The EGF repeat and discoidin domain protein,SED1/MFG-E8, is required for mammary glandbranching morphogenesisMichael A. Ensslin and Barry D. Shur*

Department of Cell Biology, Emory University School of Medicine, Atlanta, GA 30322

Edited by William J. Lennarz, Stony Brook University, Stony Brook, NY, and approved December 13, 2006 (received for review November 24, 2006)

SED1, also known as MFG-E8, is a secreted protein composed oftwo EGF repeats (the second of which contains an RGD motif) andtwo discoidin/Factor V/VIII C domains. SED1 is expressed by a widerange of cell types, where it participates in diverse cellular inter-actions, such as sperm binding to the egg coat and macrophagerecognition of apoptotic lymphocytes. Although SED1 was origi-nally identified as a milk protein, its function in the mammarygland remains unclear; suggested functions include inhibition ofviral infection and clearance of apoptotic cells during mammarygland involution. We report here that SED1 has an unexpectedobligatory role during mammary gland development. Unlike thatseen in WT glands, SED1-null glands show severely reducedbranching from epithelial ducts and from terminal end buds, whichare thin and poorly developed. SED1 is expressed by both luminaland myoepithelial cells in the developing epithelial duct, and bindsto �v integrin receptors on myoepithelial cells leading to MAPKactivation and cell proliferation. The absence of SED1 leads togreatly reduced levels of activated MAPK and a concomitantreduction in cell proliferation and branching throughout the epi-thelial tree. These results suggest that SED1 contributes, at leastpartly, to the intercellular signaling between luminal and myoep-ithelial cells that is required for branching morphogenesis.

cell adhesion � epithelia � integrins

The mammary gland is one of the most exhaustively studiedmodels of branching morphogenesis (1). A host of cell

surface receptors and extracellular matrix components havebeen identified that aid in the outgrowth and remodeling of thebilayered mammary epithelial tube into a highly branched organwith terminal alveolar units capable of secreting milk in responseto lactogenic hormones. Among the many proteins identified inmilk is SED1, also known as lactadherin and MFG-E8, amongother names (2–5). Although the function of SED1 in milkremains obscure, SED1 secretion is markedly elevated in breastcancer (3), and serum levels of SED1 are diagnostic of tumorburden suggesting that SED1 may play a role in oncogenesis (6).SED1 is also thought to serve a protective function duringlactation by inactivating rotaviruses (7) as well as facilitating theclearance of apoptotic cells during mammary gland involution (8,9). Herein, we report that SED1 has an unexpected role duringmammary gland development by participating in the intercellu-lar signaling between luminal and myoepithelial cells that isrequired for branching morphogenesis of the mammary gland.

ResultsReduced Epithelial Branching During Morphogenesis of SED1-NullGlands. Comparison of the developing ductal tree in WT andSED1-null mammary glands illustrates a requirement for SED1during gland development. The developing ducts are composedof a double-layered epithelial tube, with inner luminal epithelialcells surrounded by outer myoepithelial cells. During the first2–3 weeks after birth, gland development is slow and propor-tional to overall body growth but progresses rapidly duringpuberty in response to ovarian hormones. In WT glands, the

growing ends of the epithelial ducts possess a characteristicmultilayered terminal end bud (TEB), which gives rise to bothepithelial cell types and undergoes continual bifurcation andoutgrowth (Fig. 1A, 3 weeks, arrowheads). This process contin-ues until it reaches the limits of the mammary fat pad at �8–10weeks after birth.

SED1-null neonates possess a rudiment of the early epithelialductal tree, but the glands grow very slowly with truncated ductsand minimal end buds. SED1-null TEBs are reduced to a fewpoorly organized cell layers, unlike the prototypical multilayeredcellular cap seen in WT (Fig. 1 A, 5 weeks, brackets). Further-more, there is a dramatic decrease in ductal branching inSED1-null glands, including branching that results from TEBbifurcation, as well as secondary branching along the ducts. WTglands show a 5.7-fold increase in branch number between 3 and8 weeks of development, whereas SED1-null glands show only a2-fold increase in branching (P � 0.004 for all ages assayed; Fig.1B). By 8 weeks, epithelial branching in SED1-null glands is only23% as frequent as in WT glands.

