DEVELOPMENTAL

BIOLOGY

148,63-74 (19%)

Epithelial-Mesenchymal Interactions in Uterus and Vagina Alter the Expression of the Cell Surface Proteoglycan, Syndecan EUGENIE L. BOUTIN,*~~RALPH D. SANDERSON,? MERTONBERNFIELD,$

ANDGERALD R. CIJNHA*

*Department of Anatomy, University of Califwnia-SF. 500 Parnassus Avenue, San Francisco, California %&.3-0452; TDepartment qf Pathology and The Center for Musculoskeletal Research, University of Arkansas for Medical Sciences, &‘Ol West Markham Street, Little Rock, Arkansas i%?O5;and $J&t Program in Neonatology, Harvard Medical School, 300 Lmgwood Avenu,e, Bostoq Massachusdts 0211.5 Accepted July 19, 1991 The cell surface proteoglyran, syndecan, exhibits molecular and histological dimorphism in the mouse uterus and vagina. In the mature vagina, syndecan is localized at the surfaces of the basal and intermediate cells of the stratified epithelium and has a modal molecular mass of ca. 92 kDa. The uterus expresses a larger form of syndecan (ca. 110 kDa) which is detected at the basolateral surfaces of the simple columnar epithelial cells. We have investigated whether epithelial-mesenchymal interactions influence the expression of syndecan in these organs by analyzing tissue recombinants composed of mouse epithelium and rat mesenchyme or vice versa with monoclonal antibody 281-2, which recognizes mouse syndecan. In tissue recombinants composed of newborn mouse uterine epithelium and rat vaginal stroma, t,he uterine epithelium was induced to form a stratified vaginal epithelium which expressed syndecan in same the pattern and mass typical of vaginal epithelium. Likewise, rat uterine stroma induced newborn mouse vaginal epithelium to undergo uterine development, and this epithelium exhibited a uterine pattern of syndecan expression. Although stromal cells normally express little syndecan in most adult organs, analysis of recombinants composed of mouse stroma and rat epithelium revealed that both uterine and vaginal mouse stromata synthesized syndecan that was larger (ca. 170-190 kDa) than the epithelial syndecans. A quantitative increase in the amount of stromal syndecan was evident when stroma was grown in association with epithelium in comparison to stroma grown by itself. These data suggest that epithelial-mesenchymal interactions influence the amount localization, and mass of both epithelial and stromal syndccan. I 1891 Academic Press, Inc.

transmembrane proteoglycan, syndecan, a molecule which can function as a matrix receptor by binding cells to interstitial collagens, fibronectin, and thrombospondin (for review see Bernfield and Sanderson, 1990). A monoclonal antibody to syndecan 281-2 stains newborn urogenital sinus vaginal epithelium intensely but Mtillerian-derived vaginal epithelium weakly (Boutin ef al., unpublished observation). In the adult, the basal and/or intermediate layers of the entire stratified vaginal epithelium stain intensely, while the apical cell layer(s) generally lacks stain (Hayashi et ah, 1988; Boutin and Cunha, unpublished observation). In contrast, in neonatal uterus lateral epithelial membranes stain uniformly for syndecan, while staining in the adult epithelium is localized predominantly to basolateral cell surfaces. Differences in the relative molecular mass of syndccan have been reported that correlate with epithelial organization (stratified or simple) (Sanderson and Bernfield, 1988). Syndecan from stratified epithelial tissues migrates electrophoretically at ca. 92 kDa, while that from organs lined by simple epithelia migrates at ca. 160 kDa. These distinct relative molecular masses

INTRODUCTION The reproductive tract of the female rodent is undifferentiated at birth. The O-day uterus consists of an undifferentiated, homogeneous mesenchyme and a simple epithelial tube which lacks the glandular elements present in the adult. The vaginal epithelium is composed of Mtillerian-derived pseudostratified epithelium cranially but caudally is composed of a solid epithelial cord derived from the urogenital sinus (Forsberg, 1973; Cunha, 1975). During postnatal development the epithelium lining the female rodent genital tract undergoes a variety of gross morphogenetic changes. In the uterus the epithelium differentiates into luminal and glandular epithelia, both of which are simple columnar. In the cervix and vagina, the epithelium becomes stratified squamous. As the epithelia differentiate, cell surface changes are evident morphologically (Lamb, 1977) and biochemically (Forsberg and Nord, 1969). In particular there are major alterations in the expression of the

1 To whom all correspondence

should be addressed. 63

0012-1606/91 $3.00 Copyright All rights

(0 1991 by Academic Press, Inc. of reproduction in any form reserved

64

DEVELOPMENTAL BIOLOGY

can be entirely accounted for by differences in the number and size of glycosaminoglycan chains (Sanderson and Bernfield, 1988). Prior to Day 5, vaginal epithelium can be induced by uterine stroma to develop as a simple columnar uterine epithelium. Similarly, vaginal stroma can induce postnatal simple uterine epithelium to form a stratified vaginal epithelium whose differentiation fluctuates normally through the estrous cycle between cornified and mucified states. This ability to respond to these heterotypic inductors is lost by 10 days postpartum in the mouse (Cunha, 1976). The changes in epithelial cytodifferentiation during these uterine and vaginal inductions suggest a complete switch in biochemical and functional differentiation, but this has not been addressed rigorously (Cunha et al., 1983). Because syndecan expression differs between these epithelia, we asked whether syndecan expression would be altered coordinately during this induction, We have assessed this question in both homospecific (mouse-mouse) and heterospecific (ratmouse) uterine and vaginal tissue recombinants. MATERIALS

AND

METHODS

Tissue Recomb&ants

Uteri (Ut) and vaginae (Va) were isolated from neonatal or adult Balb/c mice (m, Cancer Research Center, UC-Berkeley) and/or neonatal Sprague-Dawley rats (r, Bantin Kingman; Fremont, CA) and rinsed in Dulbecco’s phosphate-buffered saline (PBS2; Cell Culture Facility, UCSF). Tissues were incubated in 1% trypsin (Difco 1:250) in calcium, magnesium free Hank’s (Cell Culture Facility, UCSF) at 4°C (1 hr for vaginae, 1; hr for uteri) and rinsed once with 10% fetal bovine serum in Hank’s balanced salt solution (both from Cell Culture Facility, UCSF) containing 0.1% deoxyribonuclease I (Sigma) and twice with 10% serum in Hank’s without deoxyribonuclease I. Vaginal tissues were separated into epithelial (E) and mesenchymal components (S) by gently teasing with watchmaker’s forceps. Uterine tissues were separated either by gentle teasing or more commonly by sucking the tissues into a drawn pipet (Bigsby et al, 1986). All tissues were stored at 4°C in 10% serum until used. Epithelial and mesenchymal tissues were combined either homotypically or heterotypically on solidified agar medium (0.5% agar (Difco) in Dulbecco’s modified Eagle’s medium H-16 containing 0.1% glucose and supplemented with 10% fetal calf serum, 2 mMglutamine, ’ Abbreviations used: PBS, Dulbecco’s phosphate-buffered saline; BSA, crystalline bovine serum albumin; VaS, vaginal stroma; VaE, vaginal epithelium; UtE, uterine epithelium; UtS, uterine stroma; m, mouse; r, rat.

