**3. Results**

#### *3.1. Case Details and Signalment*

One hundred and eighty-two (182) cases of CGC were identified from the archives of VPG Histology and the University of Lleida and submitted to the Department of Pathobiology and Population Sciences at the Royal Veterinary College. Following the initial histopathological review, 149 were suitable for inclusion in the study based on sample quality, presence of gastric carcinoma, and amount of tumor present in the sample. Cases included 85 (57.0%) endoscopically obtained (mucosa only) samples, and 64 (44.0%) fullthickness biopsies collected by surgical excision. Of the 149 dogs, 69 (46.3%) were female, 75 (50.3%) male and 5 (3.4%) of unreported sex. Neuter status was known in 118 cases, with 88 (74.6%) neutered. Twenty-nine non-tumor control cases were included, 24 (82.8%) endoscopic and 5 (3.4%) full-thickness biopsies. The control group comprised 15 males (51.7%), 13 females (44.8%), and one of unreported sex (3.4%). For the carcinoma group the mean and median ages were 9.1 and 9 years, respectively (range 1–16 years), and for the control group were 7.5 and 7, respectively.

The Staffordshire bull terrier (22/149, 14.8%) was the breed most frequently affected by gastric carcinoma. Of the other non-cross breeds, Labrador retrievers (12, 8.1%), golden retrievers (10, 6.7%), Boxers (9, 6.0%), Border collies (8, 5.4%), rough collies (6, 4.0%), and Belgian shepherd dogs (6, 4.0%) were also commonly affected (Table S1). A similar range of breeds was seen in the control population, with Staffordshire bull terriers, Labrador retrievers, Border collies and Belgian shepherd dogs all represented. No statistical correlation with breed, sex and age was found with regards to each CGC subtype.

#### *3.2. Histopathological Subtypes*

Undifferentiated carcinoma was the most frequent carcinoma type overall, comprising 59/149 (39.6%) samples (Table 1). SRC carcinoma made up 47 (31.5%), tubular 32 (21.5%), and mucinous 10 (6.7%). The papillary type was rare, with only a single case (0.7%) (Figure 1) There were no significant associations identified between breed or sex and any of the five classifications of gastric carcinoma.


**Table 1.** Canine gastric carcinoma subtype by sex.

**Figure 1.** Gastric carcinoma subtypes and local lymph node metastasis. (**a**) Papillary adenocarcinoma with characteristic fibrovascular stalks supporting neoplastic cells that form fingerlike projections (bar = 500 μm). Inset shows increased mitoses (bar = 20 μm). (**b**) Tubular adenocarcinoma with neoplastic tubules lined by pleomorphic cells (bar = 20 μm). (**c**) Signet-ring cell carcinoma with characteristic signet-ring cells diffusely replacing gastric mucosa (bar = 20 μm). (**d**) Mucinous adenocarcinoma with abundant lakes of mucin and occasional small numbers of signet-ring cells (bar = 50 μm). (**e**) Undifferentiated adenocarcinoma with neoplastic cells replacing and dissecting through muscularis layers (bar = 200 μm). Inset showing entotic cell-in-cell (CIC) patterns in an undifferentiated adenocarcinoma (bar = 20 μm). (**f**) Lymph node metastasis of a tubular adenocarcinoma with multifocal intravascular neoplastic emboli (bar = 200 μm). Inset shows a high-power magnification of neoplastic tubule formation within the neoplastic emboli (bar = 20 μm).

Of the 85 endoscopic samples, 36 (42.4%) were SRC carcinoma, 25 (29.4%) were undifferentiated, 19 (22.3%) tubular, and 5 (5.8%) mucinous. Of the 64 full-thickness samples, 34 (53.1%) were undifferentiated carcinoma, 13 (20.3%) tubular, 11 (17.2%) SRC, and 5 (7.7%) mucinous. The single papillary adenocarcinoma was recorded amongs<sup>t</sup> the full-thickness samples (1.5%) (Figure 2). The prevalence of SRC among endoscopic biopsies was significantly higher than among full-thickness biopsies (*p* = 0.0002). Comparisons between other subtypes in endoscopic and full-thickness biopsies did not show any statistical significance.

**Figure 2.** WHO classification of 149 canine gastric carcinomas separated by biopsy technique. Full-thickness surgical (FT), endoscopic (E).

#### *3.3. Inflammation and Presence of Helicobacter* spp.

Mean total chronic inflammation score (TCIS) for the control group was 0.5/9 and for the carcinoma group was 2.8/9. Mean TCIS for the carcinoma group was significantly higher than that of the control group (*p* = 0.0001).

