Cohesins form chromosomal cis-interactions at the developmentally regulated IFNG locus.

Cohesin-mediated sister chromatid cohesion is essential for chromosome segregation and post-replicative DNA repair. In addition, evidence from model organisms and from human genetics suggests that cohesin is involved in the control of gene expression. This non-canonical role has recently been rationalized by the findings that mammalian cohesin complexes are recruited to a subset of DNase I hypersensitive sites and to conserved noncoding sequences by the DNA-binding protein CTCF. CTCF functions at insulators (which control interactions between enhancers and promoters) and at boundary elements (which demarcate regions of distinct chromatin structure), and cohesin contributes to its enhancer-blocking activity. The underlying mechanisms remain unknown, and the full spectrum of cohesin functions remains to be determined. Here we show that cohesin forms the topological and mechanistic basis for cell-type-specific long-range chromosomal interactions in cis at the developmentally regulated cytokine locus IFNG. Hence, the ability of cohesin to constrain chromosome topology is used not only for the purpose of sister chromatid cohesion, but also to dynamically define the spatial conformation of specific loci. This new aspect of cohesin function is probably important for normal development and disease.

One model of CTCF function is the formation of chromatin loops, which have been demonstrated at the H19/IGF2, beta globin and MHC class II loci15-17. Whether cohesin controls chromosome conformation is important for understanding its contribution to insulator and boundary element function18 and for elucidating the significance of cohesin binding to genomic locations unlikely to constitute insulators or chromatin boundaries, including sites within the coding regions of active genes10,11.

Cohesin binding within a developmentally regulated gene is exemplified by IFNG, which encodes the cytokine interferon-γ (Ifn-γ). A conserved cohesin site is located within the first intron of the mouse Ifng and the human IFNG coding region10. The Ifng locus harbors numerous conserved non-coding sequences and putative cis-regulatory elements that are located at considerable distances from the coding region19-21 and can function as enhancers or insulators20. Genomic tiling arrays10 showed major binding sites for the cohesin subunit Rad21 and CTCF at −73 and +66 kb in mouse lymphoid cell lines (figure 1a). ChIP and real time PCR analysis of the human 293T cell line and of primary CD4 T cells sorted from human peripheral blood demonstrated strong binding of cohesin and CTCF at orthologous sequences located 63 kb upstream of IFNG and at evolutionary conserved regions (ECR, www.decode.org) at +1.5 and +119 kb in the human IFNG region (figure 1b, see legend for details). Analysis of genome-wide ChIP-sequencing data22 confirmed −63, +1.5 and +119 kb as the major CTCF sites in human CD4 T cells and showed that features of canonical CTCF sites, namely the histone H2A variant H2AZ and trimethylated histone H3 lysine-4 (H3K4me3) were present at −63 and +1.5 kb (not shown).

Figure 1

Developmentally regulated cohesin and CTCF binding at IFNG

a) ChIP and genomic tiling array data for the mouse B3 pre-B cell line10 show log2 enrichment of the cohesin subunit Rad21 (black) and CTCF (gray) at Ifng. Schematic representation of the mouse Ifng region (m chr.10) and the human IFNG region (h Chr.12), a flash indicates genomic rearrangement at Il26, a pseudogene in rodents20.

b) ChIP and real time PCR mapping of Rad21 (top) and CTCF (bottom) at human IFNG in 293T cells (white), non-polarized CD4 T cells (grey), CCR5+ Th1 (red) and CRTh2+ Th2 (black) effector memory cells. Star symbols indicate statistical significance (Student’s T test) *** p<0.005, ** p<0.05, * p<0.05. Primer positions are indicated relative to the IFNG transcription start site. Human/mouse identity at the −63 kb site is 59% over 291bp surrounding a canonical CTCF consensus motif. Two evolutionary conserved regions (ECR, www.dcode.org) are found at +1.5 kb (70.2% and 69.1% identity over 104 and 217bp) and three at +119 kb (70.0%, 71.0% and 62.2 and 69,1% identity over 120, 100 and 45bp). Positive control sites at Chr11:118283kb, Chr11:118333kb, CD8 Cluster I and ZPF54 were included in all experiments (supplementary figure 1) and data are normalized to Chr11:118283kb (mean±SD, n=4).

c) Isolation (left) of human effector memory CCR5+ Th1 and CRTh2+ Th2 cells and cytokine expression assessed by intracellular staining and flow cytometry (middle) or real time RT-PCR analysis before (contr.) and after activation with plate bound anti-CD3 and anti-CD28 (activ., right).

