Asymptomatic oral carriage of Candida albicans in patients with HIV infection.

OBJECTIVE: To assess the relationship of asymptomatic carriage of Candida albicans and clinically apparent thrush in patients with HIV infection. DESIGN: Prospective, longitudinal, controlled study. SETTING: The HIV clinic at St. Michael's Hospital, University of Toronto. PARTICIPANTS: One hundred and twenty-seven patients with HIV-infection were divided into 3 groups according to the CD4+ lymphocyte count, and 37 healthy volunteers served as controls. INTERVENTIONS: Determination of blood type, baseline CD4+ lymphocyte count in patients with HIV infection, and immunophenotyping. Samples of saliva (2 mL) were obtained from each patient and control. MAIN OUTCOME MEASURES: Carrier status, clinical presence of thrush, the association between carriage of C. albicans and blood type, secretor status and history of oral infection. RESULTS: In patients with HIV infection and C albicans colonization no correlation was found with blood type or secretor status of blood group antigen in the saliva. The frequency of oral carriage of yeast was greater in patients infected with HIV than in controls, but the difference was not significant for asymptomatic subjects with a CD4+ lymphocyte count greater than 500/microL. Persistent carriage of yeast and development of clinical thrush were associated with lower CD4+ counts. Clinical thrush developed only in patients with persistent asymptomatic carriage of C. albicans and CD4+ counts less than 500/microL. CONCLUSION: The greater risk of oral colonization with C. albicans in patients with HIV infection partly explains the high prevalence of thrush found in this group.

Squamous cell carcinomas of the head and neck cultured in floating collagen gels: 1. The maintenance of stromal and epithelial elements in vitro without fibroblast overgrowth.

The study of the molecular biology of head and neck squamous cell carcinomas has been heavily reliant on the analysis of cell lines. This is largely because the maintenance of primary cell cultures is difficult. However, being monoclonal, cell lines are not representative of the primary tumor because of the loss of tumor cell heterogeneity. We report a technique for primary culture of squamous cell carcinomas with maintenance of epithelial and stromal cell components without overgrowth of the fibroblast cells. Phenotypic markers for fibroblasts and squamous cells were present up to 45 days after initiation of culture, and expression of epidermal growth factor receptor and involucrin in cultures paralleled that in the primary tumor. In vivo, tumor stromal elements are thought to play an important role in the support of epithelial cell growth. In the collagen gel system the preservation of the stromal cell component likely improves culture viability and growth. More importantly, this culture system allows the in vitro tumor to more accurately reflect the tumor from which it was derived, and it permits the study of primary squamous cell carcinomas under in vitro conditions.

Myoepithelial cells actively proliferate during atrophy of rat parotid gland.

Despite limited supporting evidence, salivary gland myoepithelial cells are said to be differentiated cells with little or no capacity to replicate; they presumably develop from stem cells. This study investigated the proliferative potential of myoepithelial cells with an antibody to proliferating cell nuclear antigen and a rat model. This model involved clamping of the parotid duct causing atrophy of the gland and then releasing the duct followed by gland regeneration. Rats were sacrificed at time points during atrophy and regeneration phases and the number and location of cycling myoepithelial cells assessed. Cycling myoepithelial cells were identified with double immunohistochemical staining, cycling cells with proliferating cell nuclear antigen-positive nuclei within muscle-specific actin-positive cytoplasm (the latter identified with antibody HHF35). The results show that baseline proliferative rates of myoepithelial cells in both the resting and fully regenerated gland ranged from 0.3% to 2%, similar to rates for other major cell types in the normal rat gland. A peak myoepithelial cell proliferative rate of 23% occurred at day 5 during the atrophy phase. Rates during the regenerative phase were not significantly different than the baseline levels. Similarities of rat and human parotid gland and the definite proliferative capacity of myoepithelial cells indicates that these specialized cells must be considered one of the potential progenitor cells for human salivary gland tumors.

Differentiating myoepithelial and acinar cells in rat neonatal parotid gland and histogenetic concepts for salivary gland tumors.

Histogenetic concepts for salivary gland tumors are predicated on the presence of reserve or undifferentiated cells in normal glands, presumably the source for cell renewal and induction of tumors. Developing rat parotid gland, which remains fetal-like at birth, provides the opportunity to study differentiation and observe whether cytologically undifferentiated cells do or do not have functional indicators of specific differentiation pathways. Immunohistochemistry and immuno-electron microscopy, when applied to parotid gland at birth, at 12 days of age and in the adult gland, indicate that commitment to myoepithelial cell differentiation occurs prior to development of structural changes characteristic of these cells. Conversely, secretory granules are evident in differentiating acinar cells prior to synthesis of amylase. The results suggest that an appearance of undifferentiation does not confer reserve cell status either in the normal salivary gland or their tumors.

