ANNALS OF CLINICAL AND LABORATORY SCIENCE, Vol. 14, No. 1 Copyright © 1984, Institute for Clinical Science, Inc.

Immunohistopathology of Papillary Cystadenoma Lymphomatosum (Warthin’s Tumor) CHARLES K. FOULSHAM, II, M.D.,* GEORGE S. JOHNSON, M.D., Ph.D.,ft G. GORDON SNYDER, III, M.D.,*§ ROBERT J. CARPENTER, III, M.D.,* and NELOFAR Q. SHAFI, M.D.fi *Department of Otolaryngology and Facial Plastic Surgery, fDepartment of Laboratory Medicine, University of Connecticut Health Center, Farmington, CT 06032 and fDepartment of Laboratory Medicine, Veterans Administration Medical Center, Newington, CT 06111

ABSTRACT A retrospective study of various benign salivary gland neoplasms was designed to study the frequency distribution and density of immunoglob­ ulins (Igs) located within their various tissue compartments. Buffered- formalin fixed, paraffin embedded sections of Warthin’s tumor, cystade­ noma, oncocytoma, autoimmune thyroiditis, normal salivary gland, and reactive peripheral lymph node were obtained and processed via the per- oxidase-antiperoxidase method following trypsinization to re-expose anti­ genic sites. The subepithelial and parafollicular zones of Warthin’s tumor show an Ig density distribution (IgA IgG) >> IgM, but the relative densities in the germinal center are (IgG = IgM) >> IgA. In contrast, the normal salivary gland displays almost exclusively IgA positive cells with only an occasional IgG or IgM positive cell. In general, the three salivary gland neoplasms considered all display a relatively similar fre­ quency distribution with (IgA 5= IgG) >> IgM and much greater densities of all Igs than encountered in the normal salivary gland; however, the densities of IgA and IgG positive cells are greatest in Warthin’s tumor. A similar frequency distribution of immunocytes in the parafollicular zone of autoimmune thyroiditis and Warthin’s tumor is noted with an increased density of all Igs occurring in autoimmune thyroiditis. A significant in­ crease in the density of IgA immunocytes is noted in the germinal centers of autoimmune thyroiditis. A comparison of Warthin’s tumor with a reac­ tive peripheral lymph node shows a marked increase in the density and

§ Send reprint requests to G. Gordon Snyder, III, Surgery, University of Connecticut Health Center, M.D., Division of Otolaryngology and Facial Plastic Farmington, CT 06032. 47 0091-7370/84/0100-0047 $02.00 © Institute for Clinical Science, Inc. 48 FOULSHAM, JOHNSON, SNYDER, CARPENTER, AND SHAFI frequency of IgA in the parafollicular zone in the former lesion, and sig­ nificant elevations of IgG and IgM in the latter process both in the para­ follicular zone and in the germinal centers. No data generated by this study support an autoimmune etiology of Warthin’s tumor. Histopathologically, the oncocytoma appears to repre­ sent a pathologic entity distinctly different from papillary cystadenoma lymphomatosum. Our data tend to support the hypothesis that Warthin’s tumor arises within ectopic salivary gland elements trapped within para- parotid or intraparaotid lymph nodes during embryogenesis. The cystad­ enoma appears to represent a similar pathologic process arising in salivary gland elements independent of lymphoid tissue.

