The generation of a 7.5× dog genome assembly provides exciting new opportunities to interpret tumor-associated chromosome aberrations at the biological level. We present a genomic microarray for array comparative genomic hybridization (aCGH) analysis in the dog, comprising 275 bacterial artificial chromosome (BAC) clones spaced at intervals of approximately 10 Mb. Each clone has been positioned accurately within the genome assembly and assigned to a unique chromosome location by fluorescence in situ hybridization (FISH) analysis, both individually and as chromosome-specific BAC pools. The microarray also contains clones representing the dog orthologues of 31 genes implicated in human cancers. FISH analysis of the 10-Mb BAC clone set indicated excellent coverage of each dog chromosome by the genome assembly. The order of clones was consistent with the assembly, but the cytogenetic intervals between clones were variable. We demonstrate the application of the BAC array for aCGH analysis to identify both whole and partial chromosome imbalances using a canine histiocytic sarcoma case. Using BAC clones selected from the array as probes, multicolor FISH analysis was used to further characterize these imbalances, revealing numerous structural chromosome rearrangements. We outline the value of a combined aCGH/FISH approach, together with a well-annotated dog genome assembly, in canine and comparative cancer studies.
Canine and comparative disease studies will benefit greatly from the recent publication of a high-quality, annotated 7.5× genome assembly for the domestic dog (Lindblad-Toh et al. 2005). This key resource has served to increase further the value of the dog as a model for genetic traits, as well as advancing the study of canine diseases in their own right. The restricted gene pool associated with populations of pure-bred dogs has great potential to facilitate identification of novel disease genes by reducing the “background noise” caused by the high level of heterogeneity within typical human populations. Comparative studies of human and dog counterparts of the same disease therefore have immense potential for improving the health and welfare of both species through the development of more sophisticated approaches for diagnosis, prognosis, and therapy.
Particular interest in canine and comparative cancer studies has stemmed from recognition that many human and dog tumors share extensive pathophysiological features. Human tumors are widely known to display recurrent chromosome aberrations, many of which are hallmarks of particular tumor subtypes. A proportion of aberrations have been associated with clinical behavior, leading to the identification of key genes involved in tumor initiation and progression (see e.g., Mitelman et al. 2005). We have demonstrated previously that recurrent chromosome abnormalities are also present in a range of canine tumors and have identified genomic features that appear to be associated with specific tumor subtypes (see e.g., Thomas, Smith, et al. 2003; Thomas et al. 2005; Modiano et al. 2005, 2006). A comparative approach, involving detailed investigation of the nature and distribution of recurrent chromosome abnormalities among human and canine tumors, may provide the additional clues needed to further our understanding of the complex genetic processes involved in tumorigenesis. In order to gain full benefit of the dog genome assembly for cytogenetic studies of canine cancers, it is essential to be able to translate chromosome abnormalities directly into DNA sequence information so that we may elucidate the mechanisms involved in tumor development and develop improved therapies. We present preliminary integration of cytogenetic and genome sequence data for the dog through the molecular cytogenetic anchoring of a low-resolution framework of 275 bacterial artificial chromosome (BAC) clones, spaced uniformly at ∼10-Mb intervals throughout the canine genome sequence (version 2.0). We have used this panel of BAC clones to generate 2 important resources for cytogenetic evaluation of canine cancer; 1) 40 chromosome-specific multicolor BAC sets, each with ∼10-Mb resolution for direct molecular cytogenetic evaluation of chromosome aberrations in canine tumor cells and 2) a genome-wide microarray containing all 40 chromosome-specific clone sets, which may be used for first pass array comparative genomic hybridization (aCGH) analysis of chromosome aberrations in canine tumors. We demonstrate the application of this new integrated resource by presenting aCGH data from a canine histiocytic tumor followed by aCGH-directed multicolor single-locus probe (SLP) fluorescence in situ hybridization (FISH) analysis of tumor chromosomes. These data illustrate how the nature and distribution of chromosome copy number aberrations (CNAs) within a tumor cell population may be investigated using this resource.
