Menstrual Cycle-Dependent Localization in Endometrium · HumanEndometrial Mucin 169 AJPJanuary 1995, Vol. 146, No. 1 Sialidase (from Clostridium perfringens) and a-N- acetylgalactosaminidase

American Journal of Pathology, Vol. 146, No. 1, January 1995

Copyright C) Amenican Societyfor Investigative Pathology

A Mucin-Like Glycoprotein Identified by MAG(Mouse Ascites Golgi) Antibodies

Menstrual Cycle-Dependent Localization inHuman Endometrium

Harvey J. Kliman,*t Ronald F. Feinberg,ttLisa B. Schwartz,t Michael A. Feinman,§Ehud Lavi,*t and Erika L. Meaddough*From the Departments ofPathologv* and Obstetrics andGynecology,t Yale University School ofMedicine, New

Haven, Connecticut; the University ofPennsylvania MedicalCenter,* Philadelphia, Pennsylvania; and the PacificFertility Center, 5 Westlake Village, California

Human endometrial glands synthesize and se-crete a high molecular weight mucin-like glycop-rotein in a menstrual cycle-dependentfashion. Anovel moiety within this Golgi-associatedglycop-rotein is strongly reactive with IgG antibodies innumerous murine ascites, and has been termedMAG (mouse ascites Golgi). Immunohistochemi-cal staining of201 endometrial biopsies revealedthefolowing patterns: MAGflrst appeared in theGolgi on cycle day 5, peaked on day 15, waspresent on the surface ofthe luminal epitheliumbetween days 17 and 19, and was no longer de-tectable after day 19. MAG was also present incervical, prostate, seminal vesicle, and lacrimalglands, pancreatic acinar ceUs, gaU bladder andbile duct epithelium, and certain ceUls ofthe sali-vary and sweatglands. Interestingly, only tissuesfrom blood group A individuals exhibited thisstaining. As a common link among aU these ceUtypes is the expression ofmucins, we speculatedthat the MAG epitope could be a mucin-associ-ated bloodgroup A-related epitope. This hypoth-esis was tested by absorption experiments with avariety ofglycoconjugates and etythrocytes andby immunoblots of MAG-rich material The ab-sorption studies demonstrated that only type IIIporcine mucin (<1% sialic acid) and blood typeAor AB erythrocytes were able to absorb the anti-MAG antibody. Inasmuch as N-acetyl-galac-tosamine alone, the terminal blood group A car-

bohydrate, did not block MAG antibody binding,the MAG epitope appears to involve N-acetyl-galactosamineplus other determinants. Immuno-blots of endometrial extracts and saliva fromblood type A individuals revealed MAG-reactivematerial with a molecular weight >200 kd underreducing conditions. Because the MAG epitopeappears on the endometrial surface during thepurported implantation window, we speculatethat mucin-like epitopes could play a role inthe earliest apposition phases of conceptus-endometrial interaction. (Am J Pathol 1995,146:166-181)

The molecular basis for the apposition and adhesionphases of implantation is not precisely understoodbut may involve cell-cell recognition moieties.1 Syn-chronization between embryonic and uterine devel-opment leading to the receptive phase, or "implan-tation window," appears to be necessary forsuccessful implantation in the rodent2 and human.3-5In preparation for the implantation process, the uter-ine epithelium produces both secreted and mem-brane-bound glycoproteins, some of which are hor-monally regulated.7-9 Biochemical changes, such asa decrease in the thickness and charge of the uterineepithelial glycocalyx, have been shown to precedeimplantation.10-12 The carbohydrate complexes ofthe luminal endometrial surfaces (glycocalyx) havealso been shown to change during the adhesive

Accepted for publication September 6, 1994.

Current address of Dr. Schwartz: Division of Reproductive Endo-crinology and Infertility, Department of Obstetrics and Gynecology,New York University Medical Center, New York, NY.

Address reprint requests to Dr. Harvey J. Kliman, Developmen-tal and Perinatal Pathology Unit, Departments of Pathology andObstetrics and Gynecology, Yale University School of Medicine,310 Cedar Street, New Haven, CT 06520-8023.

166

Human Endometrial Mucin 167AJPJanuary 1995, Vol. 146, No. 1

phase,13'6 as demonstrated by changes in lectinbinding. Lindenberg and colleagues,14 15 have dem-onstrated the presence of a specific oligosaccharidedeterminant, lacto-N-fucopentaose 1, on the mouseendometrium that appears at the time of implantationand specifically binds to a lectin receptor on thetrophoblast surface of the blastocyst. These workersshowed that the addition of lacto-N-fucopentaoseinhibited blastocyst attachment to endometrialmonolayers.16 In addition to binding to specific re-

ceptor molecules, endometrial cell surface glycop-roteins may act as substrates for cell surface glyco-syltransferases. Galactosyl transferase has beenidentified on the surface of early mouse preimplan-tation embryos.17 As demonstrated in sperm-eggbinding,18 glycosyltransferases present at the em-

bryonic-endometrial interface could facilitate thecomplex heterotypic cell-cell interactions requiredfor implantation.

While characterizing the immunolocalization pat-tern of oncofetal fibronectin in human implantationsites,19 we observed highly specific Golgi staining incertain endometrial biopsies. Although initially attrib-uting this Golgi staining to the anti-oncofetal fi-bronectin antibody, we later identified IgG antibodiesin other non-fibronectin-related murine ascites thatexhibited the same specific Golgi-associated stain-ing. This fortuitous finding led us to further charac-terize the mouse ascites Golgi moiety, termed MAG.In this study, detailed analysis of multiple endome-trial samples revealed that MAG has a unique pat-tern of expression in the normal hormonally respon-sive endometrium. Furthermore, MAG is present invarious secretory epithelia and appears to be part ofa large mucin-like glycoprotein. The specific tempo-ral regulation and secretion of MAG by normal cy-

cling endometrial glands suggests an important rolefor mucin-like glycoconjugates in early embryo-endometrial interactions.

Materials and MethodsSpecimens

Endometrial tissue was examined from endometrialbiopsies, endometrial curettings, and hysterectomiesperformed at the Hospital of the University of Penn-sylvania, Yale-New Haven Hospital, and the PacificFertility Medical Center. A total of 201 specimenswere examined, including specimens throughout theentire menstrual cycle. Endometrial specimens withstromal/glandular dyssynchrony, hyperplasia, atypia,or carcinoma were not included. Occasionally, pa-

tients were biopsied in sequential cycles, and one

patient was biopsied five times during the follicularphase of one cycle. Tissues from other organs wereexamined from the files of both the Hospital of theUniversity of Pennsylvania and the Yale-New HavenHospital. Brain tissue was examined from resectionsperformed at the Hospital of the University of Penn-sylvania. The specimens were fixed in either Bouin'ssolution or 10% neutral buffered formalin overnight atroom temperature and either embedded immediatelyor stored in 70% ethanol until embedding (usuallywithin a few days of tissue collection). All human ma-terial was collected with the approval of the Institu-tional Review Boards at each institution.

Endometrial Dating

Standard hematoxylin and eosin stained sectionswere examined for dating. The endometrial sampleswere dated according to the general principles ofNoyes et al,20 as more fully detailed in Hendricksonand Kempson's21 Decision Tree for Endometrial Dat-ing. As described by these authors, only the portionsof each biopsy from the functionalis layer were usedfor dating, and the most advanced area was used toassign the final date.

Ascites, Antibodies, and Lectins

Mouse ascites were generated by first priming maleand female retired breeders with the mineral oilpristane (Sigma Chemical Co., St. Louis, MO), fol-lowed by intraperitoneal inoculation with 1 x 106 hy-bridoma cells (see Table 1) 10 to 14 days after thepristane injection. Mice were observed daily for as-cites development. When sufficient fluid had accu-mulated in the peritoneal cavity (5 ml or more) of each

Table 1. Frequency ofMAG-Positive Ascites in BalblcMice*

Mouse Sex Hybridoma MAG Reactivity

A male 9018BCDl female X18A4 +D2 +D3El female P20E2 +E3 +E4Fl male 9018F2

*Ascites were generated in mice as described in Materials andMethods. The resultant ascites were tested by immunohistochem-istry against known MAG-positive endometrial biopsies. -, no MAGstaining; ±, weakly MAG-positive; +, strongly MAG-positive. Onlyascites that were MAG-positive for endometrium were able to stainother MAG-containing tissues (see Table 3).

