Transcription and recombination factories: Common features?

Abstract

There is now substantial evidence that the eukaryotic nucleus consists of highly organized structures. Among such structures are transcription factories that consist of an ensemble of genes recruited by the RNA polymerase machinery. Here we suggest that antigen receptor variable regions are similarly organized. Specifically, we propose that the immunoglobulin heavy chain locus variable gene segments are anchored to the base of rosettes, wrapping around a cavity that contains the recombination machinery. We suggest that the folding of the chromatin fiber into rosettes underpins a critical mechanism by which antigen receptor diversity is generated.

Introduction

The chromatin fiber is not randomly organized but folds into elaborate patterns to allow high-density packing and long-range genomic interactions to occur with the appropriate frequencies. The chromatin fiber is organized into nucleosomes, consisting of 146 bp DNA elements that surround octamers of histones. Specifically, two copies of H2A, H2B, H3 and H4 form the core of the nucleosome. The nucleosomes themselves are organized into a 10 nm fiber, which in turn, folds into a 30 nm chromatin fiber.

Our knowledge about the folding of chromatin beyond the 30 nm fiber is still rudimentary. Distinct patterns for the folding of the chromatin fiber have been proposed. These involve helical and radial structures that permit packing at relatively high densities [1,2,3]. Studies using electron microscopy have suggested that chromosomes are organized as loops that are clustered as rosettes [4,5].

To describe chromatin topology in quantitative terms, polymer models that can be experimentally tested have been generated. Prominent among these are the Random Walk/Giant Loop (RW/GL), the Multi-Loop-Subcompartment (MLS) and Random-Loop (RL) models [6,7,8]. The RW/GL model describes the chromatin fiber as being confined to relatively large loops (2–5Mbp) [7,9]. The MLS configuration suggests that the chromatin fiber folds into bundles of loops [8]. The bundles consist of approximately ten loops and together span on average 1 Mbp of DNA. Flexible linkers of variable sizes have been proposed to separate the bundles of loops [8]. More recently, yet another design, the RL model has been proposed to underpin long-range chromatin topology [10]. The RL configuration allows both small and large loops to fold and unfold in a dynamic fashion [11]. Here we will discuss how novel computational, geometric and genome-wide approaches have provided new insights into long-range chromatin structure and propose that transcription and recombination factories have common structural features.

The structure of antigen receptor loci

Knowledge of how genetic loci are folded in 3D-space is still rudimentary. Perhaps the best-characterized structure involves the immunoglobulin heavy chain locus [12]. The Igh locus consists of distinct DNA elements encoding the variable (V), diversity (D), joining (J), and constant (C) regions. It is the largest known genetic locus. Fifteen partially dispersed V_H region families, encoding for approximately 195 V_H regions span approximately 3 Mbp of the murine genome. Large intergenic regions that span up to 50 kbp in size separate the individual V_H regions. Located down-stream of the V_H regions are 10–13 D_H elements, four J_H elements, and eight C_H regions encoding for the various isotypes.

Using a geometric approach, named trilateration, the mean relative 3D-positions of the V_H, D_H, J_H and C_H gene segments in pre-pro-B and pro-B cells were determined [12]. In pre-pro-B cells, the D_H-J_H region is found within close proximity of the C_H elements but away from the majority of the V_H regions. The proximal and distal V_H regions are separated from each other and do not seem to intermingle. In contrast, in preparation for recombination at the pro-B cell stage the proximal and distal V_H regions appear to have merged and juxtaposed to the D_HJ_H elements, providing equal opportunities for the entire V_H repertoire [12].

