**4. Discussion**

In this canine gastric carcinoma (CGC) study, using the WHO scheme, we have demonstrated that the most frequent subtype is the undifferentiated carcinoma, and the most common breed affected is the Staffordshire bull terrier. There is no sex predisposition, and the mean age is nine years. *Helicobacter* spp. presence was associated with increased chronic inflammation parameters and a greater chronic inflammatory score. We found that all markers showed variable expression for each subtype. CD44 and 14-3-3σ have not been previously investigated in CGC. 14-3-3σ was positive in undifferentiated, tubular and papillary carcinomas, and p16 expression was increased in mucinous and SRC carcinomas. E-cadherin and CD44 were variably expressed in all subtypes and were associated with criteria of malignancy. Within neoplastic emboli, acinar structures lacked expression of all markers, suggesting an independent molecular pathway that requires further investigation.

Of the 149 dogs included in this study from the UK and Spain, the mean and median age of 9.1 and 9 years, respectively, at the time of CGC diagnosis is in broad agreemen<sup>t</sup> with previously reported studies [3,6,7]. In this study, CGC affected male and female dogs roughly equally, with a slight preponderance towards males, as is consistent with previous reports [3]. The most commonly affected breed was the Staffordshire bull terrier, likely reflecting both the general population in the United Kingdom (only 2/22 Staffordshire bull terriers were from Spain) and previous reports of breed predisposition [7]. Breed predisposition to CGC is also reported in less common breeds, including Belgian shepherd dogs and rough collies, which also appeared with an increased frequency in this study [3,6].

Using the WHO classification for domestic animals, cases were divided into five categories based on the predominant histological features and the principal cell type of the tumor. All histological subtypes of carcinoma (SRC, tubular, mucinous, papillary and undifferentiated) were recorded, although squamous cell carcinoma, another CGC subtype, was not present in this study [44]. Undifferentiated carcinoma was the most frequent subtype, followed by SRC, contrary to the previous studies reporting an increased frequency for the tubular subtype, similar to humans [45,46]. This could likely reflect a geographic variation in gastric adenocarcinoma incidence.

Signet-ring cell carcinomas were the most frequently diagnosed type by endoscopy, whereas undifferentiated carcinomas were the most frequently diagnosed in full-thickness samples. The most likely reason behind this difference is that, histologically, the diagnosis of SRC subtype is based upon identification of the characteristic isolated or small groups of malignant cells containing intracytoplasmic mucin with an eccentric nucleus (signetring cells) within the mucosa, and hence, diagnosis from an endoscopically retrieved (mucosa only) sample is possible. The diagnosis of other histopathological subtypes typically requires the evaluation of invasion beyond the mucosa, and thus, given the relatively high number of full-thickness biopsies in this study, this may have influenced the predominant subtype. Additionally, signet-ring cells were also present in other subtypes, i.e., mucinous and undifferentiated carcinoma; however, the predominant histological features and pattern did not favor a diagnosis of SRC carcinoma. This suggests that endoscopic samples alone can lead to diagnostic pitfalls between different subtypes [47–50]. In humans, novel techniques such as endoscopic surgical dissection are recommended in subtypes like SRC that spread subepithelially in the margins [51]. In combination with future novel imaging techniques, these findings should be taken into consideration to determine the appropriate sampling and therapeutic approach [52].

Cancer-related inflammation is one of the hallmarks of neoplasia, and once cancer develops, there is evidence of a substantial shift in the microenvironment affecting the immune response [53,54]. A simplified histopathologic scoring system was adapted to reduce variability in the diagnostic interpretation [37]. This scoring system captured and quantified chronic inflammatory changes in the gastric mucosa [55]. The TCIS for the carcinoma group, and individually for each CGC subtype (independent of helicobacter presence), was significantly greater when compared to the control group's TCIS. Thus, there is a clear correlation between chronic inflammation and carcinoma for most CGC subtypes.

