Chromosomal Contact Permits Transcription between Coregulated

Stephanie Fanucchi,1 Youtaro Shibayama,1 Shaun Burd,1 Marc S. Weinberg,3,4 and Musa M. Mhlanga1,2,* 1Gene Expression and Biophysics Group, Synthetic Biology Emerging Research Area, Biosciences Unit, Council for Scientific and Industrial Research, Pretoria, Gauteng 0001, South Africa 2Unidade de Biofı´sica e Expressa˜ o ´ tica, Instituto de Medicina Molecular, Faculdade de Medicina, Universidade de Lisboa, Lisboa, 1649-028 Portugal 3Antiviral Gene Therapy Research Unit, Department of Molecular Medicine and Haematology, School of Pathology, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, Gauteng 2193, South Africa 4Department of Molecular and Experimental Medicine, The Scripps Research Institute, La Jolla, CA 92037, USA *Correspondence: [email protected] http://dx.doi.org/10.1016/j.cell.2013.09.051

SUMMARY 2012). These highly sensitive assays can target either DNA or nascent mRNA and have revealed the colocalization between Transcription of coregulated genes occurs in the FISH foci in a fraction of the population (Schoenfelder et al., context of long-range chromosomal contacts that 2010; Papantonis et al., 2010). This suggests that a subset of form multigene complexes. Such contacts and cells within the population may be enriched for specific chromo- transcription are lost in knockout studies of tran- somal interactions. are large and constrained in scription factors and structural chromatin . their ability to roam the entire nuclear volume (Strickfaden To ask whether chromosomal contacts are required et al., 2010; Barbieri et al., 2013). Thus, it is reasonable to surmise that the topological arrangements after each cellular for cotranscription in multigene complexes, we division shuffle chromosomal proximities such that their three- devised a strategy using TALENs to cleave and dimensional (3D) arrangements are altered in one-dimensional disrupt gene loops in a well-characterized multigene (1D) space (Gibcus and Dekker, 2013). This may lead to every complex. Monitoring this disruption using RNA FISH cell in the population possessing unique spatial arrangements and immunofluorescence microscopy revealed that of its chromosomes (Orlova et al., 2012). perturbing the site of contact had a direct effect on Enhancer-promoter interactions utilize chromatin looping to transcription of other interacting genes. Unex- trigger dynamic changes in transcription initiation (Osborne pectedly, this effect on cotranscription was hierar- et al., 2004; Deng et al., 2012). An example of this is the well- chical, with dominant and subordinate members of established model between the locus control region (LCR) and b the multigene complex engaged in both intra- and the promoter of the -globin gene (Osborne et al., 2004). In a tis- interchromosomal contact. This observation reveals sue-specific manner, the LCR has been shown to physically con- tact the promoter of the b-globin gene and initiate transcription the profound influence of these chromosomal con- (Deng et al., 2012). These LCR-mediated chromosomal interac- tacts on the transcription of coregulated genes in a tions have been shown to result in variability in b-globin gene multigene complex. transcript levels or variegated gene expression across the pop- ulation (Noordermeer et al., 2011). In an otherwise identical pop- INTRODUCTION ulation of cells, presumably through chromosomal interactions, such ‘‘jackpot’’ cells display higher levels of b-globin transcrip- Transcription is replete with proximal and distal chromatin loop- tion (Noordermeer et al., 2011). Accordingly, the specific set of ing interactions whose formation represents the basic organizing chromosomal interactions (and consequent gene expression principle of nuclear architecture and gene activity (Miele and that may depend on LCR-mediated interactions) will vary Dekker, 2008; Palstra et al., 2008; Tan-Wong et al., 2012). between cells across the population (Strickfaden et al., 2010). Loop-mediated chromosomal contacts are usually identified This heterogeneity reveals the absolute requirement of single- on a genome-wide scale using population-based ‘‘ cell analysis in global interactome and gene loop studies. conformation capture’’ (3C) technologies (Dekker et al., 2002; Looping also brings distal genes into close proximity, enabling Fullwood et al., 2009; Lieberman-Aiden et al., 2009; Li et al., chromosomal contact in ‘‘multigene complexes’’ at a single 2012). Analyses of 3C-based data reveal a large heterogeneity focus of multiple RNA polymerases (Schoenfelder et al., 2010; in global chromatin interactions (Fullwood et al., 2009; Noorder- Papantonis et al., 2012; Li et al., 2012). Numerous studies have meer et al., 2011; Li et al., 2012). Therefore, interacting DNA demonstrated that the formation of loop-mediated contact coin- elements identified by 3C-based technologies are verified at cides with alterations in gene expression (Spilianakis et al., 2005; the single-cell level using fluorescent in situ hybridization Apostolou and Thanos, 2008; Fullwood et al., 2009; Schoen- (FISH) assays (Schoenfelder et al., 2010; Papantonis et al., felder et al., 2010). Indeed, chromosomal contacts in multigene