168 Kliman et alAJPJanuary 1995, Vol. 146, No. 1

mouse, it was tapped by inserting an 18-gauge 1 -inchneedle intraperitoneally. Mice were repeatedlytapped in this way as long as the ascitic fluid reac-cumulated and the animal remained healthy. A varietyof ascites from both private and commercial sourceswas also tested (Table 2). Mouse ascites were usedat 1:1000 dilutions. Anti-A, -B, and -O (H) monoclonalantibodies from Dako Corp. (Carpinteria, CA) wereused, the former at a dilution of 1:1000, the latter twoat 1:250. Biotinylated anti-mouse a, y, and p specificsecondary antibodies from Vector Laboratories (Bur-lingame, CA) were used at a final concentration of2.25 pg/ml, as instructed by the manufacturer. Bioti-nylated lectins from Vector were used at the followingdilutions: Ulex europaeus 1, 1:250; Dolichos biflorus,1:500; Bandeiraea simplicifolia, 1:200; and Vicia vil-losa, 1:200.

Immuno- and Lectin Histochemistry

Sections (5 p) from paraffin-embedded tissue wereplaced on glass slides previously coated with a filmof 1 % poly-d-lysine, 30,000 to 70,000 molecularweight (Sigma), dried for 30 minutes at temperaturesno greater than 60 C, and stored at room temperatureuntil used. Immunoperoxidase staining was per-formed by the avidin-biotin detection method with kitsfrom Vector with diaminobenzidine (Sigma) as thechromogen. The temperature at which the immu-noperoxidase procedure was performed proved tobe critical. If the slides were incubated for as little as10 minutes at 37 C in the blocking, primary, or sec-ondary antibody steps, as much as 80% of the stain-ing intensity was lost. The best protocol was achievedwhen blocking, secondary, and avidin-biotin complexsteps were performed at 22 to 24 C for 45 to 60 min-utes and the primary antibody step at 4 C overnight.Biotinylated lectins were incubated for either 1 hour atroom temperature (20 to 22 C) or 4 C overnight, fol-lowed directly by avidin-biotin complex and then dia-minobenzidine. Slides were counterstained with he-

matoxylin. Commercially available ascites, ascitesacquired as gifts, and ascites generated by pristanepriming followed by intraperitoneal hybridoma inocu-lation were tested for MAG activity (Tables 1 and 2).Those ascites that intensely stained select human en-dometrial gland Golgi were used to test the remainingspecimens. Control slides were incubated with eitherundiluted P3X63Ag8 mouse myeloma cell line super-natant (American Type Culture Collection, Bethesda,MD) or control ascites that were unreactive to knownpositive endometrial samples. All immunohistochemi-cal studies included known positive control slides.

Absorption Studies

For IgG absorption, 10 pi of MAG-positive and-negative ascites, diluted 1:10 in phosphate-bufferedsaline-0.1% bovine serum albumin, were added to10 pl of protein G Sepharose beads (Pharmacia, Pis-cataway, NJ), incubated overnight at 4 C with gentleagitation, centrifuged, and used for immunohisto-chemistry. For glycoconjugate absorption, MAG-positive ascites was diluted to an IgG concentrationof approximately 5 pg/ml. Between 2 and 4 p1 ofglycoconjugate were added to 100 p1 of antibodysolution to make a 1:1 molar ratio of IgG to glyco-conjugate (Sigma), incubated overnight at 4 C, andcentrifuged in a microcentrifuge at maximal speedfor 5 minutes, and the resultant supernatant wasused to perform MAG immunohistochemistry. Ab-sorption studies with red blood cells RBCs were per-formed in a similar manner. Whole blood from types0, A, B, and AB individuals were washed two timeswith cold phosphate-buffered saline (150 mmol/LNaCI and 10 mmol/L phosphate buffer, pH 7.4) bycentrifugation and pelleted, and the RBCs were di-luted 1:1 with cold phosphate-buffered saline. A totalof 50 pl of this 50% suspension was added to 100 plof diluted MAG antibody solution, incubated over-night at 4 C, and centrifuged, and the supernatantwas used to perform MAG immunohistochemistry.

Table 2. MAG Reactivity ofAscites From Various Sources*

Designation

H4C4H4.8Al 37Pooled ControlPooled ControlPooled ControlNS-1, pooledSP2/0NMA

Strain :Sex

CB6F1/J x C57Black/BALB/c:unknownBalb/c:femaleNot knownBalb/c:femaleBalb/c:femaleNot knownBalb/c:femaleBalb/c:femaleBalb/c:male and female

Source

J. Thomas August, Johns Hopkins UniversityStephen Warren, Yale UniversityAdeza Biomedical Corporation, Sunnyvale, CAZymed, San Francisco, CAImmunovision, Springdale, ARInternational Bio Sphere, Springdale, ARBiomakor, Rehovot, IsraelHarlan Bioproducts for Science, Indianapolis, INSigma Chemical Coop St. Louis, MO

MAGReactivity

*Ascites acquired as gifts and commercial sources were tested for MAG reactivity by immunohistochemistry against known MAG-positiveendometrial biopsies. -, no MAG staining; +, strongly MAG-positive.


Sialidase (from Clostridium perfringens) and a-N-acetylgalactosaminidase (from chicken liver; OxfordGlycosystems, Rosedale, NY)-treated ovine sub-maxillary mucin (OSM; a generous gift from V. RBhavanandan, Hershey Medical Center, Hershey,PA) was prepared according to the manufacturer'sinstructions. This treatment sequentially convertedthe OSM carbohydrate side chains from sialyl-(2'6)-N-acetyl-galactosamine (GaINAc)22 to GaINAc tocore protein free of carbohydrates.23 GaINAc(Sigma) was used at a concentration of 200 mmol/L.

Immunoblotting

Immediately after hysterectomy, a uterus from a pre-menopausal 33-year-old gravida 1, para 1 woman(day 17 by dates) with a history of dysmenorrhea andmenorrhagia was opened sterilely, and endometrialsamples were removed and fixed for later histologicaldating. The remaining endometrial tissue was curet-ted and placed in 10 ml of Dulbecco's modifiedEagles' medium, containing 25 mmol/L glucose and25 mmol/L HEPES supplemented with gentamicin (50pg/ml), glutamine (4 mmol/L), and 20% (v/v) heat-inactivated fetal calf serum, and transported to thelaboratory. Pathological examination of the uterus re-vealed two small leiomyomas and day-15 endome-trium by histological criteria. The freshly removed en-dometrium, 0.9 grams, was homogenized with amodel PT 15/35 Brinkman homogenizer (Brinkman In-struments, Westbury, NY) for 3 minutes at a setting of3.5 in 2 volumes of room temperature 4X solubilizingbuffer (250 mmol/L Tris, 280 mmol/L sodium dodecylsulfate, 160 mmol/L dithiothreitol, 40% glycerol (v/v),and 0.02% bromphenol blue, pH 6.8) and stored at 4C until used. Saliva from healthy donors was collectedinto sterile 50-ml centrifuge tubes, immediately solu-bilized, 1:1 (v/v), with 2X solubilizing buffer with di-thiothreitol and used immediately or stored at -80 Cuntil used. From 25 to 100 pl of the solubilized en-dometrial tissue or saliva was electrophoresed on 6%

sodium dodecyl sulfate polyacrylamide gel electro-phoresis gels under reducing conditions. Rainbowprotein molecular weight markers (Amersham, Arling-ton Heights, IL) were run on each gel. The gels wereelectrotransferred to nitrocellulose (Schleicher andSchuell, Keene, NH) overnight at 0.5 volts at 4 C, in-cubated with MAG-positive ascites as primary anti-body (1:1000 dilution), undiluted P3X63Ag8 mousemyeloma cell line supernatant, or MAG-negative as-cites as the negative controls, with immunodetectionusing a biotinylated anti-mouse IgG or y-specific sec-ondary antibody (ABC Vectastain, Vector Laborato-ries) and developed with the substrate diaminoben-zidine.