VOLUME 148.1991

100 IU/ml penicillin, and 100 pg/ml streptomycin (all from Cell Culture Facility, UCSF)). The tissues were allowed to readhere overnight and grafted under the renal capsule as in Cunha and Donjacour (1987). Most recombinants consisted of epithelium and stroma isolated from animals that were 0 to 5 days old. In addition, some recombinants were constructed of 7 day old or older epithelium and O-day stroma. This older epithelium is not competent to respond to heterotypic induction (Cunha, 1976), and hence, these recombinants allow analysis of the expression of epithelial and stromal syndecan in heterotypic recombinants in which the prospective fate of the epithelium remains unchanged. As controls, uterine and vaginal stromata were grafted without being combined with epithelium in order to assess both the cleanliness of the epithelial-mesenchymal separation and the expression (if any) of syndecan in stroma in the absence of epithelium. Homospecific mouse recombinants were grafted under the kidney capsule of adult Balb/c female mice (Cancer Research Center, UC-Berkeley) and allowed to develop for 4 to 5 weeks. Heterospecific (mouse + rat) and homospecific rat recombinants were grafted under the kidney capsule of adult female nude mice (Animal Care Facility, UCSF and Harlan Sprague-Dawley, Indianapolis, IN). Immunocytochemistry

Grafts to be examined by immunocytochemistry were fixed by freeze substitution and embedded in paraffin (Brody and Cunha, 1989). Six-micrometer sections were deparaffinized, rinsed for 20 min in 0.1% Tween 20 (Sigma) in PBS, and blocked for 30 min at room temperature using 1.3% goat serum (Zymed) and 1% bovine serum albumen (BSA; GIBCO) in Dulbecco’s PBS. Each subsequent rinse consisted of at least three changes of 0.1% Tween 20 in PBS. Sections were incubated for 1 hr at room temperature in PBS containing 1% BSA and 250 rig/ml of the rat IgG,, monoclonal antibody 281-2, which is specific for syndecan (Jalkanen et al., 1985). Control sections were incubated in 250 rig/ml of either rat IgG (Sigma) or the rat IgG,, monoclonal antibody Mel-14 (Gallatin et ah, 1983) with 1% BSA in PBS. All sections were rinsed l-2 hr, incubated 30 min in a 1:lOO dilution of biotinylated sheep anti-rat IgG (Amersham) in 1% BSA in PBS, rinsed 30 min, and exposed to the Vectastain ABC horseradish peroxidase reagent (Vector Laboratories, Burlingame, CA). Following a final 30-min rinse, the sections were incubated for 3-5 min in a 0.05% diaminobenzidine tetrahydrochloride (Aldrich) solution containing 0.03% CoCl, * 6H,O and 0.015% hydrogen peroxide (Brody and Cunha, 1989), rinsed, dehydrated, and coverslipped. Immunocytochemical data were pooled from at least five replicate experiments.

BOUTINETAL.

Induced Changes in Syndecan Expression

65

All homotypic uterine recombinants (mUtS + mUtE) were lined by a simple columnar uterine epithelium. Moderate to weak staining for syndecan was evident at the basolateral junctions between adjacent epithelial Molecular Weight Analysis cells (Fig. lc, arrowhead) in most tissue recombinants. Western blot analysis was used to assess the molecu- The basal surface of most epithelial cells was not lar weight of syndecan. Tissues were cleaned as thor- stained. Stromal staining was observed in over half of oughly as possible of extraneous tissue, frozen and the recombinants and was most pronounced in the constored in liquid nitrogen, and processed according to the nective tissue immediately subadjacent to the epitheprocedures of Sanderson and Bernfield (1988). Briefly, lium (Fig. lc, asterisk). Considerable variability was evithe specimens were extracted with 1% Triton, 10 mM dent in the amount and the intensity of epithelial and Tris, 0.15 MNaCl, 0.5 MKCl, pH 7.4, containing protease stromal syndecan staining both within and between inhibitors, and then incubated with 281-2 bound to Seph- these tissue recombinants, but this same variability is arose-4B beads. The bound material was eluted from the detected in intact mouse uteri (Boutin et al., unpublished beads with SDS-PAGE buffer and run on a 3.8 to 20% observation) and is probably due to hormonal fluctuaSDS polyacrylamide (7.5% bis) gel containing boric acid tions during the estrus cycle. and urea (Koda et ah, 1985). Gel lanes were loaded on the Inductive stroma was able to induce these same disbasis of equivalent amounts of total extracted protein. tinct patterns of epithelial syndecan staining in heteroProteins were blotted to Gene-Trans (Plasco, Inc., Wo- typic epithelium (grafts of mUtS + mVaE and mVaS burn, MA) and probed with lz51-labeled 281-2 (Sander- + mUtE). In tissue recombinants composed of mVaS son and Bernfield, 1988). Between 5 and 11 recombinants + mUtE, the vaginal stroma induced the uterine epitheor 5 to 18 stroma grafts were homogenized in each set, lium to differentiate as a stratified vaginal-like epitheand at least two replicate experiments were analyzed lium, whose basal and intermediate cell layers were infor each recombinant type except the homotypic, homo- tensely stained with 281-2 (Table 1, Fig. lb). Similarly in specific tissue recombinations. In addition, 1 to 3 recom- the majority of mUtS + mVaE recombinants, the uterbinants were set aside at the time the grafts were har- ine stroma induced the vaginal epithelium to form a vested and processed for immunohistochemistry in simple columnar, uterine-like epithelium which stained order to assess the completeness of induction. moderately to weakly at the basolateral junctions of the epithelial cells (Table 1, Fig. Id). These latter grafts generally also exhibited stromal staining (Fig. Id). RESULTS Isolated uterine and vaginal stromata were grafted to The Pattern of Syndecan Staining in Uterine and assess the cleanliness of the epithelial-mesenchymal Vaginal Tissues separation for these initial experiments. Contaminating epithelium was observed in 4 of these 32 control grafts. The pattern of syndecan staining in the homotypic tissue recombinants (mouse vaginal stroma + mouse The remaining grafts were recovered as small masses of vaginal epithelium (mVaS + mVaE) or mUtS + mUtE; connective tissue and muscle which lacked staining for permissive inductions) was similar to that found in the syndecan (see Fig. 5b for a typical example). corresponding intact mouse organs (Hayashi et al., 1988; Boutin et al., unpublished observation). Mouse VaS + mVaE recombinants developed as cysts lined by a Localization of Epithelial Syndecan stratified epithelium that was identical to that lining In the above experiments which utilized epithelium the host vagina. Syndecan was detected at the cell surfaces of the basal and intermediate epithelial cell layers and mesenchyme from the mouse, the tissue source (epi(Fig. la, Table l), but the apical cell layer(s) which var- thelium or stroma) of syndecan cannot be ascertained. ied histologically from keratinized, squamous to muci- However, since monoclonal antibody 281-2 binds mouse, but not rat, syndecan (cf. Figs. 2a and 5d, and Figs. 3a fied, columnar (as is consistent with the functional changes which are normally observed in the mouse va- and 3c), the molecules detected in heterospecific recomgina at different stages of the estrous cycle), did not binants constructed with mouse epithelium and rat stain (Fig. la, asterisk). The basal surface of the epithe- stroma should be epithelial in origin. Thus, the localizalial cells in contact with the basement membrane nor- tion and structure of epithelial syndecans can be determined independent of stromal syndecan. In grafts made mally was not stained (but see Fig. lA, arrowhead). Syndecan was not detected in the stroma nor was stain- with rVaS and either mVaE (permissive inductions) or ing observed with control antibody in any of the recom- neonatal mUtE (instructive inductions), a stratified vaginal epithelium was induced in which syndecan binants (not shown). Results from the individual except where noted.