Materials and Methods
Unless otherwise stated, all clones described in the present study were selected from the CHORI-82 dog BAC library (http://bacpac.chori.org/library.php?id=253; Children's Hospital Oakland Research Institute, Oakland, CA), derived from a female Boxer, and from which the 7.5× dog genome assembly was constructed (Lindblad-Toh et al. 2005). For each of the 39 different chromosomes in the female dog, the BAC paired-end read placements in the genome assembly (version 2.0, May 2005; Lindblad-Toh et al. 2005) were used to identify clones located close to each centromere. Due to the presence of highly repetitive sequences at the centromere that are intractable to genome sequence assembly, the most proximal clone available on each of the dog autosomes was typically located approximately 3 Mb distal to the centromere (mean 3.17 Mb, range 3.01–3.33 Mb). From this point, BAC clones were selected from the assembly that mapped at approximately 10-Mb intervals along the length of the genome sequence of each dog chromosome. To ensure that both ends of the available genome assembly sequence data for each chromosome were represented in our cytogenetic panel, the most distal (telomeric) clone available was also selected.
BAC DNA was isolated from 2.5 ml bacterial cultures of each clone using a Qiagen REAL Prep 96 kit (Qiagen, Valencia, CA). Cytogenetic mapping was performed according to our routine multicolor FISH protocols (Breen et al. 2004). Briefly, 500 ng of DNA from each BAC clone were labeled in a nick translation reaction to incorporate 1 of 5 spectrally resolvable fluorescent nucleotides, Spectrum Red/Orange/Green- deoxyuridine triphosphate (dUTP) (Vysis, Downers Grove, IL), diethylaminomethylcoumarin-5-dUTP or Cyanine5-dUTP (Perkin Elmer Life Sciences, Boston, MA). Groups of up to 5 BAC clones assigned at consecutive 10-Mb intervals on the same chromosome were differentially labeled and hybridized simultaneously as described elsewhere (Breen et al. 2004), beginning with the 5 most proximal clones. For chromosomes represented by more than 5 clones, the most distal clone in the initial reaction was then included as the most proximal clone in the subsequent reaction to provide continuity of probe ordering along the length of the chromosome. Cytogenetic assignments for each clone were made according to the 4′,6-diamidino-2-phenylindole (DAPI)-banded nomenclature of Breen et al. (1999). Chromosome-specific tiling panels were then designed by pooling all BAC probes for each chromosome and cohybridizing them in a single multiprobe 5-color FISH reaction to tile the chromosome.
Construction of a 10-Mb Resolution Genomic Microarray for CGH Analysis
Clones represented in the 10-Mb BAC set were developed for use as targets on a BAC microarray as described previously (Thomas et al. 2003, 2005). Briefly, BAC DNA was amplified by degenerate oligonucleotide-primed PCR (DOP-PCR) using 3 different degenerate primers in separate reactions, the resulting products were pooled and amplified using a 5′ amino–linked PCR primer, and then arrayed in duplicate onto amine-binding slides (CodeLink slides, GE Healthcare, Piscataway, NJ) using a MicroGrid II arrayer (Genomic Solutions, Ann Arbor, MI) as described previously (Fiegler et al. 2003; Thomas, Fiegler, et al. 2003; Thomas et al. 2005). Also included on the array were BAC clones representing dog orthologues of 31 cancer-related genes, of which 25 have been reported previously (Thomas, Bridge, et al. 2003) and the remaining 6 (ABL1, BCR, BCL2, NF2, p16/INK4A [also known as CDKN2A], and PTEN) represent new cytogenetic assignments. With 4 exceptions (ABL1, BCR, NF2, and p16/INK4A from the CHORI-82 library), clones representing cancer genes were derived from the RPCI-81 male dog BAC library (Li et al. 1999) and were assigned a location within the dog genome assembly (version 2.0) by aligning BAC end sequence data to the genome using BLAT (Kent 2002). Due to the absence of CFA Y clones from the female CHORI-82 library, 5 clones previously identified from the RPCI-81 library, and assigned to CFA Y by FISH analysis, were also included in the data set (Thomas et al. 2005). Details of all clones represented on the array are listed as online supplementary data at http://www.cvm.ncsu.edu/mbs/breen_matthew.htm.
