Feasibility of circulating tumor DNA analysis in dogs with naturally occurring malignant and benign splenic lesions

Abstract

Comparative studies of naturally occurring canine cancers have provided new insight into many areas of cancer research. Development and validation of circulating tumor DNA (ctDNA) analysis in pet dogs can help address diagnostic needs in veterinary as well as human oncology. Dogs have high incidence of naturally occurring spontaneous cancers, demonstrate molecular heterogeneity and clonal evolution during therapy, allow serial sampling of blood from the same individuals during the course of disease progression, and have relatively compressed intervals for disease progression amenable to longitudinal studies. Here, we present a feasibility study of ctDNA analysis performed in 48 dogs including healthy dogs and dogs with either benign splenic lesions or malignant splenic tumors (hemangiosarcoma) using shallow whole genome sequencing (sWGS) of cell-free DNA. To enable detection and quantification of ctDNA using sWGS, we adapted two informatic approaches and compared their performance for the canine genome. At the time of initial clinical presentation, mean ctDNA fraction in dogs with malignant splenic tumors was 11.2%, significantly higher than dogs with benign lesions (3.2%; p = 0.001). ctDNA fraction was 14.3% and 9.0% in dogs with metastatic and localized disease, respectively (p = 0.227). In dogs treated with surgical resection of malignant tumors, mean ctDNA fraction decreased from 11.0% prior to resection to 7.9% post-resection (p = 0.047 for comparison of paired samples). Our results demonstrate that ctDNA analysis is feasible in dogs with hemangiosarcoma using a cost-effective approach such as sWGS. Additional studies are needed to validate these findings, and determine the role of ctDNA to assess burden of disease and treatment response in dogs with cancer.

Figure 1

Plasma cell-free DNA fragment size (base pairs) in nine healthy dogs. Individual samples are shown in grey, all samples combined are shown in red.

Figure 2

PlasmaSeq infers a larger proportion of shorter segments compared to ichorCNA, although the depth variation of bins within segments is approximately the same between PlasmaSeq and ichorCNA. (A) Distribution of segment size in base pairs across all samples using each approach. (B) The distribution of standard deviations of log₂ depth across all bins assigned to the same segment.

Figure 3

Circulating tumor DNA fraction for benign lesions vs malignant tumors in pre-treatment plasma samples (p = 0.001). Means are represented by diamonds for each group.

Figure 4

Genome-wide copy number variation plots from the metastatic HSA canine patient, dog no. 26, representing the plasma and tumor biopsy sWGS from day of surgery, and plasma sWGS from days 16 and 19, both after the splenectomy procedure.

Figure 5

Assessment of significant recurrence for observed copy number gains (top) and copy number losses (bottom), using the two approaches, ichorCNA (left) and PlasmaSeq (right). Copy number deletions observed on chromosomes 11, 32, 33 and 36 were consistent across the two approaches.