Results

Identification and Characterization ofMurine Ascites Containing Anti-MAGAntibodies

Our initial identification of MAG staining in normal cy-cling human endometrium followed a detailed immu-nohistochemical analysis with A-137,19 an ascitesgenerated with the anti-oncofetal fibronectin hybri-doma FDC-6 (see Table 2). Initially, the highly specificstaining pattern described below led us to concludethat oncofetal fibronectin was synthesized and se-creted by the endometrium and that this adhesiveprotein was present on the endometrial surface at theinitiation of implantation. However, studies comparingA-137 ascites with pure FDC-6 antibody obtainedfrom hybridoma supernatant demonstrated that MAGimmunostaining was not due to the FDC-6 antibodyper se (Figure 1) but due to some other componentin the ascites. We have also confirmed that the pureFDC-6 specifically identifies trophouteronectin, animplantation site oncofetal fibronectin isoform derivedfrom trophoblasts.24 In addition, the pure FDC-6 doesnot react with endometrium (Figure 1 B). These results

Figure 1. MAG immunohistochemistry with ascites versuLssupernatant fbrms of FDC-6. Serial sections of Jbrnmalin-fixed, paraffin-embedded normal human endometriumwas stained uwith FDC-6 monoclonal antibody made as anascites (A) or as a s.iperniatanit (B). Only the ascites form ofthe antibody (A) restulted in supranuclear immuinoreactiv-ity (MAG reactivity), suggesting that the anti-oncofetal fi-bronectin antibodies were not responsible for the MAGstaining. Note the supranuclear, cap shape of the MAG-reactive material (large arrowhead). Also note the distancebetween the MAG-reactive material atnd the apical portionof the endometrial gland cells (small arrowheads). Magni-fication, x350.


suggest that the anti-oncofetal fibronectin antibodieswere not responsible for the MAG staining.When y-, p-, and a-chain-specific secondary anti-

mouse antibodies were used in the immunohisto-chemical staining procedure against known MAG-positive tissue specimens, only y-chain-specificsecondary antibodies yielded Golgi staining with theMAG-positive ascites. This result suggested that theMAG staining component in the ascites was one ormore IgG antibodies generated in the mice before orduring the ascites manufacture. We confirmed thatthe MAG-reactive material is an IgG by demonstratingthat protein G sepharose beads were able to absorbout the MAG-reactive material from known positiveascites (data not shown).

To evaluate whether the generation of the MAG-active IgG was specific to the sex or strain of mice, orpossibly related to the hybridoma used to generatethe ascites, we injected male and female Balb/c micewith a variety of hybridomas and tested the resultantascites for MAG activity (Table 1). MAG-positive as-cites was generated in 50% of the female mice but innone of the male mice tested. MAG antibody produc-tion appeared to be independent of the hybridomainjected to promote the ascites. In addition to our ownascites, we tested ascites from a variety of sources(Table 2). Analogous to our result with A-137, we iden-tified a number of ascites with MAG immunoactivity,especially pooled ascites in which both MAG-reactiveand nonreactive ascites were likely to have beenblended together. This survey of MAG immunoreac-tivity demonstrated that MAG antibodies are not a uni-versal phenomenon but are produced only by certainmice.

MAG Expression in the Normal HumanMenstrual Cycle

Examination of 201 endometrial biopsies revealedthat MAG expression was linked to the specific cyclephase of each sample. Of the 201 biopsies examined,MAG staining was demonstrated in 81 (40.3%). Theearliest expression of MAG was on day 5 of the men-strual cycle (Figures 2 and 3). The staining first ap-peared in the endometrial glands in an apical pe-rinuclear location consistent with a Golgi stainingpattern (Figure 2B, C). Biopsies on day 6 began toshow similar perinuclear staining in the luminal sur-face glands. Between days 6 and 15, the Golgi stain-ing became more diffuse throughout the specimensexamined, and more intense in each individual gland.Maximal MAG staining for both endometrial glands

'and surface epithelium was on day 15 of an idealized28-day cycle (Figure 2C). Beginning on day 16, whensubnuclear vacuoles could be identified in greaterthan 50% of the glandular epithelial cells,21 the MAGstaining began to spread throughout the apical por-tion of the glandular cells, suggesting movement ofthe MAG epitope from the Golgi to secretory vesicles(Figure 2D). At this time, the MAG staining in the sur-face epithelium still remained within the Golgi. Be-ginning on day 17, MAG staining could be identifiedwithin the gland lumens (Figure 2E) and associatedwith scattered patches of apical surfaces on the lu-minal epithelial cells possibly deposited there inpart by secretion of the MAG-positive material out ofthe glands onto the luminal surface of the endome-trium (Figure 2F). On days 18 and 19, almost no MAGactivity remained in the Golgi of the gland cells butcould be identified more frequently in the Golgi of thecells of the luminal epithelium. MAG staining couldstill be identified on luminal apical surfaces on day 19.After day 19, no MAG staining could be identified inthe cells of the functionalis or surface epithelium, pre-sumably because it had all been secreted from theepithelial cells. Occasionally, deeper basalis glands,which appeared morphologically to be either inactiveor proliferative, could be MAG-positive during thelater part of the menstrual cycle. Menstrual endome-trium, although negative for MAG in the functionalislayer, also occasionally contained MAG-positivebasalis glands.

Occasionally an endometrial biopsy showed vari-ability of morphological date from one part of the bi-opsy to another. Interestingly, the MAG staining re-flected the date of each individual gland, not anaverage of the entire biopsy. For example, in biopsyspecimens that contained glands that spanned sev-eral days by histological dating, MAG staining alsospanned these dates (Figure 4). These results sug-gest that MAG expression is related to the specificstate of biochemical differentiation of each individualcell and not to the ambient conditions of the endo-metrium as a whole.

MAG Tissue Survey

We next examined a variety of other tissues to deter-mine whether the MAG epitope was specific to theendometrium or was more generally distributedthroughout the body (Table 3). MAG staining wasfound in a wide variety of tissues. It was mostlypresent in epithelial cells-especially secretoryepithelium-and was not detectable in mesenchymal


.44..vD

AJl1

I.A

4.4w I hr

h;

Figure 2. MAG expression throughout the menstrual cycle. Endometrial biopsies werefixed and processedfor MAG immunohistochemistry as de-scribed in Materials and Methods. A: No staining was noted in menstmual to day-4 endometria. B: Patchy, light Golgi staining (arrowheads) wasfirst identified in glands on day 5. C: Golgi staining progressively increased in intensity during the proliferative phase to day 15 and could be iden-tified in the Golgi of both enidometrial glands antd the suirface epithelial cells. The staining was seen only close to the nuclei in a supranuclear dis-tribution, leaving the apical cytoplasm free of staininlg (arrowheads). D: Day-16 atnd -17 endometria exhibited a loss of tight supranuclear glan-dular Golgi stainiing with an increase in diffuise apical staining (arrouheads), suggesting transport ofMAG to apical secretory vesicles. E: Glandlumens on days 17 anid 18 exhibited positively staininig material, suiggestitng secretion of the MAG epitope. F: Day-19 endometria exhibited almostnio glanidiular Golgi staining, had gland lumen secretion positivity, retained surftice epithelium Golgi staining, and also exhibited occasional strongstaining of the apical surf/ice o)f luminal epithelial cells (arrowheads). Beyond day 19, no staining could be identified in either the glands or sur-face epithelial cells. Jlagnifications, X 180 (A, C, and E) anid x450 (B, D, and F).

cells. One exception to this pattern was the stainingof the perivascular astrocytes of the central nervoussystem. One of the common links between all of thesites that were MAG-positive was the ability of thesecells to produce and secrete mucins. Mucins are highmolecular weight glycoproteins containing peptidecores surrounded by numerous oligosaccharidechains.25 The protein cores (ranging in molecularweight between 0.3 x 106 to 0.5 x 106 kd) are con-nected to each other via disulfide bonds, resulting incomplex supramolecular structures that range in mo-lecular weight between 106 and >45 x 106 kd.26 Toexamine whether the MAG antibody was recognizing

a mucin-like glycoprotein, we next performed im-munoblotting of MAG-containing extracts.

MAG Immunoblotting

When the endometrium from the day-15 hysterectomyspecimen was electrophoresed under reducing con-ditions and immunostained with either control anti-body or MAG-positive ascites, a high molecularweight smear (>200 kd) was seen only in the MAG-incubated blot (Figure 5). This smear is typical ofhighly glycosylated proteins and suggests a marked

....,....k. ;.:. T _......10011kOX ...... >.fi"

AP VV


Menses E uly | Late LuteaE Phase tI IFollicularI Folhicula I Folliculae LaeaPas

1 YiLLIL[1'tb4 Secretion

Kiti/TIIA HILIILL[\IILLLI5

Figure 3. MAG variation throughout the nor-mal menstrual cycle. A total of201 endometrialbiopsies were immunohistochemically stainedfor MAG. Relative expression (from no stainingto maximal staining) of MAG in glandularcells, surface epithelium, and lumens and onthe apical surface of the luminal epithelium inthe 81 biopsies that contained any MAG stain-ing was plotted throughout an idealized 28-dayovarian cycle. Data for the top three graphs(basal vacuolation, secretion, and pseudode- 1 2 3 4 5 6 7cidual reaction) were adapted from Noyes etall 7 and matched to the histological dating seen Menses Earlyin the analyzed biopsy specimens. FI F

heterogeneity of glycosylations on each polypeptidecore. Although many nonspecific lower molecularweight bands were noted in both the control andMAG-positive ascites blots, an additional band at ap-proximately 35 kd was also seen in the MAG-treatedmembrane. This band may either represent a proteo-lytic fragment of the higher molecular weight band ora distinct MAG-reactive protein.