aCGH analysis was performed as described previously (Thomas et al. 2005). First, a “self–self” hybridization was performed by differentially labeling and cohybridizing a constitutional DNA sample from a clinically healthy dog onto the 10-Mb array. Subsequently, as an example of the application of this microarray to the study of canine cancer, we performed aCGH analysis of a histiocytic tumor from a female flat-coated retriever (FCR). Tumor DNA was labeled with cyanine3-dCTP (Perkin Elmer) using a BioPrime Array CGH Labeling System (Invitrogen). A sex-mismatched reference sample representing equimolar pools of DNA isolated from the peripheral blood of 10 clinically healthy male FCRs was similarly labeled with cyanine5-dCTP (Perkin Elmer), and aCGH analysis was performed as described previously (Thomas et al. 2005). A dye-swap analysis was also carried out, where the test and reference samples were labeled with cyanine5-dCTP and cyanine3-dCTP, respectively. Arrays were scanned using a ScanArray 4000 at 10 μm resolution and quantified using ScanArray Express version 3.0 (Perkin Elmer). Each spot position was automatically located, and manual adjustments were made as necessary. Spots with confidence levels <0.1, and/or with >0.3 standard deviations between replicates, were excluded from further analysis. Fluorescence intensities were calculated for each spot after local background subtraction, block normalized to a median 1:1 ratio on the autosomal clones, and ratios of normalized values established. The mean fluorescence ratio of each replicate was then converted to a log2 ratio in order to weight chromosomal gains and losses equally and plotted graphically in genomic order. Regions of genomic imbalance were identified using the aCGH-smooth algorithm (Jong et al. 2004), with threshold limits for gain and loss set at log2 values equivalent to 1.15:1 (gain) and 0.85:1 (loss) of tumor versus reference ratios, respectively.
Evaluation of Genomic Imbalances by SLP FISH Analysis
Chromosome preparations were generated by direct harvest of tumor cells using conventional techniques of colcemid arrest, hypotonic treatment, and methanol/glacial acetic acid fixation. For consistency, these cells were isolated from the same biopsy specimen from which the tumor DNA was isolated. FISH analysis of chromosome and interphase preparations from the tumor sample was carried out using SLPs representing selected clones from the 10-Mb panel, chosen from regions showing a range of normal and aberrant copy number ratios. Images were acquired from a minimum of 30 representative cells in each instance. The copy number status of each probe was scored by 2 independent investigators with no prior knowledge of the corresponding aCGH data.
Chromosome Assignment of the 10-Mb BAC Set
A total of 275 clones, each positioned within the 7.5× dog genome assembly (version 2.0, Lindblad-Toh et al. 2005), were selected from the CHORI-82 library to represent a 10-Mb BAC panel spanning all dog autosomes and CFA X. The number of clones per chromosome ranged from 4 (CFA 35, 36, 37, and 38) to 14 (CFA X). An example of the FISH data for the 10-Mb clone set for CFA 1 is shown in Figure 1, in which all 13 CFA 1 BAC clones have been simultaneously hybridized as a 5-color chromosome-specific tiling set. Based on current genome assembly data, the mean spacing of markers on each chromosome ranged from 7.75 Mb (CFA 38) to 10.63 Mb (CFA 19), with an overall, genome-wide mean of 9.71 Mb. The relative order of clones along each chromosome was consistent with that predicted from the genome assembly, and with few exceptions, the cytogenetic assignment of clones to the 450-band-stage DAPI-banded ideogram of the dog (Breen et al. 1999) demonstrated the expected uniform distribution throughout the karyotype. We observed a subset of physical intervals whose size appeared to be inconsistent with that predicted from the sequence assembly. The most extensive example of such a discrepancy is shown in Figure 2. According to the dog genome assembly, BAC clones 313-K05 and 330-A23 and 330-A23 and 188-A17 (located at the centromeric end of CFA 9) are each separated by intervals of ∼10 Mb (10.31 and 9.84 Mb, respectively). FISH mapping of the CFA 9 BAC set showed that the cytogenetic interval between the 2 most proximal of these clones (313-K05 and 330-A23) was significantly greater than the assembly data would imply. Additional discrepant intervals were evident on CFA 3, 4, 8, 10, 11, 26, 27, 28, 30, 32, 35, 36, and Xp (Figure 3). Allowing for the highly repetitive sequences at chromosome termini, the mapping of clones to the proximal and distal ends of the genome assembly for each chromosome showed excellent cytogenetic coverage of all autosomes, although CFA 12 was remarkable due to underrepresentation of the centromeric band, whereas CFA X was overrepresented near the centromere.