606 Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc. complexes appear to be the main modality of transcription in engaged in intrachromosomal contact at distances >48 Mbp, metazoan cells, as they are associated with >95% of transcrip- as well as interchromosomal interactions. Furthermore, restora- tional activity (Li et al., 2012; Sandhu et al., 2012). In a compara- tion of a disrupted gene loop re-established both chromosomal ble manner to enhancer-promoter interactions, specific chro- contacts and transcription of interacting genes in a sequence- matin interactions in multigene complexes are detected in a independent manner. The unexpected hierarchical organization subset of cells within the population (Schoenfelder et al., 2010; within the TNFa-induced multigene complex reveals the unprec- Papantonis et al., 2010). Genome-wide chromatin interaction edented level of influence of these chromosomal contacts on the analysis with paired end tags (ChIA-PET) uncovered a multigene transcription of coregulated genes in a multigene complex. complex, including the GREB1 locus and three other genes (Li et al., 2012). Of the four interacting genes, only GREB1 transcrip- RESULTS tion is activated by the estrogen receptor-a (ERa)(Li et al., 2012). Intriguingly, despite the fact that this multigene complex may not The Transcriptional Response of Coregulated Genes in a assemble in every cell in the population, siRNAs targeting ERa Multigene Complex Is Asymmetric disrupted all four interacting genes (Li et al., 2012). Therefore, TNFa has been shown to rapidly and synchronously shape the even though these chromosomal contacts may only occur in a transcriptional response in early passage human umbilical vein fraction of the population, they clearly play a significant role in endothelial cells (HUVECs) by systematically inducing the forma- gene regulation. Moreover, this data supports a model of syner- tion of a large variety of different multigene complexes (Papan- gistic transcription in which chromosomal contact influences the tonis et al., 2012). Interacting genes in TNFa-induced multigene transcription of the interacting genes. This would connote that complexes are activated by NF-kB(Papantonis et al., 2010). The the topological framework for transcriptional regulation is phys- SAMD4A, TNFAIP2, and SLC6A5 genes associate in a NF-kB- ical contact via chromosomal looping in multigene complexes. dependent multigene complex (Figure 1A), which has been Current siRNA and 3C-based experimental approaches extensively characterized by 3C, 4C (3C capture-on-chip), tiling cannot be applied to multigene complexes in which all interact- microarray, and FISH and thus is an ideal model to interrogate ing genes are activated by the same transcription factor. Tumor the role that chromosomal contacts play in cotranscription at a necrosis factor alpha (TNFa) has been shown to induce the for- single-cell level (Wada et al., 2009; Papantonis et al., 2010; mation of such multigene complexes in which all interacting Papantonis et al., 2012). HUVECs were arrested in the G0 phase genes are activated by NF-kB(Papantonis et al., 2012). Ten of the cell cycle by serum deprivation (Larkin et al., 2012). Impor- minutes after TNFa stimulation, the promoters of genes located tantly, the transcription of SAMD4A, TNFAIP2, and SLC6A5 is on the same chromosome (SAMD4A and TNFAIP2) and on a rapidly induced 10 min post-TNFa stimulation (Figure 1B; different chromosome (SLC6A5) associate to form part of a Papantonis et al., 2010). Concurrent to their expression at NF-kB-dependent multigene complex (Papantonis et al., 2010). 10 min, prior 3C data, including our own (data not shown; Papan- RNA FISH assays targeting the approximate sites of interaction tonis et al., 2010), indicate that chromosomal contacts between identified by 3C suggest an association between the formation these genes occur at sites 1.5 kbp downstream of the TSS (Fig- of this NF-kB-regulated multigene complex and the cotranscrip- ure 1B). To interrogate whether the formation of these contacts is tion of interacting genes (Papantonis et al., 2010). However, both associated with cotranscription, we designed RNA FISH probes 3C and FISH approaches fail to reveal the necessity of chromo- to target the intronic sites where these contacts occur 10 min somal contacts for cotranscription of these interacting genes. post-TNFa treatment (Figure 1B). As these introns are typically Therefore, to accurately interrogate a model of synergistic regu- spliced and degraded cotranscriptionally (Wada et al., 2009), lation, a discrete perturbation of a single site within a gene loop these probes label the transcriptional start site (TSS). Each set while monitoring the transcriptional status of other members of of gene-specific RNA FISH intronic probes was conjugated to the multigene complex is required. Importantly, owing to varie- spectrally distinct fluorophores. By simultaneously performing gated gene expression (Noordermeer et al., 2011), this can RNA FISH on the three NF-kB-regulated genes, we were able only be achieved with a single-cell approach. to investigate the frequency of cotranscription across a popula- Here, we devise a single-cell strategy using TALE nucleases tion of cells. Consistent with previous studies (Papantonis et al., (TALENs) to discretely perturb sites within gene loops that are 2010, Papantonis et al., 2012), analysis of overlapping RNA FISH established to engage in chromosomal contact in the well- foci revealed colocalized foci in only a fraction of all alleles in the characterized NF-kB-regulated multigene complex (Papantonis total population (5%) (Figure 1B). Chromosomal interactions et al., 2010). This enabled us to address the long-standing ques- between transcribed genes have been shown to occur at tion of the requirement of loop-mediated contact for transcrip- discrete foci of active RNA polymerase II (RNA Pol II) (Pombo tional coregulation in a multigene complex. Using RNA FISH et al., 1999; Schoenfelder et al., 2010), as well as at nuclear and immunofluorescence (IF), we imaged the site of the disrup- speckles (Brown et al., 2008). Overlapping SAMD4A, TNFAIP2, ted loop simultaneously with the transcriptional activity of other and/or SLC6A5 RNA FISH foci consistently colocalize with the interacting genes in the NF-kB-regulated multigene complex. active, poised form of RNA Pol II (phosphorylated at Ser5) (Fig- This unique single-cell perspective revealed that perturbing ure S1A available online). Collectively, these data suggest that loop-mediated contact between the NF-kB-regulated genes the cotranscription of these genes at RNA Pol II foci may only altered the transcriptional status of interacting genes. In addi- occur in a small fraction of the HUVEC population. tion, this effect on cotranscription was hierarchical, with domi- Chromosomes in primary cells occupy distinct territories in the nant and subordinate members of the multigene complex nucleus (Levesque and Raj, 2013). As HUVECs are primary cells,

Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc. 607 Figure 1. The Transcriptional Response of Coregulated Genes in a Multigene Complex Is Asymmetric (A) The promoters of genes located on the same chromosome (SAMD4A and TNFAIP) and on a different chromosome (SLC6A5) associate via RNA Pol II (orange) to form part of a NF-kB-dependent multigene complex. BMP4 and RCOR1 are nonresponsive to TNFa and do not interact with SAMD4A (blue), TNFAIP2 (green), or SLC6A5 (red); therefore, they serve as controls in this study. (legend continued on next page)

608 Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc. they always display two spatially distinct DNA FISH foci, both tionship between the transcriptional status of SAMD4A and the before and after TNFa treatment (Figure S1B). Within territories, transcriptional activation of TNFAIP2 and SLC6A5. With respect chromosomes are segregated further into topologically associ- to coexpressed alleles, SAMD4A and TNFAIP2 were colocalized ating domains (TADs) (Cremer and Cremer, 2010; Dixon et al., at 26% of coexpressed alleles (3% in the total population 2012). As a consequence of this compartmentalization, interact- [T/P]), whereas SAMD4A, TNFAIP2, and SLC6A5 were colocal- ing genes in multigene complexes may be confined within much ized at 40% of all coexpressed alleles (5% in the T/P) (Fig- smaller genomic neighborhoods. By simultaneously performing ure 1D). Collectively, these data suggest a hierarchical mode of DNA FISH on these three coregulated genes, we were able to regulation between these genes, in which chromosomal contact investigate whether the DNA of these three genes were in close favors cotranscriptional activation. vicinity prior to activation by TNFa. Interestingly, analysis of over- lapping DNA FISH foci revealed proximal foci in both the unsti- Visualization of TALEN-Mediated Gene Loop Disruption mulated and the TNFa-treated HUVEC population (Figure S1C). Hierarchical regulation dependent upon chromosomal contact This data suggests that, in a fraction of the HUVEC population, was revealed in the GREB1 multigene complex (Li et al., 2012). the DNA of the SAMD4A, TNFAIP2, and SLC6A5 genes may qPCR analysis revealed that siRNAs targeting ERa not only dis- be constrained within a TAD prior to TNFa induction. rupted GREB1 transcription, but also were sufficient to cause a Through an enrichment of chromosomal interactions in TADs >2- to 4-fold reduction in the transcription of interacting mem- and higher levels of transcription, ‘‘jackpot’’ cells may contribute bers. siRNA approaches cannot be applied to the TNFa-induced to variable, or stochastic, effects in gene expression (Noorder- multigene complex, as these three genes are transcriptionally meer et al., 2011). As these NF-kB-regulated genes respond sto- activated by NF-kB(Papantonis et al., 2010). Toward this end, chastically to TNFa (Papantonis et al., 2010), we assessed the we developed a single-cell assay allowing for the discrete monoallelic and biallelic expression of the three genes. Nine of disruption of individual gene loops at the sites of chromosomal the 33 possible phenotypes were observed in the majority of contacts (Figure 1B). In parallel, to decrypt the role of a single in- the population (84%) (Figure 1C). With the exception of cells dividual gene loop on cotranscription in the same multigene displaying no foci (29%), all cells in this category displayed complex, we visualized transcriptional activity of other genes in either a single or dual SAMD4A foci. With respect to the individ- the multigene complex using highly sensitive RNA FISH (Fig- ual expression of each of the three genes, not all cells respond ure 1B; Raj et al., 2008). We constructed our microscopy-based similarly to TNFa, with approximately half of the alleles express- assay upon TALENs, the orthogonal and robust well-established ing SAMD4A, a lower proportion expressing TNFAIP2, and genome editing tool derived from TAL effectors of Xanthomonas approximately one-fifth expressing SLC6A5 (Figure 1C). As sp.(Christian et al., 2010, Li et al., 2011). Within the TALE struc- chromosomal contact occurs between these three genes (Fig- ture, a central repeat domain mediates highly specific DNA ure 1B; Papantonis et al., 2010), the asymmetric transcriptional recognition (Boch et al., 2009; Moscou and Bogdanove, 2009; response of these genes to TNFa suggests a hierarchical regula- Christian et al., 2010). Fusion of this domain to FokI endonu- tion in the assembly of this multigene complex. Supporting this clease enables TALENs to induce site-specific double-strand hypothesis, TNFAIP2 transcription occurs predominantly when breaks (DSB) (Christian et al., 2010, Li et al., 2011). As the TAL- SAMD4A is also transcribed (86%), and SLC6A5 transcription ENs used in this study induce a DSB at the approximate site of occurs mainly when SAMD4A (92%) or both SAMD4A and chromosomal contact that occurs 10 min post-TNFa treatment TNFAIP2 are transcribed (62%) (Figure 1C). (Figure 2A), we reasoned that, should loop-mediated contact As SAMD4A, TNFAIP2, and SLC6A5 are not always coex- be required for cotranscription, the DSB would serve to rupture pressed (Figure 1C), we also represented the colocalization chromosomal contact between these genes. To ascertain frequencies relative to cells coexpressing SAMD4A, TNFAIP2, whether the TALEN had successfully disrupted a gene loop, and/or SLC6A5 (Figure 1D). Allelic coexpression analyses be- we stained for induction of a DSB using histone variant H2A.X tween the three genes revealed that coexpression of SAMD4A (Figures S2A and S2B). Rapidly phosphorylated at Ser139 within and TNFAIP2, as well as triple expression of the three genes, minutes following DNA rupture, this modification persists was most prevalent across the population (12% and 14%, throughout the entire DNA repair process (Rogakou et al., respectively) (Figure 1D). SLC6A5 was rarely expressed (4%) 1999). To initially test our TALEN system and its ability to disrupt in the absence of TNFAIP2 expression at the corresponding a single gene loop and concurrent assembly of the multigene allele (Figure 1D). Moreover, cotranscription of TNFAIP2 and complex, we introduced into HUVECs, by high-efficiency micro- SLC6A5 in the absence of SAMD4A transcription was extremely poration, our TALEN vectors that were able to disrupt the gene rare (1%) (Figure 1D). This provides further evidence for a rela- loop of the longest of the three genes, SAMD4A. We used

