We analyzed homologous regions between the human genome and those of the rat, mouse, cattle, pig, cat, horse, dog, and chicken. By using the HSBs described by Murphy and coworkers 5 and adding data from the human/chicken and human/dog whole-genome sequence assemblies, we were able to identify 1,638 syntenic blocks in the human genome (Additional data file 4). (The dog radiation hybrid genome map data used by Murphy and coworkers 5 was replaced by the dog whole-genome assembly, which is now available.) The analysis of the human/chicken and human/dog whole-genome sequence assemblies revealed a total of 550 syntenic blocks among the three compared species (Additional data file 4). The homologous chromosomal segments of the seven mammals and the chicken were plotted against the 550 band human ideogram (Additional data file 1). We excluded the human chromosome Y from our study of evolutionary breakpoint regions (see Materials and methods, below).

In addition we identified the chromosomal position of 1,152 evolutionary breakpoint regions of 4 Mb or less in size (Additional data file 5) in the human karyotype and their corresponding evolutionary breakpoints (n = 2,304; Additional data files 1 and 5). The 2,304 evolutionary breakpoints grouped within 352 evolutionary chromosomal bands, which represents 67.77% of the human genome (2,217.46 Mb of the 3,272.19 Mb of the total human genome, NCBI35; Additional data file 5). See Figure 1 for a schematic representation of evolutionary breakpoint regions, evolutionary breakpoints and evolutionary chromosomal bands, as well as the Materials and methods section (below) for definitions of these terms. Approximately 45% (159 out of 352) of the evolutionary chromosomal bands contain evolutionary breakpoints in three or more of the eight species compared herein (Additional data file 6). These data clearly show that the distribution of the evolutionary breakpoints and breakpoint regions is concentrated in specific bands and/or chromosomes.

An analysis of the distribution of evolutionary breakpoints among the evolutionary chromosomal bands using JMP software (see Materials and methods, below) revealed a mean of six evolutionary breakpoints per evolutionary chromosomal band. Out of the 352 evolutionary chromosomal bands that were identified, 296 contain between one and ten evolutionary breakpoints, whereas 16 human chromosomal bands contain 20 or more evolutionary breakpoints each (10p11.2, 10q11.2, 15q13, 15q24, 15q25, 17p13, 17q24, 1q42.1, 22q11.2, 2p13, 2q14.3, 3p25, 3q21, 4p16, 7q22 and 8p23.1; Additional data file 6). Otherwise stated, 4.21% of the human genome (137.9 Mb of 3,272.19 Mb) accumulates 17.79% of all evolutionary breakpoints (410 of the 2,304 identified). Similarly, not all human chromosomes have been equally affected by the evolutionary process. Human chromosomes 1, 2, 3, 4, 7, 8, 10, 15, 17, and 22 carry most of the evolutionary breakpoints, whereas human chromosomes 14 and 21 are the least frequently involved.

Given the distribution of evolutionary breakpoints outlined above, we proceeded to determine whether there is a significant correlation between the position of evolutionary breakpoints and the known location of fragile sites. We mapped all fragile sites (both rare and common) and evolutionary breakpoint regions (regions ≤ 4 Mb; Table 1 and Additional data file 1) to their location on the human ideogram at the 550 band resolution. Our examination reveals that 147 chromosomal bands express fragile sites (both common and rare). A contingency analysis shows that those bands that express fragility (they contain either rare or common fragile sites) have a tendency, although not significantly so (P = 0.09), to concentrate evolutionary breakpoints as compared with bands that do not express fragile sites. In fact, we observed 104 bands that contain fragile sites (rare and common) and evolutionary breakpoints, in contrast to the 95.4 bands expected if the distribution were random. A more refined analysis was subsequently conducted in which four categories of chromosomal bands (those that contain common fragile sites, those with rare fragile sites, bands with both common and rare fragile sites, and finally bands with no fragile sites) were examined using contingency analysis. There is a significant tendency (P = 0.01) for bands with rare fragile sites to accumulate evolutionary breakpoints (22 of the 24 bands known to express rare fragile sites contain evolutionary breakpoints versus the 15.6 bands expected if the distribution were random). The same tendency does not hold in the case of common fragile sites, where 73 of 111 bands that express common fragile sites contain evolutionary breakpoints (72.2 expected), or bands that contain evolutionary breakpoints but no fragile sites (248 observed versus 256.3 expected).