The overall resolution of the clone set was increased by the inclusion of BAC clones representing 31 cancer-associated genes, which were distributed across 20 of the 38 dog autosomes. When combined with the 10-Mb BAC set, the mean spacing of markers throughout CFA 1–CFA X is reduced to 8.77 Mb (ranging from 5.96 Mb on CFA 26 to 10.63 Mb on CFA 19). The 5 clones on CFA Y cannot be integrated into the female genome sequence assembly but are included as part of the genome-wide BAC panel. The total number of clones represented on the array described in the present study is thus 311, each of which has been assigned to a unique chromosomal location by FISH analysis (Figure 3). Six canine cancer-associated genes were newly assigned by FISH analysis in the present study: BCL2 (CFA 1q13.1prox), ABL1 (CFA 9q25dist-q26.1prox), p16/INK4A (CFA 11q15dist-q16prox), NF2 (CFA 26q22), BCR (CFA 26q24), and PTEN (CFA 26q25).
Detection and Investigation of Recurrent Chromosome Aberrations in a Canine Tumor
In contrast to the results of the self–self hybridization (Figure 4A), aCGH analysis of DNA isolated from the histiocytic tumor demonstrated the presence of a range of whole and partial chromosome CNAs. Most striking was amplification of a single clone (126-D04) on CFA 20 (Figures 4B and C). aCGH data indicated whole chromosome copy number increases for CFA 3, CFA 13 (including the regions harboring the KIT and MYC oncogenes), and CFA 37 (Figure 4B). Partial chromosome loss was observed for the distal regions of CFA 16 and 23 and also the telomeric end of CFA 26 including the PTEN locus (Figure 4C). Numerous less-extensive genomic imbalances were observed throughout the genome.
Chromosome preparations derived from tumor cells (Figures 5A and B) showed obvious deviations from the normal 2n = 78 karyotype, including numerous abnormal biarmed structures. SLP analysis was conducted using BAC clones from CFA 20 and 26, representing chromosomes for which the aCGH data indicated the presence of regions of normal, increased, and decreased copy number. SLP analysis showed the expected concordance with aCGH data. Further details are provided in the legend accompanying Figure 5.
We present a 10-Mb resolution panel of cytogenetically assigned BAC clones for use in cytogenetic studies in the domestic dog as a resource that will complement the increasing range of resources available for canine genome analysis. The development of this resource presented an opportunity to compare the cytogenetic distribution of an anchored framework of BAC clones positioned uniformly at ∼10-Mb intervals throughout the version 2.0 dog genome assembly. Overall, the results indicate that the genome sequence represents excellent cytogenetic coverage (centromere-to-telomere) of each dog autosome and the X chromosome. The uniform cytogenetic distribution of most of the chromosome-specific BAC sets also indicates close correlation between cytogenetic and assembly-derived distances. The most significant exception was the centromeric region of CFA 9, which showed uneven clone spacing and which is known to contain highly repetitive sequences (Thomas et al. 2001; Thomas, Smith, et al. 2003). The challenges associated with constructing contiguous sequence reads at sites of highly repetitive DNA suggest that CFA 9q11.2-q13 may be underrepresented in the dog genome assembly, leading to an underestimation of the physical extent of this region. It is possible that the other deviations from the expected uniform intervals between clones may be attributed to the nucleotide composition and secondary structure of the DNA sequence at those locations. Similar discrepancies have been observed between human cytogenetic maps, chromosome banding patterns, and genome sequence assemblies (for a discussion, see Furey and Haussler 2003).