Of all subtypes, tubular adenocarcinoma had the highest TCIS, 3.4, followed by mucinous—3.1, SRC—2.6, and undifferentiated carcinoma—2.5. Papillary adenocarcinoma had the lowest TCIS of 2.0. All subtypes (except for papillary) had a consistently significantly greater TCIS when compared with the control group (*p* = 0.0001) (Figure 3).

**Figure 3.** Mean total chronic inflammation score (TCIS) by carcinoma histological subtype. \*\*\*\* *p* < 0.0001.

Fifty out of 149 (33.5%) gastric carcinoma samples had observable *Helicobacter* spp., of which 19 (38%) had large numbers present. The TCIS of *Helicobacter* spp. positive samples was 3.2, versus 2.6 for *Helicobacter* spp. negative samples. These two means were found to be significantly different (*p* = 0.039) (Figure 4).

**Figure 4.** Mean total chronic inflammation score (TCIS) by helicobacter status for carcinoma and control groups. H+, helicobacter positive samples, H<sup>−</sup>, helicobacter negative samples.

Thirteen out of 29 (44.8%) of the control samples had observable *Helicobacter* spp., but only 2 (15.4%) had large numbers present. The TCIS of *Helicobacter* spp. positive samples was 0.58, and helicobacter negative samples was 0.50. These two means were not significantly different.

#### *3.4. Immunohistochemical Assessment of E-Cadherin, CD44, 14-3-3σ and p16*

Twenty-two cases of gastric carcinoma, randomly selected from each histopathological subtype, were available for immunohistochemical assessment as follows: SRC (8, 36.4%), undifferentiated (7, 31.8%), tubular (5, 22.7%), papillary (1, 4.5%), and mucinous (1, 4.5%).

#### *3.5. E-Cadherin Expression*

In normal gastric surface and glandular epithelium, E-cadherin labeling was strong (3/3 intensity score) and membranous (Figure 5a). Seventeen (of 22) tumors demonstrated positive immunolabeling for E-cadherin, with both membranous and cytoplasmic (17/22, 77.3%) and nuclear (2/22, 9.1%) labeling, and with an intensity greatest in intravascular emboli, where present. Five tumors did not show immunopositivity for E-cadherin (2 undifferentiated, 1 SRC, 1 mucinous and 1 tubular carcinoma). Where dysplastic epithelium was present, E-cadherin expression decreased in intensity and distribution (compared with normal), and labeling became progressively cytoplasmic (from membranous) (Figure 5b). The TIS E-cadherin for each tumor subtype was as follows: papillary—6.0; undifferentiated— 4.6; SRC—4.3; tubular—3.2; and mucinous—0 (Table 2). Aberrantly increased expression and TIS of E-cadherin in intralymphatic/intravascular emboli, compared with both normal epithelium and immunopositive neoplastic cells, was observed in seven tumors (7/17; 3/7 undifferentiated, 1/8 signet-ring cell and 3/5 tubular carcinoma). E-cadherin was not expressed by neoplastic emboli composed of well-differentiated acinar structures (Figure 5c).

**Figure 5.** Canine gastric carcinoma immunohistochemistry for E-cadherin, CD44, 14-3-3<sup>σ</sup>, p16. (**a**) Expression of E-cadherin in normal gastric surface and glandular epithelium (bar = 100 μm). (**b**) Reduced E-cadherin expression (progressively cytoplasmic from membranous) in neoplastic tubules in a tubular adenocarcinoma (bar = 50 μm). (**c**) Poorly differentiated neoplastic cells in emboli showing cytoplasmic and membranous labeling for E-cadherin (bar = 50 μm). (**d**) Expression of CD44 in lymphocytes, macrophages and dendritic cells of normal gastric mucosa. (bar = 100 μm) (**e**) Strong membranous labeling of CD44 in sheets and chains of poorly differentiated neoplastic cells in an undifferentiated gastric carcinoma (bar = 200 μm). (**f**) Loss of membranous and cytoplasmic labeling of CD44 in tubules and strong membranous and cytoplasmic labeling in poorly differentiated cells of a neoplastic embolus (bar = 100 μm). (**g**) Normal epithelium showing no expression of 14-3-3σ (bar = 100 μm). (**h**) Strong cytoplasmic with occasional nuclear labeling of 14-3-3σ in nests of poorly differentiated neoplastic cells in an undifferentiated gastric carcinoma (bar = 20 μm). (**i**) Loss of cytoplasmic labeling of 14-3-3σ in tubules and cytoplasmic labeling in poorly differentiated cells of a neoplastic embolus (bar = 50 μm). (**j**) Normal epithelium showing no expression of p16 (bar = 100 μm). (**k**) Strong cytoplasmic labeling of p16 in neoplastic epithelium (bar = 50 μm). (**l**) Loss of cytoplasmic labeling for p16 in tubules and strong nuclear and cytoplasmic labeling in poorly differentiated cells of a neoplastic embolus (bar = 50 μm).


