2.1 Sample collection and cell lines

Blood and tissue biopsy samples from dogs were collected by veterinarians, in the course of the dogs’ medical care, with the owners’ consents, through the Cani-DNA Biobank (https://igdr.univ-rennes.fr/crb-cani-dna). This study involving dog samples was approved by the Ethics Committee (CREEA - #CE07-2020-11-CA).

For 38 dogs, between 2–5 ml of blood was taken during a routine examination or during the clinical follow-up of the dog and was collected in ethylenediamine tetraacetic acid (EDTA).

Tissue biopsy samples were collected by the attending veterinarians as part of routine veterinary care, and no animals were euthanized for the purposes of this study. Thus, anesthesia and/or analgesia methods, as well as measures to alleviate suffering, were administered by the attending veterinarian as part of the proper practice of veterinary medicine. Each biopsy was divided into two parts, one was placed in RNALater solution for genomic experiments and the other was fixed in formalin for histopathological analyses, respectively. All diagnoses were initially performed by board-certified pathologists in the context of the routine activities of veterinary pathology laboratories and were subsequently peer-reviewed and confirmed by a board-certified pathologist (JA). To fully confirm the histiocytic nature of the neoplastic cells, immunohistochemical expression of ionized calcium-binding adapter molecule (Iba1), a specific marker of the monocyte/macrophage lineage, was demonstrated in all samples from all affected dogs [13]. The age at diagnosis was calculated as the difference between the date of histological or cytological diagnosis and the date of birth.

Two histiocytic sarcoma cell lines from the sample 5472 were developed from histiocytic sarcomas tumors isolated from the lung and spleen of the same dog as previously described) [14]. These cell lines were cultured in RPMI 1640 medium (Gibco) supplemented with 10% fetal bovine serum and 0.2% primocin.

2.2 DNA extraction

Samples were treated as previously described [15]. Briefly, tumor DNA was extracted from RNAlater preserved tumor samples or cell lines with DNA Nucleospin Tissue kits while germline DNA was isolated from blood using the Nucleospin Blood L kit (Macherey Nagel, Düren, Germany), according to the manufacturer protocol.

2.3 Identification of PTPN11 mutations by ddPCR

Known hotspot mutations in the PTPN11 gene are named according to the international convention for point mutations, which specifies the original amino acid, the residue position in the protein, and the mutated amino acid. The nomenclature of the known mutations is provided in S1 Table and the primers are presented in S2 Table. The identification of PTPN11 mutations in different tumor samples was performed using using ddPCR as described in Prouteau et al. (2021). Briefly, a 20 μL reaction volume was prepared containing the droplet Supermix (Bio-Rad Laboratories, Hercules, CA, USA) with a final concentration of 500 nM of forward and reverse primers, 250 nM of HEX- and FAM-labeled probes, and 10 ng of DNA. Next, the PCR reaction mixtures were partitioned into an emulsion of approximately 20,000 droplets using a QX200 ddPCR droplet generation system (Bio-Rad). PCR was conducted using the GeneAmp PCR system 9700 thermocycler (Applied Biosystems) using the following program: 95 °C for 10 min; 40 cycles of 94 °C for 30 s, and an elongation temperature of 58°C for 60 s, and subsequently 98 °C for 10 min. The ddPCR assay included corresponding positive samples from tumours in which the mutations had previously been identified by Sanger sequencing [4], matched blood samples from the same individuals, and a blank control (water). Serial dilutions of positive DNAs in normal wild type DNA, were used to determine the sensitivity of our ddPCR assay for detecting E76K and G503V mutations in the PTPN11 gene in 10ng of DNA input. These sensitivities are 0.096% and 0.2% for hotspots 1 (E76K) and 2 (G503V), respectively.

