Abstract

DNA sequence and annotation of the entire human chromosome 7, encompassing nearly 158 million nucleotides of DNA and 1917 gene structures, are presented. To generate a higher order description, additional structural features such as imprinted genes, fragile sites, and segmental duplications were integrated at the level of the DNA sequence with medical genetic data, including 440 chromosome rearrangement breakpoints associated with disease. This approach enabled the discovery of candidate genes for developmental diseases including autism.

Chromosome 7 functional and structural features

Through comparison of the Celera mouse genome sequence to our assembly, 21,859 syntenic anchor points (14) from six murine chromosomes (5, 6, 9, 11, 12, and 13) were identified (Fig. 2;table S5, a and b). The syntenic anchor points were grouped into 36 blocks, 14 of which had not been reported before (15,16).

To generate the most complete description of genes on chromosome 7, we used computer-based annotation in conjunction with extensive laboratory experimentation (7). A team of reviewers scrutinized data and, by comparison to the reference DNA sequence, defined 1917 gene structures (Table 1). Their distribution along chromosome 7 is shown inFig. 3. The description of 297 of these gene units was either exclusive to our dataset, or present in a more complete or different form, as seen from comparison with other databases.

The gene structures were grouped into eight categories: (i) 863 known genes or human full-length cDNA sequences present in LocusLink or HUGO databases; (ii) 71 novel genes, full-length cDNA, or expressed sequence tag (EST) clusters that contain an open reading frame (ORF) (>100 amino acids) without a formal name; (iii) 40 partial genes, human cDNA, or EST clusters with an incomplete ORF (missing the start or stop codon); (iv) 481 predicted genes or predicted gene models, for which at least one exon matches supportive evidence (EST, protein homology, or mouse sequence) of nonoverlapping NCBI, RefSeq, or Ensembl entries; (v) 213 putative and noncoding RNA genes, human cDNA, or EST clusters that do not contain an apparent ORF (51 that have homology in mouse); (vi) 81 gene segments from the two T cell receptor (TCR) loci; (vii) 24TCR pseudogene segments; and (viii) 144 pseudogenes. Overall, our data suggest that 1455 potential protein coding genes (known, novel, partial, predicted) and 213 putative and noncoding RNA genes reside on chromosome 7. Extrapolating these chromosome 7 numbers to the human genome, one would predict that there are about 29,000 protein coding genes and 3700 putative and noncoding RNA genes, consistent with some other estimates (5,6,17). Of the known genes, 474 of 863 (55%) were found to have one or more alternatively spliced forms, comparable to those observed on chromosomes 14 (54%) and 22 (59%) (18,19).

The average gene size on chromosome 7 was 69.9 kb, exceeding what was reported previously (5,6). There were 18 genes greater than 500 kb in size (Fig. 3); the largest,CNTNAP2, spanned 2300 kb. The q22 Giemsa light band had the highest gene density. At one site (coordinates 98.2 to 99.2 Mb), 56 genes were found; that exceeds the mean of 10.7 genes/Mb for the rest of chromosome 7. If the 1749 annotated genes (excluding pseudogenes) are considered, they cover a total of 72.9 Mb of sequence (intragenic regions), which suggests at least 46.5% of chromosome 7 is transcribed (the average intergenic distance was 42.4 kb). A total of 1335 CpG islands were identified on chromosome 7, of which 63% (541 of 863) resided in the 5′ end of a known gene.

Overlapping genes (total 100) were identified through comparison of the sequence coordinates for each known, novel, or partial gene. The pairs were then categorized on the basis of the type of sequence overlap and transcriptional orientation (7). After excluding splice variants, 38 sense-antisense gene pairs (with direct sequence overlap) were identified, 8 and 18 of which were in head-to-head or tail-to-tail orientation, respectively (table S6). The median size of sequence overlap between these transcripts was 238 bp, and for 23 out of 38 (61%), this occurred in the coding sequence. There were also 18 sense-sense (same strand) and 43 sense-antisense overlapping gene pairs that did not share sequence overlap, but occupied the same genomic domain. We found no bias in the number or position of CpG islands near overlapping transcripts. We did, however, observe that a disproportionate number of known imprinted genes, 5 of 8 on chromosome 7 (Fig. 2), had an overlapping transcript (table S6) relative to the total number of transcripts on chromosome 7.