experiments did not vary

66

DEVELOPMENTALBIOLOGYV0~~~~148.1991

FIG. 1. Inductive stroma instructed changes in the pattern of syndecan staining in competent epithelium. Sections of mouse tissue recombinants stained with the monoclonal 281-2 for syndecan. (a) Mouse vaginal stroma + mouse vaginal epithelium. The basal and intermediate epithelial cell layers of the stratified epithelium stained intensely with the monoclonal antibody 281-2. Occasionally the basal epithelial cells were stained along their basal surface (arrowhead) or had enhanced staining at their basolateral borders (arrow). The keratinized layers of the epithelium (asterisk) and the stroma were unstained. (b) Mouse vaginal stroma + mouse uterine epithelium. The epithelium in this heterotypic recombinant was induced to form a stratified, keratinized epithelium which stained identically as the homotypic recombinant seen in (a). (c) Mouse uterine stroma + mouse uterine epithelium. Weak to moderate staining was localized to the basolateral epithelial cell membranes (arrowheads) and the subepithelial stroma (asterisk). The intensely stained oval cells in the stroma were probably plasma cells which express syndecan on their cell surfaces (Hayashi et aZ., 1989; Sanderson et al, 1989). (d) Mouse uterine stroma + mouse vaginal epithelium. The epithelium in this heterotypic recombinant was induced to form a simple columnar uterine-like epithelium which stained identically as the homotypic recombinant seen in (b). Scale bars, 20 pm.

surrounded the cell surfaces of the basal and intermediate epithelial cell layers (Fig. 2a). Apical epithelial cell layer(s) were unstained. In addition, an increased intensity was sometimes found at the basolateral edges of the basal epithelial cells, and occasionally the cells in this layer stained for syndecan on their basal surface. Stromal staining was not observed. In grafts made with rUtS and either mUtE (permissive inductions) or competent mVaE (instructive inductions), a simple columnar, uterine epithelium developed which exhibited staining for syndecan at the basolateral cell surfaces (Fig. 2~). Occasionally, the basal surface of the epithelium stained. These results are identical to the distribution of epithelial staining in mouse/mouse recombinants and indicate that epithelial cells synthesize the syndecan detected at their plasma membrane and local-

ize syndecan in distinct patterns depending on their epithelial histology. In additional heterotypic tissue recombinants, rat stroma was combined with older, noncompetent mouse epithelium. In these recombinants the epithelium continued to express its original organization despite the presence of heterologous stroma, and the pattern of syndecan staining correlated with this epithelial morphology. For example, grafts of rUtS + 3- to lo-day mVaE contained a bilayered epithelium which exhibited a vaginal pattern of syndecan expression (basal epithelial cell layer stained for syndecan, while the apical cell layer exhibited weak to negligible staining) (Fig. 2b). Tissue recombinants of rVaS + adult mUtE were lined by a simple columnar, uterine epithelium exhibiting moderate to weak basolateral staining (Fig. 2d, arrow-

BOUTIN

ET AL.

67

Induced Changes in Syndecav Expessim TABLE

1

EPITHELIAL CYTODIFFERENTIATION AND PATTERN OF SYNDECAN EXPRESSION IN HOMOTYPIC AND HETEROTYPIC TISSUE RECOMBINANTS EXAMINED BY IMMUNOCYTOCHEMISTRY

Cytodifferentation

Syndecan expression

Recomb type

Age of epithelium (days)

Age of stroma (days)

No.

mVas + mVaE mVaS + mUtE mUtS + mUtE mUtS + mVaE rVaS + mVaE rVaS + mUtE rVas + mUtE rUtS + mUtE rUtS + mVaE rUtS + mVaE mVaS + rVaE mVaS + rUtE mUtS + rUtE mUtS + rVaE mUtS + rVaE

o-3 o-3 o-3 o-3 o-3 o-3 10 days to adult o-3 0 3-10 o-5 O-10 o-5 o-5 10

o-3 o-3 o-3 o-3 o-5 o-5 0 o-5 0 o-5 O-3 0 o-3 o-3 0

12 27 19 27 14 12 8 21 11 12 17 15 22 15 6

Epithelial ES* ES BL BL ES ES BL BL BL ES -

Stromal

+ +

+ + +

Stratified (vagina)

Simple (uterine)

100% 59% 7 %I 100% 50%

9% 100% 100% 73%

Mixed”

41% 100%) 63%

30% 50 VI

100 a 100% 64%~

27%

27% loo%, 100%

100%

a In these recombinants the majority of the epithelium was induced (conformed to that normally associated with the stroma) but small areas remained unchanged (remained appropriate for the source of the epithelium). *ES, entire cell surface of basal and intermediate layers stains; BL, stain restricted to basolateral membrane; -, not detected; f, present.

head), i.e., the typical uterine pattern. Thus, epithelial morphology (simple versus stratified) and not the source of the stroma (uterine versus vaginal) determined the distribution of syndecan staining. Stromal staining was not seen in any of these heterotypic recombinants. Furthermore, there was no immunocytochemical evidence (negative stained epithelium) of epithelial contamination in the recombinants, and only 1 of 31 grafts of stroma only was contaminated with epithelium for this set of experiments. Molecular

