Published studies and our unpublished data on chromosomal evolution in primates were surveyed in the search for ENCs. We identified 31 ENCs: 15 in Catarrhini (Old World monkeys (OWMs) and Hominoidea) and 16 in Platyrrhini (New World monkeys (NWMs)). The vast majority of the NWM ENCs apparently emerged at the breakpoint of a chromosomal fission or repositioned from a telomere to the other telomere (see, for instance, the evolution of chromosome 3 5). Centromeres of human acrocentrics 15 and 14 are examples of ENCs that originated at a breakpoint and at a telomere, respectively, following a chromosomal fission 3. Their short arms consist of several megabases of acquired sequences. These circumstances suggested that telomeric ENCs could represent a different ENC category, with different progression dynamics. We therefore excluded these ENCs from the analysis and focused our investigation on the ENCs that emerged inside a chromosome and were not concomitant to a disruption of the seeding region.

Fourteen ENCs met these conservative criteria: one in woolly monkey (Lagothrix lagothricha, LLA, Atelinae, NWM), eight in OWMs 6, one in white-cheeked gibbon (Nomascus leucogenys, NLE) 17, one in orangutan (Pongo pygmaeus, PPY) 18, and three in humans (Homo sapiens, HSA) 51819. The ENC that emerged on chromosome 7 (human 8) of woolly monkey (NWM) has not been previously published. The evolutionary history of chromosome 8, supporting the emergence of an ENC in this primate, is summarized in Additional data file 1; fluorescence in situ hybridization (FISH) examples shown given in Additional data file 2a, b. Bacterial artificial chromosome (BAC) clones used in the analysis are reported in Additional data file 3. The eight ENCs found in macaque (Cercopithecinae) are also present in the silvered leaf monkey (Trachypithecus cristatus, TCR, Colobinae), indicating that all ENCs originated in the Cercopithecinae/Colobinae common ancestor. The rhesus macaque was used as a representative of OWMs because its genome has been fully sequenced 20.

Reiterative FISH experiments with corresponding human BAC clones were performed in non-human primate metaphases in order to precisely map these ENCs on the human sequence used as a reference (build35 assembly, March 2004; an example is shown in Additional data file 2c). The macaque sequence was used as a reference for the three human ENCs (rheMac2 release, January 2006). The position of the human ENCs in macaque was defined using macaque BAC clones hybridized to human metaphases. The results are summarized in Table 1. In some cases a BAC generated split signals on both sides of the centromere (Table 1, entries in bold), while flanking BACs gave a single signal on the expected pericentromeric side. The sequence corresponding to the splitting BAC was flagged as the ENC seeding region. In other cases the position of the ENC was defined by two overlapping BACs mapping on opposite sides of the ENC.

The human regions orthologous to the sequence domains where the non-human ENCs were seeded were investigated for evolutionary conservation against mouse and dog genomes by visually inspecting the University of California Santa Cruz (UCSC) Comparative Genomics Net tracks 21. The analysis was performed in order to validate the human sequence as bona fide reference sequence with respect to the changes the ENC regions underwent during evolution. We performed a similar comparative analysis for macaque regions corresponding to the three human ENCs for which the macaque was used as a reference. In both human and macaque sequences, the analysis encompassed approximately 2 Mb on each side of the seeding point. Substantial differences were found only in mouse (breaks or inversions at regions corresponding to human chromosome 2 (85.7-86.7 Mb and 137.6-137.7 Mb), chromosome 8 (61.9-62.8 Mb), and chromosome 11 (88.4-89.2 Mb)). No rearrangements were found in the dog, with the exception of the cluster of olfactory receptor (OR) genes located at 121.5-122.3 Mb in human chromosome 9 and absent in dog. The human/dog concordance strongly suggests that these rearrangements are derivative in mouse.

As mentioned, all eight ENCs found in OWMs were present in both macaque (Cercopithecinae) and silvered leaf monkey (Colobinae) species. Therefore, all originated before the Cercopithecinae/Colobinae divergence (see Figure 1), estimated to have occurred 16 million years ago (mya) 22. The position of the centromere on chromosomes orthologous to HSA2q (MMU12), HSA13 (MMU17), and HSA18 (MMU18) is shared by Hominoidea and NWMs 23 (unpublished data). The ENC seeding on these chromosomes, therefore, occurred in OWMs (Cercopithecoidea) after their divergence from Hominoidea (Figure 1), which was approximately 23 mya 22. It was not possible to precisely define the upper temporal limit of the remaining OWM ENCs because the position of the centromere on orthologous NWM and Hominoidea chromosomes showed discrepancies 1519.

