Under normal conditions, in the absence of neoplasia, major and minor salivary glands express nerve markers such as glial fibrillary acidic protein (GFAP, one of the intermediate filament proteins), S100 protein (S100, a well-known marker for glial and Schwann cells), and neuron-specific enolase (NSE) [1-6]. These neural markers, as well as protein markers for smooth muscle (e.g., alpha-smooth muscle actin [α-SMA]) are usually used for classification of salivary gland tumors [7-9]. Another marker for myoepithelial cells, p63 (a homologue of the tumor suppressor p53), which is also a well-known marker for Müllerian duct derivatives in the urogenital organs [10-13], has recently been used for classification of salivary gland tumors [7, 14]. Thus, immunoreactivity for p63 does not indicate origin in a specific cell. In addition, cytokeratins, a group of intermediate filament proteins, are used as markers for salivary gland epithelium. Cytokeratin19 (CK19) is commonly used as a marker for the neuroendocrine and gastrointestinal tumors [15]. Cytokeratin14 (CK14) is also a marker for oral epithelial cells [16, 17] and is considered a marker for maturation of the epithelial lining [17, 18].

Although these markers have demonstrated differential expression patterns in myoepithelial cells of various salivary glands, to the best our knowledge, criteria to distinguish between specific glands on the basis of immunohistochemistry are yet to be established. Moreover, previously reported data on expression of neuronal markers in fetal salivary glands are inconsistent; for example, Gustafsson et al. [1] and Lee et al. [2] stated that fetal parotid glands expressed GFAP, whereas Chi [19] and Ianez et al. [14] found no evidence for GFAP immunostaining in these glands at a similar gestational age. Therefore, in this study, we aimed to comprehensively examine the immunohistochemical reactivity toward the described markers not only in the major parotid and submandibular glands, but also in the minor glands such as pharyngeal gland near the palatine tonsil, the lingual gland at the dorsum of the tongue, and the palatal gland by using specimens from cadavers of elderly persons. For comparison, we also examined fetal parotid and submandibular glands (obtained from fetuses at 15-16 weeks of gestation) because according to Chi [19] and Ianez et al. [14], at this stage, myoepithelial cells of major salivary glands stain positive for α-SMA, but negative for S100.

The study was performed in accordance with the provisions of the Declaration of Helsinki 1995 (as revised in Edinburgh 2000). We examined 14 donated cadavers (8 men and 6 women; age, 78-92 years; mean age, 88 years). Causes of death were ischemic heart failure or intracranial bleeding. These cadavers were donated to the Tokyo Dental College for research on human anatomy, and their use in this study was approved by the university ethics committee. The donated cadavers had been fixed by arterial perfusion of 10% v/v formalin solution and stored in 50% v/v ethanol solution for more than 3 months. From 1 cadaveric head, we prepared 4 tissue blocks, each of which included 1) the tonsillar fossa and the dorsal part of the tongue, 2) the soft palate and opening of the pharyngotympanic tube, 3) the superficial part of the parotid gland, or 4) the superficial part of the submandibular gland. After performing routine procedures for preparing paraffin-embedded histological specimens, semiserial sections were cut at 0.2-mm intervals. One of every 10 sections was stained with hematoxylin and eosin (H&E) to screen for sections suitable for gland immunohistochemistry (see below).

We also examined the histology of 5 paraffin-embedded fetuses at an estimated gestational age of 15-16 weeks (crown-rump length, or CRL, 105-125 mm). By approbation of the families concerned, these specimens were donated to the Department of Anatomy, Chonbuk National University, Korea, and their use in this study was approved by the university ethics committee. According to the regulations, authors other than those affiliated to the Chonbuk University were not required to inform the corresponding committee in Japan about this research project. The fetuses were obtained after induced abortions, and each of the mothers were personally informed by an obstetrician about the possibility of donating the fetus for research without any attempt to encourage donation. Because of the randomized numbering of specimens, tracing the families concerned was not possible. The donated fetuses were fixed in 10% w/w neutral formalin solution for more than 3 months. After division into the head and neck, thorax, abdomen, pelvis, and the 4 extremities, all parts were decalcified by incubating in 0.5 mol/l ethylenediaminetetraacetic acid solution (pH 7.5; Decalcifying Solution B, Wako, Tokyo, Japan) at 4℃ for 1-3 days, depending on the size of the specimen. Using a routine procedure, we obtained paraffin-embedded histological sections 5 µm thick. All head specimens were processed into sagittal (3 specimens) or horizontal (2 specimens) sections at intervals of 50 µm. Most of the sections were stained with H&E, whereas some were used for immunohistochemistry (see below).