Introduction arising in such ectopic locations as the tonsillar fossa, maxillary sinus, larynx, Papillary cystadenoma lymphoma­ neck, and lacrimal glands.4,7,17,19,24,28 This tosum (PCL), better known as Warthin’s is the only parotid tumor known to occur tumor, was first described in 1910 by Al­ bilaterally and is frequently multicen- brecht and Arzt1 and further character­ tric.6,18 ized by Warthin34 in 1929. Since then it Papillary cystadenoma lymphoma­ has appeared frequently in descriptions tosum has no distinguishing features of ectopic salivary gland tissue and in the which clinically differentiate this benign analysis of its histologic components. lesion from a malignant lesion. There is Warthin’s tumor is classified as an ad- no evidence of metastasis or local inva­ enolymphoma, a subtype of the salivary sion and the facial nerve is unaffected. gland monomorphic adenoma.32 Al­ Sialography may show displacement of though initially considered a rare lesion the ductal system by a mass lesion within it now accounts for about six percent of or outside of the parotid gland, but this, all parotid tumors and three or four too, is not diagnostic. Diagnosis may be percent of minor salivary gland tu­ made preoperatively via a Technetium- mors.20,28,32 99 scan which will show increased uptake Clinically there is a preponderance of in those lesions with a preponderance of Caucasian males (3:1) in the fifth to sev­ oncocytes — namely oncocytomas and enth decade. The tumor shows a definite Warthin’s tumor.7,24 predilection for the parotid gland, pre­ senting as an asymptomatic, slow grow­ ing, nontender mass which is round or P a t h o lo g y oval, firm to cystic in consistency and is On gross examination, Warthin’s tu­ usually present for six months to sev­ mor is well encapsulated and sharply cir­ eral years prior to medical evalua­ cumscribed. Its cut surface is grayish tan tion 6>16.20.28.32 to pink in color and studded with whitish Although this lesion is frequently con­ areas representing germinal centers. tained within the superficial lobe of the Multiple cystic spaces formed by mul­ parotid gland, it has also been found to tiple papillary epithelial projections are lie outside the main substance of the usually present. These cystic spaces con­ gland. Warthin’s tumor is also known to tain a clear brownish mucoid mate­ arise in the submaxillary gland and minor rial 7,16,20,24,28 salivary glands of the palate. Scattered On histologic examination multiple ep­ reports have also documented this lesion ithelial infoldings are seen composed of WARTHIN’S TUMOR 49 two oxyphilic layers of cells, an outer The parotids, however, are encapuslated layer of tall columnar cells, and a basal later allowing an admixture of both epi­ layer of small irregular cells. The co­ thelial and mesenchymal tissues. In fre­ lumnar cells have hyperchromatic nuclei quent fetal and adult specimens, salivary and granular cytoplasm containing a ducts and acinar tissue can be found large number of mitochondria. The basal within both intraparotid and paraparotid cells have a vesicular oval nucleus and lymph nodes.3,4’5’6’71719 Salivary gland dense cytoplasm similar to oncocytes. tissue can occasionally be found in cer­ The cysts contain an eosinophilic pro- vical lymph nodes, sometimes at a dis­ teinaceous material with inclusions of tance from the parotid.4 cellular debris.2,7,18,20,24,25 The subepithe- Neoplastic proliferation of the epithe­ lial stroma is composed of lymphoid lial elements within the lymph node and tissue with numerous lymphoid follicles a resulting hyperplasia of the lymphatic which exhibit unremarkable germinal elements ultimately leads to clinical ev­ centers and ill-defined mantle regions. idence of a mass lesion.6,24 Dietert16 re­ Associated inflammatory response may ported a six percent incidence of War­ lead to infiltration by neutrophils and thin’s tumors found incidentally as part eosinophils. The stroma is dotted with of radical neck dissection specimens and large lymphoid cells consistent with stem Azyopardi and Hou3 reported a case of cells or large lymphocytes;17,24’29 plasma multiple Warthin’s tumors, one of which cells and histiocytes are abundant. The occurred within a lymph node and dem­ heterogenous nature of this cellular infil­ onstrated a transition from parotid ductal trate is similar to that found in Hashi- inclusions to an adenolymphoma. It is moto’s thyroiditis and other hypersensi­ felt by many authors that this embryo- tivity diseases. Occasionally lymphocytes logical rest theory adequately explains and histiocytes can be seen infiltrating the multicentricity and bilaterality of the epithelial lining as well as within the Warthin’s tumor. However, the quantity lumina of the cystic spaces.2 of lymphoid stroma frequently encoun­ tered is greatly in excess of that in a small O r ig in lymph node from which it presumably arises. The distribution, i.e., the lym­ The origin of Warthin’s tumor has long phoid follicles located at the tips of the been disputed. Currently two theories of papillary projections, an unlikely location origin have gained support: (1) hetero­ for preexisting follicles, also suggests that topic salivary gland tissue entrapped the stromal proliferation is in response to within lymph nodes; and (2) autoimmune the neoplastic process.32 response. Albrecht and Arzt suggested In 1971, Allegra2 proposed another or­ that proliferation of heterotrophic nests igin after his study of the immunoflu- of pharyngeal endoderm, the anlage of orescent characteristics of this neoplasm. salivary gland tissue, within parotid He noted that the morphologic charac­ lymph nodes was a likely source.1 Since teristics of the lymphoid stroma and the their original description, many authors distribution of immunoglobulin (IgG) as have agreed with this hypothesis. During demonstrated by immunofluorescence embryogenesis the anlage of the major resembled that found in delayed hyper­ salivary glands is surrounded by a lym­ sensitivity disease instead of normal or phoid stroma. The sublingual and sub- reactive lymph nodes. It was proposed maxillary glands are encapsulated from that oncocytomas, papillary cystade- this anlage early in embryogenesis and nomas and papillary cystadenoma lym- are thus devoid of lymphoid elements. phomatosum occupy a continuum of a 50 FOULSHAM, JOHNSON, SNYDER, CARPENTER, AND SHAFI specific pathologic process.17 Oxyphilic has many disadvantages.31 It requires metaplasia of the ductal epithelium oc­ fresh or specially processed tissue, pro­ curs with an associated increase in mi­ vides poor morphologic detail, and ex­ tochondria. Increasing cell volume and hibits interference with nonspecific proliferation of the epithelium leads to tissue autofluorescence. In addition, the characteristic papillary infolding and specialized microscopy is required and cyst formation followed by an accumula­ the resulting slide preparations are not tion of intraductal and intracystic secre­ permanent. The immunoperoxidase tions. Phagocytosis of the secretion com­ technique avoids these problems. The plex initiates an immune reponse which immunoperoxidase method uses the generates the lymphoid stroma.2 Fine reaction of horseradish peroxidase, con­ structural analysis of the secretion com­ jugated to an antibody, which is then im- plex has revealed it to contain degener­ munologically bound to the antigen in ating organelles. Electron microscopy question. The horseradish peroxidase is has shown a large number of lysosomes then enzymatically reacted with a chro- in the columnar epithelium closely as­ mogenic substrate such as diaminoben- sociated with the degradation of cellular zidine. The reaction product is visualized organelles in degenerating areas of epi­ at the tissue level as a brownish-black thelium. The apocrine secretion contrib­ precipitate.13,31 uting to the mucoid cystic contents has Paraffin sections react unpredictably, been observed to be intracellularly lysed however, with the peroxidase-antiperox- portions of oncocytic cytoplasm origi­ idase (PAP) procedure and are often vir­ nating from the epithelium.25 Absorption tually negative. Enzyme digestion was of this secretion complex may initiate and instituted to improve the reliability of sustain the immune response. the staining. Trypsin digestion was found Allegra’s immunofluorescent studies to decrease the nonspecific background demonstrated the presence of IgG in the activity and also unmask the immuno- secretion complex, within the free bor­ reactive sites, demonstrating consistent ders of the epithelial cells lining the cysts reactivity at much higher titers of specific and within the epithelial intercellular antiserum.10,11,23,27 It is believed that the canaliculi. Fluorescein-conjugated anti­ refined fixatives currently in use mask human IgG clearly identified plasma the tissue antigens by preserving the in­ cells abutting the basement membrane tegrity of the cell thereby reducing the and diffusely present within the lym­ permeability of the cellular membranes phoid tissues and the germinal centers.2 to the large antibody molecules, and also This autoimmune theory of origin may by biochemically modifying the antigen, more satisfactorily explain the multicen- e.g., cross linking of the antigen with tric and bilateral character of Warthin’s various proteins. The ability of trypsin to tumor. unmask the antigens may depend on its capacity to increase cell and tissue per­ meability.10,11 Tissue fixation allows ac­ B a c k g r o u n d o n M e t h o d o l o g y curate morphologic identification and preservation of cellular detail. With the Various immunohistochemical tech­ combined use of optimized enzyme niques, i.e., immunofluorescence and tissue digestion with trypsin to unmask immunofluorescence and immunoperox- antigens and the immunoperoxidase idase have been utilized to demonstrate technique, tissue immunoglobulin may various immunoglobulins in tissue sec­ be localized within sections and their tions. The immunofluorescent technique cells of origin identified unequivocally. WARTHIN'S TUMOR 51

