|Home | About | Journals | Submit | Contact Us | Français|
The adaptive immune system generates a specific response to a vast spectrum of antigens. This remarkable property is achieved by lymphocytes that each express single and unique antigen receptors. During lymphocyte development, antigen receptor coding elements are assembled from widely dispersed gene segments. The assembly of antigen receptors is controlled at multiple levels, including epigenetic marking, nuclear location, and chromatin topology. Here we review recently uncovered mechanisms that underpin long-range genomic interactions and the generation of antigen receptor diversity.
"…Let me anticipate what will be explained in much more detail later, namely, that the most essential part of a living cell -the chromosome fiber- may suitably be called an aperiodic crystal. In physics we have dealt hitherto only with periodic crystals. To a humble physicist's mind, these are very interesting and complicated objects; they constitute one of the most fascinating and complex material structures by which inanimate nature puzzles his wits. Yet, compared with the aperiodic crystal, they are rather plain and dull. The difference in structure is of the same kind as that between an ordinary wall paper in which the same pattern is repeated again and again in regular periodicity and a masterpiece of embroidery, say a Raphael tapestry, which shows no dull repetition, but an elaborate, coherent, meaningful design traced by the great master (E.W. Schrödinger, What is Life? 1944)."
The lymphocyte compartment consists of cells that express a diverse repertoire of antigen receptors, which enables organisms to mount an immune response specifically tailored to invading pathogens. Dreyer and Bennett (1965) first proposed that antigen receptor diversity is generated by DNA recombination. Later studies confirmed this original insight, revealing that antigen receptor loci are organized into distinct genomic regions that contain variable (V), diversity (D) and/or joining (J) and constant (C) coding elements (Brack et al., 1978; Seidman et al., 1978; Weigert et al., 1978). Since this early work, the understanding of the biochemical and molecular mechanisms that underpin the assembly of antigen receptors has blossomed. Excellent reviews have described the findings generated by these studies in great detail (Jung and Alt, 2004; Schatz and Spanopoulou, 2005). Here, we will briefly introduce the basics of polymer science in order to illuminate some of the physical considerations of chromatin structure that come into play upon exploring the nature of long-range genomic interactions. We then discuss how epigenetic marking, nuclear location, and chromatin topology modulate DNA recombination. Finally, we describe what has been learned about the actual topology of antigen receptor loci and how it relates to long-range genomic interactions and antigen receptor gene rearrangement. The main goal of this review is to bring together the seemingly unrelated concepts of polymer science, nuclear organization, long-range genomic interactions, and the assembly of antigen receptor loci.
The antigen receptor loci are not merely linear chromosomal structures but posses a three-dimensional configuration. They must fold into an elaborate pattern of loop arrangements to permit antigen receptor gene segments to encounter each other with the appropriate frequencies. Resolving this question requires insight into long-range chromatin structure and dynamics through polymer physics.
The unit of the chromatin fiber is the nucleosome. A nucleosome consists of a 146 bp DNA segment wrapped around an octamer that has two copies each of histones H2A, H2B, H3, and H4. The nucleosomes form a 10 nm fiber creating a structure resembling 'beads on a string'. A naked DNA fiber without any histones contains approximately 3 base pairs per nm if stretched linearly. Addition of histones compacts this value to 20 bases/nm in the 10 nm fiber. The 10 nm fiber, in the presence of histone H1, condenses into a more compact 30 nm fiber, of which the precise structure is still not completely resolved (Schalch et al., 2005). The 30 nm fiber contains ~100 bases/nm.
How the chromatin fiber is folded into higher order structures beyond the 30 nm fiber remains largely unknown. In the late 1970s and early 1980s, distinct folding patterns for chromosome structure were proposed, including topologies involving helical, radial or combined loop-helical folding (Sedat and Manuelidis, 1977; Paulson and Laemmli 1977; Rattner and Lin, 1985). Using electron microscopic analyses of chromosome spreads, Laemmli and collaborators showed that chromosomes appeared comprised of loops of ~90 kbp in size. It was postulated that such loops interact with a putative nuclear matrix during mitosis and cluster further into rosettes containing on average ~18 loops, yielding ~100 rosettes per mitotic chromosome (Paulson and Laemmli, 1977; Pienta and Coffey, 1984). More recently, serial thin-section electron microscopy has suggested a different topology for chromatin structure, namely a "chromonema" fiber, where a chain with a diameter of 60–130 nm, is interspersed by more loosely folded segments that have a diameter of 30 nm (Belmont and Bruce, 1994). Recent advances in technology, including structured-illumination (SIM) as well as photoactivatable localization microscopy (PALM) will permit higher resolution imaging and should increase our still-rudimentary knowledge of genome structure and the ensemble of topologies that are adopted by antigen receptor, olfactory, globin, and Hox genes and other large regions of the genome.
The chromatin fiber is in continuous motion. To what extent is this motion random and to what degree is it directed? Is it shaped and how can such a structure be described? In essence, the words 'shape and structure' are only attempts to provide order to a wide spectrum of conformations that are adopted by the chromatin fiber. Thus, the shape of a chromatin fiber can best be discussed in terms of its average properties, rather than the precise location of each nucleotide. A chromatin fiber, because of its repetitive nature, resembles a polymer, and many of its physical properties can be analyzed in terms of random walk models, established by the illustrious Kuhn, Flory, and de Gennes (Morawetz, 1985). In a random walk model, the chromatin fiber can be imagined as a series of rigid and non-flexible segments, connected by flexible hinges. Several types of random walk models have been used to describe the structure and dynamics of polymer chains (Figure 1). In a freely jointed chain model, the hinges connecting adjacent segments, named Kuhn segments, are free to rotate, and the polymer segments are allowed to overlap with each other, that is, the orientation of one segment is independent of the orientation of its two adjacent segments (Figure 1A). A more realistic model, frequently applied to describe polymer chain behavior, is the self-avoiding chain (Figure 1B). A self-avoiding chain is similar to the freely jointed chain, except that the chain cannot cross its own path or that of another chain, which is to say that the Kuhn segments cannot intersect with each other (de Gennes, 1979). Yet another model is the worm-like or Kratky and Porod chain, which considers the polymer as a continuously flexible chain rather than freely jointed discrete segments (Kratky and Porod, 1949) (Figure 1C). Recent studies have described the dynamics of the yeast chromatin fiber in terms of the worm-like chain (Bystricky et al., 2004).
