The unique natural history of the domestic cat, coupled with its intensive veterinary surveillance and growing popularity, offers unrivaled opportunities to model key biological processes, particularly those that are seldom encountered outside human and feline medicine. A vast spectrum of naturally occurring genetic disorders and other heritable pathologies has been described in the domestic cat, of which more than 160 are recognized as potential models of a human counterpart. As many as one third of these are considered rare in other common model systems such as the domestic dog and mouse [1
]. Despite these many attributes, the cat has yet to occupy a prominent role in comparative and translational biomedical studies, particularly for non-heritable disorders, having been constrained for some years by a significant deficit in the essential genomic resources. While a comprehensive 7x-coverage reference genome sequence assembly for the dog was released in 2005 [3
], by 2007 the feline toolbox contained only a draft ~1.9X shotgun assembly, capturing just ~65% of the euchromatic sequence [4
]. Despite its inherently fragmented nature, this resource offered a preliminary template for comparative biomedical studies and for evaluation of aberrant feline genomes, such as those associated with the development and progression of a malignant phenotype.
Numerous human cancer studies have identified recurrent genomic alterations that correlate with discrete diagnostic subtypes, therapeutic responses and/or disease outcomes, offering molecular means for clinically predictive subclassification. The same principles are also well established for the domestic dog [5
]. Previously, we utilized the draft 1.9x feline sequence assembly to construct a low-resolution microarray platform for detection of recurrent DNA copy number aberrations (CNAs) in feline cancers using large insert genomic clones, enabling detection of broad regions of tumor-associated aneuploidy using comparative genomic hybridization (CGH) analysis [7
]. The microarray revealed numerous recurrent CNAs among a cohort of 46 feline soft tissue sarcomas, indicative of extensive, non-random genomic instability. Of particular note, we identified subchromosomal regions whose copy number status differed significantly between two related tumor subtypes that share similar histopathology at diagnosis but which typically show contrasting clinical behavior and outcome. Our initial findings thus supported the potential for a molecular means to aid distinction between these tumor subtypes, which in turn has implications for defining prognosis and selection of appropriate treatment strategies. The low resolution of this first-generation array, however, limited the ability to determine the precise boundaries of these CNAs, and in turn to assess their gene content in the search for clinically predictive alterations.
With a more comprehensive feline reference genome assembly still in development, we embarked on an in-silico strategy that instead used a well-characterized dog microarray design to guide the construction of a second-generation feline CGH platform. Following the integration of our arrayed probe set into the most recent cat reference sequence assembly build [8
], this platform now provides ~350-fold higher overall resolution than our first generation array [7
], bringing feline cytogenomic resources more closely in line with those of other model systems.
We present a novel microarray platform that surveys the domestic cat genome for DNA copy number aberrations with an average density of 44 probes per Mb. This study demonstrates the ability to utilize existing cytogenomic platforms to guide the development of similar tools for species for which robust sequence assembly resources are not yet available. In the absence of a comprehensive, anchored and annotated feline genome sequence assembly at the outset of this study, we used an existing dog array with ~13kb probe spacing to aid selection of appropriate feline sequences for the design of oligonucleotide probes along the length of each cat chromosome. By its nature, this approach resulted in a reduction in relative effective resolution due to incomplete representation of all chromosomal regions within the evolving cat reference assembly, and the consequent absence of sequence data for direct orthologs of a subset of canine probe sequences. Despite these confounding factors, the resulting feline array design yielded a mean interval of 22.6 kb between consecutive probes, representing a ~350-fold increase in resolution compared to our previous feline CGH platform [7