Postirradiated submandibular gland: a potential model to study salivary gland radioprotection and tumourigenesis.

Theoretical reserve cells located in the intercalated and excretory ducts are postulated to be responsible for salivary gland tumourigenesis, with acinar cells playing no role in this process. Animal models, one using low-dose radiation to rat submandibular glands, indicate that this hypothesis is incorrect. Few human models have been devised to demonstrate and verify this theory. Submandibular glands in the field of ionizing radiation, as external-beam radiotherapy for head and neck tumours, were examined using an immunocytochemical technique and an antibody to proliferating cell nuclear antigen (PCNA), a specific marker for cycling cells. In the nonirradiated gland, nuclei positive for PCNA were seen in acinar as well as ductal cells of all types. Six months post irradiation, human submandibular glands show increased proliferative rates in both ductal and acinar cells that are significantly greater than control glands (p = .012). Based on this regenerative capacity, postirradiated human submandibular glands might serve as a model to investigate various treatment modalities for the prevention of radiation damage to acinar cells and the consequent patient morbidity that develops due to xerostomia. As well, these results suggest that even in humans, acinar cells are potential targets for carcinogenic agents and that current histogenic concepts for salivary gland tumourigenesis are incorrect.

Diagnostic criteria for neoplastic myoepithelial cells in pleomorphic adenomas and myoepitheliomas. Immunocytochemical detection of muscle-specific actin, cytokeratin 14, vimentin, and glial fibrillary acidic protein.

OBJECTIVE: Markers for normal salivary gland myoepithelium were used to determine the extent of their expression in the neoplastic myoepithelial (nonluminal) cells of pleomorphic adenomas and then in the tumor cells in myoepitheliomas and to gather information necessary to establish diagnostic criteria, especially muscle actin expression, for myoepitheliomas. STUDY DESIGN: Methanol/acetic acid-fixed and paraffin-embedded tissue was used to immunohistochemically study expression of intermediate and smooth-muscle actin filaments in nonluminal cells in 14 pleomorphic adenomas and to compare this to their expression in five myoepitheliomas. RESULTS: In routine histologic sections, the morphologic variants of nonluminal tumor cells--spindle, stellate, polygonal, angular, and plasmacytoid--in pleomorphic adenoma mirror the spectrum of tumor cells in myoepitheliomas. Immunocytochemical similarities are also apparent. Two specific markers for myoepithelial cells in the normal salivary gland, muscle-specific actin and cytokeratin 14, were both variably, independently, and never uniformly expressed in nonluminal cells of pleomorphic adenoma and tumor cells in myoepitheliomas regardless of their morphology. Cytokeratin 14 in addition labels basal cells of excretory ducts. Both muscle-specific actin and cytokeratin 14 preferentially localized to single layers of periductal cells in pleomorphic adenomas, angular, polygonal, and plasmacytoid cells preferentially expressed cytokeratin 14. Similar patterns were noted in the three myoepitheliomas with reasonable expression of the two markers. Only isolated single cells or small groups of plasmacytoid cells in four pleomorphic adenomas with a significant component of these cells and the two plasmacytoid myoepitheliomas immunostained for muscle-specific actin and cytokeratin 14. In both tumor types, vimentin was nearly uniformly expressed in nonluminal tumor cells of all morphologic types, including plasmacytoid cells. CONCLUSIONS: The range and transition of morphology of nonluminal cells in pleomorphic adenomas is reflected in myoepitheliomas. Incomplete or absent expression of the myoepithelial/basal cell markers, muscle-specific actin, and cytokeratin 14, and the general expression of vimentin is common to both tumors. Because these findings apply to the majority of plasmacytoid cells in pleomorphic adenomas, tumor cells with a similar morphology and immunoprofile are to be expected in myoepitheliomas; the term plasmacytoid myoepitheliomas is thus appropriate regardless of the presence or absence of muscle-specific actin.

Myofilament localization and immunoelectron microscopic detection of muscle-specific actin in neoplastic myoepithelial cells in pleomorphic adenomas and myoepitheliomas.