In the absence of confinement effects, the effective volume occupied by the chain is determined by its contour length (genomic distance separating the ends of a chromatin fiber), attractive and repulsive intra-chain interactions, molecular crowding effects and the intrinsic flexibility of the fiber. The most important parameter determining the flexibility of a polymer is the "persistence length", which is defined as the length of the polymer at which the ends become de-correlated, meaning, that the polymer becomes flexible. The persistence length for naked DNA is ~50 nm. The persistence length for a chromatin fiber remains controversial but depending on the experimental design by which it is measured, varies between 30–200 nm (Langowski and Heermann, 2007).
In eukaryotic nuclei, the total DNA contained in the chromosomes has a combined length of approximately two meters, which fits inside a nucleus with a diameter of the order of 10 µm, implying that the eukaryotic genome must have adopted strategies that permit its folding into a highly condensed state. Indeed, chromosomes show a confined geometry, indicated by the presence of chromosome arms as well as bands, arguing against free random walk behavior. Furthermore, distance measurements demonstrate that the spatial distance scales as a function of genomic separation over 4 Mbp, with exponents that are incompatible with that of free random walk statistics (Warrington and Bengtsson, 1994; Sachs et al., 1995; Münkel and Langowski, 1998).
As a first approach to describing the configuration and dynamics of the chromatin fiber in vertebrate nuclei, a chromatin topology named the random walk/giant loop model (RW/GL) was proposed (Yokota et al., 1995; Sachs et al., 1995). The RW/GL model assumes that the chromatin fiber is subject to random motion but is spatially confined as large loops (2–5 Mbp), tethered to loop attachment points (Figure 1D). However, measurements of spatial distances between genomic markers, spaced less than 4 Mbp apart, did not agree well with the RW/GL model (Munkel and Langowski, 1998). In the late 1990s, the multi-loop-subcompartment (MLS) model was proposed as an alternative configuration to describe long-range chromatin folding (Münkel and Langowski, 1998; Knoch et al., 2000). The MLS model assumes that the chromatin fiber is folded into 1 Mbp compartments, containing loops that are clustered as rosettes, connected by flexible linkers (Figure 1E). Most recently, yet another topology, the random-loop (RL) model, has been proposed to describe the long-range chromatin folding (Figure 1F). Whereas loops in the RW/GL and MLS models are assumed to be uniform in size, the RL model permits loops of variable sizes that dynamically associate and dissociate from loop attachment points (Bohn et al., 2007). The development of new computational methods to model chromatin topologies should help to define experimental strategies that evaluate model predictions. Furthermore, as will be discussed below, modeling of long-range eukaryotic chromatin structure has already suggested new avenues of investigation.
In each B cell, the antigen receptor or antibody consists of two heavy and light chain polypeptides, encoded on separate loci. The murine immunoglobulin heavy chain locus (Igh) which codes for the Ig heavy chain, manifests itself as a single massive stretch of DNA (3 Mbp) in length, which is divided into distinct DNA elements encoding the variable (V), diversity (D), joining (J), and constant (C) regions (Figure 2A). Each subregion displays much complexity. For example, fifteen partially dispersed V region families encode approximately 195 VH gene segments, depending on the genetic background, each of which is approximately 500 bp in size. The density of gene segments within the V region cluster is relatively low, containing large intergenic regions up to 50 kbp in size. Down-stream of the VH regions, are 10–13 DH and 4 JH gene segments, as well as 8 CH regions that encode the various Igh isotypes including: Cµ, Cδ, Cγ1 Cγ2a, Cγ2b, Cγ3, Cα and Cε.
The light chain of immunoglobulins is produced by one of two loci, Igκ or Igλ. The Igκ locus is comprised of approximately 120 Vκ gene segments that span almost 3 Mbp, a Jκ cluster, and a single constant region positioned within very close proximity (2.5 kbp) to the Jκ cluster (Figure 2B). The organization of the Igλ locus is quite distinct from that of the Igh and Igκ loci. Rather than a common set of J gene segments located upstream of the constant region(s), the four constant regions of Igλ each contain their own unique Jλ gene segment. Moreover, only two V region gene segments, Vλ1 and Vλ2 are frequently utilized. Vλ2 is located approximately 60 kbp from Jλ2 and it will generally not recombine with other Jλ gene segments. On the other hand, Vλ1, located 22 kbp from the Jλ1, will form joints with either Jλ1 or Jλ3 (Figure 2C). Thus, each chain of an antibody is produced using a similar theme of combining distinct gene segments.
The organization of genes encoding the T cell receptor (TCR) gene segments is strikingly similar to that observed for the immunoglobulin loci (Figure 2D–F). Two distinct T cell lineages, characterized by the antigen receptor expressed on their cell surface, αβ and γδ T cells, develop in the thymus from early T-lineage progenitor cells. The TCRβ locus spans approximately 650 kbp of genomic DNA. It contains 31 Vβ gene segments, of which 20 are functional and located upstream from 2 DβJβ clusters and 2 Cβ regions. Each of the DβJβ clusters contains a single Dβ and 6 Jβ gene segments.
The TCRα locus hews to a similar theme in that it is comprised of approximately 100 V gene segments located within a 1.5 Mbp region (Figure 2D). At least 200 kbp separates the Vα regions from the Jα cluster. The TCRα locus is unusual in that it contains many more J regions as compared to other antigen receptor loci, with 61 Jα gene segments that span 65 kbp. Nested within the TCRα locus is the TCRδ locus, containing numerous Vδ, two Dδ and Jδ elements and one Cδ region. Unlike the TCRα, TCRβ and TCRδ loci, the TCRγ locus is small (less than 200 kbp), containing few Vγ and Jγ gene segments (Figure 2F). Thus, the majority of antigen receptor loci are comprised of large numbers of V regions that span a vast genomic region and numerous clustered D or J gene segments.