]. This resource provides a rapid means for screening feline tumors for somatic DNA copy number imbalances, for determining their impact on gene dosage, and in turn for elucidating the underlying molecular pathogenesis of the disease. While recent developments now permit the direct identification of copy number variations from next-generation sequencing (NGS) analysis of constitutional DNA, this approach acquires considerable analytical complexity when applied to cancer genomes. Confounding factors include the need for accurate normalization to account for the relative abundance of non-balanced genomic regions in tumor specimens, DNA degradation and fixation artefacts associated with archival specimens, and the presence of structural chromosomal rearrangements, polyploidy, polyclonality and ‘contaminating’ non-tumor DNA. Each of these factors contribute to the challenge of robust read alignment to the reference assembly for tumor specimens, and of sensitive and accurate detection of tumor-associated CNAs from NGS data. By its nature, the kinetics of the CGH technique means that these factors have a more moderate impact on the detection of copy number imbalances. Consequently, CGH continues to act as the ‘gold standard’ for the diagnosis of several human genomic abnormalities, both constitutional and somatic, and for identification of clinically actionable molecular targets [17
Using our feline microarray design, competitive hybridization of constitutional DNA from clinically healthy male and female cats demonstrated the expected pattern of autosomal balance. In turn, copy number assessment along fcaX detected relative imbalances that were highly concordant with theoretical values for female versus male DNA comparisons. The sex-mismatch analysis also enabled delineation of the feline PAR boundary to a 25.7 kb interval between the SHROOM2
genes, approximately 7.3 Mb from fcaXptel. This concurs with the findings of Murphy et al. [18
], who used radiation hybrid mapping to refine the boundary to a < 200kb region between these two genes. Using a combination of FISH and in-silico mapping analyses, Young et al. [19
] mapped the canine PAR boundary to a 2 kb interval within the SHROOM2
gene, approximately 6.6 Mb from cfaXptel. The observations from oaCGH analysis of fcaX from the present study are therefore fully concordant with prior comparative studies [20
], supporting the markedly increased size of the feline PAR relative to that of human (~2.7 Mb), and the highly conserved localization of the canine and feline PAR boundary.
Two unexpected regions of DNA copy number imbalance were identified in normal sex-mismatch hybridizations. One, located on fcaA3ptel, indicated a discrete, high amplitude copy number loss of the TETY1
locus in the female reference relative to the male. An earlier study [21
] determined that the TETY1
locus is encoded in multiple copies on fcaYq. Interestingly, however, a region of degenerate sequence homology with TETY1
was found on fcaA3, suggestive of an ancestral autosomal origin for the Y-borne locus [21
]. Furthermore, cross-species alignment of the TETY1
sequence identified orthology with autosomal sequences from cfa24q25, consistent with known regions of conserved sequence between this dog chromosome region and fcaA3 [3
]. The absence of orthologous sequences on the human or domestic dog Y chromosomes was taken to suggest that the transposition of TETY1
from autosome to Y chromosome occurred during carnivore evolution after the divergence of ancestral cat and dog lineages [21
]. The presence of multiple copies of TETY1
on fcaY, the absence of Y-specific probes on feline and canine oaCGH platforms (due to the female origins of their reference genome sequence assemblies), and the existence of a highly conserved sequence on fcaA3, therefore explains the apparent copy number loss of this latter region in females versus males in sex-mismatch oaCGH analysis. A second region of relative copy number decrease on fcaF2 in the female reference showed partial overlap with the RUNX1T1
gene; however at present there is no obvious explanation for this observation, and it is possible that this merely represents a reference genome assembly artefact.