Although providing an insight into the physical coverage of the assembly, this cytogenetic resource also advances our ability to characterize chromosome defects and their resulting effects on genome imbalance and structural reorganization through the development of a genomic microarray for CGH analysis. Although cancer directed, the array may be utilized in the evaluation of any condition in which genomic imbalance occurs. This resource represents a key addition to the dog genome analysis “toolbox” because with each clone integrated into the dog genome assembly, for the first time we now have the means to translate chromosome aberrations directly into DNA sequence. This is a major advantage to our search for candidate genes involved in genetic diseases and heritable traits in both veterinary and comparative research. We have demonstrated the application of the resource to the study of a canine histiocytic cancer, presenting with a complex karyotype that is typical of this aggressive form of malignancy. Characterization of highly aberrant genomes using classical cytogenetics alone is highly labor intensive, particularly in the case of subtle deviations from the normal karyotype. SLP analysis enables a more rapid and accurate assessment of both structural and numerical abnormalities, but selection of the appropriate probes requires prior knowledge of these defects. aCGH analysis largely overcomes these limitations because it permits a genome-wide assessment of chromosome imbalances that allows targeted selection of SLPs for subsequent assessment of aberrations in individual cells. aCGH analysis with our 10-Mb resolution BAC array indicated the presence of a range of CNAs, both increases and decreases, throughout the genome. Subsequent FISH analysis using BAC probes selected from the arrayed 10-Mb set correlated with the CNA data indicated by aCGH analysis. Furthermore, this combined approach also revealed the chaotic structural organization of the tumor genome, resulting from a range of chromosome breakages and fusions. The most remarkable CNA identified by aCGH was amplification of the region represented by clone 126-D04 (CFA20; 23.12 Mb), consistent with SLP analysis showing all that 30 cells scored contained 6 copies of this locus, distributed across 4 different chromosome structures (Figure 5C). The power of the resource described in the present study, however, lies in the ability to readily translate such findings into information with potential biological significance. For example, using the Dog Genome Browser Gateway to interrogate the sequence assembly for this region of CFA 20, (http://genome.ucsc.edu/cgi-bin/hgGateway), we identified a gene, FOXP1, that lies within 1 Mb of the region represented by clone 126-D04. Structural alterations of the FOXP1 transcription factor gene have been described in several human cancers, most notably those of lymphoid origin (Wlodarska et al. 2005), and have been correlated with both up- and downregulation of the gene product. Increased expression of FOXP1 has been associated with poor survival in human diffuse large B-cell lymphoma (Barrans et al. 2004) but with an improved prognosis in breast carcinomas (Fox et al. 2004). This suggests that FOXP1 may also act as a useful prognostic aid in canine cancers and is currently undergoing further investigation. The assembly-integrated microarray described here, in association with the genome-wide BAC SLP panel, represents one approach by which the role of both existing and novel candidate genes may be investigated.
The cytogenetic mapping of 6 new dog cancer–associated genes in the present study, making a total of 31 thus far assigned via our studies, further increases the significance of the arrayed clone set for cancer studies. The inclusion of these markers into the integrated clone set revealed that a subset of genomic aberrations identified in the histiocytic tumor involved known cancer genes, particularly gains of the MYC and KIT oncogenes that we have described previously as features of several dog cancers, as well as deletions of the tumor-suppressor PTEN (e.g., Dickerson et al. 2005; Thomas, Fiegler, et al. 2003; Thomas et al. 2005). The addition of clones representing BCL2, BCR, and ABL1 in this genome-wide set has particular relevance for the study of hematological malignancies in which these genes are well known to be involved in cell cycle disregulation via chromosome translocation events (see Mitelman et al. 2005). With the availability of a high-quality dog genome assembly, the prior challenges involved in identifying dog orthologs of key human genes are now largely negated, and this is likely to have a tremendous impact on advancing canine and comparative disease studies.
Although the inclusion of known cancer-associated genes in the clone panel has obvious implications for studying cancer genome aberrations, perhaps more importantly, the application of this cytogenetic resource to large cohorts of tumor cases will highlight regions of the genome showing chromosomal imbalance that may harbor as-yet unidentified genes involved in the etiology of the malignancy or that may represent targets for novel directed therapies. Through the application and further refinement of the resource described herein, coupled with continued multidisciplinary and comparative studies of canine and human tumors, the scene is set for exciting advancements in our understanding of tumorigenesis in the dog.
Supplementary material can be found at http://www.cvm.ncsu.edu/mbs/breen_matthew.htm.
National Institutes of Health (R21 NS051190-01) and the American Kennel Club Canine Health Foundation (CHF-2214 and CHF-403 to M.B.), Wellcome Trust (to C.F.L. and P.E.), University of North Carolina Office of the President Genomics Fellowship (to A.Y.), and NSF-Integrative Graduate Education and Research Traineeship (E.S.).
We thank Anne Evans for excellent technical assistance.