(B) Chromosomal contact in a multigene complex is associated with cotranscription. Spectrally distinct RNA FISH intronic probes targeting the region of these genes involved in chromosomal contact reveal colocalization of nascent intronic RNA, as visualized by overlapping foci. Two-tailed Fisher’s exact test; ***p = 4.5 3 104; n, number of alleles. (C) The response to TNFa is asymmetric in diploid HUVECs. Allelic expression of each gene in the population of cells is shown. TNFAIP2 transcription occurs predominantly when SAMD4A is also transcribed (86%), and SLC6A5 transcription occurs mainly when SAMD4A (92%) or both SAMD4A and TNFAIP2 are transcribed (62%). N, number of cells (N = 166); n, number of alleles (n = 332). (D) Data in (C) was replotted to show the combined allelic transcriptional status of SAMD4A, TNFAIP2, and SLC6A5 in the total population (T/P). Colocalization frequencies relative to cells coexpressing SAMD4A, TNFAIP2, and/or SLC6A5 are also shown. n, number of alleles; scale bar, 5 mm. See also Figure S1.

Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc. 609 (legend on next page)

610 Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc. H2A.X staining of DSBs to carefully establish a time course of SAMD4A in this region, as no transcription of the intronic region TALEN activity (also a measure of transfection efficiency), noting of SAMD4A was ever evident beyond the DSB (Figure 3B and that nuclease activity was first evident after 6 hr and was sus- Supplemental Note 1 in the Extended Experimental Procedures). tained until 48 hr (Figure S2B). We chose to assay for nuclease IF in untransfected cells revealed that TNFa induction resulted activity after 24 hr, as there were high levels of DSBs and low not only in the robust transcription of members of the multigene levels of cytotoxicity at this time point. Discrete sites of H2A.X complex, but also in an increase in expression (Figure 3C). phosphorylation were evident in 60% of cells 24 hr posttrans- Consistent with the ability of the TALEN to abrogate transcrip- fection, with a higher proportion of these cells displaying single tion, there was a significant reduction in SAMD4A protein expres- allelic DSBs and fewer cells displaying dual allelic DSBs (Fig- sion in cells harboring dual allelic DSBs as detected by IF, ure 2B). The high specificity of the SAMD4A TALEN was demon- whereas cells harboring single allelic DSBs were still able to ex- strated by the consistent colocalization of SAMD4A DNA with press SAMD4A protein presumably from the intact allele (Fig- the DSB, visualized by DNA FISH (Figure 2C). TALEN cleavage ure 3C). Overall, this robust assay offered an unprecedented efficiency was further supported by the results of ‘‘surveyor as- perspective into the dynamics of TALEN activity, transcriptional says’’ (Figures 2D and S2C). Thus, we were able to assay status of the targeted gene, and related protein levels at single- uniquely for the disruption of the SAMD4A gene loop in its native cell resolution. Importantly, the observed disruption in SAMD4A environment at a single cell level. RNA (Figure 3B) and protein expression (Figure 3C) strengthened our assay, as it validates that the DSB occurs at the intended site TALEN-Mediated Disruption of a Gene Loop Abrogates of TALEN activity where H2A.X staining occurred. RNA and Protein Expression Upon stimulation by TNFa, RNA Pol II engages the SAMD4A pro- TALEN-Mediated Disruption of a Gene Loop Abrogates moter, triggering a wave of transcription that propagates down Transcription of Interacting Members the gene (Wada et al., 2009). RNA FISH tiling array analysis re- Satisfied that our TALEN assay was able to discretely disrupt vealed that the transcriptional cycle takes 85 min (Wada SAMD4A at the site that engages in chromosomal contact, we et al., 2009). Accordingly, RNA transcribed 1.5 kbp down- repeated the prior experiment with an important modification. stream of the TSS appears within 10 min (probe set i, Figure 3A) We used RNA FISH to visualize transcription of two other inter- and 34 kbp into intron 1 after 30 min post-TNFa stimulation acting genes that engage in intra- and interchromosomal contact (probe set ii, Figure 3B). Importantly, the DSB induced by the with the region of SAMD4A that had been disrupted. By simulta- SAMD4A TALEN occurs at a site between these two regions neously monitoring SAMD4A gene loop disruption, as well as where different sets of intronic RNA FISH probes bind. HUVECs transcription of additional members of this multigene complex, were dual transfected for 24 hr with the SAMD4A TALENs and we were able to discretely interrogate the effect of the disrupted exposed to TNFa for 10 min. This recapitulated SAMD4A tran- SAMD4A gene loop on the transcriptional status of two other scriptional activity, allowing for the first 1.5 kbp of SAMD4A members of the complex. to be transcribed. By using RNA FISH to monitor intronic RNA First, to exclude the possibility that the DSB was capable of transcription 10 min post-TNFa stimulation, SAMD4A transcrip- inducing cell-cycle arrest, thereby altering global transcription, tion was frequently detected (42%), either overlapping or in we designed a TALEN to rupture BMP4, a non-TNFa-responsive close proximity to the DSB (Figure 3A). Thus, these data indicate gene, located 600 kb 50 of SAMD4A on (Fig- that, despite disrupting the SAMD4A gene loop, the DSB did not ure 4A). We observed no change in the transcription of any of appear to influence the ability of RNA Pol II to access the the three members of the multigene complex relative to the SAMD4A promoter, permitting transcriptional initiation and elon- normal TNFa-induced transcriptional response (Figure 1C). gation up to the DSB. The half-life of SAMD4A intronic RNA tran- Therefore, inducing a DSB 50 of SAMD4A has no effect on scripts transcribed at 10 min is between 3 and 6 min (Larkin et al., transcription of members of the multigene complex. 2012). Consequently, RNA transcripts transcribed downstream We then monitored the effects of perturbation of the chromatin of the TSS at 10 min have degraded by the 30 min time point. loop of SAMD4A at its first intron. We observed a significant In a separate experiment, dual-transfected HUVECs were stimu- reduction of transcription of TNFAIP2 and SLC6A5 at the lated with TNFa for 30 min to allow for the first 34 kbp of SAMD4A TALEN-induced DSB (Figure 4B). In cells in which a SAMD4A to be transcribed (Figure 3B). Notably, the TALEN- single allele of SAMD4A was targeted, virtually all transcription induced perturbation was able to abrogate transcription of of TNFAIP2 and SLC6A5, as assessed by RNA FISH, was lost