Because we had shown by immunohistochemistrythat salivary gland Golgi stained with the MAG anti-bodies (Table 3), we postulated that saliva might con-tain MAG-reactive material. Saliva was collected fromseven healthy adult male and female donors, immu-noblotted, and stained with MAG-positive and MAG-negative ascites (Figure 6). All three male donors ex-pressed MAG-positive saliva, whereas one out of the

8 910 11 12 13 14 15 16 17 18 19 202122 23 24 25 2627 280 1 2 3 4 5 6 7 8 9 1011121314

Mid late Luteal PhaseeollicalarI Follicalar LaeaPas

four females had MAG-positive saliva. The MAG-reactive bands appeared as smears from a molecularweight of >200 kd to approximately 70 kd. Why didn'tall of the samples show MAG reactivity? When weobtained the blood types of the donors, we were sur-

prised to find that only samples from blood type Aindividuals, male and female, contained any MAG-reactive material. At first we considered that thiswould limit the usefulness of the antibody, but furtheranalysis proved that this result was actually an op-portunity to elucidate the nature of the MAG epitope.This blood type association suggested that the MAGepitope may be related to the blood group A oligo-saccharide. We therefore next set out to determinethe oligosaccharide specificity of the MAG-positiveascites.


Figure 4. MAG expression in endometrial glands is dependent on the maturity of each individual gland. Early proliferative (A and B) and earlysecretory (C anid D) enidometrial biopsies werefixed and staitnedforMAG as described in Materials and Methods. Every endometrial biopsy showsa small degree of variability throu.ghouit the specimeni. When MAG expression uwas analyzed anid compared with the specific histological pattern ofeach gland ne jbund that MAG expressioni varied in relation to the histological date of each portion of the biopsy. A: Simple, early proliferativegland with surrounding stroma containinzg neutrophils (arrows), consistent with day 6. Onlv two glandular epithelial cells show any supranuclearMAG reactivity (arrou.heads). Note dauighter cells (d) and prophase ntucleuis (p), consistent with proliferation. B: Folded, early proliferative glandconsistenzt with day 8from samne biopsy as shown in A. Virtueally all of the glanduilar epithelial cells now shou supranuclear MAG reactivity. C:Interval gland with a feu', minute subnuclear vacuoles (arrowheads), consistent with a day-15 biopsy. Note uniform, strong supranuclear MAGstaining throtughout the entire gland. D: Early secretory gland showing almost uniform suibnuclear vacuolesfrom a regioni near the gland shownin C. Thefact that the niuclei are not lined uip, but greater than 50% of the cells are vacuiolated (eg, arrouheads) makes this gland consistent uithday 16 (possibly 16.5). Utnlike the day-15 gland in this same biopsy (C), this gland shous only afew cells with residual, tightly supranuclear MAGreactivity (arrous). Somne cells show light, diffuse MAG reactivitv (triangles; see also Figuire 2D), suggestive of movement of the epitope into apicalsecretory resicles. Magnificationi, X 350.

Absorption Studies: Glycoproteins, RBCs,and GalNAc

Blood group-related oligosaccharides can be foundon erythrocytes and glycoproteins, especially the mu-cin glycoproteins. Mucins are glycosylated through0-linked carbohydrates attached to either a serine orthreonine in the protein core.27 These oligosacchar-ides are similar to the carbohydrates seen on eryth-rocytes and terminate with several sugars, includinggalactose, GaINAc, fucose, and sialic acid. To de-termine whether a carbohydrate structure is a com-ponent of the MAG epitope, we next performed a se-ries of absorption experiments with a variety of naturaland enzymatically produced glycoconjugates as wellas erythrocytes of the major blood types.We examined the ability of N-linked, 0-linked, and

erythrocyte glycoconjugates, as well as the mono-saccharide GaINAc, to neutralize MAG antibody fromknown positive ascites (Table 4), by immunohisto-chemical analysis of known MAG-positive endome-

trial biopsies. None of the N-linked glycoproteins ab-sorbed out the MAG activity. Heparin and chondroitinsulfate, two members of the proteoglycan family, alsohad no effect. Types and I-S bovine submaxillarymucins (which contain 12 and 5% terminal sialic acid,respectively) did not absorb any MAG activity. In con-trast, type Ill porcine stomach mucin, which containsless than 1% sialic acid, was able to absorb out all ofthe MAG activity. Because this mucin contains oligo-saccharides that resemble human blood group oli-gosaccharides,28 we next examined the ability oferythrocytes (RBCs) from A, B, 0, and AB donors toabsorb out the MAG antibody activity. 0 erythrocytescontain the H oligosaccharide and terminate with a,B-linked galactose. B erythrocytes express the H oli-gosaccharide with an additional terminal a-linked ga-lactose. A erythrocytes differ from B cells by termi-nating with an a-linked GaINAc in place of galactose.AB cells contain a roughly equal mixture of A and Boligosaccharides. 0 and B RBCs had no effect on theanti-MAG activity, whereas A RBCs absorbed out all


Table 3. Tissue Survey ofMAG Reactivity*

Tissue

Endometrial glands (menstrual cycledependent)

Cervical glandsFallopian tube epitheliumOvaryProstate glandsSeminal vesicle glandsLacrimal glandsPancreatic acinar cellsPancreatic isletsLiver (parenchyma)Intrahepatic bile ductsGall bladder epitheliumCommon bile duct epitheliumStomach mucosaDuodenal mucosaDuodenal submucosal glandsSmall intestine mucosaLarge intestine mucosaKidney (proximal collecting ducts only)Salivary glands (exocrine cells)Breast (lobules, lactating, or nonlactating)Breast (ducts)Sweat glands (eccrine only)Nasal mucosa glandsBronchial glandsSmooth muscleSkeletal muscleEndothelial cellsAdipose tissueLymph nodeSpleenSquamous epithelium and mucosa (skin,

oral, and vaginal)Brain (only perivascular astrocytes

Staining

+

+

+

++

+t

*Tissues only from blood group A patients were processed andstained for MAG as described in Materials and Methods.

tFor an unknown reason, occasional squamous epitheliumspecimens showed MAG reactivity in between the keratinocytes.

of the MAG activity. AB RBCs absorbed out most, butnot all, of the MAG activity.These results suggested that GaINAc must be at

least part of the MAG epitope. We therefore next in-vestigated whether GaINAc alone would be sufficientto compete with the MAG activity. When GaINAc wasadded to the primary antibody at a concentration of200 mmol/L, there was no effect, suggesting that, al-though GaINAc is necessary for MAG binding, it is notsufficient. Because the first carbohydrate in each0-linked glycoprotein is GaINAc linked to either aserine or threonine residue, we next evaluatedwhether a GaINAc linked to the protein core was suf-ficient to absorb out MAG or whether the full oligo-saccharide with GaINAc as the terminal sugar wasnecessary for MAG absorption. To test these two pos-sibilities, we used ovine submandibular mucin (OSM).OSM has the particular characteristic that it only con-tains the disaccharide sialyl-(2'6)-GaINAc linked toeither a serine or threonine. By treating OSM with ei-

A B

10 25 50 100

*-O 200

-*- 92.5- 69

-o 46

-*- 30

10 25 50 100

Figure 5. Immunoblot showing MAG expression by day-15 humanendometrium. Endometrial tissue (0.9 g) removed from a hysterec-tomy specimen was solubilized with sodium dodecyl sulfate and re-ducing agents. Amounts of 10, 25, 50, and 100 jl were loaded perlane in two sets of lanes. After electrophoresis, the proteins were elec-trotransferred from the gel to nitrocelluilose and immunochemicallystained with either (A) control mouse myeloma supernatant or (B)with an ascites with anti-MAG activity. Notice the diffuse band in gelB corresponding to the high molecular weight (>200 kd) MAGmucin-like glycoprotein. An approximately 35 kd band can also beseen in gel B, possibly representing a proteolytic fragment of themucin-like glycoprotein or a distinct MAG-reactive protein. Molecularweight markers are in kd.

ther sialidase alone or sialidase with a-N-acetyl-galactosaminidase, one can produce OSM with onlyGaINAc attached or with no sugars attached. Allforms of OSM, including native and enzymaticallytreated forms, were unable to absorb out the MAGactivity (Table 4), suggesting that a single GaINAcattached to a mucin protein core is not the MAGepitope. It appears, therefore, that the MAG epitopeis closely related to the blood group A oligosaccha-ride that is found on type A RBCs and material con-taining mucin oligosaccharides that terminate with aGaINAc.