Pluripotent hematopoietic stem cells, which are capable of self-renewal, develop into lymphoid-primed multipotent progenitors (LMPPs) that lack long-term self-renewal capacity and have myeloid- or lymphoid-restricted differentiation potential. LMPPs have the ability to develop into common lymphoid progenitors, which, in turn, have the potential to differentiate to pre-pro-B cells. Pre-pro-B cells, in turn, develop into committed pro-B cells, that initiate and complete Igh V(D)J gene rearrangement. Rearrangement of antigen receptor loci is mediated by RAG-1 and RAG-2, which act to cleave DNA at recognition sites that flank the V, D, and J gene segments (Schatz and Spanopoulou, 2005). At the pro-B cell stage, DHJH joining precedes that of VHDHJH gene rearrangement. Once a productive VHDHJH gene rearrangement has been generated, a pre-B cell receptor (pre-BCR) is assembled that acts, in turn, to inhibit the expression of RAG-1 and RAG-2, and promotes the survival and expansion of developing large pre-B cells. This proliferation phase is followed by exit from the cell cycle, during which RAG gene expression is re-induced to enable Igκ gene rearrangement. At the pre-B cell stage, Igκ VJ gene rearrangement is initiated. If the rearrangement is non-productive or results in Igκ deletion, B cells can undergo Igλ rearrangement. B cells that express auto-reactive receptors maintain RAG expression and promote secondary Igκ VJ rearrangements, a process termed "receptor editing". These secondary rearrangements can lead to replacement of rearranged Igκ loci with secondary productive rearrangements. Alternatively non-functional rearrangement or deletion of both Igκ alleles would permit Igλ VJ locus rearrangement to ensue. Predominantly, B cells expressing Igλ can be generated from pre-B cells with two non-productive Igκ rearrangements or via receptor editing from Igκ+ cells (Nemazee, 2006). Thus, the Igh and Ig light chain loci are assembled sequentially during early B cell development.
The development of αβ and γδ T cells in the thymus is a process characterized by the sequential rearrangement of the gene segments of the T cell antigen receptor loci (TCR). Shortly after arriving in the thymus, T cell progenitors initiate TCRβ, TCRγ, and TCR loci rearrangement. The rearrangement of the TCRβ locus is initiated and completed at a developmental stage that lacks the expression of the co-receptors for the TCR, CD4 and CD8, a population of cells commonly referred to as the double negative stage. Upon rearrangement and expression of a productive TCRβ chain and its assembly into a pre-TCR complex, RAG expression is suppressed and thymocytes undergo developmental progression, characterized by rapid cellular expansion. During this phase, thymocytes begin to express CD8, followed by CD4. Thymocytes that express CD4 and CD8, referred to as double positive cells, undergo cell cycle arrest, and initiate TCRVJ gene rearrangement. VαJα rearrangements can be initiated multiple times, such that secondary TCRα rearrangements progressively can replace primary VαJα joints. The primary rearrangements predominantly utilize the Vα gene segments positioned towards the 3' end of the locus and the most 5' Jα elements. Double positive thymocytes then progress through the processes of positive and negative selection, allowing the maturation of only those cells that express TCRs with moderate affinity for self-major histocompatibility complexes expressed by thymic epithelial cells. Positively selected thymocytes decrease expression of either CD4 or CD8 to develop into mature CD8 or CD4 single positive (SP) progeny.
Two distinct mechanisms have been described that ensure mono-allelic antigen receptor rearrangement. First, antigen receptor rearrangement is mono-allelically activated, which has been well characterized for the Igκ locus (Cedar and Bergman, 2008). Second, once a productive Igh or TCRβ V(D)J gene rearrangement has been generated, signaling mediated by the pre-BCR or pre-TCR antagonizes continued rearrangement by a feed-back mechanism (Jung and Alt, 2004; Krangel, 2007). As a result, only one copy of a functional antigen receptor gene is produced in a single lymphocyte. Thus, the adaptive arm of the immune system is generated by distinct cell types, which undergo ordered gene rearrangement, to provide each lymphocyte with a single and unique antigen receptor.
The antigen receptor chromatin fiber, both DNA and its associated histones, is epigenetically marked during developmental progression. The Igκ locus is mono-allelically demethylated at the DNA prior to the onset of VκJκ gene rearrangement and the demethylated Igκ allele is also selectively targeted in germinal center B cells by activation-induced deaminase (AID) (Cedar and Bergman, 2008). The amino-terminal tails of the core histones are also marked. These tails can be modified by lysine acetylation, arginine and lysine methylation, lysine ubiquitination as well as serine and threonine phosphorylation. Histone acetylation correlates well with chromatin accessibility. Acetylation is catalyzed by histone acetyltransferases, whereas deacetylation is mediated by the histone deacetylases. Methylation, in turn, serves to promote interactions of histones with factors involved in chromatin remodeling (chromodomains) as well as with plant homeodomain-containing proteins. The methylation state of histone lysine residues is mediated by methyltransferases and histone demethylases.
Recent studies have demonstrated that the regulation of ordered and lineage-specific Igh locus assembly correlates well with chromatin structure and DNA recombination. Isolated lymphoid nuclei, upon incubation with the recombinase, show cleavage of recombination signal sequences in a lineage and developmental stage specific fashion, linking chromatin structure, RAG1 and RAG2 activity and V(D)J gene rearrangement into a common framework (Stanhope-Baker et al., 1996). Although the precise mechanism remains to be determined, the common view is that enhancer elements and/or promoter regions liberate recombination signal sequences from a repressive chromatin environment to permit RAG1/RAG2-mediated cleavage (Golding et al., 1999; Kwon et al. 1998).
The ordered rearrangement of antigen receptor loci also correlates well with temporally-restricted epigenetic marking (Chowdury and Sen, 2004). Three patterns of epigenetic marks are particularly interesting: First, prior to DHJH rearrangement, the DH-CH region becomes acetylated. Second, VH gene segments are not associated with acetylated histones in progenitor B cells that have not undergone DHJH rearrangement. Third, Igh loci prone to undergo VH-DHJH rearrangement manifest elevated levels of VH histone acetylation (Chowdury and Sen, 2004). The pattern of histone acetylation across the DH–JH region, however, is not uniform and is mainly restricted to the most 5' and 3' located DH elements. Interestingly, these DH gene segments are used most frequently in DHJH junctions (Chakraborty et al., 2007). Thus, it appears that Igh locus histone acetylation is regulated in a sequential manner. The DH–JH domain is hyperacetylated first, followed by DHJH rearrangement. Subsequently, VH gene segments become activated by acetylation to permit VHDHJH joining.