Size of Epithelial

Syndecans

It has been reported that syndecan isolated from organs lined by a simple epithelium is larger in size than syndecan isolated from organs lined by a stratified epithelium (Sanderson and Bernfield, 1988, and Fig. 3a). Similar results were obtained in mouse uterine and vaginal homotypic recombinants (Fig. 3b). However, in these experiments, it is impossible to determine whether the syndecans detected are epithelial or stroma1 in origin. Analysis of heterospecific homotypic tissue recombinants made with mouse epithelium and rat stroma (ruts + mUtE, rVaS + mVaE) revealed that simple columnar uterine epithelial cells express a llO-kDa form of syndecan (Fig. 4b), while vaginal epithelial cells express a smaller syndecan with a 92 kDa MW (Fig. 4a). In addi-

tion, following instructive inductions (rVaS + mUtE, rUtS + mVaE), the MW of syndecan synthesized by the epithelium correlated with the newly induced epithelial phenotype. Uterine epithelium that had been induced to develop as a stratified epithelium now produced the 92kDa form of syndecan (rVaS + mUtE; Fig. 4a), and vaginal epithelium that had been induced to form a simple epithelium now synthesized the larger llO-kDa form of syndecan (ruts + mVaE grafts; Fig. 4b). Furthermore, in heterospecific recombinants constructed with noncompetent mouse epithelium and rat stroma, the molecular weight of syndecan also correlated with epithelial morphology. Thus, recombinants constructed with rUtS + 7-10 d mVaE contained the smaller 92-kDa syndecan typically detected in stratified vaginal epithelium (Fig. 4c), while the simple uterine epithelium of rVaS + adult mUtE grafts expressed the llO-kDa form of syndecan (data not shown). As controls, Western blots of intact rat uterus and vagina (Fig. 3c), homotypic rat tissue recombinants (rVaS + rVaE or rUtS + rUtE, Fig. 3b), and grafts of rat stroma only (not shown) were analyzed, and were consistently negative. Nude mouse kidney tissue contained a small amount of llO-kDa MW antigen, but at a level much lower than that observed in the tissue recombinants (not shown). Since care was taken to remove kidney tissue adhering to the tissue recombinants prior to extraction, it is unlikely that this contributed signifi-

68

DEVELOPMENTALBIOLOGY

vOLUME148,1%1

FIG. 2. Syndecan localized on epithelial cell surfaces was produced by the epithelial cells and its pattern of staining correlated with the epithelial cytodifferentiation. Sections of tissues recombinants constructed with rat stroma and mouse epithelium stained with monoclonal antibody 281-2. (a) Rat vaginal stroma + mouse (0 day) uterine epithelium. The epithelium was induced to form a stratified, mucified epithelium which stained intensely for syndecan in its basal and intermediate layers. The apical mucified layer did not stain. Occasionally an enhanced intensity of staining was observed at the basolateral border of the basal cells (arrowheads) and occasionally the basal surface was stained (arrow). The rat stroma was not stained. An identical pattern of staining was seen in rat VaS + mouse VaE tissue recombinants (not shown). (b) Rat uterine stroma + mouse (10 day) vaginal epithelium. The epithelium was not induced to form a simple columnar epithelium, rather it developed as a stratified epithelium whose basal layer (b) stained prominently for syndecan. The apical layer (a) exhibited weak staining. (e) Rat uterine stroma + mouse (0 day) vaginal epithelium. The epithelium was induced to form a simple columnar, uterine-like epithelium which stained at its basolateral borders for syndecan (arrowheads). An identical pattern was seen in rat UtS + mouse UtE tissue recombinants (not shown). (d) Rat vaginal stroma + mouse (10 day) uterine epithelium. The epithelium was not induced to form a stratified epithelium, rather it developed as a simple columnar epithelium which stained most intensely for syndecan at its basolateral borders (arrowheads). Scale bars, 20 pm.

cantly to the syndecan isolated from the tissue recombinants. Furthermore, immunocytochemistry was performed on representative tissue recombinants to verify the epithelial morphology in grafts used in the MW analyses. Stromal Syndecan

The presence of syndecan in the stroma of recombinants containing mUtS (Figs. lc and Id) and its absence in recombinants containing rUtS (Figs. 2b and 2~) suggested that stromal cells produced syndecan. Therefore,

heterospecific tissue recombinants constructed with mouse uterine stroma and rat epithelium were examined. The majority (86%) of the tissue recombinants constructed with mouse UtS exhibited significant stromal staining with 281-2 (Fig. 5a). Syndecan was most pronounced in the connective tissue immediately beneath the simple uterine epithelium and generally was associated with stromal cell membranes. Staining intensities diminished with progression toward the myometrium which itself was unstained. In no case was the epithelium or the basement membrane of these recombi-

BOUTIN ET AL.

Induced Chmnyes in Syndecan

Eqvessiov

69

Of the ‘75 heterospecific tissue recombinants examined by immunocytochemistry in these experiments, only 2 had evidence of epithelial contamination (positive staining epithelium); these tissue recombinants were not included in the analysis. A similar low rate of contamination was evident in 68 grafts of isolated mVaS or mUtS. In contrast to the immunocytochemical data, Western blot analysis of mouse stroma and rat epithelium grafts indicated that both mUtS and mVaS contained significant quantities of syndecan (Figs. 6a and 6b). While the average MW varied somewhat (1’70 to 190 kDa), it was always higher than that expressed by mUtE (110 kDa) or mVaE (92 kDa). Interestingly, even though equivalent amounts of total extracted protein were loaded in

FIG. 3. Syndecans in homotypic uterine and vaginal tissue recombinants mimicked those expressed in the intact organs. Western blots of syndecan immunoisolated with monoclonal antibody 281-2. (a) The predominant form of syndecan isolated from intact mouse uterus had a larger modal size (110 kDa) than syndecan from intact mouse vagina (modal size 92 kDa). In addition there was some evidence of a small amount of larger syndecan, modal size 170 kDa, in intact uterus. (b) Homotypic uterine and vaginal recombinants constructed with mouse tissues produced syndecans with modal sizes the same as the respective intact mouse organs, while comparable homotypic tissue recombinants constructed with rat tissues produced no 281-2 detectable syndecans. (c) Intact rat uterus and vagina did not produce syndecans recognized by monoclonal antibody 281-2.

nants stained with the monoclonal 281-2. In grafts of mouse UtS by itself, staining for syndecan could not be distinguished from background (Fig. 5b). These results suggest that the presence of epithelium enhanced the amount of syndecan in uterine stroma. To determine if this effect required homologous epithelium, tissue recombinants were constructed with mUtS and noncompetent rVaE. All six tissue recombinants were lined by an unstained, bilayered vaginal epithelium. Nonetheless, four of the six tissue recombinants exhibited distinct staining for stromal syndecan (Fig. 5~) suggesting that heterologous epithelium could increase the amount of syndecan in uterine stroma. In most (72%) heterospecific tissue recombinants containing mVaS and rat epithelium, stromal staining for syndecan was not observed (Fig. 5d). Nonetheless, a small percentage of the grafts exhibited faint stromal staining, restricted to the two to three cell layers immediately beneath the overlying rat epithelium (data not shown). Staining was not detected in grafts of mVaS only.