O bje c t iv e s U n l a b e l e d A n t ib o d y P e r o x id a se -A ntiperoxidase This study is designed: to investigate retrospectively the distribution of im­ Method:10'14,20'30’31 Briefly, serial 4- munoglobulins in Warthin’s tumors and (xm sections were deparaffinized with xy­ to make some inferences about the lene and processed to water (table I). All etiology of this lesion by comparison with sections were routinely trypsinized to a known hypersensitivity disease (au­ unmask the antigen in the tissue sec­ toimmune thyroiditis) and normal and tions.10,11,26,27 After being brought to reactive lymph nodes. Included in the water, the sections were washed in two study are also cases of oncocytoma, cyst- changes of phosphate buffered saline adenoma, papillary cystadenoma to de­ (PBS, pH 7.6) at room temperature for termine if a developmental pattern can ten minutes per change, and then incu­ be identified which would indicate bated at 37°C for 30 minutes in 0.05 M whether these lesions and Warthin’s Tris-saline buffer (pH 7.8) containing 0.1 tumor form a continuum of a specific percent (W/V) trypsin (Difco 1:250) and pathologic process that culminates in 0.1 percent (W/V) calcium chloride. The Warthin’s tumor. enzyme digestion was terminated by washing with two changes of cold (4°C) Materials and Methods PBS (pH 7.6) for 15 minutes each. The trypsinization procedure had been pre­ T is s u e S e c t io n s viously optimized to enhance specific tissue staining. The adhesive used to coat Paraffin embedded tissue blocks from 13 cases of Warthin’s tumor, which had TABLE I been fixed in ten percent formaldehyde Unlabeled-Antibody Peroxidase- in phosphate buffered saline (pH 7.4), Antiperoxidase (PAP) Method were obtained from the various institu­ tions listed in the Acknowledgment sec­ Stage Procedure tion. In addition, two cases of oncocy­ 1. Four urn paraffin sections: xylol - absolute toma (oxyphilic adenoma), two cases of ethanol ... HjO. 2. Trypsinization for 30 min at 37°C. autoimmune thyroiditis (Hashimoto’s 3. Wash in PBS (pH 7.6) for 30 min (2 changes) disease), two cases of cystadenoma, one at 4° C. 4. Block endogenous peroxidase with absolute case of papillary cystadenoma, four cases methanol containing 0.6% (V/V) hydrogen of normal salivary gland, and one reac­ peroxide for 30 min. ^ 5. Wash in PBS (pH 7.6) for 20 min (two 10 min tive lymph node were also selected for changes) at 25°C. 6. Normal swine serum (1:20) for 20 min; shake the study. Serial paraffin sections (4 (xm) off excess. were prepared routinely on slides which 7. Rabbit anti-human antiserum (IgG-l:500; IgA, IgM-l:300, 1:1,000 for light chains) for had been previously coated with an ad­ 30 min. hesive* (diluted 1:20 in distilled water). 8. Wash as in stage 5. 9. Swine anti-rabbit Ig (1:30) for 30 min. Sections for routine morphological eval­ 10. Wash as in stage 5. 11. PAP complex (rabbit, 1:200) for 30 min. uation were stained with hematoxylin 12. Wash as in stage 5. and eosin and the unlabeled antibody- 13. Diaminobenzidine (DAB; 0.06 W/V%)-H202 (0.03 V/V%) reaction for 5-8 min. enzyme method using PAP was utilized 14. Counterstain with Harris hematoxylin for 2 to demonstrate immunoglobulins (Igs). min. 15. Dehydrate, clear, mount in Permount.