The performance of the oaCGH array was evaluated further through analysis of an aggressive injection-site sarcoma, ISS-19a, which was examined at low resolution in an earlier study [7
]. oaCGH analysis identified a broad distribution of genomic imbalance, including a complex profile of interrupted segmental copy number aberrations along fcaD1, with a peak of high-level amplification evident at fcaD1ptel. This pattern shares similarities with profiles associated with chromothripsis/chromoanasynthesis observed in some human cancers. These complex forms of intrachromosomal reorganization comprise alternating segmental copy number imbalances interspersed between regions of euploidy, and are thought to manifest simultaneously from highly localized, catastrophic genomic disruption events [22
]. While the increase in probe density further delineates the discrete inflection points of these complex segmental changes, the large physical size of the high-level copy number gain on fcaD1ptel challenges assessment of the potential pathogenic significance of copy number amplification. Pertinent candidate targets of this ~11 Mb amplification include the platelet-derived growth factor D (PDGFD
) gene on fcaD1:2.37 Mb and members of the matrix metalloproteinase (MMP
) gene cluster on fcaD1:1.36 Mb, due to their varied roles in regulation of inflammation, wound healing, cell proliferation, migration, invasion, differentiation, angiogenesis and apoptosis [24
]. Of note, multiple MMP
genes have been shown to be upregulated in feline ISS, supporting an underlying role for inflammation in tumor pathogenesis [25
]. Given the intimate association between increased gene dosage and transcriptional/translational upregulation in human cancer studies [26
], it will be interesting to establish whether this complex reorganization of fcaD1 is conserved in other ISS cases. Combined molecular analysis showed that ISS-19a also presented with KIT
amplification, apparent homozygous deletion of PTEN
, and grossly balanced CDKN2A/B
copy number. These same alterations were also identified in ISS-19b, a second mass obtained from the same patient 11 months after surgical resection of the first tumor. Evaluation of additional cases will be required to determine whether these represent potentially clinically relevant aberrations that may drive ISS pathogenesis, or whether they represent the accumulation of passenger alterations. Regardless, the close conservation in the genome-wide oaCGH profiles of ISS-19a and ISS-19b provide strong support for a common origin.
Prior low-resolution CGH analysis also detected complex amplification and structural reorganization involving fcaE1ptel in ISS-19a [7
]. In the present study, oaCGH determined that this amplification extends 13.2 Mb from the fcaE1p telomere of both ISS-19a and ISS-19b. Additionally, FISH analysis with a probe representing fcaE1ptel identified multiple hybridization sites resembling satellite-like sequences on 3–5 derivative chromosomes in each cell scored (Figure 5
). It has been proposed that amplification of satellite DNA may be responsible for the induction of structural and numerical chromosome instability in feline fibrosarcoma, through aberrant kinetochore formation [27
]. The detection of a similar high-level amplification in additional cases would provide further support for its potential biological significance. A small number of additional studies have used conventional chromosome banding techniques to examine feline fibrosarcomas. These have revealed complex karyotypic instability comprising a wide variety of structural abnormalities, and aberrant chromosome numbers ranging from mild hypodiploidy to marked hyperdiploidy (summarized in [28
]). In combination, these findings show that feline fibrosarcomas exhibit a propensity towards major karyotypic disruption, which is a hallmark feature of many human soft tissue sarcomas, especially those associated with an aggressive phenotype [29
The resource described here offers a series of key advantages over our first-generation microarray platform [7
], most notably the ~350-fold increase in resolution. The increased probe density detected discrete relative DNA copy number imbalances in female versus male reference DNA hybridizations. Their identification will allow these regions to be characterized as natural polymorphisms in future studies, distinct from somatic alterations of tumor specimens. The in-silico design approach, using existing canine microarray probes to develop feline orthologs, can be exploited for direct comparison of evolutionarily-conserved genomic regions in both species. This, in conjunction with the increased resolution of our new microarray, provided accurate delineation of the feline PAR boundary, and revealed close conservation with that of the dog. These factors also allowed us to ‘drill down’ into large tumor-associated CNAs to reveal compelling gene targets within these intervals, such as the high-level amplification of fcaD1ptel in ISS-19. The increase in effective resolution also revealed the remarkably conserved profiles of complex, interrupted genomic gain and loss along fcaD1 in ISS-19a and ISS-19b, supporting the origin of the latter as a regrowth of the former, primary lesion. Interestingly, however, we also identified subtle differences between these profiles that demonstrate the potential for using this microarray platform to study tumor progression. Furthermore, our new microarray utilizes oligonucleotide probes that offer superior sensitivity and specificity for CNA detection compared to the large insert genomic clones used in our first-generation platform [17
]. The use of oligonucleotide probe technology will also enable rapid and cost-effective production of custom iterations of the array design, to address the specific goals of future studies.