The observations described above indicate a correlation between epigenetic marking and antigen receptor assembly. Recently, a direct link between these two processes has been established in a series of experiments that are particularly illuminating. Specifically, methylation of histone H3 (H3K4me3 and H3R2me2) upon interacting with RAG-2 acts to enhance DNA binding and enzymatic activity of the recombinase (Liu et al., 2007; Matthews et al., 2007; Ramon-Maiques et al., 2007; Shimazaki et al., 2009). This interaction requires a noncanonical PHD domain that is located within the noncore domain of RAG-2. At first glance these data raise the possibility that the histone code may act to promote the targeting of the recombinase to the recombination signal sequences. However, epigenetic marking of H3K4 and H3R2 by methylation is associated with transcriptional initiation and definitely not limited to H3 residues that are positioned within close proximity to the recombination signal sequences. Additional specificity likely is provided by the interaction involving RAG1 and recombination signal sequences.
How are histone marks established and removed during developmental progression? Two factors, PAX5 and EZH2, are potential players. PAX5 is a paired-homeodomain containing protein that plays a central role in B cell commitment. It is dispensable for DHJH and proximal VH-DHJH rearrangements, but it is absolutely required for distal VH-DHJH joining (Fuxa et al., 2004). H3K9 methylation epigenetically marks VH gene segments in a repressive state in hematopoietic cells that are not committed to the B cell lineage (Johnson et al. 2004). Prior to VHDHJH rearrangement, H3K9 residues become demethylated in the VH locus in a Pax-5 dependent manner. EZH2 is a polycomb group protein acting as a H3K27 histone methyltransferase (Su et al., 2003). EZH2-deficient pro-B cells show a defect in distal VH-DHJH rearrangements. It has yet to be determined how EZH2 mechanistically acts to promote distal VHDHJH rearrangement.
Recent studies demonstrate a role for noncoding RNAs in modulating histone tails. Among the first noncoding RNAs to be identified were germ-line transcripts, also named sterile transcripts (Yancopoulos and Alt, 1985). A direct role for noncoding RNAs in modulating epigenetic marking and antigen receptor assembly has been demonstrated for the TCRα locus. As mentioned earlier, the TCRα locus is unusual in that it permits multiple rearrangements to occur, in which secondary VαJα rearrangements can progressively replace primary VαJα joints through recombination of 5' Vα segments with 3' Jα elements, until an αβ TCR has been generated that is capable of interacting with members of the major histocompatibility complex. The targeting of RAG proteins to the 5' Jα regions requires the transcription of noncoding RNAs (Abarrategui and Krangel, 2006). Furthermore, suppression of transcriptional elongation substantially interferes with the trimethylation of H3K4 at Jα gene segments localized down-stream, possibly interfering with the targeting of the recombinase to the recombination signal sequence. These studies are of interest as they directly link noncoding RNAs, epigenetic marking and antigen receptor assembly.
The localization of genes varies during developmental progression (Schneider and Grosschedl, 2007). Such repositioning of loci during differentiation is often conserved throughout evolution. A prominent example involves multiple genes in yeast that appear to cluster at the nuclear membrane-nuclear lamina (Akhtar and Gasser, 2007). A subset of genes that are undergoing active transcription is associated with nuclear pores whereas others, which are associated with the inner nuclear membrane, appear to be transcriptionally silenced. In mammalian cells, the expression of the β-globin locus is initiated at the nuclear membrane prior to its movement towards more centrally located domains (Ragoczy et al., 2006). A second nuclear compartment involves the pericentromeric heterochromatin. The association of genes with pericentromeric heterochromatin is often correlated with transcriptional silencing.
Evidence indicating that the assembly of Ig loci is also regulated by nuclear location is accumulating. The Igh locus in hematopoietic progenitors associates with the inner nuclear lamina (Kosak et al., 2002). The distal VH region cluster is tethered to the nuclear membrane whereas the DHJH elements are located away (Figure 3) (Roldan et al., 2005). Notice that in such an arrangement, the spatial orientation of the Igh locus may allow the RAG-1 and RAG-2 proteins access to the DH–JH domain while the VH cluster remains closed. The presence of DHJH rearrangements but the complete absence of VHDHJH joints in hematopoietic progenitors is consistent with such a topology.
Upon commitment to the B-cell lineage, the Igh locus undergoes relocation away from the nuclear periphery and large-scale contraction, followed by VH-DHJH gene rearrangement (Kosak et al., 2002, Fuxa et al., 2004; Roldan et al., 2005; Sayegh et al., 2005). How are conformational alterations in antigen receptor loci established during developmental progression? Plausibly, this will involve the introduction of constraints in antigen receptor loci, promoted by tethering. However, we suggest that in differentiating cells, induced changes in the flexibility of the chromatin fiber also contribute. For example, a decrease in the persistence length will affect the frequency of interaction between distant genomic regions. Localized reduction in the persistence length may be achieved by histone modifications spanning the fiber. It is possible that germ-line transcription, observed when antigen receptor loci are active, may induce such modifications. Rather than targeting histone modifications at a regulatory sequence, germ-line transcription may introduce epigenetic marks across a large region as the polymerase scans the fiber. Such changes in flexibility would affect the ensemble of conformations adopted by the chromatin fiber.
Once a productive VH-DHJH rearrangement has occurred, the Igh locus undergoes decontraction. The non-functional Igh allele relocates to pericentromeric heterochromatin where it transiently interacts with an Igκ allele (Roldan et al., 2005; Hewitt et al., 2008) (Figure 3). The interactions between the Igh and Igκ alleles are intriguing. Interphase chromosomes are not dispersed throughout the nucleus but rather are positioned within distinct areas. Small chromosomes appear to preferentially localize to internal locations while large chromosomes seem to localize mainly in peripheral areas (Bolzer et al., 2005). However, individual chromosomes are not entirely separated given that different chromosomes intermingle, best characterized in nucleoli, heterochromatic regions, and now including the Igh and Igκ loci.
The de-contraction of antigen receptor loci has been postulated to suppress further rearrangement and to enforce the allelic exclusion mechanism (Roldan et al., 2005). However, the spatial distances separating the distal and proximal VH regions from the DHJH or JH gene segments have not been directly determined in cells that express a productive rearrangement (Figure 3). Thus, we are faced with the question: . does de-contraction takes place within the entire VH repertoire or only within the distal VH regions? If it is the latter scenario, then de-contraction is not an attractive mechanism to ensure allelic exclusion given that after feedback inhibition the majority of proximal VH regions would have the same probabilities of encountering a DHJH or JH gene segment as prior to pre-BCR expression. Perhaps a more efficient mechanism to ensure the allelic exclusion process is the recruitment of the non-productive Ig allele to the repressive pericentromeric region (Goldmit et al., 2005; Roldan et al. 2005; Hewitt et al., 2008).