Figure 2. Visualization of TALEN-Mediated Disruption of a Gene Loop (A) The SAMD4A TALEN targets the region in intron 1 involved in chromosomal contact 10 min post-TNFa stimulation. TALE repeat domains are colored to indicate the identity of repeat variable diresidues (RVD); each RVD is related to the cognate targeted DNA base by the following code: NI = A, HD = C, NN = G/A, NG = T. (B) Successful detection of TALEN transfection at a single-cell level. Discrete sites of H2A.X pSer139 were evident in 60% of HUVECs transfected with the SAMD4A TALEN. A higher portion of successfully transfected cells displayed single allelic DSBs than displayed dual allelic DSBs. (C) H2A.X immunofluorescence accurately labels the sites of the disrupted SAMD4A gene loop. SAMD4A DNA FISH foci consistently colocalize with SAMD4A TALEN-induced DSBs. (D) Gel showing the surveyor nuclease result from the SAMD4A TALEN pair. NT, un-transfected control; GFP, GFP-transfected cells; L, SAMD4A left TALEN only; R, SAMD4A right TALEN only; L+R, cells transfected with pcDNA left and right TALENs. Red arrow, 246 bp; blue arrow, 114 bp. Cells were counterstained with DAPI. Scale bar, 5 mm. See also Figure S2.

Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc. 611 (legend on next page)

612 Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc. at the corresponding allele or DSB, with multigene transcription site (Figure S6A). Remarkably, we observed that the disruption confined to the intact allele (Figure S3A). Importantly, as the of the TNFAIP2 loop had no significant effect on SAMD4A tran- SAMD4A protein is still observed in cells harboring a single allelic scription, yet transcription of both TNFAIP2 (Figure S6B) and DSB (Figure 3B and Supplemental Note 2 in the Extended Exper- SLC6A5 was reduced (Figure 4C). Transcriptional loss was imental Procedures), this observation excludes the possibility observed at either a single or both targeted alleles (Figure S3B). that the SAMD4A protein is required for the transcription of When we repeated the experiment with a TALEN targeting the TNFAIP2 and SLC6A5. Supporting the observed reduction in SLC6A5 gene loop on chromosome 11 (Figure S6E), the hierar- transcription at single allelic DSBs, in cells in which both chical effect was more pronounced, with transcription unaf- SAMD4A alleles were targeted, virtually all transcription from fected in both SAMD4A and TNFAIP2 (Figures 4D and S5C). TNFAIP2 and SLC6A5 was lost at both alleles (Figure S3). However, at all successfully targeted single and dual alleles of Furthermore, protein levels of TNFAIP2 and SLC6A5 were also SLC6A5 , transcription was lost (Figure S6F). This result lends severely reduced, as assessed by IF, in cells in which both credence to the hypothesis that the disrupted site of chromo- SAMD4A alleles were successfully targeted (Figure S4). Hence, somal contact precludes the ability of the gene loop to access in a manner analogous to GREB1 (Li et al., 2012), SAMD4A critical transcriptional machinery due to a requirement for loop- appeared to be influencing the transcription of other coregulated mediated chromosomal contacts. This unique single-cell and interacting genes within the same multigene complex. perspective reveals a highly unexpected hierarchical organiza- We sought to determine whether the loss of transcription of tion within the TNFa-induced multigene complex, providing the TNFAIP2 was due to the DSB abrogating the transcription of first direct evidence that chromosomal contacts play a central genes located between SAMD4A and TNFAIP2 on chromosome role in supporting transcription and determining the hierarchy. 14. RCOR1 is a gene located 400 kb 50 of TNFAIP2 and displays transcriptional activity comparable to GAPDH (Papantonis et al., Disrupted SAMD4A Gene Loops Can Be Successfully 2012). We repeated the TALEN-mediated disruption of SAMD4A Repaired gene loop while monitoring transcription of RCOR1 (Figure S5). If the integrity of the gene loop topology and chromosomal con- Transcription remained unaffected at the RCOR1 locus (Fig- tacts was essential for cotranscription, then we hypothesized ure S5). This observation concurred with the hypothesis that that restoring the chromosomal contacts would restore tran- loop-mediated chromosomal contact of SAMD4A specifically scription of interacting genes in the multigene complex. DSBs impacts the transcription of other members of the multigene are generally repaired in most cells by two highly conserved complex but has no effect on genes outside of the complex. mechanisms: rapid but error-prone nonhomologous end-joining This extended to genes on the same chromosome that were (NHEJ) or the slower but highly precise homology-directed repair interspersed between two genes in the multigene complex. (HDR) (Hoeijmakers, 2001). We generated a repair construct We sought to determine whether the presence of other gene designed to span the SAMD4A TALEN target site in order to loops in the multigene complex were equally required for cotran- exploit HDR to restore SAMD4A gene loop integrity by inserting scription, as all genes in the multigene complex were bound by an exonic eGFP sequence into intron 1 of SAMD4A (Figure 5A). the NF-kB transcription factor (Papantonis et al., 2010). We hy- We included an internal ribosome entry sequence (IRES), as pothesized that, if gene loops were equally required for cotran- well as splice sites flanking the IRES-eGFP, to facilitate the scription, then the disruption of any other gene loop would cotranscriptional splicing and independent translation of the have a similar effect to that observed for SAMD4A. Alternatively, repair construct (Figure 5A). Once repaired, through the activa- an asymmetrical relationship between gene loops infers that the tion of SAMD4A, TNFa stimulation would then be able to induce disruption of one gene loop may have no bearing on the tran- SAMD4A/IRES-eGFP transcription. We used the identification of scriptional status of other genes in the complex. To differentiate eGFP-positive cells to establish a time course for TNFa treat- between these two hypotheses, we designed TALENs targeted ment. Due to HDR being rare and given that the IRES-eGFP is to TNFAIP2 and SLC6A5 (Figure S6). The sites for these TALENs under the control of the endogenous SAMD4A promoter, we were identified by 3C (Papantonis et al., 2010), indicating where sought to enhance the efficiency of the repair. To ensure that the chromosomal contacts occurred between these genes. equimolar quantities of both left and right TALENs were synthe- Using an identical approach to our initial assay, we delivered a sized in the same cell throughout the DSB-induction process, we TALEN to TNFAIP2 while monitoring transcription of SAMD4A constructed a bidirectional TALEN system that is able to and SLC6A5. The TNFAIP2 TALEN targets the site of chromo- combine both targeting arms of the TALEN into a single bidirec- somal contact (1.5 kbp downstream of the TSS), and our tional promoter plasmid (pDT; Figure 5B). HUVECs dual trans- RNA FISH probes interrogated a region downstream of this fected with the pDT and the repair plasmids demonstrated

