Copy number abnormalities in sporadic canine colorectal cancers

Abstract

Human colorectal cancer (CRC) is one of the better-understood systems for studying the genetics of cancer initiation and progression. To develop a cross-species comparison strategy for identifying CRC causative gene or genomic alterations, we performed array comparative genomic hybridization (aCGH) to investigate copy number abnormalities (CNAs), one of the most prominent lesion types reported for human CRCs, in 10 spontaneously occurring canine CRCs. The results revealed for the first time a strong degree of genetic homology between sporadic canine and human CRCs. First, we saw that between 5% and 22% of the canine genome was amplified/deleted in these tumors, and that, reminiscent of human CRCs, the total altered sequences directly correlated to the tumor's progression stage, origin, and likely microsatellite instability status. Second, when mapping the identified CNAs onto syntenic regions of the human genome, we noted that the canine orthologs of genes participating in known human CRC pathways were recurrently disrupted, indicating that these pathways might be altered in the canine CRCs as well. Last, we observed a significant overlapping of CNAs between human and canine tumors, and tumors from the two species were clustered according to the tumor subtypes but not the species. Significantly, compared with the shared CNAs, we found that species-specific (especially human-specific) CNAs localize to evolutionarily unstable regions that harbor more segmental duplications and interspecies genomic rearrangement breakpoints. These findings indicate that CNAs recurrent between human and dog CRCs may have a higher probability of being cancer-causative, compared with CNAs found in one species only.

Figure 1.

(A) Cross-species comparison for causative (or driver) aberration identification. Once we demonstrate that the same types of cancer from the human and the dog share similar molecular and genetic pathways of cancer development and progression, we will consider abnormalities recurrent between the two species as driver candidates (solid gray area), and those found in only one species and falling in evolutionarily unstable sites (EIN sites) as passenger candidates (small squared areas). Those in the parallel-lined areas need further studies. (B) The advantage of the human–dog comparison strategy over the human-only strategy for cancer driver gene identification. The cross-species comparison strategy can make use of the difference in the genomic location of orthologous genes between the human and the dog, a result of evolutionary genomic rearrangements that occurred since the two species diverged more than 75 million years ago. This shows that two genes, which are nearby in the human genome but distant in the dog genome, are both disrupted in the human cancer (boxed with broken lines). However, in the dog cancer, only one gene is disrupted, which will be considered as driver, and the other is intact (boxed with unbroken lines), which will be deemed as passenger.

Figure 2.

Cryosectioning and H&E staining of dog colon tumor and normal tissue samples. The images represent a normal colon tissue (top), an adenoma (middle), and an adenocarcinoma (bottom).

Figure 3.

(A) CNAs identified in the genome of a dog adenocarcinoma (T11). The identified CNAs (3396 gains and 5711 losses) amount to 551 Mb (22% of the dog genome) (Table 1). Each line represents a dog chromosome with its chromosome number indicated on the right. Vertical lines shown above/below the chromosomes represent gains/losses, respectively, with their length calculated based on , where l and m are the total probe number and the mean log₂ ratio of the CNA. Except for CNAs that are larger than 1 Mb in size, the width of the vertical lines is not drawn to scale with the chromosome length. (B) Mapping dog CNAs onto the human genome. A total of 9107 CNAs and 541.6 Mb (98.2% of the total) of the same dog tumor shown above were mapped onto the human genome, amounting to 609 Mb on the human genome.

Figure 4.

MSI assay of the dog tumors. A total of eight dog microsatellite loci were analyzed as described in the text, with the top five being homologs of frequently used human MSI markers (Boland et al. 1998) and the bottom three being dinucleotide markers used by McNiel et al. (2007) to determine MSI status in canine mammary gland neoplasia. The extra band that the tumor displayed for the locus CPA5 was indicated by an arrow. T, tumor; N, its matching normal.

Figure 5.

(A) Clustering of the dog tumors. The tree was constructed by MST or top-down clustering as described in the text (both strategies yielded the same tree), with the sample information for each tumor shown on the right. The number for each branch represents the distance d(X, Y) between the two clusters X and Y involved, calculated by , where d_ij is the distance between a tumor T_i of cluster X and a tumor T_j of cluster Y calculated as described in the text, and |X| and |Y| are the total number of tumors inside clusters X and Y, respectively. (B) Clustering of tumors from both humans and dogs. The tree was constructed as described above, using the overlapping information of CNAs either identified on (for the human tumors T2551 and T3912) or mapped onto (for the dog tumors, see Fig. 3) the human genome. MSI-L: MSI-low; MSI-H: MSI-high.