Note: All Stages are carried out at room tempera­ * Elmers Glue-All Borden Inc. ture (25°C) except for Stages 2 and 3. 52 FOULSHAM, JOHNSON, SNYDER, CARPENTER, AND SHAFI the slides was extremely useful in as­ minutes), dehydrated in a graded series suring adherence of the tissue sec­ of alcohol, cleared in xylene, and tions and did not provide any difficulty mounted in Permount. with regard to nonspecific background Specificity and Controls:10'29'31 Non­ staining. specific reactions were minimized by in­ The slides were subsequently trans­ cubation with normal swine serum in ferred to absolute methanol containing Stage 6 and use of the highest (optimum) 0.6 percent (V/V) hydrogen peroxide for titre of the primary antibody in Stage 7. 30 minutes to block endogenous peroxi­ Controls for specificity included the sub­ dase activity. Nonspecific background stitution of normal rabbit serum for the staining activity was minimized by incu­ specific primary antiserum in Stage 7, bating with normal (non-immune) swine and omissions of Stages 7, 9, or 11. serum (1:20) for 20 minutes prior to ap­ Enumeration of Ig-containing cells: plication of the primary antisera. The ex­ The density of each immunocyte class cess normal swine serum was shaken off was determined as previously described and the tissue sections were then treated for tonsils.812 Briefly, a minimum of with rabbit anti-human IgA, IgG, and three high power fields (x40 objective) IgM heavy chains, or rabbit anti-human of several different arbitrarily defined k or X. light chains for 30 minutes. tissue compartments for each of the var­ After appropriate checkerboard titra­ ious types of specimens were counted by tions, dilutions of 1:500 for gamma and two independent observers for positively 1:300 for alpha and mu heavy chains, re­ stained cells. The tissue compartments spectively, and 1:1000 for each light for Warthin’s tumor included germinal chain were found to be optimal. Swine center, the parafollicular zone, and the anti-rabbit Ig (1:30) was then applied subepithelial zone. For lymph nodes, the for 30 minutes, followed by rabbit per- compartments included mantle, parafol­ oxidase-antiperoxidase (PAP) complex licular zone, and germinal centers. For (1:200) for 30 minutes.* Finally, incuba­ Hashimoto’s thyroiditis, the tissue com­ tion with freshly prepared diaminoben- partments were germinal center, mantle, zidine (DAB; 0.06 W/V percent)-hy- parafollicular zone, and the subepithelial drogen peroxide (0.03 V/V percent) in zone. For normal salivary gland, onco­ 0.05 M Tris-saline buffer (IpH 7.8) for cytomas, and cystadenomas the subepi­ five to eight minutes was used to localize thelial compartment was evaluated. sites of specific peroxidase activity. Each Specimens of salivary gland tumors in stage (7,9,11) lasted 30 minutes and was which polymorphonuclear leukocytes followed by a 20 minute wash in PBS (pH were noted in the stroma or cystic con­ 7.6) consisting of two ten minute tents or in which there was disruption of changes. the tumor epithelium with direct expo­ The tissue sections were then count­ sure of the lymphoid stroma to the cystic erstained with Harris hematoxylin (two contents were excluded from this study. Great care was taken in the interpre­ * Reagents for Immunoperoxida.se Staining: tation of immunohistochemical staining Rabbit monospecific antibodies to various human im­ reactions to eliminate false positive am­ munoglobulins (gamma, alpha, and mu heavy chains, and kappa and lambda light chains), swine antirabbit biguous staining secondary to nonspecific Igs, and soluble horseradish peroxidase— rabbit an­ uptake of exogenous immunoglobulins tiperoxidase complexes were purchased from Dako- by lymphoid cells utilizing the criteria patts A/S, Copenhagen, Denmark (Dakopatts: 10- 090, 10-MAT, 10-091, 10-9K2, 10-9L2, 21-090, and outline by Mason and Biberfeld.26 A Z 113). The United States agent is Dako Corporation quantitative record was maintained of Santa Barbara, CA 93103. the density of Ig positive cells and of WARTHIN’S TUMOR 53 their relative distribution in the various with a green filter for black and white. tissue compartments. The reproduc­ Immunofluorescence photomicrographs ibility of the enumeration method was were taken on a Zeiss Photoscope II uti­ good giving a coefficient of variation of lizing Kodak Ektachrome ED135 (ASA 2.5 percent. 200) utilizing a blue filter and appro­ Statistical Analyses: Immunocyte priate emission and absorption filters for enumerations were expressed as arith­ FITC conjugates. metic means of absolute counts of posi­ tive cells ± S.D. and as a percentage of Results the total number of cells staining for Ig in the various tissue compartments. The The class distribution, including both data exhibited a fairly normal distribu­ the density and the percentage, of Ig tion and homogeneity of variance; con­ containing plasma cells in Warthin’s sequently group comparisons were made tumor is shown in table II. The subepi­ by the Student’s t-test. Differences thelial zone of Warthin’s tumor contains showing P =£ 0.05 were accepted as sta­ a preponderance of IgA positive plasma tistically significant. cells (49.1 percent) with a lesser number Immunofluorescence: Immunofluores­ of IgG positive plasma cells (38.9 per­ cence was also used initially for the eval­ cent) (figures 1 and 2). IgM positive cells uation of specific tissue IgG staining to (12.0 percent) are also present in the sub- verify the validity of our methods by epithelial region but in much fewer num­ comparison of the immunofluorescence bers (figure 3). The density of IgA posi­ and immunoperoxidase immunoglobulin tive cells is not significantly different distributioin. The immunofluorescent re­ from that IgG; however, both the den­ sults were also compared with those re­ sities of IgA and IgG are significantly ported by Allegra2 to insure comparable greater than IgM (p < 0.01). Although staining of trypsinized, formalin-fixed, the densities of immunoglobulin positive paraffin-embedded tissues to routine cells are less in the parafollicular zone, fresh frozen tissues. Briefly, the tissue the frequency distribution of IgA: IgG: sections were processed identically to IgM is very similar. The greatest differ­ the PAP technique through the applica­ ence occurs in the germinal centers tion of normal swine serum (table I, where IgA positive cells (7.2 percent) are Stage 1 — 6). After removal of excess rare, if present at all, and IgG (62.7 per­ swine serum at the end of incubation, the cent) predominates over IgM (30.0 per­ slides were incubated with rabbit anti­ cent) in a 2:1 ratio (figures 4, 5, and 6). human IgG (1:40) (Dakopatts) for 45 min­ The results of plasma cell staining for utes then washed in PBS. Fluorescein- kappa and lambda light chains show a conjugated swine anti-rabbit IgG (1:30) kappa to lambda ratio that ranged from (Dakopatts) was applied for 45 minutes 1.1:1 in the germinal centers to 0.9:1 in in a moist chamber in the dark. Fol­ the subepithelial space (figures 7 and 8). lowing sequential washed in PBS, the The class distribution of Ig containing sections were then coverslipped using plasma cells within the subepithelial buffered glycerol. zone of various salivary gland tumors is Photography: Photomicrographs were shown in table III. All three salivary taken on a American Optical photomicro­ gland neoplasms exhibit a significantly scope (series 10 Microstar) utilizing different (p < 0.05) density and fre­ Kodak Ektachrome EPY135 (ASA 50 quency distribution of the various im- tungsten) with neutral filters for color muoglobulins within the subepithelial and Kodak Panatomic-X Fx 135 (ASA 32) zone when compared with normal sali- 54 FOULSHAM, JOHNSON, SNYDER, CARPENTER, AND SHAFI TABLE I I

Frequency Distribution and Density of Ig Containing Plasma Cells Within Various Tissue Compartments of Warthin's Tumor

Immunocyte Class Tissue Compartment IgA IgG IgM K X

Germinal Center 0.8 ± 0.7* 6.9 + 4.8§ 3.3 ± 1.6§ 7.4 ± 8.7 6.6 ± 5.3 (n = 11) (7.2%)t (62.7%) (30.0%) (52.9%) (47.1%) Parafollicular Zone 14.3 ± 19.9 10.6 ± 10.7§ 1.4 ± 0.7J 9.4 ± 12.5 13.3 ± 14.6 (n = 11) (54.4%) (40.3%) (5.3%) (41.4%) (58.6%) Subepithelial Zone 28.7 ± 16.7 22.7 ± 12.2§ 7.0 ± 9.1§ 19.3 ± 20.6 22.4 ± 11.7 (n = 11) (49.1%) (38.9%) (12.0%) (46.3%) (53.7%)

*Mean (X) ± S.D. per high power field (40X)

■(■Percentage (%) = ---- (IgX) (100)---- (IgA + IgG + IgM) < 0.05 (Student's t test, IgX compared to IgA within tissue compartments) < 0.01 vary gland tissue. The normal salivary a much more even frequency distribu­ gland shows almost exclusively IgA pos­ tion of immunocytes in general with IgA itive cells (96.2 percent) in this region » IgG >> IgM. Cystadenomas and on­ with rarely encountered IgG (2.5 per­ cocytomas have very similar immunocyte cent) or IgM (1.3 percent) positive cells. density and frequency distribution. In The various salivary gland tumors show addition, Warthin’s tumor shows a fre­ quency distribution of immunocytes fairly similar to those of both cystaden­ omas and oncocytomas; however, the