2.4 WGS sequencing of the 2 HS cell lines and somatic alteration identification

Whole genome sequencing (WGS) of the two HS cell lines was performed using the BGI sequencing platform (BGI, Shenzhen, China), using BGISEQ-500 short-read sequencing as previously described (Anais Prouteau et al. 2022.). Briefly, 1 µg of genomic DNA was quantified using a Qubit 3.0 fluorometer (Life Technologies, Paisley, UK), and was sheared using an E220 Covaris instrument (Covaris Inc., Woburn, MA, USA). Size selection was subsequently performed using the Vahstm DNA Clean beads kit (Vazyme, Nanjing, China) to obtain fragments with an average size of 200–400 bp. The selected fragments were end repaired and 3’ adenylated, and BGISEQ-500 platform-specific adaptors were ligated to the A-tailed fragments. The ligated fragments were purified and then amplified using PCR. Finally, circularization was performed to generate single-stranded DNA circles. After quantification, the libraries were loaded onto a sequencing flow cell and processed for 100 bp paired-end sequencing on the BGISEQ-500 platform.

After sequencing, the raw reads were filtered (adapter sequences, contamination, and low-quality reads were removed) according to the manufacturer’s guidelines. Sequence data were then aligned to the dog reference genome (assembly version canFam3.1) using BWA-MEM (version 0.7.17) [16], and PCR duplicate reads were removed using Picard tools (version 2.18.23) (http://broadinstitute.github.io/picard/, accessed on 1 July 2021). The read data were processed according to the GATK best practices; specifically, the base quality score recalibration was assessed using GATK4 (version 4.0.12).

The Mutect2 tool from the GATK4 software was then used to call somatic SNVs and short INDELS against a panel of normal (PON), comprising matched normal sample, and against germline variants from 722 dog genomes (https://data.broadinstitute.org/vgb/Ostrander_VCFs/722g.990.SNP.INDEL.chrAll.vcf.gz), accessed on 1 July 2021). Variant annotation was performed using the VEP program [16] with the EnsEMBL “Canis familiaris” annotation (v. 95). To confirm the common origin of the two HS cell lines, HaplotypeCaller tool from GATK4 software was used to call the germinal variants overlapping the 170K SNPs of Illumina CanineHD BeadChip. The identity-by-state (IBS) matrix was computed using the snpgdsIBS function of the SNPRelate R package, and a hierarchical clustering (average linkage) was then performed on the IBS-based distance matrix for these cell lines and germline DNA from 750 published BMDs [17].

2.5 Low pass sequencing

To confirm the tumoral status of the samples and to demonstrate that detection of divergent clones using the somatic mutational status of PTPN11 can be missed in wild-type organs, we performed low-pass sequencing on tumor samples. These included one case (dog 8074), which exhibited a PTPN11 mutation in the lung but had a wild-type status in the spleen and liver, one case with wild-type tumor pairs (dog 17092), and one case with mutated tumor pairs (dog 18716). Low-pass sequencing of tumor samples was performed by Psomagen (Rockville, MD, USA) as previously described (Anais Prouteau et al. 2022.). Briefly, the input DNA quality was verified using gel electrophoresis, and the quantity was measured using the Picogreen assay (Thermo Scientific, Waltham, MA, USA); 2 ng of DNA was prepared in 30 ul of buffer and used for library construction using the Nextera DNA Flex Library Kit (Illumina, San Diego, CA, USA), according to the manufacturer’s guidelines. The size of the final DNA libraries was then validated using the TapeStation D1000 ScreenTape (Agilent, Santa Clara, CA, USA) and D1000 reagents (Agilent, Santa Clara, CA, USA). The quantity was measured using the Picogreen assay (Thermo Scientific, Waltham, MA, USA), and the molar concentration was calculated using both sources. The libraries were normalized to 2 nM, pooled in equimolar volume, and then loaded on the flow cell from the Novaseq S4 300 cycle kit (Illumina, San Diego, CA, USA) and the XP-4lane kit (Illumina, San Diego, CA, USA). The prepared flow cell and SBS cartridge from the Novaseq S4 300 cycle kit (Illumina, San Diego, CA, USA) were inserted into the Novaseq 6000 system and sequenced using 151-10-10-151 running parameters, reaching a mean depth of 0.48 X.

After sequencing, the raw reads were processed as for the WGS (see paragraph 2.4). The somatic copy number variants (CNVs) were determined using the “R” Package DNAcopy [18]. The canine genome was split into bins with a window size of 100 kb. The number of reads was normalized to the total number of reads per sample. The log2 ratio with the corresponding germinal DNA, when available (or with a pool of normal DNAs), was estimated for each window. Several windows were merged (segmentation) into a larger segment with a significance threshold of 1 × 10⁴, and the copy number for each segment was estimated by converting the segment’s average log2 ratio into an absolute copy number, where a log2 ratio of 0 corresponds to a normal diploid state.

