Three models of the centromeric nucleosome are proposed in the literature. In the first model the centromeric nucleosome is an octamer, where Cse4/CENP-A replaces histone H3. While octameric nucleosomes with two copies of budding yeast Cse4 
or human CENP-A 
were assembled in vitro, whether only one or both copies of H3 are replaced in vivo is not known. There is evidence from different organisms for and against either of these possibilities. In HeLa cells CENP-A released from chromatin by micrococcal nuclease digestion is still associated with histone H3 even after 2M NaCl treatment resulting in dissociation of H2A and H2B, implying heterotypic tetramers with two histones H4, one H3 and one CENP-A 
. In contrast, in Drosophila
S2 and Kc cells when chromatin is digested with micrococcal nuclease and CENP-A/CID is immunoprecipitated, no H3 co-purifies with CENP-A 
. It was recently reported that Drosophila
CENP-A/CID forms homodimers in vivo, which are unexpectedly very salt-sensitive but could be crosslinked via cysteines in the four-helix bundle after a prolonged incubation 
. The authors did not exclude the formation of H3-CENP-A/CID heterodimers in addition to CENP-A/CID homodimers and it remains possible that different forms of CENP-A/CID nucleosomes are simultaneously present at the regional centromeres of Drosophila
and possibly other higher eukaryotes.
In this study we demonstrate that a budding yeast centromeric DNA fragment of only 214 bp is associated in vivo with both H3 and Cse4. We can exclude a homotypic octamer with two copies of Cse4. Our experiments suggest a very intimate spatial association between the conventional histone H3 and centromeric Cse4. This association cannot be explained if the Cse4-containing centromeric nucleosome is separated from the neighboring conventional H3 nucleosomes by spacer DNA as was proposed recently 
but rather suggests that H3 and Cse4 co-occupy the CEN DNA fragment of only 214 bp in length. We favor the Cse4-H3 heterotypic octamer model (, model 1). This octamer appears to be resistant to cysteine cross-linking, which might be due to the reduced stability of the four-helix bundle similar to the Drosophila
Models of how H3 and Cse4 can co-occupy the centromeric DNA.
The hexamer model postulates that in budding yeast the non-histone protein Scm3 replaces H2A and H2B and the nucleosome is composed of two copies each of Scm3, CENP-A and H4 
. Although it was initially proposed that the Scm3 dimer constitutes an integral part of the centromeric hexasome 
, the recent structures of budding yeast Scm3 associated with Cse4/H4 
and human HJURP in complex with CENP-A/H4 
revealed that binding of DNA as well as the (Cse4/H4)2
heterotetramer formation are incompatible with Scm3 binding. In the experiments in vitro it was demonstrated that Scm3 association with the reconstituted (Cse4/H4)2
nucleosome-like particles depends on a DNA binding domain within Scm3 
. Our results are compatible with the view that Scm3 does not form a part of the centromeric nucleosome. Under our experimental conditions we were able to co-immunoprecipitate minichromosomes with Cse4, H4, H2A, H2B and H3 but not with Scm3, which most likely dissociated from the centromere in yeast lysate.
Finally, the hemisome model proposes a tetramer consisting of Cse4, H4, H2A and H2B histones 
. According to this model, the Cse4 hemisome is positioned mostly at CDEII 
and is expected to occupy approximately 60 bp of DNA 
. This scenario leaves approximately 77 bp on each side of our 214 bp fragment available to accommodate the H3-containing nucleosome(s). We can speculate that a hemisome with Cse4 might, for example, be incorporated into a DNA loop between the two halves of an H3-containing octamer (, model 3). This model might explain the tripartite organization of the budding yeast centromere that was observed in the micrococcal nuclease protection pattern 
. Although it is technically possible that 77 bp upstream and downstream of the hypothetical centromeric hemisome are wrapped around ½ of the flanking conventional nucleosomes (, model 4), this model will result in tethering of the excised 214 bp fragment to the rest of the minichromosome which we did not observe ( and Figure S2
) and therefore can be excluded.