F i g u r e 1. Epithelial border stained with anti­ F ig u r e 2 . Same epithelial border as in figure 1 human IgA showing intermittent specific staining of stained with anti-human IgG showing minimal spe­ epithelial cells and positively staining plasma cells in cific epithelial staining and fewer positively stained the subepithelial zone. The positive reaction product plasma cells (black) in the subepithelial space (PAP, for Ig is a black precipitate. (PAP, hematoxylin coun­ hematoxylin counterstain, magnification X112, be­ terstain, magnification x 112, before reproduction) fore reproduction) WARTHIN'S TUMOR 55 The immunohistochemistry of the lymphoid stroma of Warthin’s tumor was then compared to that of a known de­ layed hypersensitivity disease, Hashi­ moto’s thyroiditis, and to a reactive pe­ ripheral lymph node. These results are shown in table IV. In the parafollicular zone, Warthin’s tumor and Hashimoto’s thyroiditis ex­ hibit a remarkably similar frequency dis­ tribution of immunocytes. However, there is a marked increase in the densi­ ties of IgA, IgG (p < .05), and IgM (p < 0.01) immunocytes in Hashimoto’s dis­ ease relative to Warthin’s tumor. In the germinal centers, there is no significant difference in the density and frequency Figure 3. Same epithelial section as in figure X of IgA, IgG, and IgM staining immuno­ stained with anti-human IgM showing no epithelial cytes between the two disease processes. staining and rare positively stained plasma cells (black) in the subepithelial space (PAP, hematoxylin In the parafollicular zone, Warthin’s counterstain, magnification X 112, before reproduc­ tumor shows a marked increase in the tion) density and frequency of IgA staining im­ munocytes relative to the reactive pe­ density of IgA and IgG staining in much ripheral lymph node; however the latter greater in Wartbin’s tumor than in the exhibits marked and moderate increases other two lesions. Germinal centers were in the density and frequency of IgG (p < an inconsistent finding in oncocytomas 0.01) and IgM (p < 0.05) respectively. and not present at all in the cystaden- This type of increase is immunologically omas. consistent with a reactive process. Simi­

F i g u r e 4. Germinal center stained with anti­ human IgA. Negligible positive staining present but mitotic figures noted (PAP, hematoxylin coun­ terstain, magnification x 112, before reproduc­ tion) 56 FOULSHAM, JOHNSON, SNYDER, CARPENTER, AND SHAFI

F i g u r e 5. Same ger­ minal center as in figure 4 stained with anti-human IgG. Prominent positively stained cells (black) in moderate numbers. (PAP, hematoxylin counterstain, magnification X 112, be­ fore reproduction)

larly, there is a marked increase in den­ the various anatomic compartments of sity and frequency of IgA staining im­ the lymphoid stroma; however, an exact munocytes within the germinal centers density of immunocytes and their of Warthin’s tumor relative to the reac­ frequency distribution in the entire lym­ tive lymph node. In the reactive node phoid stroma awaits the future applica­ however, there are marked increases in tion of morphometric analysis.8 Unfor­ both IgG (p < 0.01) and IgM (p < 0.01) tunately, the inclusion of paraparotid densities, again indicative of a reactive reactive lymph nodes in this preliminary process. study was not possible due to inconsis­ The results discussed above reflect an tencies in staining secondary to marked approximation of the number of immu­ variation in tissue preservation in those noglobulin producing plasma cells within specimens obtained. The large standard

F i g u r e 6. Same ger­ minal center as in figure 4 stained with anti-human IgM. Occasional positively staining cell present. (PAP, hematoxylin coun­ terstain, magnification X112, before reproduc­ tion) WARTHIN’S TUMOR 57 which were uncontrollable in this retro­ spective study. Other retrospective and prospective studies show similarly large standard deviations.8,21 The epithelium of Warthin’s tumor stained prominently for IgA along the lu­ minal border, the apical cytoplasm of scattered epithelial cells, and within the intercellular canaliculi (figure 9). Min­ imal staining in the same pattern was noted for IgG, however, no staining for IgM was seen. The cystic contents were also markedly positive for IgA including both the debris and occasional intact plasma cells. It was also noted that the secretions contained within the ductal lu- mina of the normal salivary glands were Figure 7. Tissue section showing the epithelial border and lymphoid stroma stained for k light positive for IgA, but tended to stain less chains. Positive cells (black) are obvious as well as intensely than in Warthin’s tumor; no im­ pale specific epithelial staining. (PAP, hematoxylin munoglobulin positive cells were noted counterstain, magnification X 112, before reproduc­ within the lumina. tion) The morphology of the germinal cen­ ters in Hashimoto’s thyroiditis and the deviations for densities of positively reactive lymph node exhibited polarity, staining cells in many specimens are not i.e., the germinal centers were accom­ due to differences in scoring by the two panied by a mantle which was wider to­ observers, but to variations in tissue ward the area of antigenic exposure. The staining secondary to fixation differences germinal centers in these cases are divi-

Fig u r e 8. Tissue sec­ tions stained for X light chains. When compared with figure 7, a greater in­ tensity of specific staining is noted along the luminal border as well as an in­ creased number of posi­ tively staining plasma cells (black) within the subepi- thelial space. (PAP, he­ matoxylin counterstain, magnification X112, be­ fore reproduction) 58 FOULSHAM, JOHNSON, SNYDER, CARPENTER, AND SHAFI

T A B L E III

Frequency Distribution and Density of Ig Containing Plasma Cells Within the Subepithelial Zone of Various Salivary Gland Neoplasma

Imaunocyte Class Specimen IgA IgG IgM K X

Warthin's Tumor 28.7 ± 16.7* 22.7 ± 12.2 7.0 ± 9.1 19.3 ± 20.6 22.4 ± 11.7 (n = 11) (49.1%)f (38.9%) (12.0%) (46.3%) (53.7%) Cystadenoma 19.7 ± 3.4 9.6 + 4.5 5.0 + 2.3 15.7 ± 9.0 24.2 ± 7.5 (n = 2) (57.4%) (28.0%) (14.6%) (39.3%) (60.7%) Oncocytoma 15.8 + 9.7 11.8 ± 8.6 4.6 ± 4.1 19.0 + 13.2 28.2 ± 25.9 (n = 2) (49.0%) (36.6%) (14.3%) (40.3%) (59.7%) Normal Submandibular 7.6 + 2.8 0.2 ± 0.2 0.1 ± 0.2 4.1 ± 2.5 10.3 ± 3.0 Gland

*Mean (X) ± S.D. per high power field (40X)

•^Percentage (%) = ______Ü222__ IgA + IgG « IgM sible into two zones as described by containing large cells with less basophilic Curran and Jones;12 zone A, in the basal nuclei, fewer mitotic figures, and im­ portion of the germinal center, con­ munoglobulin positive cells along the pe­ taining a small number of large immu­ riphery. Germinal center polarity was noglobulin staining cells with moderately not routinely present in the Warthin’s basophilic, vesicular nuclei and nu­ tumor specimens; however, a fairly well- merous mitoses, and zone B, occupying defined mantle was occasionally present. the remainder of the germinal center, The germinal centers regularly showed

TABLE IV

Frequency Distribution and Density of Ig Containing Plasma Cells Within the Germinal Centers and Parafollicular Zones of Various Specimens