The primary antibodies used for immunohistochemistry were 1) rabbit polyclonal anti-human GFAP (1:100, #Z0334, Dako Cytomation, Kyoto, Japan), 2) mouse monoclonal anti-human α-SMA (1:100, #M0760, Dako, Glostrup, Denmark), 3) mouse monoclonal anti-human S100 (1:100, #Z0311, Dako), 3) mouse monoclonal [BC4A4] anti-human p63 (1:100, #ab735, Abcam, Cambridge, UK), 4) mouse monoclonal anti-human CK14 (1:50, #LL002, Novo, Newcastle upon Tyne, UK), and 5) mouse monoclonal anti-human CK19 (1:100, #sc-6278, Santa Cruz Biotechnology, Santa Cruz, CA, USA). With the exception of the samples used for detection of S100 and SMA, antigen retrieval was performed using microwave treatment (500 W, 15 minutes, pH 6). Sections were incubated with a secondary horseradish peroxidase (HRP)-labeled antibody (1:1,000, Histofine Simple Stain Max-PO, Nichirei, Tokyo, Japan) for 30 minutes, and immunoreactivity was detected using an HRP-catalyzed reaction with diaminobenzidine (Histofine Simple Stain DAB, Nichirei) for 3-5 minutes. All samples were counterstained with hematoxylin. Negative controls consisted of samples processed without incubation with primary antibodies. Adult specimens were examined using all 6 markers, but only 3 of these markers (α-SMA, GFAP, and S100) were tested in the fetal specimens because difference in reactivity for these markers was evident between the major and minor adult glands. In addition, the Dako anti-α-SMA antibody has been shown to react strongly with the endothelium of blood vessels [20].

In this study, we investigated 2 major and 3 minor adult salivary glands; therefore, our results appeared to be more comprehensive than those of previous studies. However, we were unable to include observations of the adult sublingual gland. We observed the specific morphology of the fetal sublingual gland after complete preparation of adult specimens. Possibly because of age-related degeneration, adult tissues obtained from cadavers of elderly individuals have not been studied extensively. However, the present results seemed to be consistent with those of Gustafsson et al. [1], who reported co-expression of GFAP and CK in some myoepithelial cells in parotid and palatal glands, but not in submandibular glands. In regard to S100 and GFAP expression in the major glands, our results were mostly consistent with those of Nakazato et al. [4], Murakami et al. [3], and Okura et al. [5], who observed uniform negative staining for these markers in clinical materials obtained from patients of different age groups. We demonstrated the differences in immunoreactivity among various types of salivary glands; had adult sublingual glands also been examined, the difference would have been even more distinct. Heterogeneity of salivary glands is poorly understood. Lee et al. [2] described the details of stage-dependent changes in immunoreactivity toward S100, GFAP, and NSE by using 100 human fetal salivary glands; however, they did not discriminate among the gland types, and therefore, their results probably reflect type- and not stage-specific differences.

Many previous studies addressed an important question of whether salivary gland cells, especially myoepithelial cells, express S100. In his excellent immunohistochemical study of the major salivary glands, Chi [19] observed that S100 reactivity was limited to fetuses at 19-40 weeks of gestation and was absent in adults. Likewise, Mori et al. [22] and Dardick et al. [23] reported that myoepithelial cells of adult major salivary glands were negative for S100. The occasional positive reaction for S100 in myoepithelium was attributed to staining of unmyelinated nerve endings, and unmyelinated nerve endings were suggested to make direct contact with myoepithelial and acinar cells [24]. In fact, in the present study, we did not observe S100-positive glandular cells in the major glands, although these were specifically evident in fetal submandibular glands abundant the thin nerves that encircled each acinus. In the present study, we demonstrated the presence of S100-positive glandular cells in the minor salivary glands in both adults and fetuses. Accordingly, we believe that some myoepithelial cells may express S100 in salivary glands, which is consistent with the statement that S100 is expressed in the intercalated duct and myoepithelial cells of the major salivary glands [25].

The most striking finding of the present study is a distinct difference between the major (parotid and submandibular) and the minor (pharyngeal, lingual, and palatal) salivary glands in their immunoreactivity toward the nerve markers GFAP and S100. Observations in fetuses suggested that adult sublingual glands were likely to show immunoreactivity similar to that of the minor glands, unlike the nearby submandibular glands. In fetuses, the mesenchymal tissue around the sublingual gland was continuous with that around the submandibular gland. Thus, the epithelial-mesenchymal interaction did not seem critical for the expression of nerve markers. However, given the rich nerve supply in fetuses, nerves were likely to induce specific morphology in the submandibular gland. Were there neural crest remnants persisting in myoepithelial cells of the adult minor glands? It is necessary to bear in mind that expression of GFAP and S100 is not strictly nerve-specific and can also occur in the cartilage in both fetuses and adults [26, 27]. Thus, the S100 or GFAP expression does not always indicate a neural origin. Because fetal major glands were negative for α-SMA, the fetal myoepithelial cells did not seem to have reached sufficiently differentiated stage for contraction in contrast to the pharyngeal small gland. Therefore, expression of these nerve markers does not depend on the stage of differentiation, but rather on other factors. If immunoreactivity for nerve marker proteins depends on the physiological status of the gland, such as age-related degeneration, then the mosaic expression pattern we observed may be thought to reflect the functional differences among adjacent glandular cells. However, such differences are unlikely to occur in a single acinus. Finally, the most basic observation in this study-the small glands expressed nerve markers, whereas the large glands did not-should also be noted. Therefore, immunoexpression may be determined by the size of the gland or the degree of branching in the ductal system. Notably, immunohistochemistry of the branching morphogenesis has been investigated in the small salivary glands, but not in the major glands [28].