and measured luciferase activity in paxillin knockdown as well as control HepG2 cells. Strikingly, paxillin knockdown resulted in an approximately threefold increase in luciferase activity when driven by the H19 promoter-enhancer, whereas a 0.8-fold decrease was observed when it was driven by the IGF2 P3 enhancer (Fig. 3C). Interestingly, this result is consistent with the observed expression changes in expression of the respective endogenous genes in paxillin-knockdown cells (see Fig. 1A).

Thus, the lack of paxillin increases the activity of H19 promoter through its enhancer and decreases the activity of the IGF2 promoter. In other words, normal paxillin levels are likely to have a suppressive effect on the H19 promoter and a positive effect on the IGF2 promoter. Importantly, paxillin does not regulate these promoters directly, but through the distant shared enhancer.

Next, we investigated whether the regulation of IGF2/H19 promoters by paxillin occurs through an interaction of paxillin with the regulatory DNA elements of both genes, including the IGF2 and H19 promoters, and the shared enhancer or ICR. Therefore, we performed chromatin immunoprecipitation (ChIP) to detect paxillin–DNA complexes in HepG2 cells. ChIP analysis revealed that paxillin mostly interacted with the H19 promoter. The interaction of paxillin with IGF2 P3 promoter and the shared enhancer was weaker, and no binding to the ICR region was detected (Fig. 4B). For the enhancer sequence (+6033 to +10,972 bp), we used three different primer sets (A, B and C in Fig. 4A) to cover the whole region. Set A amplified the region at the 5′ end (+6189 to +6363) of the enhancer, set B the region approximately in the middle (+7814 to +8038) and set C the region close to the 3′ end (+9978 to +10,217). Paxillin was enriched only at the region amplified by set A, suggesting that paxillin interacts with the very beginning of the enhancer sequence (Fig. 4B). Taken together, the results indicate that paxillin binds both key transcription elements – promoter and enhancer – and, therefore, might regulate their mutual interaction.

As we have shown that paxillin regulates the activation of both promoters through the shared enhancer and also occupies these regulatory elements, we speculated that paxillin might stimulate formation of long-range chromatin loops between the IGF2 P3 or H19 promoter and the enhancer. To test this hypothesis, we used chromatin conformation capture (3C) technology to study physical interactions between chromosomal regions (Dekker et al., 2002). Instead of mapping all the interactions at the IGF2/H19 locus, we focused on comparing the interaction frequency between the IGF2 P3 or H19 promoter and the enhancer in control and paxillin-knockdown HepG2 cells. The interaction frequency was examined using the promoter as an anchor and the enhancer as a test region (schematically shown in Fig. 5A). The intensity of the amplified band corresponds to the interaction frequency between the tested regions (supplementary material Fig. S2). In agreement with the previous data, the interaction frequency between the H19 promoter and the enhancer increased in the paxillin-knockdown cells (Fig. 5C). Correspondingly, the interaction frequency between the IGF2 P3 and the enhancer was significantly lower after paxillin knockdown (Fig. 5B). In both cases, the interaction was detected at the 5′ end of the enhancer sequence. This result clearly confirms that paxillin does indeed regulate the activation of the IGF2 P3 and H19 promoters through the same enhancer, and that it does so in an opposite manner. In other words, paxillin promotes a contact of the IGF2 P3 promoter with the enhancer while simultaneously blocking the activation of the H19 promoter.

Finally, we searched for protein complexes involved in the regulatory action of paxillin exerted on the IGF2/H19 promoters. We established a HEK293 cell line stably expressing EGFP–STrEP-paxillin (PXN) and identified proteins pulled down by the tagged protein by liquid chromatography tandem mass spectrometry (LC-MS-MS) (Fig. 6A). Proteins identified in the experiment are summarized in supplementary material Table S1. In addition to many known paxillin-binding proteins, our search also identified several new nuclear gene expression regulatory proteins that interacted with paxillin. Two proteins of cohesin family were repeatedly identified in paxillin pulldown experiments, namely the structural maintenance of chromosomes 1A protein (SMC1A) and the structural maintenance of chromosomes 3 protein (SMC3). Furthermore, we also identified three subunits of the mediator complex (MED), specifically MED15, MED24 and MED23. Although the number of peptides found was very low, no peptides were detected in the control indicating that the interaction is specific.