Pregnancy induces another round of epithelial proliferationand increased branching (Fig. 1C, 5 days postcoitum), and theonset of lactogenic hormones eventually leads to secretory acini[Fig. 1C, 1 day postpartum (dpp)]. Despite the greatly reducedmass of SED1-null mammary glands, they respond to hormonalstimulation during pregnancy and lactation as evidenced by asmall increase in branching and duct extension; however, theresulting acini are sparse, thin-walled, and distended. Neverthe-less, SED1-null females have near-normal-size litters (10) andare capable of nursing their pups to at least some degree,although we have not made a systematic study of the lactation orpostweaning phenotype in this study.

SED1 Expression During Gland Morphogenesis. SED1 expression inmilk is presumed to result from apical secretion from the acinarsecretory epithelium (2–4). It was, therefore, surprising that theabsence of an apically secreted milk protein resulted in greatlyimpaired mammary gland development. To gain insight intoSED1’s mode of action, its distribution during normal mammarygland development was determined by immunocytochemistry.As expected, SED1 is heavily expressed in the secretory epithe-lium and milk fat globules of lactating acini (Fig. 2A, 1 dpp).SED1 is also detectable in the ductal epithelium during devel-opment, being localized in association with both the luminal andmyoepithelial cell populations (Fig. 2 A). Double-label immu-

Author contributions: M.A.E. and B.D.S. designed research; M.A.E. performed research;M.A.E. contributed new reagents/analytic tools; M.A.E. and B.D.S. analyzed data; andM.A.E. and B.D.S. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS direct submission.

Abbreviations: TEB, terminal end bud; dpp, days postpartum; SMA, smooth-muscle actin;rSED1, recombinant SED1.

*To whom correspondence should be addressed. E-mail: barry@cellbio.emory.edu.

© 2007 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0610296104 PNAS � February 20, 2007 � vol. 104 � no. 8 � 2715–2720

DEV

ELO

PMEN

TAL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

July

14,

202

0

nofluorescence confirms SED1 immunoreactivity associatedwith both luminal and myoepithelial cells, by using smooth-muscle actin (SMA) as a marker of myoepithelial cells (Fig. 2B).There is no detectable SED1 immunoreactivity in SED1-nulltissues. In situ hybridization confirms the expression of SED1during gland development, which, as expected, increases duringlactation (Fig. 2C). Although it is difficult to identify cell-type-specific expression, the double layer of hybridization in theductal epithelium suggests that both luminal and myoepithelialcells synthesize SED1.

Defective Branching Morphogenesis Results from Reduced Prolifera-tion Rather Than Increased Apoptosis. The reduced development ofSED1-null glands is not a consequence of a selective loss ofmyoepithelial cells, as seen by the expression of SMA in SED1-null glands (Fig. 2B). However, myoepithelial cells acquire acharacteristically abnormal morphology as SED1-null glandsmature. Unlike WT cells, which maintain a uniform squamousphenotype surrounding the luminal epithelium and alveoli,myoepithelial cells in SED1-null glands appear rounded, pro-ducing a noncontiguous, dispersed cell layer (Fig. 3A). Differ-ences in myoepithelial architecture become most evident whenWT acini become secretory, at which time SED1-null myoepi-thelial cells are highly dysplastic and show no evidence of atraditional squamous phenotype (3 days postweaning).