Imiaunocyte Class Germinal Centers IgA IgG IgM K X

Warthin’s Tumor 0.8 ± 0.7* 6.9 ± 4.8 3.3 ± 1.6 7.4 ± 1.6 6.6 ± 5.3 (n = 11) (7.2%)f (62.7%) (30.0%) (52.9%) (47.1%) Hashimoto's Thyroiditis 2.6 ± 0.8 7.5 ± 0.5 4.4 ± 1.5 8.0 ± 5.9 11.9 ± 0.6 (n = 2) (17.9%) (51.7%) (30.3%) (40.2%) (59.8%) Reactive Lymph Node 0.2 ± 0.4 61.8 ± 21.4 11.2 ± 2.7 34.0 ± 7.8 32.5 ± 8.0, (Peripheral) (n = 1) (0.3%) (84.4%) (15.3%) (51.1%) (48.9%)

Parafollicular Zone

Warthin's Tumor 14.3 ± 19.9 10.6 ± 10.7 1.4 ± 0.7 9.4 ± 12.5 13.4 ± 14.6 (n = 11) (54.4%) (40.3%) (5.3%) (41.4%) (58.6%) Hashimoto's Thyroiditis 52.5 ± 68.8 44,3 + 37.1 5.0 ± 2.8 48.3 ± 37.7 66.6 ± 36.4 (n = 2) (51.6%) (4Q.5%) (4.9%) (42.0%) (58.0%) Reactive Lymph Node 3.4 ± 2.8 41.0 ± 10.1 4.3 ± 2.9 20.0 ± 11.3 21.2 ± 7.8 (Peripheral) (n = 1) (7.0%) (84.2%) (8.8%) (48.5%) (51.5%)

*Mean (X) ± S.D. per high power field (40X) fPercentage (%) - ______(100) _ (IgA + IgG + IgM) WARTHIN’S TUMOR 59

F i g u r e 9. Epithelial component stained specif­ ically with anti-human IgA shows cytoplasmic stain­ ing in the apices of some of the epithelial cells, along the luminal border of the epithelium, within the cyst contents and in the intercellular cana- liculi. (PAP, hematoxylin counterstain, magnifica­ tion x 280, before repro­ duction)

mitotic figures and tangible bodies indi­ logic analysis.27 These factors were un­ cating their reactive nature. controllable owing to the varied origin of the selected specimens and, as such, Discussion three specimens were discarded after processing owing to poor tissue preser­ As previously mentioned, the origin of vation. The present authors agree with Warthin’s tumors has long been dis­ Huang et al21 that the type of fixative, puted. Allegra2 first proposed the au­ duration of fixation, and age of the spec­ toimmune theory in 1973 which contrasts imens may be responsible for the incon­ with the more widely held theory that sistencies sometimes encountered. these lesions arise from salivary gland In hematoxylin and eosin sections, the elements trapped within intraparotid or lymphoid infiltrate in Warthin’s tumor paraparotid lymph nodes during embryo- appears diffuse as well as follicular in na­ genesis. The rapid advances in immu- ture. Postcapillary venules and scattered nohistochemistry have only recently histiocytes and immunoblasts were allowed extensive retrospective investi­ noted; however, no subcapsular sinuses gation into the pathogenesis of this were appreciated in our specimens but neoplasm. In this study, numerous tissue have been reported by other au­ blocks were obtained from four institu­ thors.915,24 Although the cytoarchitec- tions dating back to 1962. Of necessity, ture of the lymphoid stroma is similar to all specimens were treated with trypsin that of a reactive lymph node the quan­ to enzymatically expose the antigenic tity of lymphoid tissue in the fully de­ sites under investigation. The method veloped lesion is greatly in excess of that and duration of fixation influences both found in the paraparotid nodes from the required time of exposure to trypsin which it supposedly develops. Germinal for optimal results and the quality of centers are regularly found at the tips of tissue preservation for optimal morpho­ the papillary projections in Warthin’s 60 FOULSHAM, JOHNSON, SNYDER, CARPENTER, AND SHAFI tumor, an unlikely location for performed utilizing monoclonal T-cell antibody with germinal centers to be found.32 the PAP method, however, has yet to be Coosman et al.9 examined the surface accomplished in this study. From table markers of the lymphatic elements of II it is apparent that the predominant im­ Warthin’s tumor in frozen section, and munoglobulin present in the various concluded that the lymphoid stroma was tissue compartments of Warthin’s tumor predominantly composed of B-cells in a is not IgM as noted in the cell suspension pattern similar to that observed in reac­ studies but IgA. This is consistent with tive lymph nodes. They thought that the the mucus secreting minor salivary lymphoid follicles, which are composed glands incidentally seen by Curran and of B-lymphocytes, could be residual Jones12 in their study of tonsillar tissue. lymph node cortex with a superimposed They noted that in contrast to the large reactive mononuclear cell infiltrate di­ numbers of IgG positive cells seen in the rected against the epithelial component tonsil, the stroma of the minor salivary of the tumor. Diamond and Braylan,15 on glands contained a small number of IgA the other hand, using a cell suspension positive plasma cells and only an occa­ of the lymphoid stroma concluded that sional cell which stained positively for the majority (52.5 percent) of the lym­ IgG or IgM. They also observed that the phocytes present were of T-cell lineage lumina of the minor salivary glands con­ based on the formation of rosettes with tain IgA positive secretions and con­ sheep red blood cells. The B-cells pres­ cluded that the IgA is of local origin, pro­ ent in the lymphoid stroma had a poly­ duced by the IgA positive cells in the clonal distribution based on an analysis stroma. The normal salivary glands of their surface immunoglobulins. The studied in our series showed only IgA majority of the B-cells in their series positive cells in the subepithelial region; were of the IgM class (17.5 percent) with very rare IgG or IgM positive cells were fewer cells of the IgG (3.5 percent) and seen. IgA (7.5 percent) classes. The ratio of In tonsillar tissue, the largest concen­ kappa to lambda light chains was 1.5:1. tration of immunoglobulin positive cells Tubbs et al33 similarly noted a predomi­ is found within the basal layers of epi­ nance of T-cells (60 percent) in a cell sus­ thelium and in the subepithelial region.12 pension study, but also found a reduction Our study shows a similar distribution for in the percentage of T-lymphocytes as IgA in Warthin’s tumor. Hsu et al22 also compared with normal lymphoid tissue note the preponderance of IgA positive (70 percent). Their studies also showed a cells within the subepithelial region in preponderance of IgM (38.8 percent) formalin-fixed Warthin’s tumors. Unfor­ among positively staining cells with IgD tunately their study did not include an (34.7 percent) as the second most analysis of the parafollicular zone and the common immunoglobulin. They con­ germinal centers. The results of our cluded that the polyclonal follicular study agree with theirs in that IgA IgG staining and an apparent reduction in the >> IgM positive cells in the subepithe­ number of T-lymphocytes tended to sup­ lial region. port the concept that the lymphoid infil­ The large discrepancy between the trate in Warthin’s tumor represents a cell suspension studies and the PAP tech­ reactive cellular proliferation. nique as applied to tissue sections cannot In our study, the majority of lymphoid yet be explained. Huang et al23 reported stromal cells did not stain for immuno­ the survivability of immunoglobulins in globulin and, as such, may possibly be of tissue sections undergoing enzymatic T-cell origin. Specific staining for T-cells, digestion to be best for IgG and IgA and WARTHIN'S TUMOR 61 poorest for IgM. They did not, however, within normal salivary gland epithelial note any decrease in activity for up to 60 cells. Allegra,2 utilizing an immunoflu- minutes of incubation with a 0.1% orescent technique, noted the presence trypsin solution at 37°C. In both our of IgG in the cystic contents, in moderate study and that of Hsu, the incubation numbers of plasma cells in the subepi­ times were 30 minutes or less. Future thelial space, along the luminal border morphometric analysis of our tissue sec­ and within the intercellular canaliculi. tions may help to resolve these con­ Based on these observations and on elec­ flicting results. tron microscopy of the lymphoid stroma, The cystic contents of Warthin’s he proposed a delayed hypersensitivity tumor, as well as the cytoplasm of a small origin for Warthin’s tumor. Our immu- number of epithelial cells and the inter­ nofluoresence results (figure 10) agree cellular canaliculi of the epithelium, with those of Allegra, but the immuno- were found to be moderately to strongly peroxidase results show the IgG to be IgA (alpha heavy chain plus secretory only faintly positive in the cystic contents component) positive. Hsu21 have dem­ and in the epithelial component. Unfor­ onstrated that the secretory component tunately, the distribution of IgA was not of IgA occurs only within the salivary examined in Allegra’s study. gland epithelial cells and not in the IgA On a histopathologic basis, the onco­ positive cells beneath the basement cytoma does not appear to form a con­ membrane. These cells in the subepithe- tinuum with Warthin’s tumor as a single lial region stain only for alpha heavy disease entity as previously proposed by chains. This data indicates IgA produced Allegra.2 In no instance did the epithelial by the plasma cells in the subepithelial component exhibit papillary projections space is being transported across the ep­ but only a compact mass of oncocytes ithelium via the secretory piece found with blind duct elements. In those cases