3.1 Cohort characteristics

Thirty-eight individuals with abdominal and thoracic HS were collected between 2005 and 2019 and included in this study (Table 1). This cohort include 34 Bernese Mountains Dogs (BMDs; 89.5%), 3 Rottweilers (Rott; 7.9%) and one Flat Coated Retriever (FCR; 2.6%). This proportion reflects the described predisposition of BMDs to DSH. No sex bias was observed in this cohort with 21 (55.3%) males and 17 (44.7%) females (Chi-square test, p = 0.51). The median age of HS at diagnosis in this population was 7.09 years old (range: 2.23–11.73 years).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Genotyping of PTPN11 mutations in the cohort of 38 dogs.

https://doi.org/10.1371/journal.pone.0345429.t001

3.2 Genotyping of PTPN11 mutations

Tumor DNA samples and corresponding germline DNAs extracted from blood samples of the 38 dogs were analyzed by using droplet digital PCR (ddPCR) to detect PTPN11 mutations at two hotspot regions: E76 (hotspot 1) and G503 (hotspot 2). No PTPN11 mutations were identified in germline DNAs confirming the somatic origin of these mutations.

Among the 38 dogs studied, 25 (66%) carried at least one mutation at the main hotspot 1 (E76), and 12 (32%) had a mutation at the second hotspot 2 (G503) in at least one organ (Table 1). These frequencies are consistent with previous reports of recurrent PTPN11 hotspot mutations in canine histiocytic sarcoma, particularly in Bernese Mountain Dogs (Hédan et al. 2020; Takada et al. 2019).

Mutation profiles were compared between organs to assess concordance between organs. Most dogs showed concordant profiles between organs (25/38, 66%). In contrast, seven of 38 cases (18.4%) exhibited discordant profiles between abdominal and thoracic histiocytic sarcoma (HS) lesions, including three cases (7.9%) with differing mutations and four cases (11%) with mutated versus wild-type profiles. Notably, all discordant cases occurred in Bernese Mountain Dogs. The complete results are summarized in Fig 1.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. PTPN11 mutation patterns and inter-organ concordance in 38 dogs with DHS.

(A) Venn diagram showing overlap between hotspot mutations. Among 38 dogs, 25 (66%) carried an E76 mutation and 12 (32%) a G503V mutation. Five dogs (13%) harbored mutations at both hotspots, resulting in a total of 32 (84%) dogs with at least one mutation. (B) Concordance of mutation profiles between organs. Most cases (25/38, 66%) showed concordant mutations, while 3 (8%) had discordant mutations, 4 (11%) showed mutated versus wild-type profiles, and 6 (16%) had no detectable mutations.

https://doi.org/10.1371/journal.pone.0345429.g001

3.3 Estimation of divergent clone frequency using somatic mutational status of PTPN11

As PTPN11 is frequently mutated in HS at two main hotspots (E76 and G503), these sites can be used to assess clonal divergence between different tumors from the same dog. However, in cases where both organs were wild type, no somatic mutation was available to evaluate potential clonal divergence. In addition, for cases with a mutation in one organ and a wild-type status in the other, we could not rule out that the tumor cell fraction in the wild-type organ was too low to be detected, even with sensitive techniques such as ddPCR. To illustrate this, low-pass sequencing was performed on one case (dog 8074) presenting a mutation in one organ and a wild-type status in the other, one case with wild-type tumor pairs (dog 17092) and one case with mutated tumor pairs (dog 18716). Analysis of copy number alterations (CNAs) detected by low-pass sequencing confirmed the presence of tumor tissue in all tumor tissue samples including the wild-type samples (S1-S3 Figs). In addition, Spearman correlation analysis revealed high CNA similarity only between the mutated paired samples of dog 18716, thereby identifying divergent clonal origins in the two other dogs (S4-S5 Figs; S3 Table). These results confirm the existence of divergent clonal patterns among tumors classified as wild type by PTPN11 hotspot analysis.