We therefore determined whether reduced developmentcould be attributed to alterations in the extent of cell prolifer-ation or apoptosis that normally accompany mammary glanddevelopment (11). Cell proliferation, as judged by BrdU incor-poration, is present throughout the epithelial compartment ofdeveloping WT ducts, as well as in the multilayered TEB andadjacent mesenchyme (Fig. 3B, 8.9% BrdU� cells of 2,310counted), similar to that previously reported (11). In contrast,BrdU incorporation is greatly reduced in SED1-null glands(2.4% BrdU� cells of 1,551 counted; P � 0.003). DecreasedBrdU incorporation is evident in both TEBs (�/�, 4.5%; �/�,

9.1%) and ducts (�/�, 0.78%; �/�, 7.3%), although the ductsshow a more dramatic decrease in proliferation. Unlike that seenwith BrdU incorporation, the low level of apoptosis character-istic of early gland development appears relatively normal inSED1-null glands (11) (Fig. 3C). These results suggest that theloss of SED1 leads to a dramatic reduction in cell proliferationduring mammary gland morphogenesis, resulting in a greatlydiminished ductal tree. The residual mammary epithelium thatremains at sexual maturity is responsive to placental and lacto-genic hormones, as evidenced by increased ductal branching andTEB-like protrusions. However, the resulting acini are distendedand disorganized.

SED1-Null Mammary Gland Organoids Show Reduced Branching Mor-phogenesis in Vitro. The behavior of epithelial organoids wasanalyzed in 3D collagen gels, in which branching morphogenesisoccurs in the absence of systemic influences (12). WT organoidselaborate a highly branched ductal tree after 1 week in culture,with a characteristic multilayered epithelium (Fig. 4A). In con-trast, organoids from SED1-null females remain small andundergo very limited branching during the same time period.Those branches that do occur are very different from WT, in thattheir epithelium is thin, composed of the minimal bilayeredcellular architecture. In WT organoids, SED1 immunoreactivityis localized to both the luminal compartment (Fig. 4A, redarrowhead), indicative of apically secreted protein, as well aswithin the adjacent collagenous matrix (Fig. 4A, black arrow-heads), which likely reflects SED1 that is secreted basally fromluminal epithelial cells and/or by myoepithelial cells. No SED1immunoreactivity is seen in SED1-null organoids. Although invitro assays of branching morphogenesis do not replicate all ofthe conditions in the developing mammary fat pad, most notablythe absence of TEBs normally seen in vivo, organoids culturedin vitro replicate the SED1-dependent phenotype of reducedproliferation and branching. Thus, the branching defect seen inSED1-null glands is autonomous to the developing epithelial

Fig. 1. Mammary glands from SED1-null mice fail to undergo normal branching morphogenesis. (A) Photomicrographs of delipidated whole-mount mammaryglands stained with Carmine to visualize the ductal tree. Bulbous TEBs occur at the growing ends of WT ducts (3 weeks, arrowheads) but are not evident inSED1-null glands. WT TEBs have multiple cell layers that undergo continued proliferation and bifurcation, unlike TEBs in SED1-null glands that are reduced toa few cell layers (compare 5 weeks, brackets). Paired SED1�/� and SED1�/� images are derived from littermate females and are shown at identical magnification.Lower-magnification images (5 weeks, 10 weeks) illustrate the overall branching phenotype not evident at higher magnifications. (B) Mammary duct branchingprogresses throughout development of WT glands, whereas branching remains dramatically reduced in SED1-null females (�/� vs. �/�, P � 0.004 for all ages).(C) During pregnancy (5 days postcoitum), WT glands undergo another round of side branching not seen in SED1-null glands. The residual alveolar tissue thatis present in SED1-null glands is responsive to lactogenic hormones, although the acini are sparse (1 dpp, whole mount). WT acini are engorged with milksecretions, whereas SED1-null acini are distended with a thin-walled epithelium (1 dpp, section). [Scale bars, 1.0 mm in all whole-mount images and 0.1 mm forhistological sections (5 weeks, 1 dpp)].

2716 � www.pnas.org�cgi�doi�10.1073�pnas.0610296104 Ensslin and Shur

Dow

nloa

ded

by g

uest

on

July

14,

202

0

duct and does not depend upon systemic factors that couldcontribute to the phenotype (13).