FIG. 4. Syndecan molecular size correlated with epithelial cytodifferentiation. Western blots of mouse syndecan immunoisolated from heterospecific tissue recombinants constructed with rat stroma and mouse epithelium. (a) The stratified epithelium within recombinants of rat vaginal stroma + 0 day mouse vaginal or uterine epithelium expressed syndecan with a modal size of 92 kDa. (b) The simple epithehum in recombinants of rat uterine stroma +0 day mouse uterine or vaginal epithelium expressed syndecan with a modal size of 110 kDa. (c) The stratified epithelium in recombinants constructed with rat uterine stroma + 7 day (noncompetent) mouse vaginal epithelium synthesized syndecan with a modal size of 92 kDa.

70

DEVELOPMENTAL BIOLOGY

VOLUME 148,1991

a

d

FIG. 5. Epithelium increased the amount of syndecan detected in mouse uterine stroma. Sections of heterospecific recombinants constructed with mouse stroma and rat epithelium stained with monoclonal antibody 281-2. (a) Mouse uterine stroma and rat (0 day) uterine epithelium. The simple columnar epithelium (e) was not stained by the rat 281-2 antibody; however, the subepithelial uterine stroma was heavily stained. (b) Mouse uterine stroma grafted by itself lacked significant staining for syndecan. (c) Mouse uterine stroma and rat (10 day) vaginal epithelium. The older epithelium was not induced to form a simple epithelium but rather remained bilayered and unstained (e). Despite this, the subepithelial uterine stroma stained intensely for syndecan. (d) Mouse vaginal stroma + rat (0 day) uterine epithelium. A stratified, keratinized epithelium (e) was induced that was not stained by 281-2. The mouse vaginal stroma also lacked staining for syndecan. Scale bars, 20 pm.

each lane, tissue recombinants which contained rat epithelium gave a much stronger signal in Western blots than did grafts of stroma only (Figs. 6a and 6b). This appeared to be true for tissue recombinants constructed with either mVaS or mUtS.

of tissue recombinants composed of rat stroma + mouse epithelium proves that the epithelium is directly responsible for these differences. In epithelium that was induced to form the stratified epithelium of the vagina, syndecan had a relative molecular mass of 92 kDa and was detected on nearly all surfaces of the basal and intermediate epithelial cells. In contrast, when the simple DISCUSSION columnar epithelium of the uterus was induced, syndeInduced Simple and Stratified Epithelia Express can was detected only at basolateral cell surfaces and Syndecans of Distinct Size which are Detected had a MW of 110 kDa. These differences mimicked those in Distinct Patterns seen in the epithelium of intact organs (Sanderson and Prior work with both monoclonal281-2 and a polyclo- Bernfield, 1988; Hayashi et al., 1988; Boutin and Cunha, nal antibody to syndecan indicated that organs lined by unpublished) and in homospecific recombinants (mouse simple and stratified epithelium expressed syndecans of epithelium + mouse stroma; cf. Fig. 1). Furthermore, in tissue recombinants containing redistinct sizes that are localized in a tissue-specific manner (Sanderson and Bernfield, 1988; Hayashi et al., 1988; sponsive neonatal epithelium, the mesenchyme induced Trautman et al., 1991;Jalkanen et al., 1988). Our analysis the epithelium to change its organization (Cunha, 1976),

BOUTINETAL.

Induced Changes in Syndecan Expressim

71

In this case, the expression of a cell surface molecule was altered. is Expressed by Mature Uterine and Vaginal Stroma

Syndecan

While syndecan has been detected in many adult epithelia, its detection in stromal cells has been reported only in developing organs. The stroma of the developing tooth (Thesleff et ah, 1988), kidney (Vainio et aZ.,1989a), lung (Brauker et al, 1988), salivary gland (Bernfield, un-69 published observations), and limb (Solursh et ah, 1990) all stain for syndecan during critical periods of organo-46 genesis, yet the adult stromata of these organs do not stain. In the present work, both Western blots and immunocytochemistry of tissue recombinants constructed with mUtS revealed syndecan in mature uterine stroma1cells. The syndecan detected in these tissues was not likely to be shed by epithelial cells because immunoreactive murine syndecan was not detected in rat stroma, nor was it likely due to contaminating epithelial cells because its relative mass was always greater than that FIG. 6. Syndecan was produced by both uterine and vaginal stroma. of either mUtE or mVaE. Furthermore, stromal syndeWestern blots of stromal syndecan immunoisolated from heterospecific tissue recombinants constructed with mouse stroma and rat epican was not derived from a systemic source within the thelium. (a) Tissue recombinants containing mouse uterine stroma nude mouse host because murine syndecan was not deand either uterine or vaginal rat epithelium produced large amounts tected in rat/rat recombinants. Thus, mature uterine of syndecan with a modal size 170-190 kDa. Grafts of isolated mouse stroma synthesizes syndecan, and Western blots of tisuterine stroma expressed syndecan that was similar in size but less sue recombinants constructed with mVaS indicated the abundant. (b) Tissue recombinants containing mouse vaginal stroma and either vaginal or uterine rat epithelium produced syndecan with a same is true of mature vaginal stroma. modal size 170-190 kDa. Grafts of isolated mouse vaginal stroma exAlthough vaginal stroma synthesizes syndecan, synpressed syndecan that was similar in size but less abundant. decan was not detected by immunocytochemistry in vaginal stroma and it has not been detected in other adult stromata by this technique. In culture, 3T3 cells have and these changes in epithelial morphology were accom- only one one-hundredth the amount of cell surface synpanied by changes in both the relative molecular mass decan as epithelial cells (see Bernfield and Sanderson, of syndecan and in the pattern of syndecan staining. In 1990). Hence, the amount of cell surface stromal syndecan may normally be too low to detect by immunocytocontrast, unresponsive uterine and vaginal epithelia chemistry. It is also possible cell surface syndecan may maintained their original organization (simple or stratified) and continued to express syndecan in a manner be masked in most adult stromata due to interactions appropriate for their epithelial phenotype despite the with extracellular matrix molecules (Koda et al., 1985; presence of heterologous stroma (cf. Figs. 2a with 2b, 2c Saunders and Bernfield, 1988; Sun et al, 1989). Furtherwith 2d). Thus, once the epithelium differentiated, the more, evidence suggests stromal syndecan may be primarily intracellular (Kato and Bernfield, unpublished), nature of the epithelial organization (simple vs stratified) and not the source of the underlying stroma dic- which may be folded differently during intracellular transport, processing, and storage or associated with tated the pattern of syndecan expression. In other inductions where the epithelial phenotype other proteins which mask its antigenic sites. Along changes, stroma or mesenchyme can induce the expres- these lines, it is interesting that monoclonal 281-2 and sion of specific hormone receptors (Cunha et ah, 1981; polyclonal antibodies to syndecan do not detect the inNeubauer et al., 1983), cytokeratins (Sawyer et al., 1984; tracellular antigen that must be present in the posiMackenzie and Hill, 1984; Bigsby et ah, unpublished ob- tively stained epithelial cells (see also Hayashi et al, servations), and secretory proteins (Haffen et al., 1982; 1988). In any case, our results suggest that other adult Higgins et ah, 1989a,b) in responsive epithelium. The organs should be reexamined for the presence of strocurrent results on syndecan provide another example of ma1 syndecan by Western blotting, a technique which a mesenchyme-induced change in epithelial phenotype. can both concentrate and unmask antigen.