Elucidating the cellular characteristics of the nonluminal or myoepithelial cells of pleomorphic adenomas is one approach to establishing the diagnostic criteria for myoepitheliomas. Ultrastructural features of nonluminal tumor cells in 22 pleomorphic adenomas and of tumor cells in 9 myoepitheliomas were assessed from micrographs of routinely fixed and epoxy resin-embedded samples. Recognizable myofilaments were only moderately prominent in 1 myoepithelioma. In the rest of the cases, irrespective of whether nonluminal cells of pleomorphic adenomas or tumor cells of myoepitheliomas were spindle, angular, round, or plasmacytoid in form, myofilaments were noted only in one third of the cases and were present even in these in a small proportion of the tumor cells. Intermediate filament accumulations and basal lamina were more frequent findings associated with nonluminal tumor cells. Six pleomorphic adenomas and 2 myoepitheliomas had been fixed in half-strength glutaraldehyde and embedded in LR White resin for immunoelectron microscopic detection of muscle-specific actin. In 3 (2 pleomorphic adenomas and myoepitheliomas) of these 8 cases, readily visualized bands of filaments in many tumor cells were strongly labeled by the colloidal gold probe detecting muscle-specific actin even when myofilaments were minimal and infrequent in 2 cases and undetectable in the third by routine transmission electron microscopy. Lack of myofilament detection by immunocytochemistry or routine electron microscopy does not exclude a diagnosis of pleomorphic adenoma or myoepithelioma when growth patterns and cytology indicate such diagnoses. Immunoelectron microscopy, in fact, shows that muscle-specific actin can be detected even when myofilaments or muscle actin are apparently absent or minimal by routine electron microscopy or immunohistochemistry, respectively. Because examples of pleomorphic adenoma and myoepithelioma each with similar histologic and cytologic features of the myoepitheliomatous cells can have variable degrees or complete absence of expression of myofilaments or muscle-specific actin, the time-honored term myoepithelial for the nonluminal cells of pleomorphic adenomas and the term myoepithelioma are legitimate even in the absence of those markers that are specific for normal myoepithelial cells.

Detection of proliferating cell nuclear antigen in paraffin-embedded specimens is dependent on preembedding tissue handling and fixation.

Immunohistochemical detection of proliferating cell nuclear antigen (PCNA), a cell cycle-related protein used to estimate tumor growth fraction, is variable in formalin-fixed compared with methanol-fixed tissue specimens. This is assumed to result from conformational changes in the antigenic epitope induced by formaldehyde; therefore, to be susceptible to retrieval in archival specimens. In this study, formalin fixation reduced the intensity of staining and the number of positive cells to approximately 25% of those in methanol-fixed material. The washing of tissue specimens prior to methacarn fixation also reduced PCNA staining. Loss of staining was not restored after use of a commercial retrieval kit recommended for PCNA immunohistochemistry. Immunoblotting of formalin fixatives and saline washings after removal of tissue specimens consistently demonstrated the presence of PCNA-like activity in solution. We conclude that the exceptional solubility of PCNA is responsible for reduced immunostaining in formalin-fixed material, that the loss is irreversible, and that methanol or methacarn is the fixative of choice for PCNA immunohistochemistry.

Collagen gel cultures of normal salivary gland: conditions for continued proliferation and maintenance of major cell phenotypes in vitro.

The presence of three functionally and phenotypically distinct epithelial cell populations--acinar, duct, and myoepithelial cells--in major salivary glands creates problems when developing physiologically appropriate culture systems for the study of these tissues in vitro. Previous attempts to establish cultures of rat submandibular gland resulted in continued proliferation and maintenance of glandular architecture, but loss of distinct features of differentiation of the three epithelial cell types. The present study describes an ultrathin free-floating collagen gel culture technique (mantle gels). Using this method, immunohistochemical and ultrastructural studies indicate that rat submandibular gland continues to cycle, and secretory activity and phenotypic markers for acinar, duct, and myoepithelial cells are all demonstrable after 4 weeks in culture, suggesting that this constitutes the ideal system for in vitro investigation of the pathobiology of the salivary gland.

Obstructive sialadenitis: a rat model.

