Cell type specific chromatin distributions on the nuclear level have been described for over a century [, p.100]. Differences between cell types have also been described for the distribution of heterochromatin detected with antibodies against methylated histones , for the radial distribution of gene rich and gene poor chromosomes [[37, 40, 45] this study] and the occurrence of clustering between specific chromosome territories . Here we show an example were large-scale chromatin organization of chromosome territories changes during differentiation, and thus add a new feature to the list of nuclear architectural properties that can differ between cell types.
To quantify chromatin dispersal of labeled chromosomes in cells of various differentiation stages, we counted the number of separate, labeled chromatin objects to which the chromosome territories disintegrated at increasing thresholds. In the investigated chicken cell types, chromosome territories of further differentiated cell types disaggregated into more objects. An increase in object number during differentiation may indicate that a significant number of compact chromatin domains with silent genes separate from each other into several, more decondensed, "open" chromatin domains. This would increase the accessibility to transcription factor complexes from the interchromatin compartment by increasing the available chromatin surface area of the chromosome territory. In human lymphoblasts gene rich chromosome 19 territories were found more decondensed than gene poor chromosome 18 territories  and electron microscopic evidence suggests that active genes are exposed at chromatin domain surfaces in a zone called the perichromatin region, a transitional zone that marks the transition form the chromatin domain periphery to the interchromatin compartment . A caveat of this interpretation is that so far no unequivocal proof for a profound influence of higher order chromatin compaction on gene activation and gene silencing has been presented. A further possibility is that inactive loci in the more differentiated cells do not require a tight spatial silencing by chromatin compaction anymore because the set of available molecular activators and repressors has changed. At present we can only speculate whether the correlation of increased dispersal of chromosome territories with differentiation state is a widespread feature or restricted to a few chicken blood cell types. At the highest, nuclear level of chromatin organization it was described for mammalian nuclei that heterochromatin shows distinct patterns of large blocks in terminally differentiated cells but not in blood stem cells and tumor cells [47, 48]. This indicates a compaction of chromatin in differentiated cells rather than in their precursors, unlike in our current data on the chromosome territory level. It is possible, that heterochromatin (consisting mainly of repetitive sequences) and the bulk of labeled chromosome territories behave differently in these aspects. Due to their suppression with unlabeled repetitive DNA, repetitive sequences are underrepresented in chromosome territories detected by FISH as in the present study. Also, the rather small amount of repetitive sequences and heterochromatin in the chicken genome (genome size ~1.2 Gbp according to , 1.1 Gbp according to ) as compared to mouse and human genomes (~3.2 Gbp each, ) may lead to differences in nuclear organization.
Multipotent myeloid precursor cells have the smallest nuclei of the cell types investigated here. Myeloblasts have on average larger nuclei than proerythroblasts. If the observed disaggregation of chromosome territories were based on a nuclear volume increase, the larger myeloblast nuclei should have a stronger dispersion of chromosome territories than proerythroblasts. However, the opposite is true (Figure 5). We thus conclude that chromosomal dispersion is not related to nuclear size. In general, we observed larger nuclear volumes for further differentiated cell types. Increasing nuclear size was also observed during maturation of nerve ganglia cells  while a volume decrease was described during the maturation of lymphocytes . Accordingly, unlike recently suggested , a decrease of nuclear size does not appear to be a phenomenon generally associated with terminal differentiation events.
The lysozyme gene domain is positioned inside the chromosome 1 territory in multipotent myeloid precursor cells where the lysozyme gene is inactive, but on the surface or outside in most of the territories in activated macrophages where the gene is strongly expressed. We thus did find a tendency to more exterior regions of the chromosome territory for the highly expressed gene from in activated macrophages although actual looping-out (without visible contact to the territory) was observed in only about 6%. Interestingly, while the radial distribution of the lysozyme gene domain within the nucleus is about the same in all cell types, the harboring chromosome 1 territories show differences. The finding that in erythroblasts the cLys signal is more exterior than the chromosome 1 territory signal median but in unstimulated macrophages the opposite is true also argues for a cell type specific organization of chromosome territories. A similar observation has been described for a IL-3 induced differentiation of human leukemic K562 cells where the β-globin gene cluster does not change nuclear position but the harboring chromosome 11 territory does . However, in human hematopoietic cells a relocation of a gene to a different preferential radial position  or to or away from heterochromatic nuclear compartments has been observed for some genes, correlated with transcriptional regulation at different developmental stages [e.g. [33, 53]]. Unfortunately, the harboring chromosome territories were not labeled in these studies.
While we can exclude a tethering of the inactive lysozyme gene to the centromere, at first glance this result seems compatible with the hypothesis that inactive genes are stored away in internal regions of the chromosome territory and active genes are on their surface or even looped out. However, several aspects suggest an alternative explanation. (i) Embedded in the chicken lysozyme gene domain is a second gene, cGas41, which, albeit on a low level, is expressed in all cell types used in this study, including multipotent myeloid precursor cells . Thus we found an example of an active gene with a location inside the territory as it was described previously for some mammalian genes . (ii) Although the position of the lysozyme gene domain is most peripheral in activated macrophages where the expression is highest, we also found a shift towards more external positions from multipotent myeloid precursor cells to further differentiated proerythroblasts, both non-expressing cell types. (iii) In addition to the lysozyme gene domain, we investigated the chromosome 1 centromere. Surprisingly, both loci showed a very similar distribution in all cell types investigated. Transcription from centromeres has been observed in yeast [reviewed in ] and from a human neocentromere . Formally, we thus cannot fully exclude that centromeric transcription may occur in chicken. We regard it as extremely unlikely, however, that tissue dependent differences in centromeric transcription play a role in the cell type specific spatial positioning found here. The observed modification in the morphology of chromosome territories during differentiation rather invites to hypothesize that the positional changes observed for the lysozyme gene domain are not restricted to this particular chromatin loop or only to those chromatin loops which harbor genes that become activated during cell differentiation. Instead, these positional changes may reflect a more general, differentiation dependent change in large-scale chromatin structure. Differentiation processes may thus have a more global impact on chromatin structure than previously suspected.