Fig u r e 10. Immuno- fluorescent photomicro­ graph of the epithelium showing distribution of specific staining for IgG as described by Allegra, i.e., within the cystic space, along the epithelial border and the intercellular can­ aliculi, and in plasma cells within the subepithelial space. (FITC conjugated rabbit anti-human IgG, magnification x 155, be­ fore reproduction) 62 FOULSHAM, JOHNSON, SNYDER, CARPENTER, AND SHAFI in which lymphoid follicles were present, the acquisition of a lymphoid infiltrate in­ they were located at the periphery of the cluding the development of lymphoid lesion. A mononuclear cell infiltrate, follicles. The distribution of immuno­ consisting of approximately 40 percent B- globulins closely resembles that reported lymphocytes, was present beneath the in perialimentary tract lymphoid follicles basement membrane of the epithelial undergoing reactive hyperplasia.12 An component and may represent a super­ autoimmune pathogenesis of Warthin’s imposed inflammatory response. As in tumor is unlikely. The tissue destruction, normal salivary gland, IgA predominated resulting fibrosis and loss of parenchymal in the subepithelial space (table III). The tissue seen in autoimmune diseases such frequency distribution of immunocytes as Hashimoto’s thyroiditis was not noted in oncocytoma is remarkably similar to in Warthin’s tumor; however, the im­ that of Warthin’s tumor; however, the munoglobulin distribution in this series density of IgA and IgC containing im­ is very similar between the two lesions munocytes is much greater in the latter (table IV). In addition, specific cellular lesion. or humoral immune responses to the The cystadenoma, on the other hand, components of salivary gland secretions is remarkably similar morphologically to or ductal epithelium have not been iden­ Warthin’s tumor without the lymphoid tified.15,33 It may be possible to demon­ follicles. In these lesions, the frequency strate antibodies to salivary gland epithe­ distribution of immunoglobulin classes is lium in the saliva; however, if these very similar to that seen in Warthin’s antibodies do exist, they are likely to be tumor; however, as is the case with on­ an epiphenomenon and not a causative cocytomas, the density of IgA and IgG agent. producing immunocytes is much greater The stimulus resulting in oncocytic hy­ in Warthin’s tumor (table III). It is pos­ perplasia is unknown. Many factors may tulated that the cystadenoma may arise be involved in the development of the in ectopic or normal salivary gland tissue lymphoid infiltrate including germinal with no associated lymphoid tissue by a centers: possibly luminal antigens, but process similar to that producing War­ this is not likely since the cystic cavities thin’s tumor within a paraparotid lymph do not freely communicate with the sal­ node. As such, these lesions may result ivary duct system or oval cavity; mito­ from the same basic pathologic process gens; hormones; and unknown micro­ occurring at different sites of origin. This environmental factors. The neoplastic may account for the frequency of reports epithelium may possess tumor associat­ of Warthin’s tumor arising in ectopic lo­ ed antigens and produce hormones or cations, the majority of which contain a lymphokines which may cause homing, lymphoid stroma but no germinal cen­ redistribution, or accumulation of IgA ters. precursor or plasma cells or may enhance the proliferation and/or maturation of Conclusion IgA precursor cells to plasma cells. Oncocytomas represent a pathologic The results of this study support the entity distinctively different from War­ hypothesis that Warthin’s tumor may thin’s tumor. Cystadenomas, on the arise in heterotopic salivary gland tissue, other hand, and Warthin’s tumors may entrapped within intraparotid or para­ represent different extremes of the con­ parotid lymph nodes during embryogen- tinuum of a single pathologic process esis, by oncocytic hyperplasia of the with the former arising in ectopic salivary ductal epithelium which may result in gland tissue independent of lymphoid WARTHIN’S TUMOR 63