As stated above, resolution differences exist between cytogenetically defined fragile sites in human chromosomes and the molecular delimitation of evolutionary breakpoints. That differences in resolution may confound the association between them is clearly of concern. However, of the 12 autosomal common fragile sites that have been characterized at the molecular level (Additional data file 8), six (FRA4F, FRA6E, FRA7E, FRA7G, FRA7H, and FRA9E) were shown to span evolutionary breakpoints in at least one of the species analyzed with an additional two fragile sites (FRA3B and FRA16D) located within 1 Mb of evolutionary breakpoints (Additional data file 8). Importantly, of the four autosomal common fragile sites with the highest expression frequencies (FRA3B 28 , FRA6E 30 , FRA7H 34 , and FRA16D 37 ), two (FRA6E and FRA7H) are localized within evolutionary breakpoints, and two (FRA3B and FRA16D) lie within 1 Mb of breakpoint boundaries. With respect to the eight cloned rare fragile sites 18 19 20 21 22 23 24 25 , three (FRA10A, FRA16A, and FRA16B) are located in bands that contain evolutionary breakpoints in at least one of the species analyzed by us.

The distribution of tandem repeats in human chromosomes was analyzed using 250,000 bp search windows in order to determine whether there is any correspondence between tandem repeats, fragile sites (both rare and common), and the location of evolutionary breakpoints (Additional data files 2 and 8). The tandem repeats range from microsatellites (unit size 1 bp to 6 bp) to different types of minisatellites (from 7 bp to 300 bp). We identified a high concentration of tandem repeats in the telomeres and the pericentromeric regions of each chromosome (Additional data file 2), mirroring earlier findings (for instance, see Näslund and coworkers 49 ). The distribution of tandem repeats (1 to 300 bp) along human chromosomes showed that on average 3,738.56 bp of the 250,000 bp of genomic sequence contained in each window comprised tandem repeats (about 1.5%). Chromosome 19 is exceptional for the high number of repeats found along its length 50 , which is almost double (8,377.27 bp) the average for the whole genome (Table 2 and Additional data file 3). Additionally, chromosome 19 has been shown to be exceptional in many other genomic features, most of which (including the high number of repeats) may be due to the extremely high GC content of this chromosome 51 52 .

Evolutionary breakpoints can be defined by levels of resolution 53 . The holistic perspective of evolutionary breakpoints has traditionally been underpinned by molecular cytogenetic studies that assign regions of chromosomal homology to species of the same or different orders of mammals at the chromosomal band level. Investigations using comparative chromosome painting (ZOO-fluorescence in situ hybridization [ZOO-FISH]) involving more than 80 different species from almost all of the recognized eutherian orders have defined regions of the human genome that are implicated in chromosomal evolution (for review, see Froenicke 54 ). The integration of cross-species chromosome painting data published from 30 nonprimate species 54 , and even greater numbers of primate species 8 , clearly demonstrate that evolutionary breakpoints are not uniformly distributed along the length human chromosomes, and in some cases they are conserved during chromosome evolution.

The use of whole-genome comparisons (the reductionist view) allows for the delimitation of evolutionary breakpoints at a finer level of resolution than can be obtained by chromosome painting. By analyzing published data 5 , and adding complementary information from the human/chicken and human/dog whole-genome sequence assemblies, we were able to identify 1,152 evolutionary breakpoint regions throughout the human genome at a resolution of 4 Mb or less, which contain 2,304 evolutionary breakpoints. Plotting the evolutionary breakpoints included in our data onto the 550 chromosomal band human ideogram provided a means of combining the cytogenetic and the sequence comparisons. This identified 352 human chromosomal bands that contain evolutionary breakpoints and showed that the distribution of evolutionary breakpoints is not uniform in the human genome. Quite clearly, there are evolutionary 'hot spots', defined by chromosomal bands, which are coincidental with genomic reorganization characterizing different lineages during the evolutionary process (breakpoint reuse 5 ).