The specificity of the identified interactions was confirmed by western blotting (Fig. 6B). Both SMC1A and SMC3 proteins, as well as MED23 were found in the paxillin pulldown fraction, but not in the control, which is consistent with our mass spectrometry data. The low incidence of other mediator subunits in the pulldown prevented their identification on western blots. However, no CTCF was found in the pulldown fractions, showing that paxillin is not present in the cohesin–CTCF complexes. Thus, we present the evidence for the interaction between paxillin and gene expression regulatory proteins SMC1A, SMC3 and the mediator complex.

Next, we asked whether the cohesins and mediator complex bind to the same regions on the IGF2/H19 promoters and the enhancer as paxillin. The cohesin subunit SMC1A and the mediator subunit 23 (MED23) were immunoprecipitated from crosslinked chromatin of HepG2 cells and their occupancy on the H19–IGF2 regulatory elements tested. We found that both SMC1A and MED23 were enriched on the IGF2 promoter, the H19 promoter and specifically on the 5′ end of the enhancer, similar to paxillin (Fig. 7A). Finally, we examined whether knockdown of paxillin affected the binding of cohesin and mediator to these DNA elements. The depletion of paxillin by shRNA (shPXN) significantly decreased the amount of SMC1A and MED23 proteins on the enhancer (Fig. 7B). In accordance with our 3C data (see Fig. 5A–C), we detected less SMC1A and MED23 on the IGF2 promoter (P3) (Fig. 7B), and more SMC1A (but not MED23) on the H19 promoter (Fig. 7B). Paxillin depletion did not change the binding of SMC1A to the ICR (Fig. 7B) or the mRNA levels of SMC1A and MED23 (data not shown).

Paxillin belongs to a class of focal adhesion proteins that shuttle between focal contacts and the nucleus. Its nuclear localization has been linked to the regulation of gene expression, especially that of the H19 gene (Woods et al., 2002; Dong et al., 2009; Sen et al., 2012). Nevertheless, the mechanism of the gene cluster regulation as well as the IGF2/H19 imprinting mechanism remain unknown.

In this paper we demonstrate, consistent with previous findings (Dong et al., 2009), that the depletion of paxillin upregulates H19 gene expression in human HepG2 cells. Surprisingly, we did not detect a significant change in IGF2 expression. This was unexpected because the H19–IGF2 genes usually react reciprocally towards the regulating signals, including the depletion of CTCF or cohesin (Wendt et al., 2008; Nativio et al., 2009). Furthermore, depletion of these proteins also abrogates ICR methylation, thus disrupting monoallelic expression (Constancia et al., 2000; Wendt et al., 2008; Nativio et al., 2009). In the case of paxillin depletion, however, the ICR methylation was intact and the allelic expression of H19 unaffected. Interestingly, whereas H19 was transcribed from a single allele, IGF2 was biallelicaly transcribed in both control and in paxillin-depleted cells. As the ICR insulation barrier responsible for IGF2 imprinting was intact, there has to be another regulatory mechanism mediating this process, or alternatively, the imprinting mechanism for IGF2 is dysfunctional. Indeed, reactivation of IGF2 expression on the maternal allele has been previously reported in a number of human tumors and tumor cell lines (Li et al., 1995; Singer et al., 1995; Takeda et al., 1996; Zhang et al., 1997; Cui et al., 1998; Hofmann et al., 2002). However, IGF2 has been also found to be expressed biallelically during normal development: in the choroid plexus and leptomeninges of both mouse (DeChiara et al., 1991) and man (Ohlsson et al., 1994), and postnatal human livers (Kalscheuer et al., 1993; Ohlsson et al., 1994; Davis, 1994). Thus, IGF2 expression might be controlled by diverse regulatory mechanisms. Unfortunately, we could only examine the ICR methylation status and allele-specific expression in the SAOS2 cell line given that HepG2 cells are homozygous and, therefore, allelic expression cannot be distinguished by SNPs. However, we showed that depletion of paxillin in SAOS2 cells resulted in the same changes in H19–IGF2 expression (supplementary material Fig. S3). The fact that similar changes occur in both cell lines after paxillin depletion suggests that the conclusions drawn from experiments using SAOS2 cells can also be applied to HepG2 cells.