SED1-Null Epithelial Cells Show Defective Adhesion. Studies ofMFG-E8 purified from milk suggest that it may function as a celladhesion molecule by virtue of its two binding motifs: recogni-tion of the RGD motif within the second EGF domain by �v�5/3

integrins and binding of the discoidin/C domains to membranephospholipids of mammary epithelial cells (5, 14). Consequently,the adhesive behavior of SED1-null mammary epithelial cells

was examined by using organoid cultures as well as quantitativesingle-cell adhesion assays.

Initially, we examined the ability of organoids to adhere andgrow out on 2D substrates, on which epithelial–matrix adhesionscan be isolated from other complex morphogenetic events thatoccur in 3D cultures (12). SED1-null organoids were unable toadhere to glass substrates as did WT organoids (not shown), andconsequently, organoids of both genotypes were plated onMatrigel substrates. After 3 days, WT organoids readily grow outand spread on Matrigel substrates, whereas SED1-null organoidssurvive but fail to grow out to the same degree (Fig. 4B Upper).To assess the involvement of SED1 in promoting WT outgrowth,organoids were allowed to adhere for 6 hr, after which cultureswere supplemented with heat-inactivated anti-SED1 antiserum,purified IgG, or preimmune reagents. Anti-SED1 antiseruminhibits the outgrowth of WT organoids, thus phenocopying thepoorly adhesive behavior of SED1-null cells (Fig. 4B Lower).Purified IgG produces similar results (not shown). As expected,SED1-null cells are refractory to the addition of anti-SED1reagents. Thus, SED1 promotes the adhesion and outgrowth ofmammary gland organoids, but in its absence, organoids fail tospread and ducts fail to branch and elongate, either in vitro orin vivo.

SED1-Mediated Adhesion Is RGD-Dependent. SED1-mediated adhe-sion depends on the RGD motif present within the second EGFrepeat, as shown by the ability of RGD-containing peptides toproduce a dose-dependent inhibition of mammary epithelial celladhesion to substrates coated with recombinant SED1 (rSED1).Equal concentrations of the control peptide, RAD, have noeffect (Fig. 4C, representative images; Fig. 4D, quantitativeassay). The RGD-dependent adhesion is consistent with SED1binding to �v�5/3 integrin receptors, which serve as SED1receptors in other systems (5, 14). The presence of the �v integrinsubunit in mammary gland organoids was confirmed by immu-noblotting and is primarily associated with myoepithelial cells(Fig. 4E), consistent with earlier studies indicating that the �vsubunit is restricted to the myoepithelium (15). The involvementof �v integrin in SED1-dependent adhesion was confirmed by theability of anti-�v antiserum to inhibit cell adhesion (Fig. 4F). Thelevel of inhibition produced by anti-�v integrin antibodies isconsistent with the expected contribution of myoepithelial cellsin the dissociated organoid. However, it is unclear why RGDproduced near total inhibition of adhesion, although similarresults have been reported by others (16).

Loss of SED1 Leads to Greatly Reduced MAPK Activation. Ligation of�v�5/3 integrins by the SED1-related protein, Del1, leads toMAPK activation and subsequent cell proliferation (17). Con-sequently, we determined whether the loss of SED1 frommammary epithelium correlated with altered MAPK activation.Immunoblot analysis indicates that total MAPK protein levelsare similar in WT and SED1-null organoids, whereas the level ofactivated MAPK is reduced in SED1-null, relative to WT (Fig.5A). The decrease in MAPK activation is largely confined to the44-kDa MAPK isoform (51% of WT levels); the 42-kDa isoformis more similar to WT levels (Fig. 5A). Nevertheless, immuno-blots of organoid lysates are unable to identify the relevant celltypes, and therefore differences in MAPK activation could beinfluenced by the presence of nonepithelial cells. Consequently,we turned to immunocytochemistry to identify and quantify theextent of MAPK activation specifically in mammary epithelialcells.