The Igκ locus is also relocated during B cell development. In pre-pro-B cells, the Igκ locus interacts with the nuclear membrane (Kosak et al., 2002) (Figure 3A). In committed, pro-B cells, the Igκ loci move away from the nuclear membrane to more centrally located nuclear domains (Figure 3A). Subsequently, in pre-B cells one of the Igκ alleles becomes associated with the repressive environment of pericentromeric chromatin to favor rearrangement of the euchromatic Igκ allele (Goldmit et al., 2005). How is one of the Igκ alleles selectively recruited to the heterochromatin? Allelic replication asynchrony at the Igh and Igκ loci is established early at the time of implantation as demonstrated by early and late-replicating alleles (Mostoslavsky et al. 2001). The late-replicating Igκ locus appears to selectively associate with centromeric heterochromatin. Thus, pre-existing epigenetic marks may target one of the Igκ alleles to the centromeric heterochromatin to promote allelic asynchrony.
We are now faced with the question as to how the silenced allele associated with the heterochromatin becomes activated when the first allele is non-productively rearranged. As all associations are in flux, we suggest that the silenced allele is not permanently associated with the heterochromatin. Perhaps the equilibrium between the "free" and the "bound" state of Igκ alleles is temporally regulated during the developmental progression from the pre-B to the immature-B cell stage, allowing Igκ VJ rearrangement to proceed on the second allele if the first allele fails to generate a productive rearrangement. In principle, this could be achieved by a gradient of a transcription factor(s) that modulate the interaction of Igκ alleles with the heterochromatic environment. It is the dynamics of such interactions that are critical and it seems that in vivo imaging approaches should permit a resolution to this intriguing problem.
What are the molecular players that modulate the association of the Igκ alleles with the heterochromatic environment? One candidate is IRF-4, which directs the Igκ allele away from the pericentromeric heterochromatin (Johnson et al., 2008). Another possible player is the transcriptional regulator Ikaros. Ikaros associates with a regulatory element in the Igκ locus, named Sis. Sis is required for the repositioning of the Igκ locus to the pericentromeric heterochromatin (Liu et al., 2006).
In sum, Ig loci move around in the nucleus during B cell maturation and associate with different nuclear structures. What is the purpose of moving the Ig loci around? Has unambiguous evidence been reported demonstrating that gene activity and nuclear location are linked? The observations described above indicate a correlation between antigen receptor location and antigen receptor assembly. However, different results have been documented regarding the physiological roles for positioning of genes near the nuclear membrane versus more centrally located domains. Positioning of a reporter gene at the nuclear membrane upon interaction with lamin B did not suppress transcriptional activation upon stimulation (Kumaran and Spector, 2008). On the other hand, upon tethering a reporter construct to the inner nuclear membrane protein emerin, the expression of a reporter was modestly but significantly suppressed (Reddy et al. 2008). These differences likely reflect the different experimental strategies that were employed, but they do indicate that interaction with the nuclear periphery, despite the repressive environment, is not sufficient to completely suppress transcriptional activation. The full meaning of Ig repositioning will need to be addressed by mutational analysis of DNA elements that are responsible for the associations with distinct nuclear domains by gene targeting (Hewitt et al., 2008).
The assembly of TCR loci also appears to be regulated by nuclear location. TCRα and TCRβ loci undergo contraction in cells that are poised for DNA recombination, and this pattern is reversed upon developmental maturation (Skok et al., 2007) (Figure 3). Additionally, mono-allelic association of antigen receptor genes with pericentromeric chromatin has also been observed in TCRβ loci during thymocyte maturation. The initial observations suggested a mono-allelic interaction of the TCRβ alleles with pericentromeric heterochromatin (Skok et al., 2007). Another study, however, concludes that both TCRβ alleles associate independently of each other and not strictly mono-allelic with the nuclear lamina and pericentromeric heterochromatin (Schlimgen et al., 2008). However, an inherent bias exists for one of the alleles to localize to nuclear lamina or pericentromeric heterochromatin throughout early T cell development. The precise mechanism by which TCRβ alleles are recruited to the nuclear lamina or pericentromeric heterochromatin remains to be determined. Either mechanism would diminish the probability that TCRβ V(D)J gene rearrangements are initiated simultaneously on both alleles and may buy developing thymocytes sufficient time to induce suppression of rearrangement on the non-rearranged allele by a feed-back mechanism. In contrast to the Igh, Igκ and TCRβ loci, the TCRδ and TCRα loci do not appear to associate with pericentromeric heterochromatin, consistent with the notion that both alleles simultaneously undergo TCR assembly, in double negative and double positive cells respectively (Skok et al., 2007) (Figure 3).
What are the molecular players that contribute to the interaction of antigen receptor loci with pericentromeric heterochromatin? The helix-loop-helix protein, E47, is a good candidate. E47 dosage is rate-limiting with regard to TCRβ V(D)J rearrangement and forced E47 expression interferes with pre-TCR mediated feedback inhibition (Agata et al., 2007).
What is the mechanistic basis for silencing TCRβ VβDβ Jβ DNA recombination in a heterochromatic environment? Germ-line transcription at the TCRβ locus is bi-allelic, that is, functional and non-functional alleles are both transcribed (Jia et al., 2007). Igκ germ-line transcription is also bi-allellic (Singh et al., 2003). It is intriguing that recruitment to the peri-centromeric heterochromatin does not completely silence antigen receptor noncoding RNA transcription, counter to expectations. These observations bring into question as to how VβDβ Jβ as well as VκJκ assembly is suppressed at the pericentromeric heterochromatic environment. Perhaps histone marks that permit recruitment of the RAG proteins to the recombination signal sequences or the RAG proteins themselves are excluded from the heterochromatic environment. Clearly, substantial progress has been made but the precise mechanism that underpins mono-allelic antigen receptor activation remains unresolved.
During lymphocyte development, the rearrangement of antigen receptor loci is ordered. Igh DH–JH and TCR Dβ-Jβ rearrangement are initiated prior to that of V-DJ gene joining (Alt et al. 1984). Once a productive Igh or TCRβ V(D)J gene rearrangement has been generated, signaling mediated by the pre-BCR or pre-TCR antagonizes continued rearrangement by a feedback mechanism, ensuring allelic exclusion (Jung and Alt, 2004; Krangel, 2007). Feedback signaling, however, does not suppress rearrangement of the entire Igh region repertoire. The four most proximal VH regions escape the allelic exclusion mechanism (Costa et al., 1992). Thus we are faced with the question: what might explain these differences in V region usage?