Next, we showed that paxillin occupies promoters of IGF2 and H19 genes and their shared endodermal enhancer. Two different enhancer sequences for IGF2/H19 have been identified so far, endodermal and mesodermal. Whereas the endodermal enhancer has been described in both mice and humans (Leighton et al., 1995b; Kopf et al., 1998; Ohana et al., 1999), the mesodermal one is more elusive and a putative mouse mesodermal enhancer region at 22–28 kb downstream of H19 (Ishihara et al., 2000) has not been studied in human cells. In our experiments, we used the endodermal enhancer (+6033 to +10,972 bp relative to H19; delineated by SacI restriction sites). This sequence was originally identified in mice (Yoo-Warren et al., 1988; Arney, 2003) and later reported to stimulate the H19 promoter in human endodermal cell lines such as HepG2 (Kopf et al., 1998; Ohana et al., 1999; Long and Spear, 2004; Varrault et al., 2006).

Using luciferase reporter plasmids, we showed that paxillin regulates the effect of the endodermal enhancer not only on the H19 promoter but also on the IGF2 promoter. As expected, the activity of the H19 promoter increased after paxillin knockdown in accordance with the upregulation of H19 expression in this genomic context. Interestingly, we observed decreased activity of the IGF2 promoter. Thus, paxillin regulates both genes and has an opposite effect on the activity of the IGF2 and H19 promoters. Interestingly, nuclear accumulation of paxillin upon treatment with leptomycin B had a similar effect – downregulation of H19 expression and upregulation of IGF2 expression (supplementary material Fig. S4). This suggests that it is not the total levels but rather the nuclear levels of paxillin that are more important for its nuclear function. As the expression of the IGF2/H19 genes is determined not only by paxillin, but also by cohesins, CTCF, vigilin and ZAC1 (Varrault et al., 2006; Wendt et al., 2008; Nativio et al., 2009; Liu et al., 2014), the ratio of the regulatory proteins is apparently crucial for the resultant expression of the IGF2 and H19 genes. Thus, elevated levels of paxillin in the nucleus (regardless of the total levels of paxillin) can prevail over other regulatory pathways. Further examination of the proteins regulating the IGF2/H19 gene cluster and especially their mutual crosstalk should bring more insight on the mechanism.

We then demonstrated that paxillin regulates the formation of long-range interactions between the promoters of the H19 and IGF2 genes and their shared enhancer. Interestingly, we identified an interaction of paxillin with the cohesin complex (subunits SMC1A and SMC3), which has been established as the main regulator of long-range chromatin interactions at the IGF2/H19 cluster (Wendt et al., 2008; Nativio et al., 2009). Moreover, cohesin has been found to colocalize with the mediator complex (Kagey et al., 2010). These proteins have been shown to mediate long-range interactions between enhancers and promoters of key pluripotency transcription factors in mouse embryonic stem cells (Conaway and Conaway, 2011). In this process, the ring-shaped structure of cohesin is employed to lock together the enhancer and the promoter regions of a single chromatid once they are brought into proximity by their simultaneous binding to the activator mediator (Hadjur et al., 2009). We found cohesin, as well as subunits of mediator, specifically its tail module (Malik and Roeder, 2010), among the paxillin-associated proteins in pulldown experiments. Furthermore, cohesin and mediator bound to the IGF2/H19 promoters and the 5′ end of the enhancer, where paxillin is also enriched, and importantly, the depletion of paxillin changed their amount on these DNA elements. These data suggest that SMC1A and MED23 play a role in paxillin-dependent regulation of the H19–IGF2 gene cluster. Here, we propose a model (Fig. 8) where paxillin assembles with cohesin and mediator, and mediates the long-range chromosomal interactions between IGF2 or H19 promoter and the shared distal enhancer, thus regulating their transcription. Specifically, paxillin enhances the interaction between the enhancer and the IGF2 promoter, but blocks the H19 promoter–enhancer interaction, resulting in stimulation of IGF2 and suppression of H19 expression. Our model also explains the significant decrease in cell proliferation after paxillin depletion, as it causes upregulation of H19 non-coding RNA, thus giving rise to proliferation-repressing miRNAs (Gabory et al., 2010). Both IGF2 and H19 genes are widely expressed during embryonic development, after which they are downregulated (except in skeletal muscle; Brunkow and Tilghman, 1991; Delaval and Feil, 2004). Their proper expression is therefore crucial especially during fetal development (Delaval and Feil, 2004; Pannetier and Feil, 2007; Gabory et al., 2010), and even though their regulatory mechanisms are preserved in cultured somatic cells, we assume that the key significance of the regulatory role of paxillin lies in development.