Obstructive adenitis of major salivary glands is a common entity. Although the pathologic features are well recognized, the various cell types involved in the atrophy and subsequent regeneration of the obstructed salivary gland have been controversial. For this reason, an animal model of obstructive sialadenitis that induced atrophy in the rat parotid gland was developed. A clamp was devised that would occlude the main excretory duct of the gland inducing atrophy, but that, following removal, left the duct patent and allowed regeneration to occur. Both atrophy and regeneration were studied, and the results presented. The number and location of cycling acinar, intercalated, striated, and excretory duct cells were identified, using an antibody to proliferating cell nuclear antigen (PCNA). All cell types were induced to cycle during both the atrophy and regeneration phase, but the degree of cycling was more pronounced during regeneration. Three days after release of the duct obstruction, cycling acinar cells had increased 40-fold, while striated and intercalated ducts increased 6.5- and 5.0-fold, respectively. More importantly, peak cycling of acinar cells preceded, but paralleled, the increase in acinar cell composition of the gland. This model of obstructive sialadenitis indicates that all cell types are responsible for regeneration in the obstructed rat parotid gland.

The pathobiology of salivary gland. III. PCNA-localization of cycling cells induced in rat submandibular gland by low-dose x-radiation.

Application of ionizing radiation to adult rat major salivary glands tested tenets of the bicellular reserve cell hypothesis for the induction of salivary gland tumors, namely, that stem cells preferentially located to luminal cells of the intercalated duct and basal cells of the excretory duct in normal salivary glands. The effect of a single, low dose (3000 cGy) of x-radiation administered to the parotid and submandibular glands was followed with the use of immunocytochemistry and an antibody to the cell cycle-related protein proliferating cell nuclear antigen to detect the kinetics and localization of cycling cells up to 15 days postirradiation. Maximal responses occurred in acinar cells (12.6-fold increase) of submandibular glands on day 7 postirradiation. Similar but less dramatic concurrent increases in proliferating cells were evident in intercalated (3.4-fold) and striated (2.2-fold) duct cells, but little response was seen in basal or luminal cells of submandibular gland excretory ducts. A limited but maximal proliferative response again occurred on day 7 in the parotid gland. Neither in the steady state nor irradiated submandibular gland was there evidence of specific stem ("reserve") cells associated with the intercalated or excretory ducts. It appears unnecessary to invoke stem cells in a model of cellular proliferation in salivary glands. Therefore current concepts of salivary gland tumorigenesis require modification because all cell types, including acinar cells, are at risk in the carcinogenic process.

Pathology of the salivary glands: the contribution of electron microscopy.

Electron microscopy has a limited role in the diagnosis of primary salivary gland tumors, although it can be helpful in metastatic lesions of possible salivary gland origin. The diversity of subtypes in salivary gland tumors, as well as the range of histomorphology within any one subtype, is unparalleled in any other human tumor. This and their relative infrequency causes diagnostic problems for pathologists. Ultrastructural techniques have been of major importance in determining the inter-relationship of these tumors for classification purposes, revealing the subtle variations in common cellular differentiation pathways, determining the organization of tumor cells, and displaying the importance of extracellular matrix materials in establishing diagnostic criteria for each of the many subtypes. Electron microscopy has also been valuable in non-neoplastic salivary gland disease and has an increasing role in experimental studies involving tissue from human and animal salivary parenchyma.

Immunohistochemical analysis of the proliferative capacity of duct and acinar cells during ligation-induced atrophy and subsequent regeneration of rat parotid gland.

To study the proliferative capacity of salivary gland, an animal model of regeneration was developed. A clamp, which induced atrophy in parotid gland by obstructing the main excretory duct but allowed restoration of duct patency following removal, was implanted in a series of rats. When it was removed (Day 7), the weight of the glands was reduced by 50% and acinar cells had decreased from 93.8% to 8.2% of total cell population. Regeneration occurred rapidly following removal of the clamp. The number and location of cycling intercalated, striated, and excretory duct cells and acinar cells were monitored using an antibody to proliferating cell nuclear antigen (PCNA). All cell types were induced to cycle but the predominant cell to cycle was the acinar cell. During regeneration the number of PCNA+ acinar cells increased 38.7-fold from steady-state values. Results demonstrate that acinar cells have a significant potential for cycling, contrary to current histogenetic theories of salivary gland tumourigenesis which exclude acinar cells as potential progenitor cells on the grounds of their putative limited cycling capacity.

Pathobiology of salivary glands. IV. Histogenetic concepts and cycling cells in human parotid and submandibular glands cultured in floating collagen gels.