Given the link between the presence of helicobacter, inflammation and neoplasia in humans we aimed to quantitatively assess helicobacter presence in CGCs. In previous studies *Helicobacter* spp. have been reported in both neoplastic and non-neoplastic canine stomachs [56,57]. In this study, the TCIS of the carcinoma group with a concomitant presence of helicobacter was significantly greater compared to the helicobacter-negative carcinoma group. Furthermore, larger numbers of helicobacter were found in carcinoma cases than in control cases. Although the helicobacter presence does not prove a definite role in CGC pathogenesis, there is a clear association between increased chronic inflammation and higher numbers of helicobacter in the tumor cases versus controls. Whether the presence of *Helicobacter* spp. is associated with tumor development is not exactly clear, and thus further studies, perhaps to examine particular species of helicobacter in association with CGC, would be needed.

To understand the molecular pathways and investigate the cellular origin of CGC, the expression of cellular adhesion molecules (E-cadherin and CD44) and cell cycle regulators (p16 and 14-3-3σ) were investigated in a representative proportion of cases. CD44 and 14-3-3σ have not previously been studied in CGC.

E-cadherin controls cell motility and suppresses tumor growth and metastasis [58]. Immunohistochemical examination of E-cadherin in all tumor subtypes identified abnormalities of expression and localization. Dysplastic surface and glandular epithelia in immunopositive cases revealed a reduction in E-cadherin expression. Pleomorphic cells in the undifferentiated subtype showed E-cadherin expression, albeit in decreased intensity compared to the normal epithelium. Subcellular localization of E-cadherin was observed in the majority of neoplastic cells. Intracytoplasmic sequestration and the accumulation of E-cadherin have been previously associated with abnormalities in the intracytoplasmic transport and reuptake mechanisms of the molecule [59,60]. Aberrant nuclear expression of E-cadherin in humans has been previously described in several tumor types including gastric and colorectal carcinomas [19,61]. Similarly, increased intensity intracytoplasmic and nuclear expression was noted in the neoplastic emboli composed of pleomorphic neoplastic cells forming solid nests. Surprisingly, intravascular acinar structures did not express E-cadherin. Furthermore, tubular adenocarcinoma, which is composed of acini and tubules, had the lowest TIS compared to other subtypes. This could represent an inverse association between the reduction or loss of E-cadherin with an increasing degree of differentiation, where well-differentiated structures switch off and pleomorphic cells switch on or increase the expression of E-cadherin. Previous studies in human gastric adenocarcinoma showed that abnormalities of E-cadherin localization (internalization) in neoplastic cells lead to decreased adhesion, thus favoring invasion [59,60]. Neoplastic cells lacked E-cadherin expression in discohesive subtypes like mucinous adenocarcinoma, which is in accordance with a recent study [20].

In the context of these findings, we postulate that the reduction or loss of E-cadherin expression on the cell membrane could potentially facilitate invasion, and that the reexpression of E-cadherin, including on the cell membrane, in neoplastic emboli could lead to the development of solid cohesive intravascular structures, enhancing their survival. These findings reinforce the notion that E-cadherin in CGC is a context-dependent adhesion molecule that can either be up- or down-regulated or re-expressed, depending on the stage of tumor progression [62]. Further studies analyzing ligands of E-cadherin and possible mutations are required to clarify its precise role in CGC.

CD44 serves as a signaling platform, with transmembrane and cytoplasmic domains as a co-receptor for various types of cell surface receptors that modulate cell adhesion, migration, proliferation, differentiation and survival [63–66]. In humans, its role in the

adaptive plasticity and survival of cancer cells in processes like epithelial to mesenchymal transition, invasion and metastasis has been widely studied [67,68]. In CGC, CD44 was variably expressed in all tumor subtypes. Similar to E-cadherin, the expression of CD44 was enhanced in pleomorphic neoplastic cells and was absent in well-differentiated acinar structures within neoplastic emboli. In the context of these findings, we postulate that CD44 expression could potentially facilitate invasion and expression in neoplastic emboli and, alongside E-cadherin, could lead to the development of cohesive intravascular structures, enhancing their survival. In previous studies in humans, CD44 positive cells in gastric cell lines were associated with increased chemoresistance and invasiveness [66]. Additionally, CD44 positive gastric tumors were associated with larger tumor size, a lower grade of differentiation, tumor relapse, lymph node invasion, distant metastasis and reduced survival [63–66]. It seems that CD44 may represent an important biomarker and a promising therapeutic target in canine gastric carcinomas.