Thus, to estimate the frequency of divergent clonal patterns, we focused only on cases with different PTPN11 mutations in the abdominal and thoracic HS lesions among cases with mutated tumor pairs, i.e., three dogs out of 28. However, only a subset of divergent situations can be detected with this approach, because the same somatic mutations are recurrent and frequent across different dogs—E76 in 30.2 to 48.6% of cases and G503 in 9–10% [4,5]. Consequently, identical mutations observed in two organs in the same dog may have arisen independently, particularly at the E76 hotspot. Thus, the frequency of HS originating from different clones, based only on cases in which distinct PTPN11 mutations, is underestimated and must be corrected to account for situations that are not detectable with our two-hotspot based approach.

In a scenario where two HS lesions from the same dog acquire independent PTPN11 mutations, the probability of detecting a mismatch between the two tumors (P_detected) corresponds to the probability that the lesions harbor different hotspot mutations (one at E76 and the other at G503).Let p₁ and p₂ denote the frequencies of mutations at hotspot 1 (E76) and hotspot 2 (G503), respectively, calculated among PTPN11-mutated cases. A detectable mismatch can arise only in two configurations:

1. E76 in the abdominal lesion and G503 in the thoracic lesion, or
2. G503 in the abdominal lesion and E76 in the thoracic lesion.

Each configuration occurs with probability p₁ × p₂. Therefore, the probability of detecting a mismatch among independently mutated pairs is:

Because our current cohort is limited (32 mutated tumor pairs), the statistical estimate of p1 and p2 may be unstable. To obtain more robust estimates, we applied published PTPN11 hotspot frequencies from a larger Bernese Mountain Dog cohort (Hédan et al. 2020; Takada et al. 2019), where p₁ (E76) = 0.756–0.852 and p₂ (G503) = 0.148–0.244. Using these values, the proportion of detected mismatches are:

In our cohort, 3/28 (10.7%) mutated tumor pairs showed discordant PTPN11 mutations. Adjusting for the limited detection based on hotspot-based analysis, the estimated proportion of tumor pairs truly arising from acquire independent PTPN11 mutations, (P_present) is:

Thus, since only a subset of potential divergent PTPN11 mutations can be detected using the two hotspots, we estimate that in this cohort, the theoretical proportion of tumors arising from independent PTPN11 clones could be up to three times higher than the number directly observed (10.7%).

3.4 Exploration of the HS clonality by analyzing cell lines from a DHS case harboring different mutations

To further investigate the distinct mutational profiles observed across organs affected by DHS, we performed whole-genome sequencing (WGS) on two cell lines derived from different tumor sites (spleen and lung), as well as on germline DNA (blood) from dog 5472, a case of disseminated HS (Table 1). Germline DNA and DNA from both tumor-derived cell lines were sequenced at an average depth of 30 × . Whole-genome copy number alteration (CNA) profiling revealed extensive genomic rearrangements in both cell lines (S6 Fig). SNPs from the Illumina 173K Canine HD array were extracted from the whole-genome sequencing data of the germline germinal DNA and the spleen and lung cell lines. Hierarchical clustering, based on a genome-wide genetic distance matrix derived from identity-by-state (IBS) similarity and calculated from SNP genotypes (including dog 5472 and 749 additional samples from the cohort [17]) using SNPRelate, confirmed that the two 5472-derived cell lines originated from the same dog 5472 (S7 Fig). Then, the Mutect2 program was used to catalog a total of 33332 and 26475 SNVs in the lung and the spleen derived cell lines, respectively (Table 2, S4-S5 Tables). Mutations in genes frequently altered in HS, such as TP53, or PTPN11 [4,5,7,8], were detected in both cell lines (K128N, E169K, G256E for TP53 and E76K, G503V for PTPN11). Among all identified variants, only one SNV (Chr7: 5963091 C > G) was shared between the two cells lines. This variant is a predicted deleterious coding variant in MRPL45, a gene encoding a mitochondrial ribosomal protein. With only one common variant – representing 0.003% and 0.0037% of variants in the lung and spleen cell lines, respectively – we can conclude that the two cell lines originated from different clones.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Summary of SNVs detected in the two cell lines Dog_HS_5472_Lung and Dog_HS_5472_Spleen and identification of 1 deleterious variant common to both cell lines.

https://doi.org/10.1371/journal.pone.0345429.t002