72

DEVELOPMENTALBIOLOGYV0~~~~148,1991

Alternately, mature uterine and vaginal stromal cells may be quite different from other adult stromata. The uterus and vagina undergo continual cyclic changes in phenotype through the estrous cycle, changes that are believed to be mediated by epithelial-mesenchymal interactions similar to those experienced by other tissues during development. The distribution of syndecan staining detected in uterine stroma (heaviest staining in the stromal cells adjacent to the epithelium) and the fact that stromal syndecan is increased in the presence of epithelium also implicate an involvement in epithelial-mesenchymal interactions. Stromal cell syndecan might facilitate these interactions by localizing various growth factors (Hook et al., 1986; Kardami et al., 1988; Maciag et al, 1984; Roberts et ah, 1988) or permitting the interaction of fibroblast growth factors with their receptors (Krufka et ah, 1990; Yayon et al, 1991).

which is similar to the larger syndecan detected here in the uterine stroma of Balb/c mice. The llO-kDa form of uterine epithelium was not previously observed, possibly because the tissues were at different stages of the estrous cycle which may influence the relative amounts of stromal and epithelial syndecans, or it may be due to strain differences. In mature tissues, the specific polymorphic forms of syndecan arise from variations in number and size of the glycosaminoglycan chains and not from differences in the core protein (Sanderson and Bernfield, 1988). This could result from tissue-type-specific intracellular environments that control the extent of glycosylation of the core protein. However, while the functional significance of syndecan’s polymorphic forms is unclear, it is not unreasonable to speculate that the size and arrangement of glycosaminoglycans alter the function of proteoglycans (Klebe and Mock, 1982; Fransson et al, 1984). The polymorphic forms of syndecan might vary in their Strmal Syndecan is Regulated by Epithelia ability to localize growth factors and to facilitate the Epithelium increased the amount of syndecan deinteractions of these growth factors with their receptected in uterine and vaginal stroma. This was evident tors. Both these properties have been linked to heparin in Ut stroma by immunohistochemistry with both Ut and heparan sulfate proteoglycans (Hook et ak, 1986; and Va epithelia and did not require homologous epitheKardami et ah, 1988; Maciag et aZ.,1984; Roberts et al., lium (cf. Figs. 5a and 5c with 5b). Furthermore, in1988; Krufka et ah, 1990; Yayon et ak, 1991). creases were detected in both Ut and Va stroma in WestThe different forms of syndecan might also differ in ern blots. Since the Western blot samples contained their ability to act as matrix receptors. The glycosaminoequivalent amounts of total extracted protein, this glycan chains of syndecan can interact with extracellushould have favored the detection of stromal syndecan lar matrix molecules (Saunders and Bernfield, 1988) in grafts of stroma only (where the stromal syndecan such as fibronectin, collagen types I, III, and V, and would not be diluted by rat epithelial proteins); howthrombospondin (Koda et al., 1985; Saunders and Bernever, the opposite result was observed. Recombinants field, 1988; Sun et al., 1989), and syndecan has the potenthat contained rat epithelium contained greater tial to transmit these cell-matrix interactions intracelamounts of stromal syndecan than grafts of mouse lularly via its interaction with the cytoskeleton stroma only. Therefore, it appears that Ut and Va epi(Rapraeger et ab, 1986). Perhaps the increased glycosthelium induced a true increase in the amount of stroaminoglycan chain size in stromal cell syndecan facilima1 syndecan and not merely an unmasking of the syntates cell-matrix interactions. Indeed, TGF-/3 has been decan epitope. This property of mature Ut and Va epitheimplicated in epithelial-mesenchymal interactions and lia duplicates the embryonic epithelia of the kidney induces NMuMG mammary epithelial cells to increase (Vainio et al., 1989a), tooth (Vainio et al., 1989b), and the extent of cell spreading concomitant with an inlimb (Solursh et al., 1990), which induce increases in emcrease in syndecan glycosaminoglycan chain size and bryonic mesenchymal syndecan. the amount of the subcellular fibronectin matrix (Rasmussen and Rapraeger, 1988; Rapraeger, 1989). The Distinct Patterns of Syndecan Size, Localization, The polymorphic forms of syndecan with the smallest and Tissue Distribution May Rejlect Distinct glycosaminoglycan chains are found in vaginal (92 kDa) Functions and uterine (110 kDa) epithelia primarily at epithelial The analysis of rat/mouse tissue recombinants al- cell-cell interfaces where interstitial matrix molecules lowed the detection of organ- and tissue-specific forms would not be expected (Hayashi et al., 1988; and this of syndecan. The predominant form of syndecan de- report). This syndecan might mediate epithelial celltected in the vagina of both Swiss Webster (Sanderson cell interactions either directly or indirectly (Fitchett et and Bernfield, 1988) and Balb/c mice is the 92-kDa epi- al., 1990; Jalkanen et ah, 1990; Kato and Bernfield, 1990) thelial form. However, syndecan isolated from the as has been proposed for other heparan sulfate proteouterus of Swiss Webster mice had a modal molecular glycans (Cole et al., 1985; Farach et aZ.,198’7;Fransson et mass of ca. 160 kDa (Sanderson and Bernfield, 1988), al., 1986).

BOUTIN ET AL.