Figure 3. TALEN-Mediated Disruption of the SAMD4A Gene Loop Abrogates SAMD4A RNA and Protein Expression (A) SAMD4A TALEN-induced DSB does not alter transcription up to the DSB. Nascent intronic SAMD4A RNA (detected by probe set i) transcribed 50 of the DSB was evident in 42% of HUVECs displaying DSB, as evidenced by H2A.X staining. n, number of alleles (n = 194); NS, no significant difference. (B) The SAMD4A TALEN abrogates transcription downstream of the DSB. Nascent intronic SAMD4A RNA (detected by probe set ii) transcribed 30 of the DSB was never observed. n, number of alleles (n = 84); two-tailed Fisher’s exact test; ***p < 0.001. (C) Disrupting the SAMD4A gene loop is sufficient to abrogate protein expression. TNFa induces SAMD4A protein expression 16 hr post-TNFa stimulation. Cells harboring dual allelic DSBs, as detected by H2A.X, displayed a significant reduction in protein expression, whereas cells harboring single allelic DSBs still express SAMD4A protein. R.F.U., relative fluorescent units; mean ± SD; *p < 0.01; two-tailed unpaired Student’s t test; cells were counterstained with DAPI; scale bar, 5 mm.

Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc. 613 Figure 4. TALEN-Mediated Disruption of a Single Gene Loop and the Associated Chromosomal Contacts in a Multigene Complex Alters the Transcriptional Status of Other Genes Occupying the Same Complex (A) A DSB induced in the non-TNFa-responsive gene, BMP4, did not alter transcription of SAMD4A, TNFAIP2,orSLC6A5 relative to the normal TNFa response (Figure 1C). (B) The disruption of the SAMD4A gene loop abrogates TNFAIP2 and SLC6A5 transcription and colocalization. SAMD4A loop disruption detected by H2A.X was simultaneously monitored with transcription of TNFAIP2 and SLC6A5 by RNA FISH. (C) Disruption of the TNFAIP2 gene loop does not affect SAMD4A gene expression but alters SLC6A5 transcription and SAMD4A/SLC6A5 colocalization. (D) Disruption of the SLC6A5 gene loop does not alter SAMD4A/TNFAIP2 transcription or colocalization. Two-tailed Fisher’s exact test; *p < 0.05, **p < 0.01, and ***p < 0.001; n, number of DSBs; cells were counterstained with DAPI; scale bar, 5 mm. See also Figures S3, S4, S5, and S6. poor repair efficiency in the transfected cells, as evidenced by DSB to induce efficient HDR (Figure 5B) (Orlando et al., 2010; the lack of eGFP signal (data not shown). To modify the repair Bedell et al., 2012). Notably, eGFP protein signal was observed construct, we amplified the IRES-eGFP sequence with associ- in 1% of HUVECs dual transfected with the pDT plasmid and ated flanking splice sites and included short 50 20 bp and 30 PCR product for 72 hr and stimulated with TNFa for 24 hr (Fig- 18 bp homologous extensions on either side of the SAMD4A ure 5B). However, eGFP mRNA transcripts, which can only arise

614 Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc. following successful HDR, were observed in 10% of cells in the by constraining these genes to areas permissible to long-range population (Figure S7). Notably, eGFP-positive, or ‘‘green,’’ cells interactions (Figure 7). Indeed, analysis of DNA in unstimulated coincided with a large number of RNA FISH foci (Figure S7). As HUVECs revealed that, despite occupying distal genomic loca- the SAMD4A promoter does not constitutively express the tions, in a fraction of the HUVEC population, the SAMD4A, IRES-eGFP mRNA, it is reasonable that very few cells in the TNFAIP2, and SLC6A5 DNA are in close proximity prior to population would transcribe sufficient IRES-eGFP mRNA to TNFa induction (Figure S1C). The enrichment of chromosomal enable the detection of a green cell. Therefore, detection of interactions in these ‘‘jackpot’’ cells may contribute to hierarchi- IRES-eGFP mRNA was the more sensitive approach to identi- cal SAMD4A, TNFAIP2, and SLC6A5 gene expression. However, fying successful HDR and was used to interrogate the effects current biochemical and imaging technologies lack the spatio- of a repaired gene loop. temporal resolution to interrogate whether loop-mediated cotranscription between these genes can only occur in these Restoration of the SAMD4A Gene Loop Restores ‘‘jackpot’’ cells or alternatively, given more time, whether these Transcription of Interacting Members in a Sequence- chromosomal interactions may occur in most cells across the Independent Manner population. Satisfied that the repair experiment was fully functional, we Chromosomal translocations are natural perturbations in chro- sought to investigate whether the re-establishment of contact mosome structure that alter the spatial positioning of DNA within was sufficient to restore transcription of interacting genes. We chromosome territories (Harewood et al., 2010). Interestingly, stimulated HUVECs for 20 min with TNFa, recapitulating tran- these discrete perturbations in chromosome structure not only scription of the first 1.5 kbp of SAMD4A and IRES-eGFP. We influence genes located near to the breakpoint, but also are suf- detected transcription of eGFP using RNA FISH probes that ficient to modify gene expression in cis and trans (Harewood bind to its RNA (Figures 6A and S7) and related the position of et al., 2010; Levesque and Raj, 2013). Therefore, through the re- eGFP transcription to members of the multigene complex. positioning of chromosomes and relocation of DNA into different Distinct eGFP foci were evident in 10% of transfected cells, TADs, translocations may alter transcription by disrupting intra- and these foci overlapped with SAMD4A intronic mRNA (Fig- and interchromosomal interactions. We were able to reveal that ure 6A). In the instances in which we had successful HDR-medi- perturbing the SAMD4A gene loop has a direct effect on the tran- ated repair, we observed no change in the transcription of scriptional status of TNFAIP2 and SLC6A5 (Figure 4B). In addi- TNFAIP2 and SLC6A5 relative to the normal TNFa-induced tran- tion, corresponding to the transcriptional response (Figure 1C), scriptional response (Figure 6B). Moreover, there was no signif- the effect on cotranscription was hierarchical, with the perturba- icant difference in the colocalization frequencies between eGFP/ tion of TNFAIP2 altering SLC6A5 expression but having no influ- SAMD4A, TNFAIP2, and SLC6A5 in the repaired and mock- ence on SAMD4A (Figure 4C). Furthermore, perturbing SLC6A5 transfected cells (Figure 6B). This result indicates that re-estab- did not impact either SAMD4A or TNFAIP2 transcription (Fig- lishment of an intact SAMD4A loop, in a sequence-independent ure 4D). This observation raises the question: if all of the factors manner, restores chromosomal contacts as well as transcription necessary for transcription are present, why does transcription of TNFAIP2 and SLC6A5 in this multigene complex. of these interacting genes not occur? A possible explanation is that the recruitment of the multitude of repair proteins to the DISCUSSION DSB acts as an obstruction to the normal chromatin contacts between these gene loops. Alternatively, contacts between Here, we show that disrupting sites within gene loops that interacting genes may still be present but are ‘‘bridged’’ by the engage in chromosomal contact significantly impacts the tran- repair machinery. Although this bridging maintains the contact scription of interacting genes in a multigene complex. We between gene loops, it is still inadequate for transcriptional activ- revealed this level of gene regulation by implementing a single- ity to occur. It is important to note that the DNA of these three loci cell strategy that allows discrete perturbation within chromatin appear to be in close proximity in unstimulated HUVECs despite loops. Initial evidence for this regulation was our single-cell- the absence of 3C contact (Papantonis et al., 2010). Thus, it based observation of the hierarchical transcriptional response would be very difficult to definitively show that the disruption of of these three genes to TNFa induction (Figure 1C). Despite the a gene loop abrogates contact between other interacting genes, clear asymmetric transcriptional response, colocalization of at least as measured by diffraction-limited colocalized DNA FISH RNA FISH foci was only observed in a subset of the population foci. Another possible explanation could be that the DNA repair (Figure 1B). The functional relationship between chromosomal machinery is occluding the entry of RNA Pol II to the other inter- contact and cotranscription of interacting members of a multi- acting genes. However, the observed unidirectional loss of tran- gene complex remains opaque. However, if loop-mediated scription between these three genes (Figure 4) excludes this contact is indeed a prerequisite for cotranscription, the RNA possibility. Thus, taken together, the most likely conclusion is FISH colocalization data suggest that only certain cells in the that the DSB serves to disrupt chromosomal contact between population may possess the correct spatial arrangements of interacting genes. The repair experiment reveals that restoration their chromosomes to permit such regulation (Figure 1B). Such of an intact SAMD4A gene loop in a sequence-independent variegated gene expression is suggested to occur in TADs, high- manner is sufficient to restore cotranscription, as well as coloc- ly conserved compartments within mammalian chromosome alization, of TNFAIP2 and SLC6A5 (Figure 6B). Therefore, we territories (Dixon et al., 2012). We speculate that TADs may speculate that, by disrupting chromosomal interactions, the to- ensure robust SAMD4A, TNFAIP2, and SLC6A5 gene regulation pological framework (comprised of gene loops and RNA Pol II)

Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc. 615 (legend on next page)

616 Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc. Figure 6. Repairing the Disrupted SAMD4A Gene Loop Restores Transcription of Genes in a Multigene Complex (A) HUVECs were stimulated for 20 min with TNFa, recapitulating transcription of the first 1.5 kbp of SAMD4A as well as the IRES-eGFP. Distinct eGFP foci were evident in 10% of transfected cells, and these foci overlapped with SAMD4A mRNA. n, number of cells (n = 118); scale bar, 5 mm. (B) There was no significant difference in the colocalization frequencies between eGFP/ SAMD4A, TNFAIP2, and SLC6A5 and mock- transfected cells. Cells were counterstained with DAPI; N, number of cells (N = 448); n, number of eGFP foci (n = 45); NS, no significant difference; scale bar, 5 mm.

complex that ‘‘organizes’’ transcription through loop-mediated contact. Recent published data in live cells reveals that RNA Pol II is mobile and clustering pre- cedes transcriptional elongation and is linked to transcriptional initiation (Cisse et al., 2013). Therefore, we speculate that the SAMD4A gene loop provides a topological platform that serves as a scaffold on which a focus of many RNA Pol II molecules can cluster and engage in transcription of subordinate members of the multigene complex. When TNFAIP2 or SLC6A5 cannot engage in chromosomal contact, their ability to ac- cess the focus of Pol II is limited. The strict hierarchical relationship between the interacting members further suggests a hand-off, or ‘‘collector’’ process, whereby dynamic chromosomal contacts with the dominant gene loop are formed (Figure 7). This could occur by a mecha- nism in which TNFAIP2, via chromosomal contact, collects its Pol II from an intact is unable to assemble, thus disrupting transcription. Collectively, SAMD4A gene loop and, in turn, SLC6A5 from an intact TNFAIP2 these data provide strong evidence that intact chromatin is a loop. Therefore, these data argue neither in favor of nor against requirement for loop-mediated cotranscription. putative ‘‘transcription factories’’ as they are currently defined One way that enhancer-promoter interactions are proposed to (Cook, 1999) but suggest that such factories may be dynamically enhance transcription is by bringing protein complexes to the assembled rather than immobile structures. A recent study re- promoter (Deng et al., 2012; Gibcus and Dekker, 2013). In an vealed that sequences within the promoter might drive the coloc- analogous manner, through interchromosomal interactions, alization between these NFkB-regulated genes (Larkin et al., NF-kB has been shown to be delivered to the promoter of induc- 2013). Therefore, an alternative model might involve sequence- ible genes (Apostolou and Thanos, 2008). Similarly, we propose specific elements within the promoters that facilitate the accu- SAMD4A to be the dominant member of the NF-kB multigene mulation of different thresholds of general transcription factors

Figure 5. The SAMD4A Gene Loop Was Successfully Repaired (A) Graphical representation of the repair strategy. The repair construct consists of an IRES and eGFP, flanked by splice sites. (B) TNFa-induced activation of SAMD4A induces eGFP expression. eGFP signal was observed in 1% of HUVECs dual transfected with the pDT vector and PCR product for 72 hr and was stimulated with TNFa for 24 hr. Cells were counterstained with DAPI; n, number of cells; scale bar, 5 mm. See also Figure S7.

Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc. 617 Figure 7. Hypothetical Model of Hierarchical Transcription between Coregulated Genes in a Multigene Complex TADs constrain genes to compartments in the nucleus that are admissible to long-range chromosomal interactions. Upon induction by TNFa, NF-kB-responsive genes engage in chromosomal interactions. As two of these genes reside on the same chromosome but almost 50 Mbp apart and another resides on a different chromosome, these interactions most likely occur within a TAD. Single-cell analysis 10 min post-TNFa stimulation reveals that the transcriptional response of interacting genes in a multigene complex is asymmetric (Figure 1C). Allelic analysis revealed that the following four categories were most prevalent: (a) no transcription of any gene, (b) SAMD4A expression only, (c) SAMD4A, SLC6A5, and TNFAIP2 combined expression, and (d) combined expression of SAMD4A and TNFAIP2. Despite the low frequency of these interactions (Figure 1D), disrupting the sites within the gene loops that engage in chromosomal contact revealed a hierarchical organization between interacting genes in this TNFa-induced multigene complex. for the three genes. Therefore, only upon the creation of a nu- deletions, chromosomal translocations, or silencing of transcrip- clear subcompartment of sufficient activity by SAMD4A can tion factors, all of which disrupt loop-mediated contact—may TNFAIP2 and then SLC6A5 be activated. inadvertently result in unintended consequences to transcription Our observation shifts the general paradigm of how transcrip- of other members of a given multigene complex. tional regulation in three dimensions occurs. Although these chromosomal interactions are rare and stochastic (Noordermeer EXPERIMENTAL PROCEDURES et al., 2011), our single-cell view strongly suggests that chromo- somal contact between genes engaged in multigene complexes Cell Culture have a significant impact on cotranscription of interacting genes. Early passage HUVECs from pooled donors (Lonza) were grown to 80% Such long-range interactions of cotranscribed genes could confluence in EGM-2 media with supplements (Lonza), serum-starved in EGM-2 + 0.5% FBS, and treated with TNFa (10 ng/ml; Sigma) for up to serve to organize transcription in nuclear space, using hierarchi- 30 min. cal relationships between gene loops. Without gene loops of dominant members, subordinate members of the multigene TALEN Synthesis complex cannot engage in long-range chromosomal contacts, TALENs were generated using the protocol by Sanjana et al., 2012. HUVECs nor can they participate in transcription. Finally, as looping were then microporated with the respective TALENs using the Neon Trans- enables stochastic chromosomal contact between genes in fection System (Life Technologies) according to the manufacturer’s multigene complexes, this study provides supporting evidence instructions. Nuclease activity was assessed by the surveyor assay and the to prior work showing that gene looping is a fundamental re- consistent colocalization of DNA FISH with DSBs, as assessed by H2A.X immunofluorescence. quirement of transcriptional activity (Tan-Wong et al., 2012). Importantly, such chromatin looping within multigene complexes Surveyor Assay (Li et al., 2012; Sanyal et al., 2012; Djebali et al., 2012) may be Genomic DNA of transfected cells was extracted using QuickExtract DNA governed by similar hierarchical regulation. Perturbation of extraction solution (Epicenter). For the ‘‘surveyor’’ or T7E1 assay, DNA was de- members of multigene complexes—through knockouts, gene natured at 95C for 5 min, slowly cooled down to room temperature to allow for

