The organization of a eukaryotic nucleus reflects its specific expression profile1,2. On the genomic scale, this translates to preferred tissue-specific chromatin folding and chromosome positioning within interphase nuclei, including intermingling of looped segments within and between chromosomes3,4. Consequently, co-regulated genes tend to localize near each other at activity hubs. However, the description of colocalized and coexpressed genes has been anecdotal and limited to a few genes. What had been missing until recently5,6 was information on how extensive and consistent these associations are at the genomic level. On page 53 of this issue, Peter Fraser and colleagues7 use an array of biochemical and cytogenetic approaches to elegantly describe such a genomic transcription interactome by exploring the gene associations of the mouse α- (Hba) and β-globin (Hbb) loci in erythroid cells.

Globins cast a wide net

The globin genes are highly expressed in developing erythrocytes, and they consistently associate with transcription factories, which are foci of hyperphosphorylated RNA polymerase II (RNAPII) and are sites of active transcription. Using double-label RNA fluorescence in situ hybridization (FISH), Schoenfelder et al.7 discovered that several expressed erythroid-specific genes (of a set of 33 examined loci on 15 chromosomes) frequently localize to the same transcription factories as Hba or Hbb. To expand on this initial observation, Schoenfelder et al. used a modified chromosome conformation capture technique (enhanced ChIP-4C, or e4C) to establish the genomic interactome of the globin loci. They precede the assay with a chromatin immunoprecipitation (ChIP) against hyperphosphorylated RNAPII and a bait-specific (globin) enrichment step involving a biotin pulldown. These clever adaptations ensure that the detected associations are restricted to actively transcribed genes and thus result in a highly improved signal-to-noise ratio. Consequently, the results provide the first truly robust and reproducible large-scale detection of long-range and interchromosomal associations between active genes. Microarray analysis of the e4C material confirms known interactions between Hbb and erythroid-specific genes (for example, Eraf and Uros) and hundreds more. Strikingly, around 90% of these detected interactions were in trans for both Hbb and Hba. The authors performed elaborate controls to demonstrate that virtually all ligation products obtained via the e4C protocol originated from individual cross-linked complexes and thus represent bona fide long-distance interactions. Moreover, paired-end tag sequencing (PET) of phosphorylated RNAPII ChIP material confirmed that between 80% and 90% of e4C hits overlapped with at least one annotated gene, though a correlation with actual expression data might have provided further confirmation that interacting genes were indeed transcriptionally active. Multicolor RNA FISH also revealed that up to three different RNA species (limited to three only by technical constraints) could colocalize at a common transcription factory, strongly suggesting that multiple genes can be transcribed at a single factory (Fig. 1). Importantly, the data reveal that the Hba and Hbb networks are gene specific, as close to 80% of associated genes were unique to either Hba or Hbb and only a minority was common to both.

Figure 1: Gene associations at transcription factories in the erythroid nucleus.
figure 1

The Hbb locus (like Hba) associates with other active genes both in cis and in trans at transcription factories (left). Specialized transcription factories containing the transactivator Klf1 mediate the preferential interaction between Klf1-regulated genes to boost their expression (right). Hba and Hbb may interact with members of their own network and/or each other at the same factory. RNAPII, red; Klf1 protein, blue; genes, striped regions. Chromatin loops from the same chromosome associating in cis at a factory are depicted in the same color (for example, blue, green).

Full size image

Searching for common motifs among Hba- and Hbb-interacting genes, the authors discovered that many share binding sites for the erythroid-specific transactivator Klf1. As it turns out, in erythroid nuclei, Klf1 is present in relatively few foci that overlap with a subset of transcription factories. It is these Klf1-containing RNAPII foci that associate prominently with Klf1-regulated genes and are thus best described as specialized transcription factories (Fig. 1). Supporting evidence comes from the analysis of Klf1 knockout cells, in which associations between Hbb and other Klf1-regulated genes, as well as their colocalization with transcription factories, are disrupted. Non–Klf1 target associations, on the other hand, remained largely unaffected.

Clearly, the demonstrated interactions of Hbb and Hba with other transcribed genes do not occur simultaneously in each erythroid nucleus. Instead, given the dynamics of nuclear organization and the stochastic nature of gene transcription, at any given time a smaller and varied subset of interactions is to be expected. Still, the study reveals a genomic view of preferred active interaction partners of two network hubs (Hba and Hbb), as well as the existence of long-suspected, but until now not demonstrated, specialized transcription factories that boost the expression of erythroid-specific genes.

Anchors away

Since the description of the 3C method by Dekker et al.8, variations of this technique have been used to probe the local conformation of complex gene loci as well as genomic regions of ever-increasing size9,10,11,12. Almost all of these studies, including the work by Schoenfelder et al.7, have been anchored to a particular locus with specific bait primers, which increases the specificity of the assay but also biases it. Interactions between other loci of the network (such as that between Uros and Eraf here) are not registered. Another recent study describes the interactome of estrogen receptor–bound chromatin in a human epithelial adenocarcinoma cell line by ChIP and PET without the use of an anchor6. The results revealed an unbiased map of estrogen receptor binding sites and their interactions. However, the interactome failed to be truly genomic, as interchromosomal interactions between binding sites were not reproducibly detected, possibly in part due to the variable and abnormal karyotype of these cells. A promising alternative for probing genomic interactions may be the recently published Hi-C method5. As this method is also sequencing based and unbiased, it is capable of reliably uncovering highly detailed chromosomal interactions both in cis and in trans. It could easily be adapted to include a ChIP step to restrict the analysis to tissue-, pathway- or activity-specific chromatin-gene networks. Regardless, an important remaining question is whether the preferential gene associations at common transcription factories (in particular the specialized type) or other nuclear bodies occur by self-assembly or active recruitment.