These findings have raised the question as to whether all antigen receptor loci are organized in a similar fashion. Recent studies that involved the TCRα locus have indicated that not all antigen receptor loci are spatially organized as the Igh locus [13]. The TCRα locus encodes for approximately 100 Vα regions that span a 1.5 Mbp genomic region. The distal Vα regions are initially separated by relatively large spatial distances from the Jα gene segments but are juxtaposed to Jα gene segments during progressive rearrangements deleting proximal Vα regions [13]. Within the TCRα locus is embedded another locus encoding for antigen receptors, termed TCRδ. The TCRδ locus undergoes rearrangement in thymocyte progenitors, whereas the TCRα locus recombines in maturing thymocytes. Distal Vα regions are in a contracted state in thymocyte progenitors, but become de-contracted upon maturation. It has been proposed that the contracted conformation of the TCRα/δ locus permits efficient rearrangements of Vδ variable gene segments in early progenitors while the de-contraction in the TCRα locus initially restricts rearrangements only to the most proximal located Vα regions [13]. Thus, the more distally located Vα gene segments would only be positioned into close spatial proximity of the Jα regions upon progressive deletions of the proximal Vα elements. Overall, these studies indicate the presence of highly ordered and developmental regulated topologies that permit encounters between V, D and J regions to occur with the appropriate frequencies.

Rosettes and Anchors

As aforementioned, using a geometric approach it was recently revealed that during the transition from the pre-pro-B to the pro-B cell stage, the Igh locus undergoes large-scale conformational changes [14]. These data bring into question how the Igh fiber is folded. Comparison of experimental and simulated spatial distances as a function of genomic separation predicts that Igh locus topology is organized into clusters of loops consistent with an MLS configuration [12]. The organization of the Igh locus as clusters of loops is not unique to antigen receptor loci. Rosette-like structures have been observed in mitotic as well as interphase chromosomes, including regions such as the T helper type-2 cytokine locus as well as the bithorax complex and a genomic region involved in the development of Prader-Willi syndrome [15,16,17,18].

What are the components that promote a well-ordered yet dynamic structure of rosettes? How are the large-scale structural changes established during developmental progression? It seems likely that protein tethers mediate this function. Binding sites for putative anchors, including YY1 and CTCF, have been identified throughout the genome. YY1 deficient pro-B cells showed significant abnormalities in Igh locus rearrangement that appear to involve the distal V_H regions. Further scrutiny revealed that YY1 in pro-B cells acts to promote Igh locus contraction [19]. CTCF is a multiple zinc-finger containing protein that binds to its target sites with particularly high affinity. Genome-wide studies have been illuminating in a sense that they provided a global view of CTCF occupancy and its partners named cohesins [20]. The cohesins consist of four core subunits termed Smc1, Smc3, Rad21 and Scc3. During DNA replication, the cohesins interact with sister chromatids by forming a ring-like structure, surrounding the two strands. Thus, the emerging view is that cohesins and CTCF act in concert to promote the assembly of loops. CTCF forms multimers between putative anchors and cohesins act to stabilize loop formation. Recent data have indicated that cohesins perform critical roles during transcription as well by stabilizing loop formation between enhancer and promoter elements that involve the mediator complex [21,22].

CTCF binding sites also span antigen receptor loci. Up to fifty binding sites span the entire Igh locus [23]. The CTCF binding sites span the V_H region cluster but are absent within the D_H - J_H cluster. Other factors that have been found to affect Igh locus topology, possibly involved in tethering, are Ikaros and Pax-5. Ikaros and Pax-5 play critical roles in early B cell development. They are dispensable for the induction of D_HJ_H and proximal V_H-D_HJ_H gene rearrangement, but are absolutely required for DNA recombination involving the distal V_H regions [24,25,26,27]. Thus, it now seems settled that there are multiple molecular components that play a critical role in chromatin topology. Likely an array of additional players will be identified that perform this function. The critical challenge will be to determine how they act together to modulate chromatin structure as well as chromatin dynamics. Are the participants that appear to modulate Igh locus topology restricted to the antigen receptor loci or are they involved genome-wide? The latter possibility seems more likely. YY1 and CTCF are expressed ubiquitously and clearly perform a universal function.

Clearly much progress has been made in identifying factors that promote the folding of chromatin into rosette-like configurations. Genome-wide occupancy studies have provided a global view of the binding patterns of such factors. It should be mentioned that genome-wide binding patterns are based on populations of cells. However, it is the differences in locus topology between single cells that are particularly intriguing and perhaps more fundamental to our understanding of chromosome function than the average trajectories taken by the chromatin fiber.