Functional chromosomal domains may provide a means by which regions undergoing DNA recombination are distinguishable from regions not undergoing DNA rearrangement. Such functional domains have been described for the chicken β-globin cluster, which is marked by the presence of boundary elements, containing CTCF binding sites (Felsenfeld et al., 2004). To determine whether such functional domains exist within antigen receptor loci, VH regions, normally distally located and subject to the allelic exclusion mechanism, were inserted in a chromosomal position immediately 5' of the D–J cluster (Bates et al., 2007). Interestingly, the targeted VH regions appear to rearrange with substantially higher frequency than their endogenous counterparts. Furthermore, targeting of the VH region to a location immediately 5' of the DH–JH region reveals that the ordered rearrangement process is perturbed as well (Bates et al., 2007). As suggested previously, these data are consistent with a model in which the majority of VH and DH domains are located in functionally separate territories (Bates et al., 2007).
Compartmentalization of antigen receptor loci into functional domains may also explain the underlying mechanism that permits the stage-specific rearrangement of TCRα and TCRδ loci. The TCRδ elements are embedded within the TCRα locus. The TCRδ locus undergoes DNA recombination in double negative thymocytes, whereas TCRα-gene rearrangements are not induced prior to the double positive stage. The mutually exclusive rearrangement patterns of TCRα and TCRδ loci suggest that they are located in separate chromatin territories that become accessible at distinct stages of thymocyte development.
The process of ordered rearrangement correlates well with the expression of noncoding RNAs. noncoding RNAs initiate from V-region promoters following DHJH rearrangement (Yancopoulos and Alt, 1985). Furthermore, in pro-B cells antisense noncoding RNAs traverse across the entire DH–JH region prior to DH–JH rearrangement (Bolland et al., 2007). Once DH–JH rearrangements have been completed, bi-allelic anti-sense transcription is initiated across the VH region cluster (Bolland et al., 2004). How does anti-sense transcription promote ordered Igh locus DNA recombination? It is conceivable that anti-sense transcription across the Igh locus acts to promote chromatin accessibility in a manner similar as described above for the TCRα VJ gene rearrangement. Alternatively, germ-line transcripts may alter the three dimensional structure of distinct chromatin territories to promote localized accessibility, independent of epigenetic marking. Interesting work has recently shown an architectural role for a noncoding RNA that is required for the formation of paraspeckels, nuclear structures whose precise function remain to be resolved (Clemson et al., 2009). We think that the potential role of noncoding RNAs in modulating chromatin topology is of particular interest. It is conceivable that they modulate the higher order folding of the antigen receptor loci into separate domains to promote spatial proximity and to establish boundaries. Targeting of noncoding RNAs combined with geometric approaches may reveal their role in the organization of genome structure.
Fluorescence in situ hybridization studies suggest the presence of chromatin domains within the Igh locus (Jhunjhunwala et al., 2008). In pre-pro-B cells, each allele appears as two to three clusters of fluorescence that are connected by linkers, whereas in the large majority of pro-B cells only one such cluster is detectable. However, visualization of distinct clusters by fluorescence is limited by the resolution of the technique and does not necessarily imply functional chromatin compartments. An alternative strategy would be to define functional compartments in terms of epigenetic histone marks, for example present at boundary elements, as described for the chicken β-globin locus (Felsenfeld et al. 2004). But this also may not be a general strategy to define chromatin domains. Chromatin compartments can also be viewed as physical units. Two genomic markers are localized in one compartment if they coordinately motion towards or away from an anchor localized in a separate compartment (Figure 4). Here we would like to consider three distinct possible arrangements, named the V, L and O configurations (Figure 4). In a V-configuration, two DNA segments are located in one compartment whereas an anchor is localized in a separate compartment. In a V-arrangement, two markers move coordinately away from or towards the anchor (Figure 4). In an L-configuration, the anchor is located in a compartment together with another genomic marker, whereas the second marker is positioned in a separate compartment. In an L-configuration both markers motion independently towards or away from the anchor (Figure 4). In an O-configuration, the anchor and both genomic markers are located in one compartment or alternatively in three distinct compartments and motion independently from each other (Figure 4). Thus, chromatin territories could be viewed as physical units in which DNA regions move coordinately towards or away from an anchor located in another unit.
Using triple point spatial distance measurements, as outlined above, it should be possible to determine whether genomic markers within the most 5' VH regions show coordinate movement towards or away from an anchor located in a different compartment. Boundaries that separate compartments could also be identified using this approach. Such a strategy would permit the resolution of a few interesting questions: Do chromatin boundaries separate VH regions that are allelically-excluded from other VH regions that are not subject to the allelic exclusion mechanism? Do these boundaries correlate with the cluster of CTCF binding sites that separate the DH from the VH domains? Are boundaries, separating the V, D, or J domains, also present in other antigen receptor loci? Are the TCRα and TCRδ loci located in distinct functional chromosomal domains? By defining compartments as physical units, it should be possible to determine to what degree antigen receptor loci are compartmentalized and where the borders are located that separate such domains.
The antigen receptor loci are more than just linear DNA sequences. Thus, we are faced with the questions: How are the antigen receptor loci organized in three dimensional space? What is the fundamental mechanism that gives rise to antigen receptor locus topology? How do the topologies of antigen receptor loci relate to function? Recent studies have provided some insight into these questions. As a first approach, spatial distances measurements as a function of genomic separation showed that the Igh locus topology could not be described in terms of the self-avoiding random walk, the worm-like chain, or the random walk/giant loop (RW/GL) model (Figure 1) (Jhunjhunwala et al., 2008). Rather, this analysis showed substantial agreement between pre-pro-B Igh locus topology and simulated multi-loop subcompartment (MLS) conformations. The spatial distances as a function of genomic distances in pro-B cells also agreed well with those predicted by the MLS model but only for genomic markers separated by less then 1 Mbp. However, beyond a genomic separation of 1 Mbp, the spatial distances levelled off in pro-B cells and did not compare well with the MLS configuration (Jhunjhunwala et al., 2008). Thus, it appears that the Igh locus topology shows substantial, but not complete, agreement with simulated MLS configurations.