14-3-3σ is a cell cycle regulator that may serve as either a tumor suppressor or an oncogene involved in tissue invasion and metastasis. Histological features of malignancy have been previously associated with the overexpression and/or neo-expression of the protein [69,70]. The role of this protein as a tissue differentiation marker and as an oncoprotein in veterinary medicine has been previously studied in canine mammary, urinary bladder, renal cell and equine penile squamous cell carcinomas [33,34,71,72]. In the current study, immunohistochemical analysis revealed the absence of 14-3-3σ immunolabeling in a normal stomach and in SRC and mucinous carcinomas. The latter finding was consistent for all cases of the two subtypes and likely demonstrates a 14-3-3σ independent molecular pathway in carcinogenesis. In contrast, tubular and undifferentiated subtypes showed strong intracytoplasmic neoexpression, and occasionally nuclear expression of 14-3-3<sup>σ</sup>. The single papillary carcinoma had weak intracytoplasmic neoexpression. Undifferentiated carcinomas demonstrated strong intracytoplasmic, and occasionally nuclear, expression in the most pleomorphic neoplastic cells. Similarly, in those tumors with neoplastic emboli formed by pleomorphic cells forming solid nests, there was strong intracytoplasmic and occasionally nuclear expression of 14-3-3<sup>σ</sup>. However, acinar structures within the intravascular emboli lacked expression of 14-3-3<sup>σ</sup>. Aberrant nuclear expression of 14-3-3<sup>σ</sup>, in a subset of cases of renal cell carcinomas, was associated with a malignant phenotype and shorter survival rate [43]. Thus, we hypothesize, in view of these findings, that expression of 14-3-3σ is associated with features of malignancy, and that neoexpression of the molecule is essential in the stage of intravascular invasion. Expression of the protein in the extracellular milieu, as previously demonstrated for other canine carcinomas, was not observed in this study [72].

The protein p16 acts as a tumor suppressor and a cell cycle regulator, by slowing progression from the G1 to S phase through the inhibition of cyclinD-CDK4/6. Its role and possible relationship to tumor progression and prognosis have been studied in a variety of human tumors [73–75]. In human gastric carcinomas, loss of p16 expression has been associated with malignant characteristics and poor prognosis [28,76,77]. In the CGCs studied here, the expression of p16 seemed to be an event common to all subtypes, and thus appears fundamental for neoplasia development, with expression most strong in mucinous and SRC carcinomas. Half of the tumors showed cytoplasmic, and 36.7% both nuclear and cytoplasmic, immunolabeling. Within neoplastic emboli, expression was cytoplasmic and only in one tubular carcinoma case was both cytoplasmic and nuclear. Similar to the other markers, when acinar structures were present within the emboli, p16 expression was absent. These results contradict a previous canine study, where lack of p16 expression in tubular, SRC and undifferenced subtypes was significantly associated with histological criteria of malignancy in 14 cases [29]. In humans, the overexpression of p16 is associated with mutations in genes encoding retinoblastoma protein (Rb) and p53, and is considered to be a mechanism to arrest the uncontrolled proliferation caused by failure of the Rb pathway [78]. Furthermore, the significance of the p16 expression within the cell is not clear, and few studies have associated cytoplasmic expression with malignant

features [79]. Currently, similar to humans, the significance of p16 expression patterns in CGC remains unclear [78].