In,duced Changes in Syndecan Expression

More experimental work will be necessary to determine the role of syndecan in the organogenesis and functional differentiation of female genital tract organs and why polymorphic forms exist. It is intriguing that embryonic epithelia also modulate the expression of another stromal molecule, tenascin (Inaguma et al., 1988; Aufderheide et al., 1987; Vainio et al., 1989a,b). Indeed, the expression of tenascin in the early developing tooth (Chiquet-Ehrismann et al., 1986; Thesleff et al, 1987) closely mimics that of syndecan (Thesleff et al., 1988), and both are induced by oral epithelium (Vainio et al., 1989b). Hence, it is possible that the expression of these stromal molecules is coordinately regulated by epithelial signals and that syndecan and tenascin work in concert to mediate cell-matrix interactions in developing organs. The authors thank Ella Battle for technical assistance. This work was supported in part by NIH Grants HD-17491, HD-07050, CA-05388, HD-06763, and the Arthritis Foundation. REFERENCES AUFDERHEIDE, E., CHIQUET-EHRISMANN, R., and EKBLOM, P. (1987). Epithelial-mesenchymal interactions in the developing kidney lead to expression of tenascin in the mesenchyme. J. CeZl Biol. 105,599608. BERNFIELD, M., and SANDERSON, R. D. (1990). Syndecan, a developmentally regulated cell surface proteoglycan that binds extracellular matrix and growth factors. Phil. Trans. R. Sot. Land. B 327,171-186. BIGSBY, R. M., COOKE, P. S., and CUNHA, G. R. (1986). A simple method for separating murine uterine epithelial and mesenchymal cells. Am. J Physiol. 251, E630-E636. BRAUKER, J. H., TRAUTMAN, M. S., and BERNFIELD, M. (1988). Glycosaminoglycans on syndecan, a cell surface proteoglycan, are modified during development in the embryonic mouse lung. J. Cell Biol. 107, 157a. BRODY, J. R., and CUNHA, G. R. (1989). Histologic, morphometric, and immunocytochemical analysis of myometrial development in rats and mice. 1. Normal development. Am. J. Anat. 186, l-20. CHIQUET-EHRISMANN, R., MACKIE, E. J., PEARSON, C. A., and SAKAKURA, T. (1986). Tenascin: An extracellular matrix protein involved in tissue interactions during fetal development and oncogenesis. Cell 47, 131-139. COLE, G. J., SCHUBERT, D., and GLASER, L. (1985). Cell-substratum adhesion in chick neural retina depends upon protein-heparan sulfate interactions. J. Cell Biol. 100, 1192-1199. CUNHA, G. R. (1975). The dual origin of vaginal epithelium. Am. J. Anat. 143, 387-392. CUNHA, G. R. (1976). Stromal induction and specification of morphogenesis and cytodifferentiation of the epithelia of the Mtillerian ducts and urogenital sinus during development of the uterus and vagina in mice. J. Exp. 2001. 196,361-369. CUNHA, G. R., and DONJACOUR, A. (1987). Mesenchymal-epithelial interactions: Technical considerations. h “Progress in Clinical and Biological Research” (D. S. Coffey, N. Bruchovsky, W. A. Gardner, Jr., M. I. Resnick, and J. P. Karr, Eds.), Vol. 239, pp. 273-282. Liss, New York. CUNHA, G. R., SHANNON, J. M., NEUBAUER, B. L., SAWYER, L. M., FUJII, H., TAGUCHI, O., and CHUNG, L. W. K. (1981). Mesenchymal-epithelial interactions in sex differentiation. Hum. Genet. 58, 68-77.

73

CUNHA, G. R., SHANNON, J. M., TAGUCHI, O., FUJII, H., and MELOY, B. A. (1983). Epithelial-mesenchymal interactions in hormone-induced development. Zrl “Epithelial-Mesenchymal Interactions in Development” (R. H. Sawyer and J. F. Fallon, Eds.), pp. 51-74. Praeger, New York. FARACH, M. C., TANG, J. P., DECKER, G. L., and CARSON, D. D. (1987). Heparin/heparan sulfate is involved in attachment and spreading of mouse embryos in nifro. Dev. Biol. 123, 401-410. FITCHETT, J. E., MCALMON, K. R., HAY, E. D., and BERNFIELD, M. (1990). Epithelial cells lose syndecan prior to epithelial-mesenchyma1 transformation in the developing rat palate. J CelZ Biol. 111, 145a. FORSBERG, J.-G. (1973). Cervicovaginal epithelium: Its origin and development. Am. J. Obstef. Gynecol. 115, 1025-1043. FORSBERG, J.-G., and NORD, L. (1969). Immunofluorescence studies on the distribution of mouse uterine epithelial antigen in foetal, immature and adult mice. J. Embqol. Exp. Morphol. 21, 85-95. FRANSSON, L.-A., CARLSTEDT, I., C~%TER, L., and MALMSTR~M, A. (1984). Binding of transferrin to the core protein of fibroblast proteoheparan sulfate. Proc. N&l. Acad. Sci. USA 81,5657-5661. FRANSSON, L.-A., CARLSTEDT, I., C&TER, L., and MALMSTR~M, A. (1986). The functions of the heparan sulphate proteoglycans. Irl “Functions of the Proteoglycans” (D. Evered and J. Whelen, Eds.), pp. 125-142. Wiley, Chichester. GALLATIN, W. M., WEISSMAN, I. L., and BUTCHER, E. C. (1983). A cell surface molecule involved in organ-specific homing of lymphocytes. Nature (London) 304,30-34. HAFFEN, K., KEDINGER, M., SIMON-ASSMANN, P. M., and LACROIX, B. (1982). Mesenchyme-dependent differentiation of intestinal brushborder enzymes in the gizzard endoderm of the chick embryo. I?{ “Embryonic Development, Part B: Cellular Aspects” (M. Weber and M. Burger, Eds.), pp. 261-270. Liss, New York. HAYASHI, K., HAYASHI, M., BOUTIN, E., CUNHA, G. R., BERNFIELD, M., and TRELSTAD, R. L. (1988). Hormonal modification of epithelial differentiation and expression of cell surface heparan sulfate proteoglycan in the mouse vaginal epithelium: An immunohistochemical and electron microscopic study. Lab. Invest. 58, 68-76. HAYASHI, K., HAYASHI, M., PAN, S., BERNFIELD, M. R., and TRELSTAD, R. L. (1989). Plasma cells have a cell surface protein which is or is related to the core antigen of a cell surface heparan sulfate proteoglycan. J. Cell Biol. 107, 803a. HIGGINS, S. J., YOUNG, P., BRODY, J. R., and CUNHA, G. R. (1989a). Induction of functional cytodifferentiation in the epithelium of tissue recombinants. I. Homotypic seminal vesicle recombinants. Development 106, 219-234. HIGGINS, S. J., YOUNG, P., and CUNHA, G. R. (1989b). Induction of functional cytodifferentiation in the epithelium of tissue recombinants. II. Instructive induction of Wolffian duct epithelia by neonatal seminal vesicle mesenchyme. Developent 106, 235-250. H&K, M., WOODS, A., JOHANSSON, S., KJELLI?N, L., and COUCHMAN, J. R. (1986). Functions of proteoglycans at the cell surface. 1~ “Functions of the Proteoglycans” (D. Evered and J. Whelen, Eds.), pp. 143-157. Wiley, Chichester. INAGUMA, Y., KUSAKABE, M., MACKIE, E. J., PEARSON, C. A., CHIQUETEHRISMANN, R., and SAKAKURA, T. (1988). Epithelial induction of stromal tenascin in the mouse mammary gland: from embryogenesis to carcinogenesis. Del:. Biol. 128, 245-255. JALKANEN, M., NGUYEN, H., RAPRAEGER, A., KURN, N., and BERNFIELD, M. (1985). Heparan sulfate proteoglycans from mouse mammary epithelial cells: Localization on the cell surface with a monoclonal antibody. J. Cell Biol. 101, 976-984. JALKANEN, M., RAPRAEGER, A., and BERNFIELD, M. (1988). Mouse mammary epithelial cells produce basement membrane and cell

