There are several reasons why long-range chromatin movements may have been difficult to detect and their functional relevance underestimated. First, the rapid, short-term motions appear to be far more frequent with the long-range movements being the rare exception. Second, long-range motions seem to be exquisitely sensitive to light damage and their observation requires low light techniques, which are not routinely used [13]**. Third, long-range motions may be very tightly controlled and may only occur as part of physiological responses. They are less likely to be captured by observation of generic cultured cells in vitro and require analysis of chromatin movements in physiological situation such as that observed during cell cycle progression in D. melanogaster spermatogenesis [2]. It may not be surprising that long-range, directed chromatin movements are not the norm but possibly the exception and only occur in response to specific physiological cues. It is not entirely trivial to envision how directionality of movement would be established and how, if the actin/myosin system mediates this movement, the starting and endpoint of the motion are determined. Intriguingly, the nucleus contains several molecular motor proteins of unknown function and it will be key to determine whether they are involved in physiological long range chromatin displacements.

These and many other observations clearly show that genes move within the nucleus and that the movement is often related to their activity. But it is less clear whether the observed movement of gene loci is essential for gene expression. In the case of the internalization of the β-globin locus during erythroid maturation, the gene becomes activated prior to its change in localization and activity does not depend on its internal nuclear position [17]**. Similarly, association with splicing factor compartments as observed for human α-globin was not seen for β-globin or for mouse globins [18]**. With respect to expulsion from chromosome territories, many highly expressed genes do not dissociate from chromosomes and while the Hoxd locus loops out from chromosomes in cells along the anterior-posterior axis, it does not do so in the limb bud [22]**. Rather than these positioning events being required for proper gene expression, it seems more plausible that the repositioning of a gene is a consequence of its activity status. This does not mean that positioning is not important. Observations in S. cerevisiae suggest that the association of an activated locus with the nuclear periphery is neither sufficient nor necessary for activity, but serves to optimally express the gene [24]**. Thus the movement of a locus is a fine-tuning step in the hierarchy of gene regulation.

Taking into consideration these various observations on chromatin motion and positioning, a plausible scenario that emerges is the following [14,25] (Fig. 2): As a gene becomes activated, its chromatin structure is altered locally by the action of chromatin remodeling machines. This leads to changes in the higher order chromatin structure and likely modulates the local accessibility of transcriptional regulators to the locus. It is possible, and likely, that transcription commences at this point, albeit possibly at a low level. Since the intranuclear mobility of the locus is probably strongly affected by its higher order chromatin structure, the locus is now increasingly mobile. Its motion is most likely still locally constrained, although its potential for long range motion may be increased. This altered mobility of the partially activated, or merely potentiated, locus now allows it to explore its nuclear environment in a stochastic and random manner by local, constrained diffusion and to find a functional compartment that corresponds to and supports its functional status, such as a transcription factory or a splicing factor compartment. While this association most likely perpetuates the gene’s activity status, it does not necessarily confer stability, since at least some genes appear to be expressed in transcriptional bursts [26**,27]. A similar sequence of events may occur during silencing, but in that case a gene locus primed for repression would undergo locally constrained motion before it stochastically associates with a heterochromatin domain for stronger and more permanent silencing (Fig. 2).

Several recent studies have addressed this issue and a consensus is emerging demonstrating that DSBs are largely immobile in mammalian cells. Two early analyses of damaged DNA in chemically fixed cells came to divergent conclusions. Induction of DNA damage using ultrasoft X-rays showed that damaged regions were held in a relatively fixed position within the nucleus [29]. In contrast, irradiation with α-particles led to the movement of damaged sites over several micrometers [30]. The discrepancy between these observations has recently been resolved by analysis of DSB dynamics in living cells. Using the core histone H2B as an in vivo marker, no significant motion of damaged DNA regions after damage by UV or γ-irradiation was detected [31]. This observation on bulk damaged DNA was extended by the development of an experimental system to visualize the dynamic behavior of broken chromosome ends in living cells after controlled induction of a single DSB [32]**. Direct tracking of the movement of the broken chromosome ends by specific fluorescent tags on either side of the break demonstrated that cut ends were unable to roam the nuclear volume over large distances and breakage of the chromosome only resulted in slightly increased local motion of the DNA break. Surprisingly, the positional stability of the broken ends was not dependent on the structural proteins H2AX and SMC1 nor on the MDC1/RAD51/NBS1 repair complex which acts as primary sensor of breaks and had been implicated in anchoring DNA ends. Increased local mobility followed by long range mobilization of broken DNA ends was, however, observed in the absence of the heterodimer Ku80/Ku70, pointing to its importance in aligning and holding broken chromosomes in place. These observations in living cells strongly suggest that broken chromosome ends are positionally stable in the mammalian cell nucleus. As a consequence, translocations must preferentially occur between already proximally positioned simultaneous DSBs rather than spatially distant chromosome lesions.