MAG Staining and the Blood Typeof the Patient

We speculated that the apparent A oligosaccharidespecificity of MAG might explain why we could iden-tify MAG staining in only 40.3% of the endometrialbiopsies examined (see above). When we sorted theimmunohistochemical results according to bloodtype, the correlation became clearer (Table 5). We

A

I< - J Mz U. 3 <= &n Li j X U X

B

udcnc U. 3.4~-- U-L ..J <Wc .4-j 9 U


< 200

*<- 92.5-69

Figure 6. Immunoblot showing MAG expressionin saliva. Saliva from seven normal adults, threemale (HJK, RFF, and CLW) and four female

------- 46 (STS, ELM, LBS, and KY4N), was immunoblottedin two parallel sets for MA,G expression as de-scribed in Materials and Methods. The sets wereseparated and reacted with either MAG-positiveascites (A) or MAG-negative ascites (B). A Sev-eral of the samples revealed a MAG-reactivesmear between >200 kd to approximately 70 kd.

< 30 Only blood type A individuals (HJK, STS, RFF,and CLW) expressed the high molecular weightsmear. Although weak in this particular blot, adistinct band was always seen with the RFFsample. ELM is blood type B, whereas LBS andKAN are blood type 0. Notice that the approxi-mately 35-kd band that was seen in the endome-trial sample (Figure 5B) was not seen here. B:No high molcecular weight smear was seen in theMAG-negative blot for any of the samples. Verylight, identical nonspecific bands were noted inboth MAG-positive and MAG-negative blots.

Table 4. Absorption Studies

AbsorbsGlycoconjugate/Sugar Linkage Terminal carbohydratest MAG?

Ovalbumin GIcNAc-Asp Gal, GIcNAc, Man Noal-Acid glycoprotein GIcNAc-Asp Gal NoBovine serum albumin GIcNAc-Asp Gal, GIcNAc, Man NoHeparin Xylose-Ser GIcNSO4, GIcUA, GIcNAc NoChondroitin sulfate Xylose-Ser GIcUA NoType l* bovine submaxillary mucin GaINAc-Ser(Thr) Sialic Acid (12%), Gal, GIcNAc, GaINAc NoType I-S* bovine submaxillary mucin GaINAc-Ser(Thr) Sialic Acid (5%), Gal, GIcNAc, GaINAc NoType III* porcine stomach mucin GaINAc-Ser(Thr) Sialic Acid (1%), Gal, GIcNAc, GaINAc >95%Type 0 RBCs GaINAc-Ser(Thr) Galf-GIcNAc NoType B RBCs GaINAc-Ser(Thr) Gala-Galf3-GIcNAc NoType A RBCs GaINAc-Ser(Thr) GalNAca-Galf3-GIcNAc >95%Type AB RBCs GaINAc-Ser(Thr) Gala-Gal3-GIcNAc -50%

GalNAca-Gal3-GIcNAcGaINAc NoOSM GaINAc-Ser(Thr) Sialyl-GaINAc- NoOSM, Sialidase Rx'd GaINAc-Ser(Thr) GaINAc NoOSM, Sialidase & GalNAcase Rx'd Ser(Thr) None No

t Gal, galactose; GIcNAc, N-acetyl-glucosamine; Man, mannose; GIcNSO4, glucosamine sulfate; GIcUA, glucuronic acid; Sialyl, N-acetylneuraminic acid29

*Sigma designation

were able to positively determine the blood types ofonly 90 of the patients examined. Of these, 29 wereA, 13 were B, 45 were 0, and 3 were AB, in generalagreement with the distribution of ABO blood types inthe United States.30 All of the biopsies from the bloodgroup A patients between cycle days 5 and 19 wereMAG (Golgi)-positive (23 of 23). In addition, 5 of the10 blood type A and 2 of the 2 blood type AB biopsiesbetween days 17 and 19 had apical surface MAGreactivity. In contrast, all the biopsies from patientswith blood groups B and 0 (n = 58) were MAG

(Golgi)-negative at all times of the cycle, as were allday-20 to -28 and menstrual biopsies from all bloodtypes. Interestingly, 3 non-A biopsies from days 17 to19, 1 from a blood group B patient and 2 from bloodgroup 0 patients, did show MAG reactivity but only inthe region of the apical surface of the luminal epithe-lium. These results suggest that, although bloodgroup B and 0 patients are not capable of generatingthe MAG epitope in their Golgi, they may be capableof generating the epitope on their surface epitheliumbetween days 17 and 19 of the menstrual cycle.


Table 5. MAG Staining and Patient Blood Type*

MAG-Positive MAG-PositivePatient Number (%) day-5 to -19 menstrual and day Day-17 to -19 endometriablood of patients endometria -20 to -28 endometria with MAG-positive surfacetype (total = 90) (No. positive/total) (No. positive/total) (No. positive/total)

A 29 (32%) 23/23 0/6 5/10B 13 (14%) 0/10 0/3 1/20 45 (50%) 0/30 0/15 2/6AB 3 (3%) 3/3 0/0 2/2

*Patient blood types of 90 of the total of 201 biopsies examined were able to be obtained from Blood Bank records. The number of MAG-positive biopsies out of the number of biopsies in each day range is expressed for days 5 to 19 and menstrual and days 20 to 28. A separatelisting of the biopsies between days 17 and 19 that exhibited any MAG epitope staining of the apical surface luminal epithelium is given in thelast column. The amount of surface epithelium in each biopsy varied greatly and may partly explain why not all of these day-17 to -19 biop-sies expressed luminal apical MAG.

Anti-Blood Group and LectinImmunohistochemistry

Because we now had evidence that the MAG epitopewas similar to the blood group A oligosaccharide, weinvestigated whether other anti-A antibodies or anti-Alectins might stain human endometrium in a fashionsimilar to the MAG antibody. Anti-A monoclonal an-tibody stained tissue from blood group A patients butnot from B or 0 patients (data not shown). When theseendometrial biopsies were examined, staining couldbe identified in the endothelial cells, RBCs, lumens ofthe vessels, surfaces of the glands, surface epithe-lium, and occasionally the Golgi of the glandular epi-thelial cells. Antibodies to blood group B and 0 an-tigens stained the endothelial cells and RBCs of thecorresponding blood type patients, stained the sur-faces and occasionally the lumens of the glands, butdid not stain the Golgi of any of the endometrial bi-opsies examined. Anti-A lectins (D. biflorus, V villosa,and B. simplicifolia) produced similar results withendothelial cells, vessel lumens, and cell surfaces ofthe endometrial glandular epithelium. Only V villosastained any of the Golgi in the endometrial epithelialcells. Although GaINAc had no inhibitory effect whenadded to the MAG-positive ascites at 200 mmol/L(Table 4), GaINAc at this same concentration com-pletely abolished the lectin staining of all the tissuesamples examined. These results suggest that MAGstaining is a specific subset of the staining seen withanti-A antibody and V villosa and that neither anti-Aantibodies nor lectins can selectively react with theMAG epitope.

DiscussionExamination of an unexpected Golgi-reactive immu-noglobulin in mouse ascites has led us to character-ize a menstrual cycle-dependent mucin-like glycop-

rotein in human endometrium. MAG first appears inendometrial gland Golgi as early as day 5, is secretedon days 16 to 18, and becomes apparent on the api-cal aspect of the surface epithelium only on days 17to 19. The antibody that recognizes the MAG sub-stance is an IgG that can be identified in high levelsin approximately 50% of female mice that are inducedto form pristane-hybridoma ascites. None of the fivemale mice we used to make ascites formed MAG an-tibodies. The menstrual cycle distribution of MAG ap-peared similar when different active ascites wereused to perform the immunohistochemical studies.Our data do not exclude the possibility that differentascites react with slightly different epitopes, or evendifferent proteins, albeit with the same endometrialstaining pattern.