74

DEVELOPMENTALBIOLOGY

surface heparan sulfate proteoglycans containing distinct core proteins. J. Cell BioL 106, 953-962. JALKANEN, M., MALI, M., H~~RK~NEN,P., and LEPP& S. (1990). Acquired epithelial phenotype for transformed mouse mammary tumor cells (S115) by syndecan gene transfection. J. Cell Biol. 111, 145a. KARDAMI, E., SPECTOR,D., and STROHMAN,R. C. (1988). Heparin inhibits skeletal muscle growth in vitro. Dev. BioL 126, 19-28. KATO, M., and BERNFIELD,M. (1990). Epithelial cells made deficient in syndecan lose cell surface expression of E-cadherin and /31-integrins. J. Cell Biol. 111, 263a. KLEBE, R. J., and MOCK, P. J. (1982). Effect of glycosaminoglycan on fibronectin-mediated cell attachment. J. Cell. Physiol. 112, 5-9. KODA, J. E., RAPRAEGER,A., and BERNFIELD,M. (1985). Heparan sulfate proteoglycans from mouse mammary epithelial cells: Cell surface proteoglycan as a receptor for interstitial collagens. J. Biol. Chem. 260,8157-8162. KRUFKA, A., OLWIN, B., and RAPRAEGER,A. (1990). Cell surface heparan sulfate proteoglycan binding of bFGF: Role in high and low affinity binding and uptake. J. Cell Biol. 111,223a. LAMB, J. C., IV, NEWBOLD,R. R., STUMPF,W. E., and MCLACHLAN,J. A. (1977). The development of the luminal surface of the vagina, cervix and uterus of the mouse. Scanning Electron Microsc. 2,355-360. MACIAG,T., MEHLMAN,T., and FRIESEL,R. (1984).Heparin binds endothelial cell growth factor, the principal endothelial cell mitogen in bovine brain. Science 225,932-935. MACKENZIE,I. C., and HILL, M. W. (1984). Connective tissue influences on patterns of epithelial architecture and keratinization in skin and oral mucosa of the adult mouse. Cell Tissue Res. 235,551-559. NEUBAUER, B. L., CHUNG, L. W., MCCORMICK,K. A., TAGUCHI, O., THOMPSON,T. C., and CUNHA, G. C. (1983). Epithelial-mesenchymal interactions in prostatic development. II. Biochemical observations of prostatic induction by urogenital sinus mesenchyme in epithelium of adult rodent urinary bladder. J. Cell BioL 96,1671-1676. RAPRAEGER,A. (1989). Transforming growth factor (type 0) promotes the addition of chondroitin sulfate chains to the cell surface proteoglycan (syndecan) of mouse mammary epithelia. J. Cell BioL 109, 2509-2518.

RAPRAEGER,A., JALKANEN, M., and BERNFIELD,M. (1986). Cell surface proteoglycan associates with the cytoskeleton at the basolatera1 cell surface of mouse mammary epithelial cells. J. Cell Biol. 103, 2683-2696.

RASMUSSEN,S., and RAPRAEGER,A. (1988). Altered structure of the hybrid cell surface proteoglycan of mammary epithelial cells in response to transforming growth factor-p. J. Cell Biol. 107,1959-1967.

VOLUME148.1991

ROBERTS,R., GALLAGHER, J., SPOONCER, E., ALLEN, T. D., BLOOMFIELD, F., and DEXTER, T. M. (1988). Heparan sulphate bound growth factors: A mechanism for stromal cell mediated haemopoiesis. Nature (London) 332, 376-378. SANDERSON,R. D., and BERNFIELD, M. (1988). Molecular polymorphism of a cell surface proteoglycan: Distinct structures on simple and stratified epithelia. Proc. N&l. Acad. Sci. USA 85,9562-9566. SANDERSON,R. D., LALOR, P., and BERNFIELD,M. (1989). B lymphocytes express and lose syndecan at specific stages of differentiation. Cell Reg. 1, 27-35. SAUNDERS, S., and BERNFIELD,M. (1988). Cell surface proteoglycan

binds mouse mammary epithelial cells to fibronectin and behaves as a receptor for interstitial matrix. J. Cell Biol. 106,423-430. SAWYER,R. H., O’GUIN, W. M., and KNAPP, L. W. (1984). Avian scale development. X. Dermal induction of tissue-specific keratins in extraembryonic ectoderm. Dev. BioL 101,8-18. SOLURSH,M., REITER, R. S., JENSEN,K. L., KATO, M., and BERNFIELD, M. (1990). Transient expression of a cell surface heparan sulfate proteoglycan (syndecan) during limb development. Deu. BioL 140, 83-92.

SUN, X., MOSHER,D. F., and RAPRAEGER,A. (1989). Heparan sulfatemediated binding of epithelial cell surface proteoglycan to thrombospondin. J. Biol. Chem. 264,2885-2889. TRAUTMAN,M. S., KIMELMAN, J., and BERNFIELD,M. (1991). Developmental expression of syndecan, an integral membrane proteoglycan, correlates with cell differentiation. Development 111,213-220. THESLEFF, I., MACKIE, E., VAINIO, S., and CHIQUET-EHRISMANN,R. (1987). Changes in the distribution of tenascin during tooth development. Development 101,289-296. THESLEFF, I., JALKANEN, M., VAINIO, S., and BERNFIELD,M. (1988). Cell surface proteoglycan expression correlates with epithelialmesenchymal interaction during tooth morphogenesis. Dev. Biol. 129,565-572.

VAINIO, S., JALKANEN, M., LEHTONEN,E., BERNFIELD,M., and SAXEN, L. (1989a). Epithelial-mesenchymal interactions regulate the stagespecific expression of a cell surface proteoglycan, syndecan, in the developing kidney. Dev. BioL 134,382-391. VAINIO, S., JALKANEN, M., and THESLEFF, I. (1989b). Syndecan and tenascin expression is induced by epithelial-mesenchymal interactions in embryonic tooth mesenchyme. J. Cell Biol. 108,1945-1954. YAYON,A., KLAGSBRUN,M., ESKO,J. D., LEDER, P., and ORNIT~,D. M. (1991). Cell surface, heparin-like molecules are required for binding of basic fibroblast growth factor to its high affinity receptor. Cell 64,841-848.