Based on our findings, we propose a new regulatory mechanism for the IGF2/H19 locus employing paxillin in a complex with cohesin and mediator; this complex mediates promoter–enhancer interactions. Similar interactions have also been observed in androgen-receptor-dependent genes (Wang et al., 2005). Considering that the androgen-receptor–paxillin complex occupies promoters of androgen-receptor-responsive genes, such as PSA (also known as KLK3) or NKX3-1 in prostate cancer cell line (Sen et al., 2012), it is worth exploring whether paxillin is employed here in a similar way. Given the fact that paxillin is a focal adhesion protein, this shows that cells might regulate gene expression by taking advantage of the physical barrier formed by the nuclear envelope and thus controlling access of regulatory factors to their target genes.

SAOS2 and HEK293 cells were grown in Dulbecco's modified Eagle's medium (DMEM), HepG2 cells in RPMI-1640 medium, both of them supplemented with 10% fetal bovine serum (FBS) and antibiotics. Cells were incubated at 37°C in a humidified 5% CO₂ and air atmosphere. Plasmid transfections were performed using Lipofectamine 2000 (Invitrogen, Carlsbad, CA) according to manufacturer's protocol. Stable cell lines were established by selection with geneticin G-418 (800 µg/ml, Gibco, Life Technologies, Carlsbad, CA) or puromycin (2 µg/ml, Gibco, Life Technologies, Carlsbad, CA).

The following antibodies were used in the present study: rabbit monoclonal against the N-terminal of paxillin (Merck Millipore, Billerica, MA); rabbit polyclonal anti-DRIP130 antibody (MED23, Abcam, Cambridge, UK); rabbit polyclonal anti-SMC1A antibody (ChIP Grade, Abcam); goat polyclonal anti-SMC3 antibody (Santa Cruz Biotechnology, Inc., Dallas, TX); rabbit polyclonal anti-CTCF antibody (ChIP Grade, Abcam); rabbit polyclonal anti-H3K4me2 antibody (Merck Millipore); rabbit polyclonal anti-GFP (Molecular Probes, Life Technologies, Carlsbad, CA); mouse monoclonal anti-actin antibody (Sigma-Aldrich, St Louis, MO); and mouse monoclonal anti-BrdU antibody (Sigma-Aldrich).

The pSTrEP-GFP-PXN vector was prepared by ligation of full-length human paxillin cDNA (U14588.1; a generous gift from Ravi Salgia, The University of Chicago School of Medicine, Chicago, IL) into pSTrEP-EGFP-C3 previously prepared by insertion of the One-StrEP-tag sequence (IBA GmbH, Göttingen, Germany) into pEGFP-C3 (Dzijak et al., 2012). Firefly luciferase reporter vectors pH19P-luc and IGF2P3-luc were constructed by inserting the H19 promoter (−819 to +13 bp relative to H19 transcription start site) or IGF2 promoter 3 (−499 to +13 bp relative to IGF2 transcript 3 transcription start site; +7858 to +8369 bp relative to the IGF2 exon 1 start) into pGL4.10 (Promega, Madison, WI) using the KpnI and HindIII sites. The H19 endodermal enhancer sequence (+6033 to +10,972 bp) was inserted into the SalI site of pH19P-luc or pIGF2P3-luc to create pH19P-luc-E or pIGF2P3-luc-E, respectively.

The lentiviral particles containing shRNAs were generated in HEK293T cells using the Addgene (Cambridge, MA) two-plasmid system (pMD2G, psPAX2) and pLKO.1 vectors containing anti-paxillin shRNA (shPXN) or non-targeting shRNA (shNON; Non-Target shRNA Control, Sigma-Aldrich), according to the manufacturer's protocol. Five clones of shPXN were purchased and screened for knockdown efficiency and the TRCN0000123136 clone was selected for further experiments. The viral supernatants were spun down, concentrated by PEG precipitation (PEG 6000, Sigma-Aldrich), aliquoted and stored at −80°C. HepG2 or SAOS2 cells were seeded on a 24-well plate 24 h before transduction. Virus-containing supernatant was added to the medium, incubated overnight and, after 24 h, replaced with fresh medium containing puromycin. Protein, RNA or DNA content was analyzed at 5 days post transduction unless otherwise stated.