Naive, non-polarized CD4 T cells are the progenitors of specialized T helper (Th) cell types, including Th1 cells (which confer immunity to intracellular pathogens) and Th2 cells (which aid antibody production and mediate responses to helminths). Th cell differentiation has become a paradigm for how genetically identical cells acquire distinct gene expression programs by the interplay of transcription factors and chromatin regulators23. The IFNG locus is prepared for expression during the differentiation of Th1 cells: DNA methylation declines20,24, cell type-specific HSS emerge19-21 and IFNG transcription becomes inducible23. Conversely, IFNG is silenced during Th2 differentiation23. To gain insight into the developmental regulation of cohesin and CTCF at human IFNG we isolated Th1 and Th2 effector memory cells (figure 1c). These cells arise in vivo, can be identified based on the expression of the chemokine receptor CCR5 and the prostaglandin D2 receptor CRTh2, respectively25, and retain their cytokine profiles when expanded in vitro under non-polarising conditions25. Real time RT-PCR showed that IFNG transcripts were highly induced upon activation of CCR5+ Th1 cells (figure 1c) and intracellular staining showed the inducible expression of IFNγ protein in CCR5+ Th1 cells, while CRTH2+ Th2 cells expressed interleukin-4 (IL4, figure 1c). The positioning of CTCF and cohesin binding sites across IFNG was similar between non-polarized CD4 T cells, CCR5+ Th1 cells, CRTH2+ Th2 cells and non-lymphoid 293T cells (figure 1b), consistent with the reported preference of CTCF and cohesin for constitutive HSS10. Interestingly, however, both cohesin and CTCF were significantly more abundant at the −63, +1.5 and +119 kb sites in CCR5+ Th1 cells (figure 1b). Increased cohesin and CTCF binding in CCR5+ Th1 cells was selective for the IFNG locus, since CCR5+ Th1 and CRTH2+ Th2 cells expressed similar levels of Rad21 and CTCF (supplementary figure 1a), and the occupancy of control sites outside the IFNG region was equivalent (supplementary figure 1b).

Analysis of activating and repressive post-translational histone modifications across the IFNG region suggested boundary element function for the −63 site, but not the +1.5 or the +119 kb sites in non-polarised human CD4 T cells26. We therefore explored whether these sites have additional functions in organising the IFNG region in 3-dimensional nuclear space by chromosome conformation capture27 (3C) experiments on CCR5+ Th1 and CRTH2+ Th2 chromatin templates (supplementary figure 2). The inducible nature of IFNG expression in CCR5+ Th1 cells allowed us to define the chromosome topology of the locus in its inducible state from changes in locus conformation that may accompany high-level transcription. Using a Hind III restriction fragment harboring the intronic +1.5 kb site as a bait, we looked for long-range interactions with sites across the IFNG genomic region. Crosslinking frequencies fell with increasing distance from the bait, but were significantly increased for the CTCF and cohesin sites at −63 and +119 kb, indicating long-range interactions at IFNG that occur selectively in CCR5+ Th1 cells (figure 2a). Nuclear dimensions were similar for CCR5+ Th1 and CRTH2+ Th2 cells (figure 2b), which rules out spatial constraints as an explanation for increased crosslinking efficiencies in CCR5+ Th1 cells. Since the IFNG loci on homologous chromosomes were separated by at least 1.5 μm in 3-dimensional fluorescence in situ hybridisation (3D FISH) experiments (figure 2b) and long-range interactions in the IFNG genomic region were readily detectable in purified G1 cells (i.e. prior to DNA replication and the formation of sister chromatids, figure 2c), the interactions recorded by 3C most likely occurred in cis, not in trans between IFNG alleles. These data establish that CTCF and cohesin sites form the topological basis for cell type-specific, long-range interactions at IFNG that occur preferentially in human CCR5+ Th1 cells.

Figure 2

Cell type-specific long-range chromosomal interactions at IFNG are based on CTCF/cohesin sites

a) Chromosome conformation capture (3C) and Taqman PCR were used to determine interactions between IFNG +1.5 kb (beacon) and primers (black arrowheads) placed within the indicated HindIII fragments (shaded blue) in non-polarized CD4 T cells (grey), CCR5+ Th1 (red) and CRTh2+ Th2 (black) cells (supplementary figure 2 and table 1). Crosslinking efficiencies are expressed relative to +119 kb in CCR5 cells and normalised to mouse Acta2 (see supplementary methods and supplementary figure 2) or to human ACTA2 with similar results (supplementary figure 3a). Cohesin sites and the approximate positions of the IFNG and IL26 genes are indicated (mean±SD, n=5 for CCR5 and CRTh2, n=4 for non-polarized CD4 T cells). *** p<0.005 (Student’s T test, Th1 vs. non-polarized and Th1 vs. Th2).

b) 3D-FISH to determine nuclear diameters of CCR5+ Th1 (red, 9.4+/−1.1 μm, n=50) and CRTh2+ Th2 cells (black, 9.5+/−1.5 μm, n=50) and distances between IFNG alleles in individual CCR5 (red, 4.5+/−1.6 μm, n=50, minimum 1.5 μm) and CRTh2 cells (black, 5.0+/−1.9 μm, n=50, minimum 1.1 μm).

c) Long range interactions at IFNG in CCR5+ Th1 cells occur prior to DNA replication. 3C analysis of IFNG in total CCR5+ Th1 cells (red) compared to CCR5+ Th1 cells in the G1 phase of the cell cycle (orange). Data are presented as in a), (mean±SE, n=2).