Localization of cells with proliferative capacity in human major salivary glands lacks extensive study. Minced fragments of human parotid (n = 3) and submandibular (n = 3) glands embedded in a floating collagen gel matrix and cultured for up to 28 days allowed maintenance of the three-dimensional relationship of the various cell types in these glands. Immunocytochemistry and electron microscopy of a time-dependent series of cultured gland fragments showed gradual cytologic modification of acinar cells so that acini became duct-like but also established that even after 28 days of culture certain cellular features allowed continued identification of acinar cells. Serial section immunostaining for amylase, cytokeratins, and proliferating cell nuclear antigen (a specific marker for cycling cells) revealed that acinar, intercalated duct, and excretory duct (both basal and luminal) cells are all capable of entering the cell cycle. At day 5 of culture, the number of cycling cells increased 16-fold in the parotid gland and 9-fold in the submandibular gland over that in the respective in situ gland. In this in vitro system, which perhaps simulates regenerative processes in human salivary glands, none of the samples showed cycling cells localized only to segments of intercalated duct or the basal cells of excretory duct as suggested by current histogenetic concepts.

Association between gastrointestinal tract carriage of Candida, blood group O, and nonsecretion of blood group antigens in patients with peptic ulcer.

Of 112 patients with peptic ulcer disease examined for oral carriage of Candida, 66 (59%) were carriers. Candida carriage was associated with blood group O (P < 0.05) and, independently, with nonsecretion of blood group antigens (P < 0.01). For each subject, the presence or absence of yeasts was found to be a constant characteristic, and only among patients positive for Candida was blood group O or nonsecretion more frequent than expected in the general population. The quantity of yeasts isolated was significantly greater in patients than in normal subjects (P < 0.002), as was the frequency of carriage in the patient population (59% vs 32%). This increase was not associated with treatment with H2-receptor antagonists. The results of paired oral and gastroduodenal aspirate cultures suggested that identifying Candida in the oral cavity was a good indicator of the presence of yeasts elsewhere in the gastrointestinal tract. Mechanisms whereby overgrowth of Candida in the upper gastrointestinal tract might contribute to the inflammatory background of peptic ulcer disease are discussed.

Current status of histogenetic and morphogenetic concepts of salivary gland tumorigenesis.

Because of their complexity and relative infrequency, salivary gland tumors commonly result in diagnostic problems. Histogenetic and morphogenetic concepts of tumorigenesis in these glands are reviewed and their relevance to routine diagnosis and classification of salivary gland tumors evaluated. Evidence is presented from animal and human studies that under steady-state and pathophysiological conditions, all cell types present in the normal gland, including acinar cells, are capable of rapidly entering the cell cycle and are, therefore, possible targets for neoplastic transformation.

Immunogold localization of actin and cytokeratin filaments in myoepithelium of human parotid salivary gland.

Myoepithelial cells of salivary gland are uniquely specialized cells; their function is unclear, but the considerable complement of muscle-specific actin suggests contractility is one function. By routine transmission electron microscopy myofilament visualization is variable. Some myoepithelial cells appear to have limited and only focal aggregates of myofilaments, while others seem to have readily appreciated myofilaments within a longitudinally oriented cytoplasmic zone at the basal portion of the cell. However, immunogold electron microscopy using the anti-muscle-specific actin antibody, HHF35, while indicating a basal distribution for the muscle-isoform of actin in a platelike fashion in certain myoepithelial cells, also reveals that others associated with both intercalated ducts and acini have a more generalized distribution of myofilaments throughout the cytoplasm. Actin was also noted within tonofilaments and double immunogold labeling using both the HHF35 and AE1/AE3 (anticytokeratins) antibodies confirmed the variable interrelationship of these two filaments. Within any one myoepithelial cell, actin and cytokeratins might colocalize in some areas of the cytoplasm containing filaments, but not in adjacent zones. These results suggest that intermediate filaments and myofilaments are complexly organized in myoepithelial cells, and that quantitative and qualitative differences exist in the expression and distribution of intermediate filaments and myofilaments. These cells are likely structurally, if not functionally, heterogeneous.

The pathobiology of salivary gland. II. Morphological evaluation of acinic cell carcinomas in the parotid gland of male transgenic (MMTV/v-Ha-ras) mice as a model for human tumours.