Transcription factories

Nascent transcripts are not randomly localized in the nucleus. Transcripts have been detected in a limited number of nuclear structures, sensitive to inhibitors that affect transcription initiation and/or elongation. Such nuclear structures were named transcription factories because of their similarities with replication factories [28]. Transcription factories range in size between 40–100 nm in diameter. They include, based on the number of nascent RNA transcripts up to eight RNA polymerases [28,29,30,31,32]. Rather compelling data has recently been accumulated indicating that DNA is recruited to and pulled through the RNA polymerase machinery. It was shown that during transcriptional elongation, distinct genes, localized in cis and in trans, are brought into close spatial proximity, and pulled through relatively immobile RNA polymerase II complexes [33]. These remarkable studies demonstrate that transcription factories contain immobile RNA polymerase II molecules, through which transcribed genes are transported.

The number of transcription factors varies greatly, between hundreds and thousands, depending on cell context. Both intrachromosomal and interchromosomal interactomes have been observed within transcription factories [32]. Using formaldehyde-cross linking approaches it was revealed that co-regulated genes are primarily transcribed in common hubs [34]. These data raise the interesting possibility that a proportion of transcription factories become specialized because of an increase in the concentration of particular transcription factors, and suggesting that subsets of genes that are co-ordinately regulated are selectively recruited to specific nuclear structures to become transcribed. Although chromosomes are folded into distinct units, intermingling of chromosomes have been observed as well documented for nucleoli. To what extent trans-interaction between chromosomes do occur within transcriptiona factories remains to be determined. Since the number of transcription factories is limited it is to be expected that a substantial portion of transcribed loci is recruited by polymerases present in transcription factories. This might be achieved by random walk behaviour. Alternatively controlled and directed motion of chromatin fibers, through mechanisms yet to be revealed, may promote the assembly of transcription factories. Finally, now we are faced with the question as to how such factories are organized in 3D-space. 3D-FISH combined with geometric approaches as well as formaldehyde cross-linking strategies should provide further insight into the topologies of these intriguing nuclear structures [12,35].

Transcription and recombination factories

As aforementioned recent data have revealed that the immunoglobulin heavy chain locus is organized as clusters of loops, similar to those proposed by the MLS model [14,35,36]. How the loops are established and maintained remains unknown. Recently it has been proposed that the formation of clusters of loops in the Igh locus involves transcription, in a manner similar as described for transcription factories [37]. Non-coding, intergenic and antisense transcription has been observed at multiple locations throughout antigen receptor loci [38,39,40]. Such transcripts are initiated by V-region promoter and/or enhancer elements and elongate across coding, intergenic and regulatory regions. Non-coding anti-sense transcription is initiated prior to antigen receptor V(D)J gene rearrangements and is developmentally regulated. It was convincingly shown that non-coding transcription plays a critical role in antigen receptor assembly at the TCRα locus [41]. Specifically, it was demonstrated that interference with elongation of non-coding RNA transcription severely interfered with VαJα gene rearrangements [41]. This then raises the interesting possibility that RNA polymerases ‘fixed’ within transcription factories, reel in antigen receptor loci to elongate non-coding transcripts. As non-coding transcripts are pulled through the RNA polymerase complex, recombination signal sequences may become accessible to the recombination machinery [37].

Rosettes providing equal opportunities for all

Is active non-coding transcription the only critical player in establishing clusters of loops that mediate DNA recombination? It seems unlikely. We have recently identified genome-wide CTCF occupancy in pro-B cells using ChIP-Seq [42]. As aforementioned, CTCF spans the entire Igh locus with the exception of the C_H-D_H region (Figure 1) [23,42]. Here we have explored further the possibility that CTCF functions to promote looping and recruit the V_H regions to the recombination machinery. Specifically, we have plotted CTCF occupancy and V_H region localization (Figure 1). A close relationship between CTCF occupancy and the positions of a large subset of V_H regions was observed (Figure 1). Remarkably, CTCF binding sites close to the proximal V_H regions immediately flank, in an asymmetric pattern, the 3’ end of recombination signal sequences (Figure 2). This pattern was not observed for pseudo V_H regions. Rather they seem to be located away from the CTCF binding sites. Pseudo V_H regions do not rearrange giving support for the idea that in order for proximal V_H regions to recombine they need to be located within close genomic proximity to CTCF binding sites.