The MLS model predicts a rosette-like configuration for the chromatin fiber. In itself, this notion is not novel. Rosette-like structures have been observed previously using electron microscopy as well as formaldehyde cross-linking approaches. They were originally observed in preparations derived from both mitotic as well as interphase chromosomes (Paulson and Laemmli et al., 1977; Okada and Comings, 1979). Aggregates of loops have also been suggested to underlie the observations obtained by chromosome-conformation capture studies in the T helper type-2 cytokine locus. Notably, SATB1 (special AT-rich-binding protein 1) is proposed to be required for a multi-loop containing structure (Cai et al., 2006). A similar structure, also based on formaldehyde cross-linking studies, has been proposed to underpin the organization of the bithorax complex (Lanzuolo et al., 2007).
The comparison of experimental with simulated data predicts that the chromatin fiber at the Igh locus in pre-pro-B cells is organized into 1 Mbp rosettes, containing on average 120 kbp loops, which are separated by 60 kbp linkers. Does this analysis imply that the Igh locus fiber is structured into evenly sized loops that fold back in a regular and fixed pattern to loop attachment points, forming a perfect rosette? Not likely. Rather we would like to suggest that the chromatin fiber folds into loops of variable sizes, that loop size is determined by bridging factors and that loop formation is dynamic. Such a configuration has been proposed, for large genomic separations (5–10 Mbp), in the random loop (RL) model (Bohn et al., 2007). The continuous dynamic folding and unfolding of loops is attractive with regard to VH region gene usage as it would permit increased flexibility for VH regions that are located within close proximity of putative loop attachment points. Clearly the precise arrangements of loops in the Igh locus are yet to be determined. However, as a working model we envison that the Igh locus folds into chromatin compartments that contain loops of variable sizes, which fold and unfold in elaborate and dynamic patterns.
All loops, including those present in rosettes, require a tether at their base. Putative tethers that span the entire Igh locus have been identified. Prominent among these are YY1 and CTCF. YY1 is a zinc-finger protein that is evolutionary conserved from the fruit fly Drosophila melanogaster to humans. VH-DHJH rearrangement is severely perturbed in YY1-deficient pro-B cells affecting mainly the distal VH gene segments (Liu et al., 2007). Interestingly, the Igh locus is de-contracted in YY1-deficient pro-B cells, raising the possibility that YY1 acts to modulate Igh locus topology to promote distal VH-DHJH rearrangement (Liu et al., 2007).
How does YY1 promote Igh locus long-range chromatin contraction? The precise mechanism is unknown, but YY1 is of interest as it has been demonstrated to interact with CTCF (Donohoe et al., 2007). CTCF is a factor previously shown to promote looping within the β-globin locus (Splinter et al., 2006). In addition to interacting with YY1, recent genome-wide binding studies have revealed that CTCF also interacts with cohesins (Parelho et al., 2008). These data are intriguing as during DNA replication, the cohesins form a ring-like structure, surrounding and stabilizing the sister chromatid strands. Interestingly in a recent elegant study it is demonstrated that the cohesins also interact in cis to confine IFNG locus topology (Hadjur et al., 2009).
Is CTCF a reasonable candidate to act as a bridging factor in antigen receptor loci? The binding pattern of CTCF in the Igh locus in pro-B cells is striking (Degner et al., 2009). A total of fifty-three CTCF binding sites span the entire VH region cluster but are largely absent within the DH–JH and CH gene segments (Degner et al., 2009). A few CTCF sites are located immediately upstream of the DHJH cluster and may form a boundary between the VH cluster and the DHJH gene segments. The binding of CTCF to sites present in the Igh locus is not pro-B cell specific, and it is unlikely by itself to account for the differences observed in pre-pro-B and pro-B Igh locus topology. However, Rad21, a component of the cohesin complex was found to preferentially bind with CTCF in pro-B cells as compared to thymocytes and pre-B cells (Parelho et al. 2008; Degner et al. 2009). CTCF expression has been conditionally ablated in developing thymocytes and peripheral T cells (Heath et al. 2008; Ribeiro de Almeida et al., 2009). Surprisingly, TCR antigen receptor assembly appears normal in the absence of CTCF (Heath et al. 2008). However, we note that it still needs to be determined that in these studies CTCF was completely ablated.
Other candidates have emerged that might be involved in directly modulating Igh locus topology. Among these is Pax-5, a homeodomain protein that plays an essential role in B cell commitment. Interestingly, the distal VH cluster in Pax5-deficient pro-B cells is de-contracted and it is this region that is severely perturbed in VHDHJH rearrangement (Fuxa et al., 2004; Roldan et al., 2005). How Pax5 promotes locus contraction remains to be established. It seems unlikely that it acts by itself given taht forced Pax5 expression in T-lineage cells does not promote Igh locus contraction (Fuxa et al., 2004). Furthermore, in pre-B cells, Pax-5 abundance is high whereas the Igh alleles are de-contracted, arguing against a direct and dominant role for Pax-5 in locus contraction (Roldan et al., 2005). Ikaros is yet another player involved in modulating Igh locus topology. Ikaros does not only activate Rag expression and Igh locus accessibility, but also appears to promote Igh locus contraction (Reynaud et al., 2008).
A protein involved in DNA repair, 53BP1 has recently been implicated in modulating antigen receptor topology. 53BP1 initially targets double strand DNA breaks upon interacting with H2AX and methylated histone H4K20 or possibly H3K79, which are constitutive histone marks exposed upon break formation. 53BP1 promotes VαJα gene rearrangement (Difilippantonio et al., 2008). Interestingly, the TCRα locus is in an extended state in 53BP1-ablated as compared to wild-type thymocytes. These data raise the question as to how 53BP1 permits DNA elements, separated by large genomic distances, to interact with relatively high probability. Two fundamentally different models have been suggested to underpin 53BP1 activity. 53BP1 may accumulate at double strand DNA breaks, and upon homo-oligomerization may act as a bridging factor (Difilippantonio et al., 2008). Alternatively, 53BP1 may directly modulate chromatin dynamics, possibly by affecting the persistence length of the chromatin fiber (Dimitrova et al., 2008).