We next explored the relationship of CTCF and cohesin with long-range chromosomal interactions. Depletion of CTCF by siRNA-mediated knockdown in CCR5+ Th1 cells resulted in reduced 3C interactions between the IFNG locus (+1,5 kb) and the −63 and +119 kb sites (supplementary figure 4), reminiscent of the role for CTCF in long-range interactions at other loci15-17. Importantly, Rad21 knockdown substantially reduced long-range interactions of the IFNG coding region with the −63 and the +119 kb sites (figure 3a), demonstrating that the conformation of the locus depended on cohesin as well as CTCF. ChIP analysis of the IFNG region in CCR5+ Th1 cells showed that Rad21 RNAi reduced the binding of Rad21 to its sites at IFNG, while CTCF remained bound (figure 3b). This result suggests that CTCF cannot maintain long-range interactions at IFNG in the absence of cohesin. Control experiments established that Rad21 knockdown did not interfere with technical aspects of the 3C methodology (supplementary figure 2) or with cell cycle progression of human CCR5+ Th1 cells (supplementary figure 5). Interestingly, basal INFG transcript levels were reduced by half (50±13%, n=5, p<0.005, one sided T test) in Rad21 depleted CCR5+ Th1 cells (figure 3c) and the inducibility of INFG transcripts upon T cell activation was reduced to 76±12% of control levels (figure 3c, n=4, p<0.05). These effects were selective since the inducibility of IL2 was not compromised (115±30%, n=3) and the expression of IGF2, which is restricted to the paternal allele by cohesin11, was elevated (247±148%, n=3, p<0.05) in Rad21 depleted CCR5+ Th1 cells (figure 3c). The basal and inducible expression of IFNγ protein were also reduced (to 48±9%, n=4, p<0.05 and 57±7%, n=4, p<0.05) by Rad21 knockdown (figure 3c). Hence, cohesin is required for the CCR5+ Th1 cell-specific conformation of the IFNG locus and contributes to its regulation.

Figure 3

Long-range chromosomal interactions at IFNG require cohesin

a) CCR5+ Th1 cells were transfected with control (red) or Rad21 (light blue) siRNA oligonucleotides. Long-range interactions at IFNG were assessed by 3C (mean±SE, n=4 for −63, −38 and +119kb, mean±SE, n=2 for −98, −28, +42, +78 and +142kb). Data are expressed relative to +119 kb in control siRNA treated CCR5+ cells after normalisation to mouse Acta2 in admixed mouse chromatin, which also served to control for technical aspects of the 3C methodology (supplementary figure 2). Normalisation to human ACTA2 gave similar results (supplementary figure 3b). *** p<0.005, ** p<0.05 (Student’s T test). The expression of CTCF and Rad21 was assessed by immunoblotting 72 hours after transfection of CCR5+ Th1 cells with Rad21 or control siRNA. Lamin is a control.

b) ChIP for Rad21 (top) and CTCF (bottom) occupancy of IFNG 72 hours after transfection of CCR5+ Th1 cells with control (red) or Rad21 siRNA (light blue). Primer positions are indicated in kb (mean± SE, n=2).

c) Real time RT-PCR analysis of CCR5+ Th1 cells 72 hours after transfection with Rad21 siRNA or control siRNA. Rad21 and CTCF are shown as knockdown controls (top panel, mean± SD, n=5). Levels of IFNG are shown one week (basal, mean± SD, n=5) or 4 hours after T cell activation (induced, mean± SD, n=4) with anti-CD3 and anti-CD28. IL2 is shown as an activation control (mean± SD, n=3) and IGF2 as known cohesin target (mean± SD, n=3). Data are normalised to GAPDH, HPRT and TBP and shown relative to control siRNA transfected CCR5 cells (middle panels). ELISA for basal (mean± SD, n=4) and activation-induced (mean± SD, n=4) IFNγ protein secretion. Data are shown relative to control siRNA CCR5 cells (bottom panels).

Taken together, these data show that the remodelling of the developmentally regulated IFNG locus during the differentiation of naive CD4 T cells into Th1 cells is accompanied by the enhanced association of CTCF and cohesin with a select set of conserved sequence elements. CTCF and cohesin - perhaps in collaboration with other factors or modifications - mediate Th1-specific long-range interactions between these sites. Recent data provide support for an ‘architectural role’ of CTCF/cohesin sites in chromosome conformation28. Our data show that long-range interactions mediated by cohesin can be dynamically regulated in a developmental context, rather than static in nature. An attractive scenario is that CTCF recruits cohesins to specific sites10-13, and that cohesins in turn mediate chromosomal interactions between these sites in cis (this study). This novel aspect of cohesin function is likely to be of general importance in defining local and global genome topology. Cohesin-mediated chromosomal cis-interactions provide one possible mechanism for the impact of cohesin on the regulation of H19/Igf2 (reference 11​, 9, and J-M Peters and A Murrell, personal communication), of cytokine loci such as IL4, IL5, IL13 and Ifng, MHC class II genes17 and of other loci9,11 that depend on CTCF/cohesin for their appropriate expression. We speculate that by constraining the physical conformation of chromatin, cohesin may affect the probability with which gene regulatory elements interact with each other, both negatively and positively, in normal development3-7 and in disease7.