Although the exact number of fragile sites described in the human genome is a matter of interpretation, a recent revision lists 119 fragile sites, 88 of which are defined as common and 31 as rare 39 . Our data show that human chromosomal bands that express fragile sites (both common and rare combined) have a tendency to contain evolutionary breakpoints (Table 1), although the association is statistically supported only in the case of rare fragile sites. This association suggests an important role for fragile sites in genome reorganization, most likely by functioning as regions of chromosomal instability.

Although the mechanisms underlying the breakage at common fragile sites are still poorly understood, rare fragile sites are associated with the amplification of repeat motifs (CCG repeats and AT-rich regions). The molecular characterization of 13 common fragile sites has revealed that there are no simple repeat sequences responsible for their instability (for review, see Schwartz and coworkers 39 and Glover 55 ). Rather they are enriched in A/T content, have the potential to form secondary structures, and contain clusters of flexible sequences (flexibility clusters). These are all features that may affect DNA replication and chromatin condensation, suggesting a common basis for fragility (presence of repeat sequences) that would characterize all fragile sites (both common and rare).

Previously, evolutionary studies involving fragile sites have attempted to address two important questions. First, because fragile sites are considered part of the chromosome structure, are the characteristics underlying their susceptibility to breakage conserved during evolution? Also, can fragile sites be considered 'targets' for evolutionary reorganization? In terms of the first question recent studies have shown that some human common fragile sites have been conserved in homologous regions in mouse and some primate species 29 56 57 , suggesting that the characteristics governing a chromatid's susceptibility to breakage are conserved during evolution. The high degree of correspondence between the location of fragile sites and evolutionary breakpoints shown by our study has a bearing on the second question posed above, namely whether fragile sites are 'targets' for evolutionary reorganization. Comparative cytogenetic studies performed in primate families such as Hominidae, Cebidae, and Cercopithecidae 7 16 58 59 60 revealed that a high proportion of chromosomal bands implicated in evolutionary reorganization, centromeric shifts, and delimiting heterochromatic regions also contain fragile sites in the human genome. By increasing the number of species analyzed (mouse, rat, cattle, dog, pig, cat, horse, and chicken), as well as improving the resolution of evolutionary breakpoints using whole-genome comparisons, we have been able to draw more precise conclusions on the distribution of evolutionary breakpoints and their correspondence to human bands that are known to contain fragile sites. Our data show that fragile sites appear to be conserved as 'fragile chromosomal bands', in which evolutionary breakpoints accumulate in much the same way that human fragile sites may be considered to signal regions of chromosomal instability observed in cancer cells 61 .

Given the 'hot spot' theory, one may question whether repetitive elements are driving chromosomal evolution by triggering reorganization in these regions (for instance, see the reports by Armengol 42 and Cáceres 62 and their coworkers) or, alternatively, that the repeats accumulate preferentially in these regions following reorganization. That our study shows that rare fragile sites in particular have a highly significant association (P = 0.01) with both evolutionary breakpoints and tandem repeats has important implications for the role of this particular type of fragile site in chromosomal instability, and hence genome evolution. The molecular characterization of chromosomal regions implicated in evolutionary breakpoints in human, mouse, and primate genomes has similarly shown that large-scale reorganization tends to occur at, or close to, regions rich in segmental duplications and some type of simple tandem repeat (for example, the dinucleotide [TA]n) 41 63 64 65 .