Cells were seeded on glass coverslips overnight, BrdU (Sigma-Aldrich) was added to the medium (final concentration 1 µM) and cells were cultured for 1 h. Subsequently, the cells were fixed in 3% paraformaldehyde, and DNA was denatured by incubation in 2 M HCl for 30 min and immediately neutralized in 0.1 M borate buffer, pH 9.0. After permeabilization with 0.1% Triton X-100 (w/v) and several washes in PBS, cells were incubated with primary anti-BrdU antibody for 1 h, followed by anti-rabbit-IgG secondary antibody for 30 min and mounted with Mowiol (Sigma-Aldrich) containing 0.1 μg/ml DAPI (Sigma-Aldrich). Images were acquired with a confocal microscope (Leica TCS SP) and ratios of BrdU-positive to all cells were obtained upon counting the cells using ImageJ.

Cells were homogenized in SDS lysis buffer (60 mM Tris-HCl pH 6.80, 10% glycerol, 2% SDS) and briefly sonicated. The lysates were cleared by centrifugation and total protein content was measured by using a Bradford assay. Equal amounts of proteins were loaded onto 10% polyacrylamide gel, proteins separated by SDS-PAGE and transferred onto a nitrocellulose membrane. The blots were incubated with appropriate primary and fluorescently conjugated secondary antibodies and scanned on a Li-cor Odyssey imager (Lincoln, NE).

Samples were prepared as follows: proteins were eluted with 50 mM Tris-HCl pH 8.00, 2.5 mM desthiobiotin, and eluates digested with 1 µg of trypsin (Trypsin Gold, Promega, Madison, WI) overnight at 30°C. Peptides were desalted using a microtrap column (MichromBioresources, CA), dried down and dissolved for LC-MS analysis performed as previously described (Stodůlková et al., 2008), with minor modifications. The mass spectrometric analysis (MALDI-TOF-TOF type) was performed using the Mascot 2.0 search engine (Matrix Science, Boston, MA) with the following search parameters: SwissProt database, taxonomy human, trypsin specificity, no fixed modifications, oxidized methionine as a variable modification, MS-tolerance of 50 ppm and MS/MS tolerance 0.5 Da. Only proteins identified on three peptides with Mascot score above 30 were considered to be positive hits.

Total RNA was isolated using GenElute Miniprep Kit (Sigma-Aldrich) according to the manufacturer's protocol. RNA was treated with RNase-free DNase I for 30 min at room temperature. The concentration of RNA was measured by spectrophotometry and RNA integrity was checked on a denaturing agarose gel. A total of 100 ng of RNA was reverse-transcribed with random hexamer primers using TaqMan Reverse Transcription Reagents (Applied Biosystems, Life Technologies, Carlsbad, CA). Quantitative real-time PCR (qPCR) was performed on ABI Prism 7300 instrument using SYBR Green PCR Master Mix (Applied Biosystems, Life Technologies) and appropriate primers. Data were evaluated with ΔΔCt method and transcript levels normalized to those of the GAPDH gene. Primers used for RT-PCR are shown in supplementary material Table S2.

Cells were harvested by trypsinization, spun down, washed twice in ice-cold PBS and resuspended in a lysis buffer [20 mM HEPES-KOH pH 7.40, 5 mM CH₃COOK, 150 mM NaCl, 0.5 mM MgCl₂, 0.5 mM DTT and 0.5% (w/v) Triton X-100] containing EDTA-free protease inhibitor cocktail (cOmplete, Roche, Basel, Switzerland) and phosphatase inhibitor cocktail (Phos STOP, Roche). The lysate was centrifuged at 16,000 g for 20 min to pellet insoluble proteins. STrEP-T-actin-coupled sepharose beads (IBA GMBH, Göttingen, Germany) were washed several times with the lysis buffer, added to the lysate and the mixture incubated for 2 h. The beads were then spun down, washed five times for 10 min with the lysis buffer and finally the bound proteins were eluted in a native form using lysis buffer or 50 mM Tris-HCl pH 8.00 (for MS analysis) containing 2.5 mM desthiobiotin. The ‘pulldown’ eluate was analyzed by SDS-PAGE or MS and proteins visualized by staining the polyacrylamide gel with SilverQuest kit according to manufacturer's protocol (Invitrogen, Carlsbad, CA). The eluates were concentrated by ultrafiltration (Ultracel 10K, Merck Millipore, Billerica, MA).