Our results suggest that MAG antibodies bind to amucin-associated moiety, on the basis of the followingobservations: 1), the tissue distribution survey dem-onstrated MAG in many cells and tissues of endoder-mal origin that are known to secrete mucins31; 2), theabsorption results that showed that only material thatcontained mucins with blood group A oligosacchar-ides were effective in absorbing out the MAG activity,specifically, an extract of porcine gastric mucins andA and AB erythrocytes (which contain the transmem-brane mucin glycophorin); and 3), the immunoblotsthat exhibited high molecular weight smears, a pat-tern typical of highly glycosylated glycoproteins suchas mucins. An interesting exception to the endoder-mal association of MAG-reactive tissues was the find-ing of MAG reactivity in the perivascular astrocytes ofthe brain. Although not normally thought of as a siteof mucin production, a high molecular weight mucinrelated to the Tamm-Horsfall protein has been iden-tified at this site32 where it may function in the blood-brain barrier.

Although we have demonstrated the critical natureof a terminal GaINAc residue in the MAG epitope, it


appears that the MAG antibody is identifying morethan just GaINAc inasmuch as a single GaINAc resi-due attached to a mucin peptide backbone or Gal-NAc alone is not sufficient to absorb or compete withthe MAG activity. In addition, anti-A antibody andGaINAc-reactive lectins do not stain in a MAG-likepattern. It appears, therefore, that MAG reactivity is asubset of more general GaINAc probes. The MAGepitope may include more proximal carbohydratesand possibly a portion of the mucin polypeptide back-bone.

Previous Descriptions of MAG-LikeActivity and Blood Group A-Like Antigens

MAG-like antibodies appear to have been noted pre-viously, sometimes without inferences being drawn bythe investigators. The first example of an endogenousantibody in mouse ascites with glycoprotein reactivitywas described by Gooi and Feizi.33 These workersdemonstrated that many samples of ascites con-tained natural antibodies that reacted with fetal gly-coproteins (meconium) and other blood group sub-stances. The first definitive characterization of anendogenous mouse ascites Golgi-reactive antibodywas made by Smith et al,34 who showed that ascitesand serum from multiple strains of mice containedendogenous IgG antibodies to Golgi-specificepitopes. They proved by monensin treatment andelectron microscopy that these antibodies reactedspecifically with the Golgi of pancreatic acinar cellsand with many other exocrine tissues in the rat. In-terestingly, they observed that the titer of these anti-Golgi antibodies increased with the length of time themice were kept in captivity, suggesting to them thatthe antigenic stimulation came from some environ-mental exposure. It is also of interest to note that mu-cins are particularly immunogenic in mice.27The next demonstration of MAG-like activity came

in a series of papers from several laboratories thatwere working on P-glycoprotein,35-39 an integralmembrane pump that mediates multidrug resistancein tumor cells. The common finding in this work wasthat ascites antibodies from two different commercialsources, purportedly specific for P-glycoprotein,stained the Golgi of endometrial glands. AlthoughP-glycoprotein was known to be a surface glycopro-tein, these workers speculated that the endometriumhad a specialized Golgi form of the protein. As thiswork developed, the staining seemed to be related tothe blood group of the patient from whom the tissuewas derived.35'36 Interestingly, Axiotis et a137 demon-strated that their P-glycoprotein staining was men-

strual cycle dependent in the endometrium and thatstaining peaked on days 15 and 16 of the cycle. Themeaning of this work became clear from the elegantstudies of Finstad et alV" who demonstrated that sev-eral of the particular lots of ascites antibodies used bythese P-glycoprotein workers contained high titers ofan endogenous IgG anti-A blood group-specific an-tibody that was responsible for the Golgi staining pat-tern described previously. The MAG antibody we de-scribe here appears to be similar, if not identical to,that described by Finstad et al.40 In spite of the cau-tion implied by that work, papers continue to be pub-lished showing Golgi staining for P-glycoprotein in en-dometrial gland Golgi with the same source ofantibodies.41

Does Endometrial MAG ProductionCorrelate with Other EndometrialMucin Epitopes?

Other mucin glycoprotein epitopes have been iden-tified in the human endometrium.42 TAG-72 (tumor-associated glycoprotein) is a mucin glycoproteinfound in malignant and normal human tissues. In thehuman endometrium it is not detectable by immuno-histochemistry during the proliferative phase, first ap-pears around or shortly after the time of ovulation,then peaks during the late luteal phase.43'44 The cel-lular distribution of TAG-72 appeared to be apical andcytoplasmic. CA-125, an epitope on a mucin glyco-protein found in many epithelial tissues of MOllerianorigin,45 has been shown to have maximal expressionduring the early proliferative and mid-secretoryphase, times when MAG levels are low, and lowestduring the early secretory phase, a time at which MAGexpression is high.4647 In addition to the temporaldifferences between CA-125 and MAG, CA-125, un-like MAG's supranuclear distribution, is present in theinfranuclear region of the epithelial cells during theproliferative phase. A mucin sialoglycoprotein iden-tified by monoclonal antibody D9B1 has also beenidentified in human endometrium.48 The D9B1epitope first appears on approximately day 15 in abasal distribution, peaks on day 20, and then is seenin the gland lumens on days 21 and 22, a patterndifferent from that of MAG (Figure 2). In addition tothese specific mucin glycoproteins, Ravn et a149 haveshown that human endometrium expresses a mucintype-3 chain A antigen in a menstrual cycle-dependent fashion. Using the HH5 antibody, theseworkers demonstrated that this A antigen was firstexpressed in mid-proliferative endometrium, peaked


during late proliferative and interval phases, but re-appeared during the mid-secretory phase, making itsimilar but not identical to the MAG epitope. All ofthese epitopes, although apparently part of high mo-lecular weight glycoproteins, do not appear to parallelexactly the expression of MAG in the human endo-metrium.

In contrast, Rye et al,50 using two different mono-clonal antibodies, have recently described the immu-nohistochemical expression of polymorphic epithelialmucin in the human endometrium. One antibody(NCRC 1 1), which reacts with a human breast cancer-associated antigen,51 showed staining only in the lu-teal phase, much like TAG-72 and D9B1. The otherantibody (HMFG 1), which reacts with the milk fatglobule membrane antigen,52 stained endometrialglands between days 5 and 18, as does MAG. Thesimilarity in menstrual cycle expression betweenthese two epitopes suggests that MAG may be re-lated or identical to what has now come to be knownas the MUC-1 family of mucin glycoproteins.27The MUC-1 mucins are a family of highly glycosy-

lated, high molecular weight (>200 kd) glycoproteinspresent on the surfaces of many epithelial cells.27'28The basic MUC-1 structure consists of a straight pro-tein backbone that contains many highly glycosy-lated regions. Carbohydrates are attached to the mu-cin protein backbone via an al,3 linkage betweenN-acetyl galactosamine and the oxygen atom ofserine or threonine. Many of the carbohydrates foundon epithelial mucin glycoproteins are similar to theblood group oligosaccharides. MUC-1 synthesis hasbeen shown to be hormonally regulated in mammaryepithelial cells.53 More recently, additional types ofMUC-1 mucin molecules have been described, themembrane-associated mucins.54 These includesuch molecules as leukosalin (CD43), glycophorin,ascites sialoglycoprotein-1 (ASGP-1), epiglycanin,and episialin. Episialin, like some of the mucins de-scribed above, is encoded by MUC-1 and has beendefined by a variety of names, including polymorphicepithelial mucin (PEM) and epithelial membrane an-tigen (EMA). Unlike regular mucins, which are se-creted by a variety of cells and combine to form largepolymeric gels in the extracellular space, mem-brane-associated mucins remain attached to the sur-faces of cells where they may act as ligands for avariety of receptor molecules.55 For example, amembrane-associated mucin (GlyCAM-1) found onthe apical surface of endothelial cells of high endo-thelial venules in lymph nodes is the ligand for lym-phocyte L-selectin,56 whereas leukosalin, a neutro-phil membrane-associated mucin, interacts withthe endothelial lectin domain-containing protein

CD62.55 57 Once these mucin-lectin interactions areinitiated, tighter, integrin-mediated interactions areestablished.58

A Role in Implantation?