14. D e L e l l is , R. H., S t e r n b e r g e r , L . A., e t . a l . : tissue and the latter arising in hetero­ Immunoperoxidase technics in diagnostic pa­ topic salivary gland tissue entrapped thology. Am. J. Clin. Path. 71:483-488, 1979. within lymph node tissue. 15. D ia m o n d , L. W. a n d B r a y l a n , R. C.: Cell sur­ face markers on lymphoid cells from Warthin’s Acknowledgments tumor. Cancer 44:580-583, 1979. Tissue blocks were obtained from the Departments 16. D ie t e r t , S . E.: Papillary cystadenoma lympho­ of Pathology: New Britain General Hospital, New matosum (Warthin’s Tumor) in patients in a gen­ Britain, CT; University of Connecticut Health eral hospital over a 24 year period. Am. J. Clin. Center, Farmington, CT; Hartford Hospital, Hart­ Path. 63:866-875, 1975. ford, CT; and Newington Veterans Administration 17. D o n a l d , P. J. a n d K r a u s e , C. J.: Papillary cyst­ Medical Center, Newington, CT. adenoma of the larynx. Laryngoscope 83:2024— Thanks are extended to J. Treffeisen and W. Kar- 2028, 1973. powich for technical assistance in the preparation of 18. H a b s , H . and H a n s e n , J. E.: Bilateral simulta­ histologic sections was provided; to Shirley R. neous Warthin’s tumors in a woman. Southern Johnson, Ph. D. for assistance with the statistical anal­ Med. J. 70:257-258, Feb. 1977. 19. H a n d l e r , S. and M i l l e r , R. H .: Warthin’s ysis; and to Lia Buccini for typing the manuscript. tumor: A rare case of a neck mass. Ear Nose References Throat J. 57:205-208, 1978. 20. H e in z , I.: The adenolymphomata. Australian N. 1. A l b r e c h t , H. and A r z t , L: Beitràge zur Frage Zealand J. Surg. 22:47-51, 1951. der Gewebsveirrung. Papillâre cystadenoma in 21. Hsu, S. M., Hsu, P. L., and N ayak, R. N .: War­ lymphorüsen. Frankfort Z. Path. 4:47, 1910. thin’s tumor: An immunhistochemical study of 2. A l l e g r a , S. R. : Warthin’s tumor: A hypersensi­ its lymphoid stroma. Human Path. 12:251-257, tivity disease? Human Path. 2:403-420, 1971. 1981. 3. A z z o p a r d i , ? and Hou,?: The genesis of adeno- 22. Hsu, S. M. and R e e , H. J.: Self-sandwich lymphoma. J. Path Baet. 88:213-218, 1964. method—an improved immunoperoxidase 4. B a d e n , E. e t . a l .: Intraoral papillary cystade­ technic for the detection of small amounts of an­ noma lymphomatosum. J. Oral Surg. 34:533- tigens. Am. J. Clin. Path. 74:32-40, 1980. 541, 1976. 23. H u a n g , S., M in a ss ia n , H ., and M o r e , J. D.: Ap­ 5. B atsakis, J. G. and R e c e z i , J. A.: Selected con­ plication of immunofluorescent staining on par­ troversial lesions of salivary tissues. Otol. Clin. affin sections improved by trypsin digestion. North Am. 10:309-328, 1977. Lab. Invest. 35:383-390, 1976. 6. B e c k , L. D. and M a g u d a , T. A.: Papillary cyst­ 24. J e n s e n , P. S. and H u s t e d , E.: Cystadenolym- adenoma lymphomatosum (Warthin’s Tumor): A phoma (Warthin’s Tumor) of the parotid gland. multicentric benign tumor. Laryngoscope Scand. J. Plast. Recon. Surg. 4:114-121, 1970. 77:1840-48, 1967. 25. K im , ? e t . a l . : Lysosomes in the epithelial com­ 7. B l a t n i k , D. S., K u r t i n , J. B., and L e h m a n , ponent of Warthin’s tumor. Arch. Pathol. 95:56- R. H.: Multicentricity of Warthin’s tumors. Wis­ 62, 1973. consin Med. J. 75:561—563, 1976. 26. M a s o n , D. M . and B ib e r f e l d , P.: Technical as­ 8. B r a n d t z a e g , L., S u r j a n , L., Jr ., and B e r d a l , P.: pects of lymphoma immunohistology. J. Histo- Immunoglobulin system of human tonsils. Clin. chem. Cytochem. 28:731-745, 1980. Exp. Immunol. 31:367-381, 1978. 27. M e p h a m , B. L., F r a t e r , W., and M i t c h e l l , 9. C o o s m a n , J., D e e g a n , M. J., and B a t s a k i s , B. S.: The use of proteolytic enzymes to improve J. G. : Warthin’s tumor: B-lymphocytes within immunoglobulin staining by the PAP technique. the lymphoid infiltrate. Arch. Pathol. Lab. Med. Histochem. J. 11:345-357, 1979. 101-,354-356, July 1977. 28. P o t d a b , G. G.: Warthin’s tumor. Arch. Surg. 10. C u r r a n , R. C. and G r e g o r y , J. : Demonstration 98:206-208, 1969. of immunoglobulin in cryostat and paraffin sec­ 29. S t e r n b e r g e r , L. A.: Immunocytochemistry, 2nd tions of human tonsil by immunofluorescence ed. New York, John Wiley and Sons, 1979. and immunoperoxidase techniques. J. Clin. 30. S t e r n b e r g e r , L. A.: The unlabeled antibody Path. 31:974-983, 1978. (PAP) method. J. Histochem. Cytochem. 27: 11. C u r r a n , R. C. and G r e g o r y , J.: The unmasking 1657, 1979. of antigens in paraffin sections of tissue by 31. T a y l o r , C. R.: Immunoperoxidase technique. trypsin. Experientia 33:1400-1401, 1977. Arch. Pathol. Lab. Med. 102:113-120, 1978. 12. C u r r a n , R. C. and Jo n e s E. L. : Immunoglob- 32. T h a c k r a y , A. C. and L u c a s , R. B.: Tumors of the ulin-containing cells in human tonsils as dem­ major salivary glands. Atlas of Tumor Pathology, onstrated by immunohistochemistry. Clin. Exp. Series II, Fascicle 10. Washington, D.C. Armed Immunol. 28:103-115, 1977. Forces Institute of Pathology, 1974. 13. C u l l i n g , C. F. A., R e e d , P. E., and Si n n o t t , 33. T u b b s , R. R., e t . a l . : Immunohistochemistry of N.M .: The effect of various fixatives and trypsin Warthin’s tumor. Am. J. Clin. Path. 74:795-797, digestion upon the staining of routine paraffin- Dec. 1980. embedded sections by the peroxidase-antiperox- 34. W a r t h in , A. S.: Papillary cystadenoma lympho­ idase and immunofluorescent technique. J. of matosum; A rare teratoid of the parotid region. Histotechnology 3:10—19, 1980. J. Cancer Res. 23:116-125, 1929.