As expected, MAPK immunoreactivity is present in both theluminal and myoepithelial cell compartments in both WT andSED1-null organoids (Fig. 5B). In marked contrast, activatedMAPK is confined predominantly to the region of myoepithelialcells (Fig. 5B, P-MAPK, arrowheads). A small percentage of

Fig. 2. SED1 is expressed in the mammary epithelium during development.(A) SED1 immunoreactivity is seen at the earliest stages of ductal developmentassociated with both luminal and myoepithelial cells. Immunoreactivity in-creases during development and reaches peak levels at �5–6 weeks, whenboth luminal and myoepithelial cells are embedded in SED1 immunoreactivity(arrowheads denote myoepithelial cells based on nuclear architecture, appo-sition to the basal lamina, and SMA reactivity). SED1 expression remains stableor decreases slightly until the onset of lactation, when it is abundantlyexpressed in the secretory acini (1 dpp). No immunoreactivity is seen withpreimmune sera (PI). (B) Double-label immunofluorescence in WT glands(�/�) illustrates SED1 immunoreactivity (Alexa Fluor 488, green) associatedwith both luminal and myoepithelial cells, as seen by colocalization with SMA(Alexa Fluor 594, red) (arrowheads). (C) In situ hybridization confirms SED1transcription during gland development (5 weeks) and is up-regulated duringlactation (1 dpp). Sense probes produced no hybridization. The double layerof hybridization is consistent with SED1 being synthesized by both epithelialcell types. (Scale bars, 0.05 mm in all images.)

Ensslin and Shur PNAS � February 20, 2007 � vol. 104 � no. 8 � 2717

DEV

ELO

PMEN

TAL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

July

14,

202

0

cells positive for activated MAPK are not clearly associated withthe epithelial duct and may reflect fibroblastic or adipose cells.

We further investigated the identity of cells displaying acti-vated MAPK by double-label immunocytochemistry by usingSMA as a marker for myoepithelial cells. Virtually all cells withactivated MAPK occur in the outer layers of the ductal epithe-lium (Fig. 5C), similar to that seen in Fig. 5B. Overlay of theactivated MAPK and SMA images shows that most of the cellswith activated MAPK lie adjacent to or overlie the SMA label(white arrowheads), consistent with their being myoepithelialcells. The few remaining cells with activated MAPK appear to lieoutside of the SMA immunoreactivity (pink arrowhead) and mayreflect fragments of myoepithelial cells from adjacent sections,nonepithelial cells, and/or progenitors of myoepithelial and/orluminal epithelia cells. In any event, MAPK activation is readilyapparent throughout the myoepithelial cell layer in WT glands(�7.6% of cells of 5,300 counted), whereas MAPK activation is

markedly reduced in SED1-null glands (�1.0% of cells of 3,640counted, P � 0.001).

These studies predict that SED1 should be able to directlyactivate MAPK when added to isolated mammary epithelialcells. This was shown to be the case, as seen by the time-dependent, transient activation of MAPK after addition ofrSED1 to primary mammary epithelial cell cultures (Fig. 5D).

DiscussionMammary gland morphogenesis is orchestrated by a wide rangeof cell surface receptors and extracellular matrix components,including, but not limited to, integrin (15) and nonintegrinreceptors (18–20), their laminin and collagen ligands (20),metalloproteinases (20, 21), epimorphin (22), and other extra-cellular effectors that aid in the remodeling of the epithelial–mesenchymal interface. The results presented here indicate thatSED1, a protein previously implicated in sperm–egg adhesion