Could similar mucin-lectin interactions take placebetween the conceptus and the endometrium duringthe earliest phase of implantation? The work ofLindenberg suggests this possibility.' 1012 Like thelacto-N-fucopentaose epitope found in the mouseendometrium, one function for mucin glycoproteinssuch as MAG may be to promote embryo-endome-trial interaction. This hypothesis proposes that lectin-like proteins exist on the human preimplantation em-bryo, as has been shown for the macrophageasialoglycoprotein-binding protein,59 and that per-haps this lectin specifically reacts with the MAGepitope. If this is so, how do type 0 and B womenacquire the terminal N-acetyl galactosamine neces-sary to make up this epitope? Our finding of the MAGepitope on the apical surface of the luminal epithe-lium in some type 0 and B patients suggests thepossibility of an enzyme or enzymes that are capableof modifying or adding to pre-existing oligosacchar-ides on the cell surface. Glycotransferases, usuallythought of as Golgi-localized oligosaccharide syn-thesizers, are also found on cell surfaces and theextracellular space.60-2 Therefore, it is possible thata cell surface glycotransferase in a non-A endome-trium may be able to convert a cell surface mucinoligosaccharide to a MAG mucin-containing oligo-saccharide. Alternatively, the presence of apical sur-face epithelial staining in type 0 and B patients mayreflect the presence of additional IgGs in the ascitesthat are specific for these two blood types. Additionalwork will be necessary to assess these possibilities.

The MAG glycoprotein, observed at first as anartifact, appears to be a hormonally regulated mu-cin-like glycoprotein expressed on the endometrialsurface during the purported implantation window.Like the lacto-N-fucopentaose epitope described inthe mouse,1 the essential GaINAc epitope recog-nized by the MAG antibody may be important for theinitial interaction between the human conceptus andthe endometrium. Interestingly, endometrial integrinsf3363 and a464 are also expressed during the implan-tation window. Thus, as has been demonstrated forneutrophil-endothelial interactions,55,57,58 sequentialinteractions between endometrial mucins and tro-phoblast lectins followed by integrin-mediated adhe-sion could contribute to a fertile, receptive endo-metrium. Prospective studies correlating MAG


expression with ovulation induction and assisted re-production technology protocols should help us un-derstand the relationship between the MAG epitopeand successful implantation in humans.

AcknowledgmentsWe thank Drs. Frederick Naftolin, Jon S. Morrow,David L. Olive, Alan S. Penzias, David Keefe, andEugene Davidson for helpful comments on this work;Minxia Liu, Edward Buchanan, and Cai-Liang Wangfor their excellent technical assistance; the histotech-nologists in the Departments of Pathology at the Hos-pital of the University of Pennsylvania and the Yale-New Haven Hospital for their continuous efforts inpreparing unstained slides for this work; Drs. J.Thomas August, Stephen Warren, and the many com-panies for supplying ascites samples; Dr. V. P. Bha-vanandan for supplying the OSM; Dr. Bernice W. Kli-man for proofreading the manuscript; and especiallySandra T. Stein for her unceasing willingness to sup-ply repeated control endometrial samples for thesestudies.

References

1. Lindenberg S: Ultrastructure in human implantation:transmission and scanning electron microscopy. Bail-liere's Clin Obstet Gynaecol 1991, 5:1-14

2. Vanderhyden BC, Armstrong DT: Decreased embry-onic survival of in-vitro fertilized oocytes in rats is dueto retardation of preimplantation development. J Re-prod Fertil 1988, 83:851

3. Navot D, Scott RT, Droesch K, Veeck LL, Liu HC, Rosen-waks Z: The window of embryo transfer and the effi-ciency of human conception in vitro. Fertil Steril 1991,55:114-118

4. Navot D, Bergh PA, Williams M, Garrisi GJ, Guzman I,Sandler B, Fox J, Schreiner EP, Hofmann GE, GrunfeldL: An insight into early reproductive processesthrough the in vivo model of ovum donation. J Clin En-docrinol Metab 1991, 72:408-414

5. Harper MJ: The implantation window. Baillieres ClinObstet Gynaecol 1992, 6:351-371

6. Aplin JD: Glycans as biochemical markers of humanendometrial secretory differentiation. J Reprod Fertil1991, 92:525-541

7. Carson D, Farrar JD, Laidlaw J, Wright DA: Selectiveactivation of the N-glycosylation apparatus in uteri byestrogen. J Biol Chem 1990, 265:2947-2955

8. Munakata H, Isemura M, Yosizawa Z: Enzymatic sulfa-tion of exogenous high molecular weight glycopep-tides by microsomal fraction of the rabbit uterine en-dometrium. J Biol Chem 1985, 260:6851-6856

9. Mani SK, Carson DD, Glasser SR: Steroid hormonesdifferentially modulate glycoconjugate synthesis andvectoral secretion by polarized uterine epithelial cellsin vitro. Endocrinology 1992, 130:240-248

10. Enders AC, Schlafke S: Surface coat of the mouseblastocyst and the uterus during the preimplantationperiod. Anat Rec 1974, 181:31-46

11. Hewitt K, Beer AE, Grinnell F: Disappearance of an-ionic sites from the surface of the rat endometrial epi-thelium at the time of blastocyst implantation. Biol Re-prod 1979, 21:691-707

12. Anderson TL, Olsen GE, Hoffman LH: Stage specificalterations in the apical membrane glycoproteins onendometrial epithelial cells related to implantation inrabbits. Biol Reprod 1986, 34:701-720

13. Chavez DJ, Anderson TL: The glycocalyx of mouseuterine luminal epithelium during estrus, the peri-implantation period, and delayed implantation. I. Ac-quisition of RCA-1 binding sites during pregnancy.Biol Reprod 1985, 32:1135-1142

14. Lindenberg S: Experimental studies on the initial tro-phoblast endometrial interaction. Dan Med Bull 1991,38:371-380

15. Kimber SJ, Lindenberg S: Hormonal control of a car-bohydrate epitope involved in implantation in mice. JReprod Fertil 1990, 89:13-21

16. Lindenberg S, Sundberg K, Kimber SJ, Lundblad A:The mil oligosaccharide, lacto-N-fucopentaose 1, in-hibits attachment of mouse blastocysts on endome-trial monolayers. J Reprod Fertil 1988, 83:149-158

17. Sato M, Muramatsu T, Berger EG: Immunological de-tection of cell surface galactosyltransferase in preim-plantation mouse embryos. Dev Biol 1984,102:514-518

18. Miller DJ, Macek MB, Shur BD: Complementarity be-tween sperm surface f-1,4-galactosyltransferase andegg-coat ZP3 mediates sperm-egg binding. Nature1992, 357:589-593

19. Feinberg RF, Kliman HJ, Lockwood CL: Is oncofetal fi-bronectin a trophoblast glue for human implantation?Am J Pathol 1991, 138:537-543

20. Noyes RW, Hertig AT, Rock J: Dating the endome-trium. Fertil Steril 1950, 1:3-25

21. Bennington JL: Surgical pathology of the uterine cor-pus. Major Prob Pathol 12:80-85

22. Graham ERB, Gottschalk A: Studies on mucoproteins.I. The structure of the prosthetic group of ovine sub-maxillary gland mucoprotein. Biochim Biophys Acta1960, 38:513-524

23. Gottschalk A, Schauer H, Uhlenbruck G: Immunologi-cal properties of ovine submaxillary glycoprotein.Hoppe-Seyler's Z Physiol Chem 1971, 352:117-124

24. Feinberg RF, Kliman HJ: Tropho-uteronectin (TUN): aunique oncofetal fibronectin deposited in the extracel-lular matrix of the tropho-uterine junction and regu-lated in vitro by cultured human trophoblasts. Tropho-blast Res 1993, 7:167-179

25. Silbergerg A, Meyer FA: Structure and function of mu-cus. Mucus in Health and Disease II. Edited by EN


Chantler, JB Elder, and M. Elstein. New York, PlenumPress, 1982, p 53-74

26. Carlstedt I, Sheehan JK: Macromolecular propertiesand polymeric structure of mucus glycoproteins. CibaFound Symp 1984, 109:157-172

27. Devine PL, McKenzie FC: Mucins: structure, function,and associations with malignancy. BioEssays 1992,14:619-625

28. Rose MC: Mucins: structure, function, and role in pul-monary diseases. Am J Physiol 1992, 263:L413-L429

29. Roden L: Structure and metabolism of connective tis-sue proteoglycans. The Biochemistry of Glycoproteinsand Proteoglycans. Edited by WJ Lennarz. New York,Plenum Press, 1980, pp 267-371

