Human chromosome 17 is implicated in a wide range of human genetic diseases. It is home to genes involved in early-onset breast cancer (BRCA1
), neurofibromatosis (NF1
) and the DNA damage response (TP53
encoding the p53 protein). The complex rearrangement and duplication structure of chromsome 17 predisposes it to non-allelic homologous recombination (NAHR), resulting in DNA rearrangements that cause several well-studied microdeletion disorders. These include herditary neuropathy with pressure palsies (HNPP)7
at 17p12, and Smith–Magenis syndrome (SMS) deletions at 17p11.2 (refs 8
). Microduplication counterparts for both of these conditions are also known, famously in the case of Charcot–Marie–Tooth disease type 1A (CMT1A) at 17p12 (refs 10
), the most common inherited peripheral neuropathy (see ref. 12
for a review), and more recently for SMS13
. Further, the complex architecture in this region, consisting of ~50-kb subunits of direct and inverted repeats able to form cruciform structures, is also responsible for susceptibility to one of the most common somatic rearrangement events characterized, isodicentric 17q, which is associated with several cancers and signifies poor prognosis14
. Finally, all cases of Miller–Dieker syndrome12
and many cases of isolated lissencephaly sequence12
are associated with haploinsufficiency effects of deletions at 17p13.3.
As part of the Human Genome Project2
, we generated a finished sequence of human chromosome 17, which comprises ~2.8% of the euchromatic genome. The finished sequence contains 78,839,971 bases and is interrupted by nine euchromatic gaps and one gap containing the centromere region (). The total size of the euchromatic gaps is estimated at 854 kb (see Methods and Supplementary Table S1). The frequency and size of gaps is slightly higher than the human genome average2
. (It is also significantly higher than for mouse chromosome 11; see below.) This is largely due to segmentally duplicated or unclonable regions on human chromosome 17 (see Supplementary Information). A clone list for the most recent version of the sequence is provided in Supplementary Table S2.
Landscapes of human chromosome 17 and mouse chromosome 11
An overview of chromosome 17 is shown in . The chromosome has features characteristic of gene-rich regions2
, including a high average G+C content (45.5%) and a collection of transposable element fossils (45.3% overall) with a relative excess of short interspersed elements (SINEs, 22.3%) and a deficit of long interspersed elements (LINEs, 14.4%). It also has a large excess of segmental duplications (defined as having >90% identity and being >1 kb in length15
). Such segmental duplications constitute 8.6% (7 Mb) of the finished sequence, in contrast to a median of 3.9% for human chromosomes and an average of 4.3% across autosomal sequences. As with other duplication-rich chromosomes, chromosome 17 has a large proportion of intrachromosomal duplication: 62.3% of the duplicated sequence is strictly intrachromosomal, 17.4% is both intra- and inter-chromosomal, and 20.3% is solely interchromosomal. We elaborate below on the segmental duplication content of the chromosome.
We have produced a manually curated catalogue of protein-coding genes on chromosome 17 (see Methods), annotating 1,266 gene loci and 274 pseudogene loci (Supplementary Table S3). This catalogue includes all of the 1,079 genes on the chromosome in the RefSeq database (see Methods and Supplementary Information). The gene density (16.2 genes per Mb) is substantially higher than the genome average2
and ranks only behind chromosome 19 (26.2 genes per Mb; ref. 3
). There is evidence of extensive alternative splicing, with gene loci having an average of five distinct transcripts, and 76.6% having at least two transcripts, a proportion comparable to recent reports3,16-19
. Of the 274 pseudogenes on the chromosome, 73.0% are processed (intronless) pseudogenes arising from retroposition. In addition to protein-coding genes, we identified 42 transfer RNA genes on the chromosome (Supplementary Table S4).
As part of the Mouse Genome Sequencing Consortium (MGSC), a finished sequence of mouse chromosome 11 has been produced (Supplementary Table S5). The sequence was derived from the C57BL/6J strain and has a finished euchromatic length of 118,816,080 bases, representing ~4.6% of the mouse genome. The finished sequence is interrupted by only two euchromatic gaps, with an estimated total size of 97 kb, and one additional gap containing the acrocentric heterochromatin (, Supplementary Table S1 and Methods). See Supplementary Table S5 for a clone list of the most up-to-date version of mouse chromosome 11.