We discovered 20 euchromatic regions each greater than 500 kb in size where no known, novel, or partial gene was found (named putative gene deserts;Fig. 3,table S7). These intervals, which were mostly (16 out of 20) located within or at the boundary of Giemsa dark bands, covered 20.5 Mb (13%) of chromosome 7; the largest was 1850 kb. They contained a low number of CpGs (0.8/Mb vs. 9.8/Mb control) and short interspersed nuclear elements (SINEs) (7.5 versus 16%), a high density of long interspersed nuclear elements (LINEs) (28.1 versus 18.8%), and a decreased mouse syntenic anchor point density (4.2 versus 7.2%). Gene-poor regions described on chromosomes 14, 20, and 21 (18,20,21) also exhibited similar characteristics (in all cases, each category yielded statistically significant results when compared with controls,P < 0.0001). Moreover, our analysis of the equivalent regions in mouse also did not yield any new genes, which suggests that these 20 deserts occurred before the divergence of the human and mouse genomes (table S7). As in humans, the mouse region contained low SINE (3.8 versus 13.6% average) and high LINE (30.2 versus 21%) content. However, the observation that 14 of 20 of these deserts were larger (average increase in size was 19%) in mouse was intriguing given the estimates of a 15% compression in overall size of the murine genome compared with the human (14,16).

As part of this project, we had shown previously that segmental duplications (duplicons) on chromosome 7 can be targets for nonallelic homologous recombination leading to genomic deletions or inversions in Williams-Beuren syndrome (WBS) (10), or gene conversion events in Shwachman-Diamond syndrome (22). Moreover, our analysis of the entire human genome (using NCBI Build 31) revealed that chromosome 7 contained the largest amount of intrachromosomal duplication (23), which suggests that there could be other disease associations. To complete a more refined analysis of chromosome 7, we compared the CRA_TCAGchr7.v1 assembly with itself to search for all recent (>90% sequence identity) and large (>5 kb) intrachromosomal duplications. Overall, 146 distinct segments were identified, which composed 5.3% (8.3 out of 157.9 Mb) of the chromosome (Fig. 4).

Large duplicons (>100 kb) were identified at 7p22, 7p14-p15, the pericentromeric region, 7q11.21, 7q11.23, 7q22, and 7q36. We identified segmental duplications on chromosome 7 contained in 37 bacterial artificial clones (BACs) (confirmed by FISH) not described before (24). Notwithstanding, sequence analysis and metaphase FISH alone would not allow detection of all segmental duplications, and our finding that duplicons were present at 4 of the 7 remaining physical gaps on chromosome 7 suggests additional complexities (Fig. 1). Near-identical duplicons situated directly adjacent to each other on the chromosome could also be missed. Using high-resolution FISH analysis, we discovered one such segmental duplication about 1 Mb in size just telomeric to the WBS region (spanningD7S2470 toD7S2545) (7). Before we can add this to the sequence assembly, the exact boundaries of duplication will need to be determined. Additional WBS-like duplicons were observed at 7q22.1, where 29 rearrangement breakpoints were mapped; 24 were involved in malignancy (Fig. 5). Finally, our observation that 12 of 35 (34.2%) of mouse synteny breaks coincide with a recent segmental duplication on human chromosome 7 (P < 0.0001) (7) supports the idea synteny breaks (genomic rearrangements) do not always occur as random events (table S8).

Medical annotation of the chromosome 7 DNA sequence

To facilitate positional cloning studies and genotype-phenotype correlation, we have incorporated all medically relevant data into the DNA sequence map (supporting online text). From this ongoing initiative, we positioned 70 additional microsatellite markers, collated 1440 clinical karyotypes, and gathered numerous structural data, based on our FISH resource, that have been distributed worldwide. We have also cloned theFRA7G,FRA7H, andFRA7E fragile sites and have identified imprinted and differentially expressed genes (7). We studied rearrangement breakpoints from patients who have chromosome 7 anomalies and could place 440 on the sequence map (Fig. 5) (7). For more than 100 of these, molecular data were not available previously. Examples of new breakpoints mapped within a genomic region that contains a disease locus (gene not yet identified) associated with the patient’s phenotype (e.g., acute myeloid leukemia at 7q22, cavernous malformation syndrome at 7p15, splenic lymphoma at 7q33) are summarized intable S9. Therefore, studies of the sequence at the rearrangement breakpoint(s) could provide insight into the regulation and function of genes, as is described below for three different developmental diseases.

