Cancer Genetics and Cytogenetics 8(3): 249-274 (Mar 1983)
Sequence of centromere separation: Occurrence, possible significance, and control
Baldev K. Vig

This review describes the existence of a phenomenon, sequential separation of centromeres, in mitotic cells of various species including both animals and plants. Critical observations at metaanaphase show that the centromeres of chromosomes in a given genome do not separate into two sister units randomly, but that there is a genetically controlled, nonrandom, species-specific sequence which is independent of the length of the chromosome or the position of the centromere. A stricter control appears to exist for late-separating than for early-separating chromosomes. At early stages of metaanaphase several chromosomes initiate onset of separation simultaneously or in rapid succession, but late-separating chromosomes are better defined in their sequential position. The effect of Colcemid on the sequence of separation is minimal. It is proposed that aneuploidy in humans and other organisms may result from out-of-phase separation of a given chromosome. With the exception of chromosome No. 16, it appears that very early- or very late-separating centromeres are involved in human trisomies more often than those in between.

Perhaps one function of centromeric heterochromatin is the control of centromere separation. The amount of such chromatin shows a positive correlation with the timing of separation of the centromeres. Superimposed upon this quantitative influence is the qualitative aspect, as discussed for various genomes. This suggestion explains a lack of extremely large quantities of heterochromatin near the centromere. Its existence in the form of homogeneously staining regions distal to the centromere, as in some cancer cells or in sex chromosomes, seemingly has no influence on the separation of centromeres.

A brief discussion of centromere separation errors in human disease is provided, and suggestions for further studies are made.


Studies carried out with three plant species also testify to the existence of sequences. In a reconstructed karyotype ACB of the broad bean (Vicia faba, 2n = 12) [19, 20] in which all pairs of chromosomes are identifiable without banding, the earliest separating chromosome was No. 5. This chromosome constituted 00% of the sample among cells in which only up to five chromosomes had separated (the expected value was 23%). However, it had separated completely in cells with more than five chromosomes showing separation [21]. The chromosomes last separating belonged to pair No. 4, whereas others were in between. A summary of the CSIs is presented in Figure 5. When subjected to a t-test, these data showed significant differences in the CSIs of several pairs [21].

In this reconstructed karyotype, a few seeds were found with a recognizable reciprocal translocation between the nucleolus-organizing chromosome No. 3 and the longest chromosome, No. 1. This provided a sample of 42 cells upon which to base analysis for CSI. It turned out to be similar to that for the parental ACB line. This comparison might indicate that a translocation involving chromatin at the noncentromeric region has no effect on the sequence of separation of centromeres. Similar conclusions are drawn from the study of another reconstructed line, EF.

In Crepis capillaris (2n = 6, Fig. 6) the largest submetacentric chromosome separates earlier then the other two subtelocentric pairs [22]. In Haplopappus gracilis (2n = 4) the large pair also separates earlier then the smaller one. However, great variation was observed in these two species in that the apparent degree of synchronous separation of the homologs is quite low compared to such synchronous separation observed in V. faba (see Refs. 21 and 22, and below).

The above studies indicate that the centromeres in the genomes of both animals and plants separate in a nonrandom, genetically controlled sequence which is dependent upon neither the size of the chromosome nor the position of the centromere. It is further suggested that such sequences may be independent of the tissue type or physiological conditions under which cells are grown.


The data on the sequence of separation make use of averages of the values for both members of a homologous pair. If nonrandom separation of centromeres is under genetic control, then these genetic elements would he more similar in homologs than in nonhomologs. In other words, homologs should show a closer time relationship in separation than nonhomologs. Besides, one should observe, to a degree, synchronous separation for homologs. The data on the human genome [12] clearly points to the fact that two homologs, e.g., chromosomes No. 2, separate much closer to each other than any nonhomolog a measureable distance away in the sequence, e.g., No. 1, 13-15, 21-22, or XY. This is also true for all other species for which such data are available, e.g., Potorus [17], V. faba [21], and Haplopappus and Crepis [22].