Fig. 3. SED1-null mammary glands show abnormal cellular architecture and greatly reduced cell proliferation. (A) Myoepithelial cells in WT glands have acharacteristic squamous morphology as they surround the ductal epithelium and acini (1 dpp, 3 days postweaning). However, myoepithelial cells in SED1-nullglands are bulbous and discontinuous around the ductal epithelium (arrowheads). Myoepithelial cells are identified by SMA immunocytochemistry; nuclei arevisualized with DAPI. (Scale bars, 0.05 mm in all images.) (B) Cell proliferation, as judged by BrdU incorporation (brown nuclei, arrowheads), is abundant in theepithelial and mesenchymal compartments of WT ducts (5 weeks shown) but is greatly reduced in SED1-null glands (P � 0.003). (Scale bars, 0.1 mm in all images.)(C) The degree of apoptosis, as assessed by TUNEL staining, is low and similar during early development (5 weeks) of both gland genotypes (red nuclei,arrowhead). No TUNEL-positive nuclei are seen in control incubations. (Scale bars, 0.05 mm in all images.)

Fig. 4. SED1 participates in RGD-dependent adhesion required for branching morphogenesis. (A) WT organoids cultured in 3D collagen gels produce highlybranched ductal trees (arrowheads, Upper) and are characterized by a multilayered epithelia invading the collagenous matrix (Lower). In contrast, organoidsfrom age-matched SED1-null females remain viable during the same culture period but fail to extend branches and possess thin-walled epithelia. SED1immunoreactivity is detected within the lumen (red arrowhead) of WT organoids and in the adjacent extracellular matrix (black arrowheads); no SED1immunoreactivity is detectable in SED1-null organoids. [Scale bars, 0.1 mm for images (Upper), 0.05 mm for Lower.] (B) WT organoids adhere and grow out on2D Matrigel substrates, whereas SED1-null cells adhere but grow out poorly, consistent with SED1 contributing to a normal adhesive phenotype. Heat-inactivatedanti-SED1 antiserum (1:100) prevents outgrowth and spreading of WT cells, thus recapitulating the SED1-null phenotype. The addition of preimmune (PI)reagents had no effect on epithelial cell outgrowth. Nuclei stained with Syto24 (green). (Scale bars, 0.2 mm for all images.) (C) Recombinant SED1 supportsmammary epithelial cell adhesion and is inhibited by competitive RGD containing peptides (0.5 mM), whereas RAD peptides have no effect. (D) RGD-dependentinhibition of single cell adhesion to SED1 matrices is dose-dependent, whereas RAD peptides fail to inhibit adhesion (10-min adhesion assay shown). (Error bars,SD.) (E) The presence of the 120-kDa �v integrin subunit was confirmed by immunoblotting lysates of WT organoids; nonimmune (ni) serum produced noreactivity. The antiserum localizes �v to the myoepithelial cells (arrowheads) surrounding the developing luminal cells. (F) The inclusion of anti-�v antiserumproduced a dose-dependent inhibition of single-cell adhesion to SED1 matrices (10- and 20-min assays shown); ni serum had no effect. (Error bars, SD.)

2718 � www.pnas.org�cgi�doi�10.1073�pnas.0610296104 Ensslin and Shur

Dow

nloa

ded

by g

uest

on

July

14,

202

0

(10, 23) and macrophage–lymphocyte interactions (24, 25), playsan obligatory role in branching morphogenesis as well.

The evidence suggests that SED1 functions as a signalingadhesive molecule, influencing interactions between myoepithe-lial and luminal epithelial cells that are known to be critical formammary gland development (21, 22). A simple working modelfor SED1 function can be formulated that is consistent withresults presented here, as well as with earlier biochemical studiesof SED1’s binding motifs (5, 14–16). The model suggests thatSED1, secreted by either myoepithelial or luminal epithelialcells, participates in branching morphogenesis by virtue of itsN-terminal RGD motif binding to �v-containing integrins onmyoepithelial cells, and intercalation of the C-terminal discoi-din/C domains into the phospholipid bilayer of luminal epithelialcells, as shown by others (5, 14) (Fig. 5E). Ligation of �v�5/3integrin leads to MAPK activation in myoepithelial cells, whichdirectly or indirectly, maintains cell proliferation throughout thedeveloping ductal tree. In the absence of SED1, MAPK activa-tion fails to occur with a concomitant reduction in cell prolif-eration, leading to grossly reduced mammary gland morphogen-esis. In this scenario, the adhesive function of SED1 could bemodulated by metalloproteinases known to regulate the activi-ties of other extracellular morphogens and adhesive molecules(20, 21). Whether the increase in SED1 characteristic of breastcancer is instructive or incidental to breast metastasis is un-known, but it is intriguing that tumorigenesis often results fromunrestrained side branching (26), a process that is regulated, atleast to some degree, by SED1.