30. Walker, RH (Ed): ABO, H and P Blood Groups andStructurally Related Antigens: Technical Manual, ed11. Bethesda, MD, American Association of BloodBanks, 1993, p 204

31. Strous GJ, Dekker J: Mucin-type glycoproteins. CritRev Biochem Mol Biol 1992, 27:57-92

32. Zalc B, Collet A, Monge M, Ollier HMP, Jacque C, Hart-mann L, Baumann NA: Tamm-Horsfall protein, a kidneymarker is expressed on brain sulfogalactosylceramide-positive astroglial structures. Brain Res 1984, 291:182-187

33. Gooi HC, Feizi T: Natural antibodies as contaminantsof hybridoma products. Biochem Biophys Res Com-mun 1982, 106:539-545

34. Smith ZD, D'Eugenio-Gumkowski F, Yanagisawa K,Jamieson JD: Endogenous and monoclonal antibod-ies to the rat pancreatic acinar cell Golgi complex. JCell Biol 1984, 98:2035-2046

35. Weinstein RS, Coon JS, Dominquez JM, Jakate SM,Lebovitz MD, Chang MA, Kluskens LF: Correlation be-tween ABO blood type and Golgi P-glycoprotein ex-pression in epithelia (letter). Lancet 1990, 336:54-55

36. Weinstein RS, Kuszak JR, Jakate SM, Lebovitz MD,Kluskens LF, Coon JS: ABO blood type predicts thecytolocalization of anti-P-glycoprotein monoclonal an-tibody reactivity in human colon and ureter. HumPathol 1990, 21 :949-958

37. Axiotis CA, Guarch R, Merino MJ, Laporte N, Neu-mann RD: P-glycoprotein expression is increased inhuman secretory and gestational endometrium. LabInvest 1991, 65:577-581

38. Axiotis CA, Monteagudo C, Merino MJ, LaPorte N,Neumann RD: Immunohistochemical detection ofP-glycoprotein in endometrial adenocarcinoma. Am JPathol 1991, 138:799-806

39. Finstad CL, Saigo PE, Rubin SC, Federici MG, Prov-encher DM, Hoskins WJ, Lewis JLJ, Lloyd KO: Immu-nohistochemical localization of P-glycoprotein in adulthuman ovary and female genital tract of patients withbenign gynecological conditions. J Histochem Cyto-chem 1990, 38:1677-1681

40. Finstad CL, Yin BW, Gordon CM, Federici MG, Welt S,Lloyd KO: Some monoclonal antibody reagents (C219and JSB-1) to P-glycoprotein contain antibodies to

blood group A carbohydrate determinants: a problemof quality control for immunohistochemical analysis. JHistochem Cytochem 1991, 39:1603-1610

41. Schneider J, Efferth T, Centeno MM, Mattern J,Rodriguez-Escudero FJ, Volm M: High rate of expres-sion of multidrug resistance-associated P-glycoproteinin human endometrial carcinoma and normal endome-trial tissue. Eur J Cancer 1993, 29A:554-8

42. Fay TN, Grudzinskas JG: Human endometrial pep-tides: a review of their potential role in implantationand placentation. Hum Reprod 1991, 6:1311-1326

43. Thor A, Viglione MJ, Muraro R, Ohuchi N, Schlom J,Gorstein F: Monoclonal antibody B72.3 reactivity withhuman endometrium: a study of normal and malignanttissues. Int J Gynecol Pathol 1987, 6:235-247

44. Lessey BA, Pidzola JA: Tumor-associated glycoprotein(TAG-72) in endometriotic implants. J Clin EndocrinolMetab 1993, 1075-1079

45. Kabawat SE, Bast RC, Welch WR, Knapp RC, ColvinRB: Immunopathologic characterization of a mono-clonal antibody that recognizes common surface anti-gens of human ovarian tumors of serous, endometri-oid, and clear cell types. Am J Clin Pathol 1983, 79:98-1 04

46. Weintraub J, Bischof P, Tseng L, Redard M, VassilakosP: CA 125 is an excretory product of human endome-trial glands. Biol Reprod 1990, 42:721-726

47. Zeimet AG, MOller-Holzner E, Marth C, DaxenbichlerG, Dapunt 0: Tumor marker CA-125 in tissues of thefemale reproductive tract and in serum during the nor-mal menstrual cycle. Fertil Steril 1993, 59:1028-1035

48. Aplin JD: Glycans as biochemical markers of humanendometrial secretory differentiation. J Reprod Fertil1991, 92:525-541

49. Ravn V, Teglbjaerg CS, Visfeldt J, Bock JE, SorensenH, Dabelsteen E: Mucin-type carbohydrates (type 3chain antigens) in normal cycling human endome-trium. Int J Gynecol Pathol 1992, 11:38-46

50. Rye PD, Bell SC, Walker RA: Immunohistochemical ex-pression of tumour-associated glycoprotein and poly-morphic epithelial mucin in the human endometriumduring the menstrual cycle. J Reprod Fertil 1993, 97:55 1-556

51. Ellis 10, Robins RA, Elston CW, Blamey RW, Ferry B,Baldwin RW: A monoclonal antibody, NCRC 11, raisedto human breast carcinoma. I. Production and immu-nohistological characterization. Histopathology 1984,8:501-506

52. Taylor-Papadimitriou J, Peterson J, Arklie J, Burchell J,Ceriani RL, Bodmer WF: Monoclonal antibodies toepithelium-specific components of the human milk fatglobule membrane: production and reaction with cellsin culture. Int J Cancer 1981, 28:17-21

53. Parry G, Li J, Stubbs J, Bissell MJ, Schmidhauser C,Spicer AP, Gendler SJ: Studies of Muc-1 mucin ex-pression and polarity in the mouse mammary glanddemonstrate developmental regulation of Muc-1 gly-cosylation and establish the hormonal basis for mRNA


expression. J Cell Sci 1992, 101:191-19954. Hilkens J, Ligtenberg MJL, Vos HL, Litvinov SV: Cell

membrane-associated mucins and their adhesion-modulating property. Trends Biochem Sci 1992, 17:359-363

55. Shimizu Y, Shaw S: Mucins in the mainstream. Nature1993, 366:630-631

56. Lasky LA, Singer MS, Dowbenko D, Imai Y, HenzelWJ, Grimley C, Fennie C, Gillett N, Watson SR, RosenSD: An endothelial ligand for L-selectin is a novelmucin-like molecule. Cell 1992, 69:927-938

57. Williams AF: Out of equilibrium. Nature 1991, 352:473-474

58. Butcher EC: Leukocyte-endothelial cell recognition:three (or more) steps to specificity and diversity. Cell1991, 67:1033-1036

59. Ii M, Kurata H, Itoh N, Yamashina I, Kawasaki T: Mo-lecular cloning and sequence analysis of cDNA en-coding the macrophage lectin specific for galactoseand N-acetylgalactosamine. J Biol Chem 1990, 265:11295-11298

60. Patt LM, Grimes WJ: Cell surface glycolipid and gly-coprotein glycosyltransferases of normal and trans-formed cells. J Biol Chem 1974, 249:4157-4165

61. Roth J, Greenwell P, Watkins WM: Immunolocalizationof blood group A gene specified al,3N-acetylgalac-tosaminyltransferase and blood group A substance inthe trans-tubular network of the Golgi apparatus andmucus of intestinal goblet cells. Eur J Cell Biol 1988,46:105-112

62. Lopez LC, Youaki A, Evans SC, Shur BD: Evidence fora molecular distinction between Golgi and cell surfaceforms of ,B1,4-galactosyltransferase. J Biol Chem1991, 266:15984-15991

63. Lessey BA, Damjanovich L, Coutifaris C, Castelbaum A,Albelda SM: Integrin adhesion molecules in the humanendometrium: correlation with the normal and abnormalmenstrual cycle. J Clin Invest 1992, 90:188-195

64. Lessey BA, Castelbaum AJ, Buck CA, Lei Y, YowellCW, Sun J: Further characterization of endometrial in-tegrins during the menstrual cycle and in pregnancy.Fertil Steril 1994, 62:497-506

Documents

Menstrual Cycle-Dependent Localization in Endometrium · HumanEndometrial Mucin 169 AJPJanuary 1995, Vol. 146, No. 1 Sialidase (from Clostridium perfringens) and a-N- acetylgalactosaminidase