As pointed out above, changes in the flexibility of the chromatin fiber may also contribute to conformational alterations. The majority of factors known to modulate Igh locus chromatin structure, including Pax5, YY1, and Ikaros, also epigenetically mark chromatin. As such marking may affect the persistence length and overall conformation, it is critical to determine the precise mechanism by which these factors act to regulate chromatin topology. Live-cell imaging and measurement of chromatin dynamics may resolve the question of localized changes in the flexibility of the chromosomal fiber at the Igh locus.
Thus, a few molecular components have been identified that modulate antigen receptor locus topology. The candidates identified likely represent only the tip of the iceberg. The challenge will be now to identify others and to find out how they collaborate to modulate chromatin structure and how they promote encounters between close and remote gene segments.
Apart from the four most proximal V-regions, VH region usage is random and unbiased (Yancopoulos et al., 1984; Malynn et al., 1990; Gu et al., 1991; Love et al., 2000). Distal VH regions rearrange often and the frequency of VH region usage scatters uniformly throughout the distal cluster (Gu et al., 1991; Love et al., 2000). Thus, with the exception of VH regions located within the immediate proximity of the DHJH cluster, no correlation has been established between Igh V region usage and genomic location. Consequently nearby and remote V regions must have equal opportunities to make a connection with a DJ or J element. Thus, we are faced with the question as to how the gigantic range of antigen receptor diversity is generated during lymphocyte development. Specifically, why do V elements scattered over a vast genomic region rearrange with similar frequencies?
Recent studies have allowed insight into this question. Spatial distance measurements were performed across the entire Igh locus and analyzed to present a statistical view of Igh locus structure. Specifically, a geometric approach, named trilateration, was applied to determine the relative average three dimensionalcoordinates of the VH, DH, JH and CH gene segments in pre-pro-B and pro-B cells (Jhunjhunwala et al., 2008). Spatial distance measurements of different genomic markers across the Igh locus from a common set of four 'reference' markers were used to determine the average 3D-coordinates of all the genomic markers. The most striking features of the topology are: (1) Large-scale Igh locus conformational changes appear to accompany the transition from the pre-pro-B to the pro-B cell stage. (2) Most telling, in pro-B cells the proximal and distal VH regions seem to merge and to be juxtaposed to the DHJH elements, enabling the entire VH region repertoire access to the DH–JH elements. It is important to emphasize that the variation in the chromatin fiber configuration is more fundamental than a statistical representation of the structure as revealed by trilateration. Ultimately, it is the ensemble of conformations that permit equal opportunities for V regions to connect with a DJ or J segment. Nevertheless, the geometric analysis leads to one cardinal conclusion: If it is assumed that spatial proximity directly relates to the probability of encounter, the entire VH region repertoire has similar access to the DHJH gene segments in cells prone to undergo DNA recombination, providing an equal playing field.
The comparison of cumulative frequency from experimentally-derived spatial distributions and theoretical distributions reveals that DH–JH gene segments undergo free random walk behavior whereas the VH regions are spatially confined (Jhunjhunwala et al., 2008). Gene-targeting approaches have suggested that the large number of V regions directly compete for close encounters with DJ elements (Bassing et al., 2008). Thus, we suggest a scenario in which DJ or J elements wander freely, searching for a connection with competing V regions, until a productive rearrangement has been generated.
Are other antigen receptor loci structured in a similar fashion? Probably. The V gene segments in the Igκ, TCRβ and TCRα loci are also scattered over a vast genomic region. Apart from a Vβ element located 3' of the TCRβ Dβ-Jβ region, usage of the Vβ region is also randomized. The TCRβ locus has been shown to undergo contraction in the double negative thymocyte compartment, but becomes de-contracted upon developmental progression into double positive cells (Skok et al., 2007). Where the contraction and de-contraction occurs within the TCRβ locus remains to be resolved but it is conceivable that the Vβ regions have merged into one compartment in cells poised to undergo TCRβ gene rearrangement, perhaps in a manner similar to that described for the Igh locus.
In contrast to the Igh and TCRβ loci, Vα gene usage is not random (Krangel, 2007). It is conceivable that the TCRα chromatin fiber folds into distinct territories to permit the developmental regulation of early versus late Vα and/or Jα regions. Alternatively, epigenetic marking may regulate or contribute to the ordered rearrangement of TCRα V gene segments. Related epigenetic marking may also underpin the non-random usage of Vκ elements (Feeney et al., 1997). Thus the fundamental point is that antigen receptor topology permits equal opportunities for all V regions that are trying to make a connection with DJ or J gene segment, but epigenetic marking may contribute to the non-random usage of V regions.
It is plausible that the topology described for the Igh locus is limited to that of antigen receptor loci, given that these loci undergo DNA recombination and may have special requirements for close spatial proximity of paired DNA elements. However, we consider this unlikely, as spatial distances also level off as a function of genomic separation in a region centromeric of the Igh locus as well as in other loci (Jhunjhunwala et al. 2008; Rauch et al., 2008; Mateos-Langerak et al., 2009). Thus, although a detailed analysis is required, we propose that the Igh topology reflects a general feature of antigen receptor loci, and possibly of large portions of the mammalian chromatin fiber. Such a topology mechanistically permits regulatory elements that are separated by large genomic distances, including enhancer and promoter elements, to be localized in close spatial proximity. This is not to say that the statistical structure described for the Igh locus is uniform throughout the genome. Likely, chromatin compartments will vary in the number of loops, the sizes of loops and linkers, the spacing between the bases of the loops, dynamics of loop formation and the number of clusters of loops. Resolving antigen receptor loci topologies at high resolution would be a major step forward. Furthermore, with respect to antigen receptor diversification and promoter-enhancer interactions, it is the seemingly aimless wandering of the eukaryotic chromatin fiber that is especially intriguing. Within a given lymphocyte population, a wide spectrum of conformations must exist that permit equal opportunities for the entire repertoire of antigen receptor gene segments. Such ‘Raphael-like trajectories’ are not completely randomly structured. Rather these patterns reflect ‘an elaborate, coherent and meaningful design’ that underpins the generation of antigen receptor diversity.
We are grateful to D. Forbes, A. Goldrath, M. Levine, G. Arya, B. McGinnis, R. Riblet, anonymous reviewers and members of the Murre laboratory for comments on the manuscript. We especially thank T. Knoch for stimulating discussions, T. Hwa for suggesting to perform triple point measurements and to define chromatin compartments as physical units and S. Cutchin for lessons in geometry. We thank D. Heermann for making us aware of the "Random-Loop" model. We thank the NIH for support.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.