Figure 1.

Proximal variable regions in the immunoglobulin heavy chain locus are frequently associated with nearby CTCF occupancy. CTCF binding, determined by ChIP-sequencing is shown for Rag-1 deficient pro-B cells (42). CTCF occupancy is closely linked with the presence of proximal V_H regions. Only few V_H regions are highlighted but a similar pattern is observed for the entire proximal V_H region cluster. Pseudo V_H regions, on the other hand, are not located within close genomic proximity of the proximal V_H regions.

Figure 2.

Cumulative frequency distributions of V_H segments located adjacent CTCF binding site. (a, b) A large fraction of proximal V_H segments are located within 100 bp of CTCF binding. (b) A clear reversal of this trend is present at the border between proximal and distal V_H segments suggesting that at least two separate mechanisms are involved in structuring the locus for recombination.

These data bring into question as to how CTCF occupancy and the presence of proximal V_H regions relate to antigen receptor rearrangement. Here we would like to propose the following: As aforementioned, the spatial distances measured in the Igh locus agree well with simulated rosette-like structures [12]. Based on genome-wide occupancy studies, we suggest that the proximal V_H regions are organized as rosettes by CTCF [42]. Specifically, we would like to propose that CTCF wraps the proximal V_H regions around a cavity (Figure 3). The cavity itself, we suggest, contains the D_H-J_H elements associated with RAG1/2, previously named the recombination center (Figure 3) [43].

Figure 3.

Model describing the potential role of CTCF in Igh locus rearrangement. Spatial distances measured within the Igh locus suggest that it is folded into rosette-like structures. While it is currently unknown what proteins anchor the bases of these rosettes, the presence of CTCF adjacent to proximal V_H segments makes it a key candidate. (a) In such a model, the V_H segments are in close proximity, surrounding an inner cavity. At the pro-B cell stage, the D_H segments bound by the recombination enzymes have the ability to move into this cavity, creating an equal probability of recombination with any of the V_H segments. Incorporation of distal V_H segments into these structures is likely CTCF independent. Note that in this model loop formation provides access of half of the proximal V regions to the recombinase located within the cavity whereas the other half becomes positioned within the loop and might not be accessible to the recombination machinery. (b) Transcription through the locus prior to recombination may disrupt long-range genomic interactions, allowing new interactions to form.

We note that in the proposed configuration, half of the V_H regions that flank CTCF occupancy are wrapped around the cavity and thus would become accessible to the recombinase machinery. On the other hand, the other half of proximal V_H regions would positioned within the loop and plausibly less likely to encounter a D_HJ_H element (Figure 3). Then how does in such a configuration the entire repertoire of proximal V_H regions encounter D_HJ_H elements with similar frequencies? It seems likely that loop formation is dynamic, permitting loops to rapidly associate and dissociate, establishing new neighbours, permitting the entire set of proximal V_H regions equal access to the recombination center.

Conclusion

While the CTCF sites flank recombination signal sequences for a substantial fraction of the Igh proximal V_H regions, such a correlation was not found for the distal V_H regions. As previously suggested, it may very well be that the distal V_H regions are recruited to the recombination center by the RNA polymerase machinery [37]. Thus, the distal and proximal recombination signal sequences may use different mechanisms in order to be organized properly in 3D-space.

How similar are transcription and recombination factories? Do they share common features in that multiple DNA regions are recruited to a given domain containing the relevant recombination machinery? If rearrangement does occur at transcription factories, how does RNA polymerase interact with CTCF and could this allow the structure of rosettes to be constantly changing? Many questions remain. Regardless of the precise mechanism, the cardinal point of the model proposed here is that the folding of the chromatin fiber, into clusters of loops that position V_H regions at the base of rosettes, permits an equal playing field for the Igh V_H region repertoire.