Things might be very different in yeast. S.cerevisiae cells have been shown to contain a limited number of nuclear sites which act as DNA damage repair centers [33]. These centers seem to serve repair of multiple DNA breaks since breaks on different chromosomes associate with a single repair site, suggesting that DSBs migrate to repair centers. This is in stark contrast to mammalian cells where multiple DSBs do not cluster around a common repair site, but rather each break creates a new repair focus [32]**. It is probably not a coincidence that the differential dynamic behavior of DSBs in yeast and mammalian cells directly mirrors the ability of chromatin in these systems to explore the nuclear volume. While the default short-range motion of chromatin over ~1um allows a yeast DSB to explore a large fraction of the yeast nucleus creating a high probability of encountering a DNA repair center, the same extent of motion limits the DSBs to a small fraction of the mammalian nucleus, thereby limiting their ability to encounter another DSB or a repair site. The differential ability of a DSB to explore varying fractions of the nucleus in yeast and mammals thus seems to have contributed to the development of very different strategies of organizing DNA repair in the nucleus of yeast and mammals.

The observed immobility of broken chromosome ends in mammalian cells has significant implications for the mechanism by which chromosomal translocations form [28] (Fig. 3). Since DSBs on different chromosomes are unable to diffuse through the nucleus, the illegitimate joining of unrepaired breaks must occur predominantly amongst neighboring chromosomes. The emerging view then is that translocations happen when two DSBs on distinct chromosomes are positioned such that they can interact with each other as part of their locally constrained motion. If DSBs are further apart than their intrinsic reach of local motion, they will not translocate. It is important to point out that the formation of translocations does not require long-range motion and the ability of a broken chromosome end to move over long distances sufficient for formation of translocations. We know this because in the absence of Ku80, broken chromosome ends separate, at times over several micrometers, but the majority of them remain unrepaired and do not undergo translocations [32]**. It thus seems that translocations occur preferentially amongst proximally positioned chromosomes.

The observations on chromatin mobility of DSBs pointing to translocations amongst proximal chromosomes are complemented by an independent body of evidence based on mapping of the 3D spatial position of frequent translocation partners. A series of studies have demonstrated that the frequently translocated chromosome regions and specific break points are on average found in preferential spatial proximity and are often direct neighbors prior to undergoing translocations [14,28]. This concept has recently been extended by the finding that translocation frequencies amongst chromosomes in human lymphocytes not only correlate with physical association of the involved chromosomes, but with the degree of intermingling of the chromatin fibers at the interface between the two chromosomes [14, 34**].

The positional stability of DSBs has direct physiological consequences. For one, it may explain the well-established observation that tumors from different tissues exhibit distinct recurrent chromosomal translocations [35]. It is well known that the 3D spatial organization of genomes is tissue specific [36,37]; thus, chromosomes and gene loci physically associate with different partners in different tissues. Since the immobility of DSBs restricts their choice of interaction partners to their immediate physical neighbors, the set of translocation partners is limited and directly determined by the tissue-specific arrangement of chromosomes.

Possibly the most important physiological consequence of the positional immobility of DSBs is its protective role against genomic instability. If broken chromosome ends were highly mobile within the nucleus, the probability of illegitimate joining would be very high every time two DSBs occur within the same cell nucleus. The positional immobility prevents illegitimate joining of spontaneously occurring distant DSBs since they cannot be brought into spatial proximity. This might afford the cell the opportunity to repair at least one of the breaks before they have an opportunity to interact with each other and illegitimately join to form a translocation. Interestingly, in yeast where any given locus has access to a higher proportion of the nucleus, and thus can interact with a larger fraction of the genome, recombination events between sites on different chromosomes are significantly more frequent than in mammalian cells [38].