Nonetheless the degree of synchronous separation of the two homologous centromeres, as measured by the value of separation (initiation of separation versus complete separation of the daughter centromeres), never reaches 100%. It also varies from species to species. Thus in Potorus, the degree of synchronous separation was 42% for chromosome No. 1, 41% for No. 2, 32% for No. 3, and so on. Synchronous separation between nonhomologous Y1 and Y2 chromosomes was only 10%, whereas for X and Y2 it was 21% [17]. Similarly, C. capillaris exhibited maximum synchrony of up to 51% and H. gracilis only 31%. This was also true of human genomes. On the other hand, V. faba showed up to 83% synchronous separation for pair No. 1 and the lowest, 72%, for pair No. 4 (Table 4 in Ref. 21). These results are in conformity with the idea of genetic control (or controlling elements), since natural self-fertilization in species such as V. faba expectedly confers a greater degree of homozygosity for such elements than in crossbred species such as Potorus, human, Haplopappus, and Crepis. Such data from inbred strains of, say, Chinese hamsters should be instructive, but no analysis of this sort has so far been performed.



The above studies indicate a direct, albeit only suggestive, correlation between the amount of CCH and the sequence of centromere separation; the larger the block of CCH a chromosome carries, the later its centromere separates. This means that an extremely large amount of CCH may delay the separation of a chromosome to a point where it would be so delayed in separation as not to be included in anaphasic memovements [sic] with reasonable success. This idea explains why a chromosome with such a large amount of CCH has not been discovered, inspite of the fact that excessively large quantities of constitutive heterochromatin at the distal end of the chromosome are common observations in both normal (e.g., human Y chromosome) and abnormal (e.g., cancerous cells) [38] cells of various types.

A direct relationship between the quantity of CCH and delayed separation at the centromere is possible only if the CCH in various chromosomes of a genome is qualitively uniform. A lack of such uniformity may mean that some qualitatively more influential type of CCH may hold the centromeres together for a longer time than predictable based only on the quantity of this heterochromatin. Perhaps such is the case with the human genome where we did not find a direct correlation of the type mentioned above. Thus, whereas chromosomes No. 1, 9, and 16 in humans carry rather large quantities of CCH, the centromeres separating last are the ones that carry ribosomal genes, namely, No. 13, 14, 15, 21, and 22, Nonetheless when two No. 1 chromosomes distinguishable due to C-chromatin polymorphism in a given cell are studied, the homolog with less CCH usually separates earlier than the other [12] (Fig. 9g). These studies may therefore show that the position of a given centromere in the sequence of separation is determined by the quantity of CCH and that superposed upon this quantitative influence is the qualitative aspect. As a matter of fact the chromosomes separating last in humans, including the Y chromosome, also have SAT IV in common [39]. Does a direct relationship between separation and CCH then mean that various chromosomes in the genome carry qualitatively uniform heterochromatin? Studies on this aspect in wood lemming and Potorus should elucidate this point. It is also possible that an asynchronous heterochromatin replication pattern, as reported in maize [40], has some bearing on the timing of centromere separation.

The cytological satellites and/or CCH in humans are shown to be heterogeneous in DNA content in that the DNA content may not be related to the size or intensity of fluorescence of the satellite [41]. An apparently uniform CCH region, as in chromosome No. 1 in humans, may actually be made up of several interdistinguishable subunits [421. Furthermore, qualitative differences are now being revealed by cytological techniques, e.g., the differences between the CCH of chromosome No. I versus No. 9 and 16 in humans demonstrable by the use of the fluorochrome D 287/170 [43]. Perhaps an important factor to be considered in support of the CCH hypothesis, especially in view of the fact that the centromeric region per se does not respond to labeling for DNA synthesis [44], is the demonstration that in Triturus lampbrush chromosomes the satellite DNA is transcribed [45]. Such transcription may involve proteins associated with the separation of centromeres, besides other functions attributed to such DNA [46].

Imprinting, Disruptive Selection, Antithetical Dominance