The analysis of the distribution of tandem repeats in human chromosomes and their spatial relationship to evolutionary breakpoints presented here highlights two important points. First, it emphasizes the high concentration of base pair repeats found at the telomeres and the pericentromeric areas (which is in agreement with previous reports on the distribution of duplicated regions; see Murphy and coworkers 5 ), and the distribution of polymorphic minisatellites 49 throughout the human genome. The second, possibly more remarkable finding is the concentration of tandem repeats at evolutionary chromosomal bands. Although this is by no means ubiquitous, the correspondence is typified by human chromosome 3 (Table 2 and Additional data file 1). Bands with the greatest number of tandem repeats in this chromosome (3p25, 3p21.3, 3p12, 3q13.1, 3q21, and 3q29) are also chromosomal regions that have been implicated in evolutionary rearrangements. It is noteworthy that the chromosomal bands 3p25, 3p21, 3p12, and 3q21 have previously been identified as breakpoints in primate evolution 66 , and that the evolutionary breakpoints at 3p25.1, 3p12.3, and 3q21.3 are associated with duplications in hominid evolution 67 68 69 .

In particular, human chromosome 7 (Figure 2a) is interesting both from the evolutionary as well as clinical perspective. Our analysis shows that there are six bands on this chromosome that contain the greatest concentration of tandem repeats in the human genome: 7p22, 7p13, 7p11, 7q11, 7q22, and 7q36. All six bands incorporate fragile sites (FRA7B, FRA7D, FRA7A, FRA7J, FRA7F, and FRA7I) and all but one of them (7p13) correspond to regions where evolutionary breakpoints tend to concentrate, as indicated by comparisons of the human genome with those of mouse, rat, cattle, pig, dog, cat, chicken (present study), and different primate species 8 . Three of these chromosomal bands (7p22, 7q11, and 7q22) appear to be the boundaries for mammalian ancestral chromosomes 7a and 7b (Figure 2a) and have been implicated in almost all mammalian species studied to date by comparative chromosome painting using human painting probes 8 54 . A recent study of the evolutionary history of human chromosome 7 70 demonstrated that this chromosome may be derived from the orangutan homolog by two inversions (one paracentric and another pericentric) that involved three chromosomal breakpoints that map to 7p22.1, 7q11.23, and 7q22.1 in human (one of these, 7q22.1, is common to both rearrangements). All three bands have the greatest number of tandem repeats (present study) and are particularly rich in segmental duplications 40 . Moreover, they are considered 'hot spots' for human diseases such as the Williams-Beuren syndrome 71 72 and leukemias 73 .

Other notable associations between tandem repeats, fragile site location, and evolutionary breakpoints include the greatest concentration of tandem repeats found in the human genome - those in bands 12q13.1 and 12q24. The band 12q13.1 contains one fragile site (FRA12A) and two evolutionary breakpoints, whereas 12q24 contains three fragile sites (FRA12C, FRA12D, and FRA12E) and seven evolutionary breakpoints (Figure 2b). Human chromosome 12 is considered to be the result of the fusion of two ancestral chromosomal segments 12a and 12b (Figure 2b) that are thought to have occurred in the Simiiformes (Catarrhini and Platyrrhini) ancestor. Chromosomal band 12q24 forms the boundary of these segments 8 , once again highlighting a chromosomal region that is characterized both by its fragility and involvement in evolutionary change.

The Ensembl genome browser of Sanger Center and EMBL 74 as well as published data 5 were used as sources for determining homologies between the human genome and those of the mouse, rat, cattle, pig, dog, cat, horse, and chicken. We used the sequence coordinates described by Murphy and coworkers 5 to delimit homologous synteny blocks (HSBs), where the data from cattle, pig, cat, and horse are based on RH maps; the homologous regions between human, rat, and mouse are based on whole-genome assemblies. To determine syntenic regions between the human genome (NCBI Build 35) and that of the dog and chicken, we used the completed human/chicken (WASHUC 1) and human/dog (CanFam 1.0) whole-genome sequence assemblies available from the Ensembl genome browser. In the case of the dog and chicken we analyzed homologous syntenic blocks that varied in size between 0.1 Mb and 84 Mb (Additional data file 4), according to the Ensembl genome browser.