Split-hand split-foot syndrome [Online Mendelian Inheritance of Man (OMIM) 183600], also known as ectrodactyly or lobster claw deformity, is a human developmental condition that is genetically heterogeneous and demonstrates variable expression of phenotype. These genetic characteristics are problematic in the clinical setting, making it impossible to predict carrier status and severity of the disease. Examples of such pedigrees (the first reported in 1908) were even used to argue against the applicability of Mendelian genetics to humans. Our analysis of chromosomal deletions in patients with the syndrome led to the mapping of a disease locus within a 1.2-Mb critical region at 7q21.3 (namedSHFM1) (Fig. 6), but in a decade-long search, no disease gene could be identified. We positioned a balanced translocation or inversion breakpoint from 12 unrelated ectrodactyly patients and found them to be scattered throughout the critical region. The breakpoints did not collectively interrupt any single gene, which suggests a “position effect” mutation might be involved. Reports that double-null knockouts of two murineDistalless homeobox genes,Dlx5 andDlx6, exhibit ectrodactyly (25,26) strongly implicate the human orthologs (which map within theSHFM1 region). In the simplest explanation, the chromosomal breakpoints located up to 1 Mb centromeric of humanDLX5 andDLX6 separate critical regulatory elements from the genes and lead to their dysregulation during development. From our characterization of other patients (represented in the Genome Browser), position-effect mutations are known for other developmental genes on chromosome 7:GLI3,TWIST,SHH, andCDK6 in Greig’s cephalopolysyndactyly (up to 15 kb-3′), Saethre-Choetzen (up to 100 kb-5′), holoprosencephaly (up to 250 kb-5′) and triphalangeal thumb (up to 1000 kb-5′), and splenic marginal zone lymphoma (up to 66 kb-5′), respectively. Further study of the breakpoints including comparative DNA analysis will guide experiments to identify candidate regulatory sequences for testing in functional assays.

We followed a line of investigation similar to that used to study chromosomal rearrangements in autism patients to assess candidate genes for the susceptibility locus (AUTS1) mapped to 7q (27). Using the sequence as a guide, we fine-mapped the chromosome 7 derivative-translocation breakpoints from three new autism patients within 7q22-q31 (the region demonstrating maximum linkage). The breakpoints in cases 16724, 18667, and 11550 (table S9) were positioned within BACs at 7q22.1, 7q31.2, and 7q31.3, respectively. The most interesting finding was that the breakpoint in case 11550 overlapped with a rearrangement from an unrelated patient (10893502_2), diagnosed with speech and language disorder (a common component of the autism phenotype). The last breakpoint was anchored to the sequence by using data from the literature (28). Both breakpoints disrupt the same, apparently noncoding, RNA transcript (TCAG_4133353; GenBankCB338058), which is composed of at least 5 exons spanning 288 kb and is not found in the mouse genome.

For autism case 18667, the translocation breakpoint mapped near to theFOXP2 gene, which was shown previously to cause a form of speech and language disorder (29). The child inherited the translocation from the mother, who had speech delay, which suggests thatFOXP2 might also be involved in autism. Seven isoforms ofFOXP2 (spanning 545 kb) were characterized, but all mapped at least 680 kb 3′ of the translocation breakpoint (in a gene-poor region), once again raising the possibility of a position-effect mutation. Finally, in autism case 16724, the breakpoint was nearest to the neuronal pentraxin 2 (NPTX2) gene, which is thought to be involved in excitatory synaptogenesis and could, therefore, be considered a functional candidate for autism. The TCAG_4133353,FOXP2, andNPTX2 are now being examined for mutations in autism families.

In the third example, for two WBS patients (17430 and 16724) who do not carry the common 1.6-Mb hemizygous microdeletion found in 95% of affected individuals, we have discovered a new 500-kb inversion variant associated with the disease (supporting online text). Unlike the majority of other WBS deletions or inversions (10), the breakpoints in these individuals do not appear to be associated with segmental duplication-mediated events because they map to unique sequences. TheCYLN2 andGTF2IRD1 genes and theSCYA24 andSCYA26 genes are closest to the centromeric and telomeric inversion breakpoints, respectively.

Future studies of human chromosome 7 and community-based annotation

Our goal of establishing a complete reference sequence benefited from data derived from both whole-genome shotgun and clone-based projects. Although some experimentation will be necessary to resolve minor discrepancies in the assembly, with the current framework in hand, the focus of work can be turned to confirming representation, testing for polymorphism, finalizing the gene map, and applying the information for disease study. For the last, we will continue to incorporate as much biomedical information as possible into the DNA sequence map. As we demonstrated, the approach of studying chromosomal rearrangements en masse enabled rapid identification of candidate genes for monogenic and complex diseases and facilitated many functional and structural studies of chromosome 7.

Throughout our study, we found differences and inconsistencies between databases, arguing strongly for the need of additional community involvement in establishing and annotating the consensus sequence of the human genome. We have established a user-friendly database and organized our results into standardized files, in the spirit that this compilation of chromosome 7 information will not only be used as a primary source, but also be incorporated into other projects.

Supplementary Material

Acknowledgments

We thank The Centre for Applied Genomics at The Hospital for Sick Children (HSC) and Celera Genomics, as well as clinical collaborators and families. Supported by Genome Canada, the Canadian Institutes of Health Research, the Canadian Genetic Diseases Network, the Howard Hughes Medical Institute International Scholar Program (to S.W.S.), and the HSC Foundation. We also thank groups worldwide for contributing genomic information to databases.

Footnotes

References and Notes

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.