In the transgenic TG.SH (mouse mammary tumour virus/v-Ha-ras) mouse, designed to develop mammary tumours, occasional spontaneous salivary gland tumours have been reported, predominantly in males. The incidence and histomorphology of salivary gland tumours in 73 TG.SH mice were surveyed and in total, 21.9% developed both overt and microscopic parotid tumours. The majority developed between 73 and 150 days of age. In 31.5% of the TG.SH mice, occasional unilateral, but more frequently bilateral exophthalmos due to hyperplasia of the intraorbital (Harderian) lacrimal gland was observed. In 70% of these animals, parotid tumours developed later. Since Harderian gland hyperplasia, occurring as early as 5 weeks of age, preceded the development of palpable salivary gland lesions, this stigma is useful for the early selection of animals likely to progress to tumour formation. Before tumour-bearing transgenic mice are considered to be suitable models of human neoplastic disease, morphological characterization is necessary to ensure that the tumours are histologically representative of the human lesions for which they are potential models. In this study, all parotid tumours consisted of acinar-like glandular structures with central lumina discernible by electron microscopy. Ultrastructurally, secretory granules evident in the apical cytoplasm of the tumour cells resembled the zymogen granules of the normal parotid acinar cell, and some cells had a prominent complement of rough endoplasmic reticulum. These features, along with focal amylase expression detected immunohistochemically in some parotid tumours, identified these neoplasms as acinic cell carcinomas that mimic the human salivary gland acinic cell carcinoma faithfully.

Immunohistochemical and ultrastructural correlates of muscle-actin expression in pleomorphic adenomas and myoepitheliomas based on comparison of formalin and methanol fixation.

The degree and range of differentiation of the cells referred to as myoepithelial-like in pleomorphic adenomas and the tumour cells of myoepitheliomas are not definitely established. This type of information is critical for establishing reliable diagnostic criteria, such as expression of muscle-specific actin and ultrastructural identification of myofilaments, in these and other salivary gland tumours. Pleomorphic adenomas (18) and myoepitheliomas (5), of which 10 cases were fixed only in formalin and 13 cases where tissues were fixed in both formalin and methanol/acetic acid, were studied. Each tumour and normal accompanying parotid was immunostained with two monoclonal antibodies for smooth muscle actin, HHF35 and MSA. Staining of myoepithelial cells was absent in certain samples of normal gland with both HHF35 (15%) and MSA (69%) when formalin-fixed tissue was used. Using formalin-fixed tissue from 15 pleomorphic adenomas/myoepitheliomas, 2 (14%) had focal positivity with HHF35, while 8 cases (57%) were positive with MSA. However, a certain degree of false positivity was suspected since in samples of normal parotid, both acinar and duct cells were frequently stained, particularly with MSA. With methanol/acetic acid-fixed tissue only 4 of 13 cases (31%) were positive with either MSA or HHF35 and 2 of these only had a minor proportion of the tumour cells expressing muscle-specific actin. Using alcohol-fixed tissue, myoepithelial cells were strongly stained in all examples of normal parotid gland with both anti-actin antibodies. In 5 cases examined by electron microscopy, there was no apparent correlation between immunohistochemical results and the presence or absence of cytoplasmic filament accumulation. The results indicate considerable tumour cell heterogeneity in muscle-specific actin expression and suggest that non-luminal cells in pleomorphic adenomas and the tumour cells in myoepitheliomas may differentiate as classical myoepithelial cells, as partially differentiated (i.e. modified myoepithelial cells) or as the counterpart of basal cells present in the intra- and interlobular ducts of normal salivary gland.

Cellular differentiation and morphologic heterogeneity in polymorphous low-grade adenocarcinoma of minor salivary gland.

Histologic diversity is intrinsic to salivary gland tumors, but it is a particular feature of polymorphous low-grade adenocarcinoma as denoted by the terminology. Application of immunohistochemistry and electron microscopy to three examples allowed the study of some aspects of tumor cell differentiation in this minor salivary gland lesion. In one case where the tumor cells were cytologically of one type, immunohistochemistry clearly identified both luminal and nonluminal tumor cells, but the latter showed no evidence of myoepithelial cell differentiation. A second case revealed differentiation only of luminal-type tumor cells, while a third example was largely differentiated as myoepithelial cells but again immunohistochemistry confirmed focal formation of duct-like structures by luminal epithelium. These cases show a considerable range of tumor cell heterogeneity as well as variations in their organization. This variation in differentiation characteristics underlies the histology of polymorphous low-grade adenocarcinomas, and likely occurs in other salivary gland tumors. To establish specific and reliable diagnostic criteria for such tumors requires awareness of this neoplastic process. The limited malignant potential and excellent survival of patients with polymorphous low-grade adenocarcinoma is apparently little affected by patterns of differentiation in this particular neoplasm.