**Table 2.** Mean total immunohistochemical score by gastric carcinoma subtype. The same score for neoplastic vascular emboli, where present, is given in parentheses.

#### *3.6. CD44 Expression*

In the normal canine stomach, CD44 was expressed on the membrane of lymphocytes, macrophages, dendritic cells, and satellite glial cells (Figure 5d). Gastric epithelium, stromal tissue, smooth muscle fibers and matrix fibroblasts were negative. Positive immunolabeling (neo-expression) occurred in all gastric carcinoma cases. CD44 expression was cytoplasmic and membranous (5/22, 22.7%) or only membranous (9/22, 40.9%), with an intensity greatest in mucinous tumors (mean intensity score 2.5). Enhanced membranous expression was observed in those neoplastic epithelial cells that showed the greatest features of malignancy (pleomorphism and invasion), with a higher mean score in papillary, undifferentiated and tubular carcinomas (6, 4.7 and 4, respectively) (Figure 5e). Intravascular neoplastic emboli, present in 4/7 undifferentiated, 4/5 tubular and 3/7 SRC carcinomas, showed increased TIS compared to immunopositive neoplastic cells. Neoplastic cells in emboli exhibited membranous, or both cytoplasmic and membranous expression, in poorly differentiated neoplastic cells. Interestingly, there were neoplastic emboli that had solid nests and welldifferentiated acinar arrangements, and, in these emboli, only the solid nests were strongly positive, while the well-differentiated embolic acini were negative (Figure 5f).

#### *3.7. 14-3-3 σ Expression*

In a histologically normal canine stomach, 14-3-3σ was not expressed in any cell types, including epithelium, stromal tissue or lymphocytes (Figure 5g). Positive immunolabeling (neo-expression) occurred in 10 of 22 (45.4%) gastric carcinomas, 7/7 undifferentiated, and 3/5 tubular. TIS 14-3-3σ for undifferentiated and tubular carcinomas was 3.4 and 2.4, respectively. SRC, papillary and mucinous carcinomas were all negative for 14-3-3<sup>σ</sup>. Cytoplasmic and/or nuclear neo-expression of 14-3-3σ was present, with an intensity greatest in undifferentiated tumors (intensity mean 2.1, and total score 3.4). One case of undifferentiated carcinoma showed both nuclear and cytoplasmic labeling in neoplastic cells (Figure 5h). Where neoplastic emboli were present, TIS was higher compared to immunopositive neoplastic cells, and 14-3-3σ labeling intensity was increased in poorly differentiated and pleomorphic cells. However, when neoplastic cells formed intravascular acinar structures, 14-3-3σ was not expressed (Figure 5i).

## *3.8. p16 Expression*

In a histologically normal canine stomach, p16 was not expressed (Figure 5j), but positive cytoplasmic and/or nuclear labeling was noted in regions of dysplastic epithelium (Figure 5k). p16 expression occurred in 19/22 (86.4%) carcinomas; 6/7 undifferentiated, 5/5 tubular, 1/1 papillary, 6/8 SRC, and 1/1 mucinous. Positive immunolabeling of pleomorphic cells and dysplastic tubules was found in all tumors, with high-grade expression in mucinous tumors (signet-ring cells in SRC and mucinous carcinomas were strongly immunopositive for p16). p16 exhibited cytoplasmic (11/22, 50%) and both nuclear/cytoplasmic (8/22, 36.36%) expression, with intensity greatest in mucinous tumors (intensity mean 3 and total score 9).

Where neoplastic emboli were present, TIS was higher compared to immunopositive neoplastic cells. Within intravascular and intralymphatic emboli, p16 intensity and cytoplasmic and/or nuclear expression was increased in poorly differentiated and pleomorphic

cells. However, when neoplastic cells formed intravascular acinar structures, p16 was not expressed (Figure 5l). Similar findings were noted in the lymph node metastasis of one sample. The mean TIS p16 for undifferentiated, tubular, papillary, SRC and mucinous carcinomas was 4.3, 6.8, 6, 6.9 and 9, respectively.