Mouse chromosome 11 can be divided into two parts on the basis of conserved synteny with human chromosomes (). Although the proximal 59 Mb shows conserved synteny with five different human chromosomes (1, 2, 5, 7 and 22), the distal ~60 Mb has conserved synteny exclusively with human chromosome 17 (see and Supplementary Fig. S1). The distal portion of mouse chromosome 11 is similar to human chromosome 17 in various respects, including having a high G+C content (45.5%), high gene density (18.7 genes per Mb), an excess of SINEs (14.4%), and a deficit of LINEs (8.2%). The ratio of SINEs to LINEs is similar to that in human, although the absolute density is lower, consistent with the fraction of detectable transposon sequence in mouse. In contrast, the proximal portion of mouse chromosome 11 has a lower G+C content (41.9%), low gene density (8.2 genes per Mb), and an opposite ratio of SINEs and LINEs (7.8% versus 19.1%). Although it is similar to human chromosome 17 in most other respects, the distal region of mouse chromosome 11 differs sharply with respect to segmental duplications: such sequences constitute only 1.1% of the distal region. Indeed, mouse chromosome 11 has the lowest rate of segmental duplication (1.4%) among all mouse chromosomes.
The block of undirected synteny (corresponding regions of orthology between the species, in which order may be disrupted by internal rearrangements) between human chromosome 17 and mouse chromosome 11 is the second largest such autosomal block between human and mouse (see Supplementary Information). Within this block, there are 23 segments of directed (collinear) synteny of >100 kb, indicating frequent inversion. To better understand this relationship, we reconstructed the chromosomal structure of the primate–rodent ancestor of human chromosome 17 (Supplementary Fig. S2) on the basis of genome sequences from human, mouse, dog and opossum (see Supplementary Information). The primate–rodent ancestor could be resolved except for one small segment, and also represents the state of the boreoeutherian ancestor except in two regions of conserved directed synteny between human and mouse that differ from dog and opossum (see Supplementary Information and Supplementary Fig. S2).
Comparison of the modern chromosomes with our ancestral reconstruction () shows that the mouse structure is virtually unchanged, whereas significant rearrangements have occurred in human. There are 20 rearrangement breakpoints in human but only three in mouse. (The terminal break and one internal break are used for different events in both species, leading to only 22 observed differences in the map of direct human–mouse conserved synteny; see Supplementary Fig S1.) This is contrary to the general pattern for the human and mouse genomes, which typically shows many more rearrangements along the rodent lineage5,6
(see Supplementary Information). The mouse-specific rearrangements are confined to a single region orthologous to the Smith–Magenis region in human, which shows an even more complex pattern of rearrangements at finer scale ().
Syntenic relationship between mouse, human and the ancestral chromosome
What accounts for the extreme rearrangement of human chromosome 17? The answer seems to be related to the extensive segmental duplication on the human chromosome. The human-specific breaks in conserved synteny are highly correlated with regions of intrachromosomal duplication (). Roughly 74% of the duplicated bases reside in the breakpoint regions between stretches of conserved synteny, which comprise only 7.3% of the bases in the chromosome sequence.
Duplication landscape of chromosome 17 and its association with breaks in conserved synteny
To further explore the nature of these duplicated sequences, we clustered the duplicons into classes on the basis of shared sequence (see Methods). After clustering, most of the duplicated sequences fall into one of three classes (1–3), which account for 51.6%, 19.9% and 3.1%, respectively, of the intrachromosomal duplications ( and ). For each duplication class, we then sought to define a ‘core element', corresponding to the maximum length of sequence occurring in the largest number of duplicons (see Methods). These core elements can serve as useful markers for identifying class members and may provide insight into the origins of the duplications20
Duplication classes on human chromosome 17
The class 1 core element (~11 kb in length) contains a TBC1 domain and is found in 12 of the 19 human-specific breakpoint regions. Several of the class 1 core elements are actively transcribed, including six very recently duplicated genes from the TBC1D3 family (at 31–33 Mb) and the USP6
oncogene (also known as TRE-2
), which is a fusion of a TBC1D3 gene and the USP32
. The TBC1D3 family genes encode proteins with TBC1 domains, and are believed to be GTPase-activating proteins specific to RAB5
, a RAS-related gene involved in vesicle trafficking. The TBC1D3 gene family has no known homologues outside the primates, and we could find no similar sequence in dog or mouse, making it impossible to identify an ancestral copy. However, the element certainly pre-dates the primate clade: human copies show substantial divergence (~34%) (Supplementary Fig. S3) and copies can be identified in ring-tailed lemur by polymerase chain reaction (PCR) (see Methods). The element is also duplicated in macaque, and a phylogeny of the elements in human and macaque (Supplementary Fig. S3) shows that at least some of these duplications pre-date the divergence of Old World monkeys, as some macaque copies have a human copy as their nearest neighbour.