The ENC on orangutan chromosome 11 is Pongo-specific 18 and is shared by both orangutan subspecies (Pongo pygmaeus abelii and Pongo pygmaeus pygmaeus). Consequently, it was seeded within the interval 4-14 mya (between Pongidae/Hominidae and PPY abelii/PPY pygmaeus splits, respectively). The HSA11 ENC is, very likely, Hominidae-specific 18. Thus, it dates within the interval 8-14 mya (after Pongidae/Hominidae split and before gorilla-pan-homo divergence, respectively). HSA3 and HSA6 ENCs are shared by great apes, so they date prior to 8 mya. Uncertainty on the ancestral position of the centromere in these chromosomes impinges on the uncertainty of the upper temporal limit of their occurrence 519. For the ENC of the woolly monkey (LLA7, NWM, Atelidae), we could define only the upper temporal limit of 22-23 mya, which is the estimated divergence time of the Atelidae (LLA) and Cebidae (CJA) lineages 24.

SD analysis was straightforward for the three human ENCs (chromosomes 3, 6 and 11) due to the high quality of the sequence assembly within these human pericentromeric regions 25. Duplications were found in the pericentromeric regions of all three human chromosomes. On chromosome 6 and particularly on chromosome 3, intrachromosomal duplications predominate. The duplication status of the sequenced macaque and orangutan genomes is less accurate with respect to humans because of the severe limitations intrinsic to the whole genome shotgun (WGS) sequencing assembly approach 26 in resolving high-identity duplications (note that whole genome sequence data are not currently available for the white-cheeked gibbon and woolly monkey).

To circumvent, at least in part, these problems, we exploited complementary bioinformatic and molecular cytogenetic techniques because they are partially 'assembly independent'. First, we examined each of the ENC regions for the presence of recent duplications in various primates using WGS sequence detection (WSSD) 27, where whole genome shotgun (WGS) reads from each primate are mapped against the human reference genome (hg17). Table 2 lists WSSD positive intervals detected for each primate species. Segmental duplications were detected, for example, on MMU4 (HSA6), MMU17 (HSA13), and PPY11 (HSA11). We then selected and tested various BAC clones by FISH. Some duplication data already resulted from experiments aimed at identifying the seeding region using human BAC clones (see above). However, split signals on both sides of the centromere could be alternatively interpreted as due to a disruption of distinct, non-duplicated sequences composing the human BAC, as a consequence of the colonization of alpha satellite DNA. Additionally, orthologous human clones may not be suitable for the analysis because of the restructuring process that could have substantially altered the pericentromeric sequences within each species. Final, new material, not represented in human BACs, may exist within these locations due to lineage-specific interchromosomal duplications.

Considering these potential biases, we also selected species-specific BAC clones identified with different approaches. For macaque we took advantage of the data on MMU BAC clones available at the Bioinformatics Research Laboratory of the Baylor College of Medicine, Houston, TX, USA 28. For orangutan (PPY) and white-checked gibbon (NLE), we queried appropriate BAC-end sequences from CHORI-276 (PPY) and CHORI-271 (NLE) BAC libraries using the Trace Archive database of the NCBI 29. A BAC library was not available for the woolly monkey (LLA). The phylogenetic distance of this NWM species coupled with the potential degenerative consequences of pericentromeric restructuring processes prompted us to discard the woolly monkey from the pericentromeric duplication analysis. Relevant FISH results of species-specific BAC clones yielding duplicated signals around the ENCs are reported in Table 3 (all tested clones are given in Additional data file 4); examples are shown in Figure 2 and Additional data file 2e, f. We discovered pericentromeric duplications mapping near the centromeres for almost all ENCs. One BAC-end of asterisked BACs in Table 3 and Additional data file 4 was identified by RepeatMasker as entirely composed of 171 bp alpha satellite repeats. No internal repeat was found truncated, and the homology with the apha satellite consensus ranged from 75% to 90%.

Two findings were of particular interest. Four nearly overlapping human BACs (RP11-543A19, -1043D14, -539I23, and -527N12) covering a region of 1.3 Mb (chr13: 61,111,769-62,699,203) around the MMU17 ENC gave duplicated signals around the centromere. Additionally, the two human BACs defining the ENC of MMU2 (HSA3) are 319 kb apart (Table 1). Three BACs spanning this interval (RP11-1089F10, -1142P11, and -10O22) failed to give any FISH signals in macaque, suggesting a deletion of the corresponding region within the macaque lineage. To exclude the possibility of a technical artifact, we mixed on the same slide human and macaque metaphases, added an excess of probe, and extended the hybridization time for three days. Again in these conditions, no signal was detected in macaque metaphases, while strong signals were present in human metaphases. We performed a BLAST sequence similarity using the human 319 kb region as query against macaque sequences deposited in the Trace Archive database 29. Only very small stretches (less than 1 kb) of homologous DNA were found externally located with respect to a central chr3:164,271,000-164,461,000 region (190 kb) in which no homology was detected (data not shown). Additionally, the macaque BAC clone CH250-91J4, identified at the Baylor College database (see above), mapping at HSA chr3:164,777,357-164,967,209, which is slightly external to the 'deleted' region, failed to yield any signal in human metaphases (data not shown). Altogether, these data strongly suggest that the region is highly rearranged in macaque.