More exotic models can be also considered. Two recent studies compared Cse4-GFP fluorescence in vivo to independent standards and found 3.5–6.0 
or even 7.6 
Cse4-GFP molecules per budding yeast centromere in anaphase. Even more surprisingly, in prolonged G1 arrest Cse4-GFP fluorescence was reduced more than two-fold 
. These observations are inconsistent with the notion of a single Cse4 nucleosome at the budding yeast centromere 
. It was proposed that the budding yeast centromere is in fact a regional centromere with additional Cse4s associated with the flanking DNA similar to the much larger centromeres of higher eukaryotes 
. However, we could not observe any Cse4 associated with the non-centromeric part of the 2.4 kb minichromosome, which is expected to assemble 10 conventional nucleosomes. Therefore no additional Cse4 nucleosomes assemble, at least at these relatively short flanking sequences. Our results are consistent with those of 
who did not detect additional Cse4 nucleosomes in centromere-flanking regions by high-resolution mapping of yeast genome. The additional Cse4 molecules at the centromere could result from Cse4 mis-incorporation which is observed in strains overexpressing Cse4 
and could potentially be caused by GFP-tagging. Alternatively, additional Cse4 molecules may not be incorporated into the centromeric nucleosome but are rather associated with it via protein-protein and/or protein-DNA interactions (, model 2). In this scenario the centromeric nucleosome can be a Cse4-H3 heterotypic octamer to which more Cse4 molecules are bound. Intriguingly, when (Cse4/H4)2
heterotetramers were reconstituted in the presence of Scm3 into nucleosome-like particles on a 207 bp-long high affinity nucleosome positioning DNA sequence in vitro, high molecular weight complexes possibly representing additional Cse4/H4 in loose association with the Cse4/H4/DNA complex were detected 
. Similar complexes were reported to be assembled in vitro on a 148 bp CEN3 DNA 
It is more than a decade now since it was proposed that H3 is replaced by the histone variant Cse4 
. Our results appear to contradict this well-established dogma. If Cse4 and H3 indeed co-localize to the centromeric DNA why wasn't it noticed before? We can offer the following explanation. We have noticed that in most publications reporting ChIP experiments at the budding yeast centromere, the absolute efficiency of ChIP of the CEN DNA with H3 and Cse4 is very similar and typically in the range of 1% 
. The claim that only Cse4 is associated with the CEN DNA is then based on an observation that non-centromeric DNA is co-immunoprecipitated with H3 at about 5 to 10-fold higher rate than CEN DNA while almost no non-CEN DNA is found associated with Cse4 (Figure S8
). We suggest that if CEN DNA were generally difficult to immunoprecipitate, for example due to cross-linking of the large number of kinetochore proteins during the in vivo cross-linking, this would explain the reduced efficiency of H3 ChIP at the centromere compared to the chromosomal arms.
Our results appear to contradict those of 
. This group could co-immunoprecipitate differentially tagged versions of Cse4 from budding yeast but did not observe co-immunoprecipitation of tagged Cse4 and H3. However, one of the tagged Cse4s was expressed from a plasmid and Cse4 overexpression was reported to result in its ectopic incorporation genome-wide into octameric nucleosomes that were not observed in the wild type strain 
. It remains possible that even in budding yeast there is a degree of heterogeneity in the composition of the centromeric nucleosomes among different chromosomes and that either a homotypic Cse4/Cse4 octamer or a heterotypic Cse4/H3 octamer can provide the essential function.
At this time we can only speculate at the function of H3 at the budding yeast point centromere. It is possible that the presence of two different nucleosomes (or hemisomes), one with Cse4 and one with H3 provides structural asymmetry which might form the basis for two separate surfaces, one facing the sister centromere and another providing the attachment site for the spindle microtubule.