The class 2 core element (~4 kb in length) is found in 9 of the 19 human-specific breakpoint regions (). The ancestral copy is probably the copy immediately upstream of the NSF
gene at 41.95 Mb, as this is the only copy with an orthologous sequence in mouse. The class 2 duplicons occur primarily in three regions of chromosome 17q, at 26–27 Mb, 40–42 Mb and 60–63 Mb. These regions are all sites of extensive, small local inversion events that have occurred since the primate–rodent ancestor. A phylogenetic tree of the core elements (Supplementary Fig. S4) shows two main clusters, one for the copies at 26–27 Mb and another for those at 40–42 and 60–63 Mb. Notably, the reconstruction of the ancestral chromosome order (Supplementary Fig. S2) places these latter regions nearly adjacent to one another, suggesting that the core elements dispersed in a local region that was then disrupted by an inversion with breakpoints at 33 and 57 Mb. This ancestral chromosomal order seems to correspond to the modern karyotype of orangutan22
, which has a paracentric inversion with respect to the human karyotype on 17q, with nearly coincident breakpoints. This would imply that the inversion in the human lineage is a fairly recent event. Notably, two class 2 core elements (the ancestral copy and a copy at 40.97 Mb) flank the boundaries of a 900-kb polymorphic inversion in the human population23
and also the boundaries of a human rearrangement with respect to the ancestral chromosome ( and Supplementary Fig. S2), indicating a reuse of these breakpoints within the human lineage (see Supplementary Information).
The class 3 duplicons all reside in the CMT1A/SMS region (13.8–20.4 Mb), which seems to be a region of remarkable genomic fragility. The class is too small to define a meaningful core element, but examination of the longest copy shows that it is a previously identified low-copy repeat, LCRA110
. This element is the result of a fusion of two distinct 3′-UTR elements from elsewhere in the human genome. Although at least one of the individual sub-elements comprising the fusion is found in mouse and dog, the fusion product is not seen in these species, but only in primates. One of the two sub-elements is actively transcribed (see Supplementary Information).
of the CMT1A/SMS region have identified large, highly similar local repeat blocks. The blocks are themselves fusions of multiple duplicons (including class 1 core elements) that do not occur together elsewhere in the genome. None of these duplicated sequences has a mouse orthologue within the region (), having all come from the q arm of 17, indicating that this region may be a site of multiple inversion. Indeed, cytogentic mapping of primate rearrangements22
shows the CMT1A/SMS region to be the site of an inversion breakpoint in orangutan and of a translocation between chromosomes 17 and 5 in gorilla24
Certain common characteristics emerge from the study of segmental duplication on chromosome 17, as well as from our recent study of segmental duplications on chromosome 15 (ref. 20
). First, the vast majority of duplicons in these classes fall in breaks of conserved synteny. Second, segmentally duplicated sequence seems to spread readily within local regions. Local clustering is evident for all duplicon classes studied on chromosomes 15 and 17. Third, segmentally duplicated sequence then seems to spread over larger distances by chromosomal inversions. In two cases (class 2 on chromosome 17 and the main class on chromosome 15; ref. 20
), such dispersal is confirmed on the basis of both sequence phylogeny and reconstruction of ancestral chromosome structure. With additional comparative data on lower primates, it should be possible to reconstruct the dispersal of most of the segmentally duplicated sequences in the human genome. Finally, the core elements underlying the main classes of segmentally duplicated sequence on chromosomes 15 and 17 all show active transcription from a large number of the duplicons10,25,26
. This observation suggests a link between active transcription, duplication and rearrangement. It is possible that the nature of transcription, the chromatin structure or the transcripts at these loci may render the regions more susceptible to breakage, duplication or non-allelic recombination.
Our comparison of human chromosome 17 and distal mouse chromosome 11 shows that a genomic region can undergo strikingly different rearrangements in different lineages. There has been remarkably little interchromosomal rearrangement in the region orthologous to human chromosome 17 across all mammals, which may perhaps reflect an inherent stability of the region or a selective constraint on contiguity. In contrast, the region shows extensive internal rearrangement in the human lineage (20 breakpoints relative to the ancestral genome) but little rearrangement in mouse (only three rearrangements relative to the ancestor). The occurrence of the human breakpoints near segmental duplications supports a mechanistic link between these two features that may explain the unusual history of this genomic region in the primate lineage. Together with our recent analysis of chromosome 15, this analysis sheds light on the forces that shape the large-scale structure of genomes and, in the process, create both evolutionary variation and the chromosomal fragility underlying human disease.