618 Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc. formation of heteroduplex DNA, treated with 5 U of T7E1 for 1 hr at 37C, and Barbieri, M., Chotalia, M., Fraser, J., Lavitas, L.-M., Dostie, J., Pombo, A., and then analyzed by 1.2% agarose gel electrophoresis. Nicodemi, M. (2013). A model of the large-scale organization of chromatin. Biochem. Soc. Trans. 41, 508–512. RNA FISH Probes Bedell, V.M., Wang, Y., Campbell, J.M., Poshusta, T.L., Starker, C.G., Krug, RNA FISH was performed according to the protocol by Raj et al. (2008) using R.G., 2nd, Tan, W., Penheiter, S.G., Ma, A.C., Leung, A.Y., et al. (2012). In vivo 48 20-mer probes (Biosearch). Following the conjugation reaction, probes genome editing using a high-efficiency TALEN system. Nature 491, 114–118. were separated and purified to enrich for dye-conjugated probes by Boch, J., Scholze, H., Schornack, S., Landgraf, A., Hahn, S., Kay, S., Lahaye, reverse-phase HPLC on a C18 column. T., Nickstadt, A., and Bonas, U. (2009). Breaking the code of DNA binding specificity of TAL-type III effectors. Science 326, 1509–1512. Immuno-RNA RNA FISH Brown, J.M., Green, J., das Neves, R.P., Wallace, H.A., Smith, A.J., Hughes, For each experiment, HUVECs on coverslips were grown to 80% confluence. J., Gray, N., Taylor, S., Wood, W.G., Higgs, D.R., et al. (2008). Association Double-strand breaks were detected with rabbit polyclonal anti-phospho- between active genes occurs at nuclear speckles and is modulated by chro- histone H2A.X (Ser139) (Sigma). Goat polyclonal anti-SAMD4A C-15, mouse matin environment. J. Cell Biol. 182, 1083–1097. monoclonal anti-TNFAIP2 F-6, and goat polyclonal anti-SLC6A5/GLYT2 N-20 (Santa Cruz) were used to detect SAMD4A, TNFAIP2, and SLC6A5 pro- Christian, M., Cermak, T., Doyle, E.L., Schmidt, C., Zhang, F., Hummel, A., teins, respectively. For RNA FISH detection, cells were hybridized overnight in Bogdanove, A.J., and Voytas, D.F. (2010). Targeting DNA double-strand a humidified chamber at 37C with 50 ng of RNA FISH probes. breaks with TAL effector nucleases. Genetics 186, 757–761. Cisse, I.I., Izeddin, I., Causse, S.Z., Boudarene, L., Senecal, A., Muresan, L., Statistics Dugast-Darzacq, C., Hajj, B., Dahan, M., and Darzacq, X. (2013). Real-time p values from two-tailed unpaired Student’s t test, Mann-Whitney U tests, and dynamics of RNA polymerase II clustering in live human cells. Science 341, Fisher Exact tests were calculated using R (http://www.R-project.org).*p< 664–667. 0.05, ** p < 0.01, and ***p < 0.001. Cook, P.R. (1999). The organization of replication and transcription. Science 284, 1790–1795. SUPPLEMENTAL INFORMATION Cremer, T., and Cremer, M. (2010). Chromosome territories. Cold Spring Harb. Perspect. Biol. 2, a003889. Supplemental Information includes Extended Experimental Procedures and Dekker, J., Rippe, K., Dekker, M., and Kleckner, N. (2002). Capturing chromo- seven figures and can be found with this article online at http://dx.doi.org/ some conformation. Science 295, 1306–1311. 10.1016/j.cell.2013.09.051. Deng, W., Lee, J., Wang, H., Miller, J., Reik, A., Gregory, P.D., Dean, A., and Blobel, G.A. (2012). Controlling long-range genomic interactions at a native AUTHOR CONTRIBUTIONS locus by targeted tethering of a looping factor. Cell 149, 1233–1244. Dixon, J.R., Selvaraj, S., Yue, F., Kim, A., Li, Y., Shen, Y., Hu, M., Liu, J.S., and M.M.M. designed the overall study with contributions from S.F. S.F. designed Ren, B. (2012). Topological domains in mammalian genomes identified by and carried out experiments, collected and analyzed data, and cowrote the analysis of chromatin interactions. Nature 485, 376–380. paper. Y.S. designed and carried out experiments and collected and analyzed data with S.F. and M.M.M. S.B. carried out experiments, adapted the rapid Djebali, S., Davis, C.A., Merkel, A., Dobin, A., Lassmann, T., Mortazavi, A., TALEN assembly protocol, and analyzed data with Y.S. M.S.W. and M.M.M. Tanzer, A., Lagarde, J., Lin, W., Schlesinger, F., et al. (2012). Landscape of designed the vector for the repair experiment. M.S.W. constructed the repair transcription in human cells. Nature 489, 101–108. vector. S.F. and Y.S. carried out the repair experiment. S.F., Y.S., M.S.W., Fullwood, M.J., Liu, M.H., Pan, Y.F., Liu, J., Xu, H., Mohamed, Y.B., Orlov, Y.L., and M.M.M. discussed and edited the paper. M.M.M. supervised this study, Velkov, S., Ho, A., Mei, P.H., et al. (2009). An oestrogen-receptor-alpha-bound designed and performed experiments, analyzed data, and wrote the paper. human chromatin interactome. Nature 462, 58–64. Gibcus, J.H., and Dekker, J. (2013). The hierarchy of the 3D genome. Mol. Cell ACKNOWLEDGMENTS 49, 773–782. Harewood, L., Schu¨ tz, F., Boyle, S., Perry, P., Delorenzi, M., Bickmore, W.A., We thank all members of the Gene Expression and Biophysics Laboratory. We and Reymond, A. (2010). The effect of translocation-induced nuclear reorgani- thank A. Lenox for experimental help and advice. We thank A. Premo for assis- zation on gene expression. Genome Res. 20, 554–564. tance with graphics and cartoons and S. Barichievy, A. Gontijo, F. Brom- bacher, P. Tomancak, and D. Henriques for comments on the manuscript. Hoeijmakers, J.H.J. (2001). Genome maintenance mechanisms for preventing S.F. and Y.S. are recipients of postdoctoral fellowships from the Claude cancer. Nature 411, 366–374. Leon Foundation, South Africa. This research is supported by grant PG- Larkin, J.D., Cook, P.R., and Papantonis, A. (2012). Dynamic reconfiguration V2KYP07 from the Council for Industrial and Scientific Research (CSIR, South of long human genes during one transcription cycle. Mol. Cell. Biol. 32, Africa) and by a grant from the Emerging Research Area Program of The 2738–2747. Department of Science and Technology (DST, South Africa) and grant Larkin, J.D., Papantonis, A., and Cook, P.R. (2013). Promoter type influences PTDC/SAU-GMG/115652/2009 from the Fundac¸a˜ o para a Cieˆ ncia e a Tecno- transcriptional topography by targeting genes to distinct nucleoplasmic sites. logia (FCT, Portugal), all to M.M.M. J. Cell Sci. 126, 2052–2059. Levesque, M.J., and Raj, A. (2013). Single-chromosome transcriptional Received: April 26, 2013 profiling reveals chromosomal gene expression regulation. Nat. Methods 10, Revised: July 18, 2013 246–248. Accepted: September 23, 2013 Published: October 24, 2013 Li, T., Huang, S., Jiang, W.Z., Wright, D., Spalding, M.H., Weeks, D.P., and Yang, B. (2011). TAL nucleases (TALNs): hybrid proteins composed of TAL REFERENCES effectors and FokI DNA-cleavage domain. Nucleic Acids Res. 39, 359–372. Li, G., Ruan, X., Auerbach, R.K., Sandhu, K.S., Zheng, M., Wang, P., Poh, Apostolou, E., and Thanos, D. (2008). Virus Infection Induces NF-kappaB- H.M., Goh, Y., Lim, J., Zhang, J., et al. (2012). Extensive promoter-centered dependent interchromosomal associations mediating monoallelic IFN-beta chromatin interactions provide a topological basis for transcription regulation. gene expression. Cell 134, 85–96. Cell 148, 84–98.

Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc. 619 Lieberman-Aiden, E., van Berkum, N.L., Williams, L., Imakaev, M., Ragoczy, Pombo, A., Jackson, D.A., Hollinshead, M., Wang, Z., Roeder, R.G., and Cook, T., Telling, A., Amit, I., Lajoie, B.R., Sabo, P.J., Dorschner, M.O., et al. P.R. (1999). Regional specialization in human nuclei: visualization of discrete (2009). Comprehensive mapping of long-range interactions reveals folding sites of transcription by RNA polymerase III. EMBO J. 18, 2241–2253. principles of the . Science 326, 289–293. Raj, A., van den Bogaard, P., Rifkin, S.A., van Oudenaarden, A., and Tyagi, S. Miele, A., and Dekker, J. (2008). Long-range chromosomal interactions and (2008). Imaging individual mRNA molecules using multiple singly labeled gene regulation. Mol. Biosyst. 4, 1046–1057. probes. Nat. Methods 5, 877–879. Moscou, M.J., and Bogdanove, A.J. (2009). A simple cipher governs DNA Rogakou, E.P., Boon, C., Redon, C., and Bonner, W.M. (1999). Megabase recognition by TAL effectors. Science 326, 1501. chromatin domains involved in DNA double-strand breaks in vivo. J. Cell Noordermeer, D., de Wit, E., Klous, P., van de Werken, H., Simonis, M., Lopez- Biol. 146, 905–916. Jones, M., Eussen, B., de Klein, A., Singer, R.H., and de Laat, W. (2011). Variegated gene expression caused by cell-specific long-range DNA interac- Sandhu, K.S., Li, G., Poh, H.M., Quek, Y.L., Sia, Y.Y., Peh, S.Q., Mulawadi, tions. Nat. Cell Biol. 13, 944–951. F.H., Lim, J., Sikic, M., Menghi, F., et al. (2012). Large-scale functional organi- zation of long-range chromatin interaction networks. Cell Rep. 2, 1207–1219. Orlando, S.J., Santiago, Y., DeKelver, R.C., Freyvert, Y., Boydston, E.A., Moehle, E.A., Choi, V.M., Gopalan, S.M., Lou, J.F., Li, J., et al. (2010). Zinc- Sanyal, A., Lajoie, B.R., Jain, G., and Dekker, J. (2012). The long-range finger nuclease-driven targeted integration into mammalian genomes using interaction landscape of gene promoters. Nature 489, 109–113. donors with limited chromosomal homology. Nucleic Acids Res. 38, e152. Schoenfelder, S., Sexton, T., Chakalova, L., Cope, N.F., Horton, A., Andrews,  Orlova, D.Y., Stixova´ , L., Kozubek, S., Gierman, H.J., Susta´ ckova ´ , G., Cherny- S., Kurukuti, S., Mitchell, J.A., Umlauf, D., Dimitrova, D.S., et al. (2010). shev, A.V., Medvedev, R.N., Legartova´ , S., Versteeg, R., Matula, P., et al. Preferential associations between co-regulated genes reveal a transcriptional (2012). Arrangement of nuclear structures is not transmitted through mitosis interactome in erythroid cells. Nat. Genet. 42, 53–61. but is identical in sister cells. J. Cell. Biochem. 113, 3313–3329. Spilianakis, C.G., Lalioti, M.D., Town, T., Lee, G.R., and Flavell, R.A. (2005). Osborne, C.S., Chakalova, L., Brown, K.E., Carter, D., Horton, A., Debrand, E., Interchromosomal associations between alternatively expressed loci. Nature Goyenechea, B., Mitchell, J.A., Lopes, S., Reik, W., and Fraser, P. (2004). 435, 637–645. Active genes dynamically colocalize to shared sites of ongoing transcription. Nat. Genet. 36, 1065–1071. Strickfaden, H., Zunhammer, A., van Koningsbruggen, S., Ko¨ hler, D., and Cremer, T. (2010). 4D chromatin dynamics in cycling cells: Theodor Boveri’s Palstra, R.J., Simonis, M., Klous, P., Brasset, E., Eijkelkamp, B., and de Laat, hypotheses revisited. Nucleus 1, 284–297. W. (2008). Maintenance of long-range DNA interactions after inhibition of ongoing RNA polymerase II transcription. PLoS One 3, e1661. Tan-Wong, S.M., Zaugg, J.B., Camblong, J., Xu, Z., Zhang, D.W., Mischo, Papantonis, A., Larkin, J.D., Wada, Y., Ohta, Y., Ihara, S., Kodama, T., and H.E., Ansari, A.Z., Luscombe, N.M., Steinmetz, L.M., and Proudfoot, N.J. Cook, P.R. (2010). Active RNA polymerases: mobile or immobile molecular (2012). Gene loops enhance transcriptional directionality. Science 338, machines? PLoS Biol. 8, e1000419. 671–675. Papantonis, A., Kohro, T., Baboo, S., Larkin, J.D., Deng, B., Short, P., Wada, Y., Ohta, Y., Xu, M., Tsutsumi, S., Minami, T., Inoue, K., Komura, D., Tsutsumi, S., Taylor, S., Kanki, Y., Kobayashi, M., et al. (2012). TNFa signals Kitakami, J., Oshida, N., Papantonis, A., et al. (2009). A wave of nascent tran- through specialized factories where responsive coding and miRNA genes scription on activated human genes. Proc. Natl. Acad. Sci. USA 106, 18357– are transcribed. EMBO J. 31, 4404–4414. 18361.

620 Cell 155, 606–620, October 24, 2013 ª2013 Elsevier Inc.