We carefully analyzed the human genome (used as reference for non-human ENCs) and the macaque genome (used as reference for the three human ENCs) for annotated genes mapping within, or in proximity of, ENC seeding regions. The analysis was performed by querying the human and macaque RefSeq-related tracks of the UCSC genome browser 21 (hg17 assembly, RheMac2 assembly). No RefSeq genes were identified within the seeding regions as defined above (Table 1). In order to assess the statistical significance of gene depletion in the regions where ENCs were seeded, we performed a gene/exon density simulation (see Materials and methods) for 14 ENC regions. We found that the gene/exon density of the 14 ENCs is significantly depleted (p < 0.0001) when compared to random simulated data (Figure 3a-c). Table 4 reports the most proximal and distal RefSeq genes with respect to the ENC seeding point. The distance between the two genes is reported in the second column. Clusters of olfactory receptor genes flank the ENCs of chromosomes MMU14 (HSA11), MMU15 (HSA9), and HSA11 (MMU14). These OR clusters were not considered because OR genes are extremely redundant and a large number of these are pseudogenic within the primate lineage. The inactivation of a few of them would unlikely have strong phenotypical consequences. It is worth noting, in this context, that more than half of OR genes became inactive in recent human evolution 30.

The precise location of some human neocentromeres has been achieved through CENP-A mapping by chromatin immunoprecipitation (ChIP)-on-chip experiments (reviewed by Marshall et al. 11). AT content has been shown to be one of the few common features shared by these neocentromeres. We calculated the AT content for the human domains corresponding to the ENC seeding regions as defined in Table 1. The results are reported in the last column of Table 1.

Metaphase preparations were obtained from cell lines (lymphoblasts or fibroblasts) from the following ape species: common chimpanzee (Pan troglodytes, PTR), gorilla (Gorilla gorilla, GGO), Bornean orangutan (Pongo pygmaeus pygmaeus, PPY), white-cheeked gibbon (Nomascus leucogenys, NLE). OWMs: rhesus monkey (Macaca mulatta, MMU), vervet monkey (Chlorocebus aethiops, CAE, Cercopithecinae), silvered leaf monkey (Trachypithecus cristatus, TCR, Colobinae). NWMs: wooly monkey (Lagothrix lagothricha, LLA, Atelidae), common marmoset (Callithrix jacchus, CJA, Callitricidae).

DNA extraction from BACs was reported previously 2. Co-hybridization FISH experiments were performed essentially as described by Lichter et al. 34. To suppress cross-hybridization signals due to repeat sequences, COT1 DNA (5 μg) was added to the hybridization mixture. Digital images were obtained using a Leica DMRXA2 epifluorescence microscope equipped with a cooled CCD camera (Princeton Instruments, Princeton, NJ, USA). Cy3-dUTP, Fluorescein-dCTP, Cy5-dCTP and DAPI fluorescence signals, detected with specific filters, were recorded separately as grayscale images. Pseudocoloring and merging of images were performed using Adobe Photoshop™ software.

BAC-end sequences were retrieved from the Trace Archive database 29. They were then analyzed using the RepeatMasker software 35 in order to identify BAC-ends partially or entirely composed of repeat sequences. The software provides information on the extension and type of repeat.

In order to identify segmental duplication content in various primates, we used the previously described assembly-independent approach (WSSD) where WGS sequence reads 27 from each query primate genome were mapped against regions from the human genome reference sequence (build35) corresponding to the ENCs. We considered regions of excess WGS read depth (≥ mean + 1.5 × standard deviation) to represent putative duplicated regions in each primate. Due to different genomic sequence divergences between each primate and the human reference sequence, we used sequence identity thresholds of ≥ 88% to map macaque reads while ≥ 94% was used for alignment of reads from chimpanzee and orangutan.

In order to statistically assess the depletion of gene/exon density in the regions where ENCs were seeded, we performed the gene/exon density simulation as follows. First, we computed the average gene/exon density for the 14 ENC regions based on their annotation within the human genome. This became our observed value for gene/exon density within ENC regions (red line in Figure 3). Next, we randomly selected the same number of gap-free base-pairs (23.2 Mbp) from the human genome and computed the average gene/exon density for these randomly selected intervals. We generated 10,000 replicates and plotted the distribution of gene/exon density based on this simulation. We computed an empirical p-value as the number of replicates with gene/exon density equal to or lower than the observed density in 10,000 replicates. We repeated the analysis excluding ENCs that had been identified within the human lineage of evolution (n = 3) and obtained similar results (data not shown). For genes, we considered the position of all human non-redundant genes (RefSeq gene n = 22,589) and their corresponding exons as determined by the UCSC genome browser 21. As a second analysis to assess transcript density, we independently mapped the location of all spliced human ESTs (n = 4,246,559) to the human genome (build35) and selected the location of the highest alignment score. If an EST/transcript mapped to two or more locations with an equivalent score, one was selected at random, such that each transcript was assigned once and only once to the human genome. As part of this analysis, we clustered exons that overlapped as a result of alternative splicing and counted each cluster as a single exon.