For all species analyzed, we follow Murphy and coworkers 5 in viewing an 'evolutionary breakpoint region' as the interval between two syntenic blocks. As did those authors, we use evolutionary breakpoint regions that are 4 Mb in size or less in order to avoid problems of low comparative coverage. 'Evolutionary breakpoints' are defined by sequence coordinates in any of the seven mammalian species compared with human plus the chicken. They serve to delimit the start and end of each breakpoint region. Likewise, the limits of each chromosomal band in the human karyotype can be defined by sequence coordinates using the Ensembl database 74 . Following this procedure, evolutionary breakpoints of each homologous segment were mapped to the human ideogram at the 550 band resolution, allowing us to identify 'evolutionary chromosomal bands' (E bands), which are defined as any band in the human ideogram that contains at least one evolutionary breakpoint in any of the eight species compared with the human genome (Figure 1). We used the JMP software (version 5.1.2; SAS Institute Inc., Cary, NC, USA) to investigate the distribution of evolutionary breakpoints.

The data reported by Schwartz and coworkers 39 were used as reference for the location, classification, and number of fragile sites described in the human genome. Human fragile sites may be classified into two groups based on frequency of occurrence and mechanisms of expression, and are generally referred to as either common or rare fragile sites 17 . In this investigation we considered a total of 119 fragile sites 39 , of which 88 are defined as common and 31 as rare fragile sites (Additional data file 7). These were mapped to specific chromosomal bands on the human ideogram at the 550 band resolution (Table 1). The evolutionary chromosome breakpoint boundaries, each identified by human reference coordinates (see above), were similarly treated in order to determine whether these fell within a specific chromosomal band region that is known to express fragility. It is important to note that in some cases a chromosomal band described as containing a fragile site in the literature can, at higher resolution (for example, the 550 band ideogram), be shown to comprise several sub-bands. For example, the common fragile site FRAJ is mapped to 7q11, which corresponds to four sub-bands in the 550 band ideogram (7q11.1, 7q11.21, 7q11.22, and 7q11.23).

We defined the chromosomal location of 12 autosomal common fragile sites that have been characterized at the molecular level by the position provided by the Ensembl 74 and NIH databases 75 for the molecular markers and/or the BAC clones described in the original papers (Additional data file 8). These fragile sites were examined to determine whether any evolutionary breakpoint spanned these regions in at least one of the species compared herein.

We analyzed the distribution of tandem repeats in the human genome sequence (NCBI Build 35) using the 'Tandem Repeats Finder' (TRF) algorithm (version 3.21 76 ) in all human chromosomes (HSA) except HSA X and HSA Y. The complete sequences of each chromosome were scanned for tandem repeats using the program TRF with the parameters established by default (+2 -7 -7 0.80 0.10 50 500).

We scrutinized each chromosome's complete sequence using moving non-overlapping windows of 0.250 Mb in order to analyze the density and distribution of tandem repeats in the human genome. Given the high incidence of repeats at the telomeres/subtelomeric and the centromeric/pericentromeric areas 49 (confirmed by our study; Additional data file 9), we excluded a 3 Mb section at each of these localities, which are referred to herein as the T (telomeric) and C (centromeric) regions. A further classification involves chromosomal bands that contain evolutionary breakpoints in at least one of the eight species compared with the human genome (E bands); all remaining bands were designated as B bands (for example, non-evolutionary chromosomal bands). Additionally, the presence/absence of a fragile site (rare or common) was recorded for each chromosomal band based on their published location 39 , as defined in the human ideogram at the 550 band resolution (Additional data file 9). Tukey-Kramer tests were used (JMP package version 5.1.2; SAS Institute Inc.) to evaluate whether tandem repeats concentrate significantly (P ≤ 0.05) in evolutionary chromosomal bands (E bands) and/or fragile sites (FS bands). In both cases, the centromeric and telomeric regions were excluded before statistical analysis because they had much higher repeat values overall.