The region covering IGF2 SNP rs680 and H19 rs2839704 was amplified from SAOS2 genomic DNA with appropriate primers (see supplementary material Table S2) and sequenced. RNA isolated from SAOS2 shNON or SAOS2 shPXN cells was reverse transcribed, amplified with the same primers and digested with RsaI or ApaI for H19 SNP or IGF2 SNP, respectively. Fragments were resolved on 3% agarose gel. Absence of DNA contamination in the RNA samples was checked by parallel controls without reverse transcription.

Genomic DNA isolated from SAOS2 cells was subjected to bisulfite conversion using the Methylamp DNA Modification Kit (Epigentek, Farmingdale, NY) according to the manufacturer's protocol. Bisulfite treatment efficiently converts unmethylated cytosine into uracil, whereas 5-methylcytosine remains unchanged (Hayatsu et al., 1970). Treated DNA was purified and amplified by PCR, with primers corresponding to the ICR region. Primers specific for bisulfite-treated DNA were designed with the Bisulfite Primer Seeker program (Zymo Research, Irvine, CA; see supplementary material Table S2). The amplified PCR product was inserted into pJET vector and transformed into competent cells. Plasmid DNA isolated from at least 30 bacterial colonies (for each shNON and shPXN variant) was sequenced.

ChIP was performed using MAGnify Chromatin Immunoprecipitation System (Invitrogen) according to the manufacturer's protocol. Around 300,000 cells were used as a starting material, and DNA was sheared with 15 sonication cycles (30 s ON, 30 s OFF; intensity HIGH) using a Bioruptor Next Gen sonicator (Diagenode, Seraing, Belgium). Antibodies against paxillin, SMC1, MED23, CTCF, control IgG (negative control) or H3K4me2 (positive control) were used. Input and immunoprecipitated DNA levels were quantified by qPCR. Enrichment was calculated as the amount in the immunoprecipitation over input and further normalized to a region where paxillin does not bind (GAPDH-coding region). ChIP primers are shown in supplementary material Table S2.

The preparation of Firefly luciferase reporter vectors pH19P-luc, pIGF2P3-luc, pH19-luc-E, pIGF2P3-luc-E (see ‘Cloning’ for details) was adopted from Ishihara et al., 2006. These constructs were co-transfected with Renilla luciferase reporter vector pGL4.74 into shPXN or shNON HepG2 cells at 4 days after lentiviral transduction. After 24 h, all cells were lysed and luciferase activity of both species measured according to Dual-Glo^®Luciferase Assay System protocol (Promega, Madison, WI), using Modulus™ II Microplate Multimode Reader (Turner Biosystems, Sunnyvale, CA). Data were normalized to Renilla luciferase activity and all assays performed in triplicate and repeated at least three times.

The technique was adopted from Hagège et al. (Hagège et al., 2007), except for the final semiquantitative analyses (Naumova et al., 2012). Briefly, cells expressing shNON or shPXN (5 days after transduction) were crosslinked with 1% formaldehyde, lysed and digested with PstI restriction enzyme. After ligation, samples were reverse-crosslinked overnight at 65°C. Isolated DNA was amplified with corresponding primers (schematically shown in Fig. 5A; see supplementary material Table S1 for the sequences) and PCR products resolved on 3% agarose gels. The amount of DNA input was first titrated and bands analyzed semi-quantitatively using ImageJ software; the background was subtracted and data normalized to a total DNA input unaffected by the restriction digest (ICR region). Three biological replicates were prepared and analyzed in three technical repeats.

The authors would like to thank Iva Jelínková and Pavel Kříž for their excellent technical assistance, and Lenka Jarolímová, Tomáš Vacík and Vlada Philimonenko for methodological help. We are grateful to Ravi Salgia (The University of Chicago School of Medicine) for sharing paxillin cDNA.