Two recent studies report that SED1-null females show ab-normalities in mammary gland involution (8, 9). We have notmade a systematic study of the involution phenotype in thisstudy, because it focuses on SED1 function during mammarygland development. Nevertheless, one study analyzed SED1function during involution by using animals backcrossed onto theC57/B6 background (9). In our laboratory, backcrossing theoriginal C57/B6:129 SED1 line onto the C57/B6 backgroundresulted in a weak developmental phenotype, whereas back-

crossing to the 129 line produced a highly penetrant develop-mental phenotype (not shown). However, SED1-null C57/B6females retain the dilated duct phenotype observed after re-peated pregnancies, as reported by others (9). The second study(8) used females maintained on the hybrid C57/B6:129 back-ground, as in this study, and reported a variable developmentaldefect similar to that reported here, particularly with regard toreduced branching and underdeveloped TEBs (see figure 3 a andb, ref. 8). The low penetrance of the developmental phenotypemay likely reflect the targeting construct used, which results inthe retention of a truncated SED1 protein (devoid of one of thetwo discoidin/C domains) anchored to the cell surface, ratherthan secreted into the lumen. Thus, the construct developed byAtabai et al. (8) may distinguish between those cell interactionsinfluenced by secreted SED1 (clearance of apoptotic cells atinvolution), as opposed to those that involve cell-associatedSED1 (signaling between luminal and myoepithelial cells).

In summary, SED1 appears to facilitate a diverse range ofcellular interactions as varied as sperm-egg binding (10, 23),macrophage clearance of apoptotic lymphocytes (24, 25) andmammary epithelial cells (8, 9), and mammary gland branchingmorphogenesis (current study). That SED1 may represent onemember of a family of similar bi-motif signaling proteins isillustrated by the homologous protein Del1, which participates inendothelial adhesion to basal lamina during vascularization (17).It is not unreasonable to expect other EGF-discoidin/C domainproteins will be identified that contribute to distinct cellularinteractions.

Materials and MethodsMammary Gland Whole Mounts and Organoid Cultures. Whole-mount mammary tissue was prepared (6) from WT and SED1-null mice (10), and duct branching was quantified from photo-graphs of delipidated Carmine-stained glands by counting eachbranch point in the entire gland. Fifty-four glands were analyzed:two (gland nos. 2 and 4) from each of 27 females; three to sixfemales of each age and genotype. Significance was determined

Fig. 5. The loss of SED1 is associated with reduced MAPK activation in myoepithelial cells. (A) The level of total MAPK protein is similar between WT andSED1-null organoids as assessed by immunoblotting, whereas the activated form of MAPK is greatly diminished in SED1-null organoids. Immunoblotting�-tubulin served as a protein-loading control. Densitometry analysis indicates that activation of the 44-kDa MAPK is reduced by nearly 50%, whereas activationof the 42-kDa isoform is closer to normal levels. (B) Immunocytochemistry reveals total MAPK (brown stain) in both the luminal and myoepithelial compartmentsin both WT and SED1-null organoids, but activated MAPK (black arrowheads) is confined primarily to the myoepithelial compartment, consistent with theexpression of the �v integrin on myoepithelial cells. In contrast, MAPK activation is greatly reduced in the epithelial compartment of SED1-null organoids (P �0.001). (Scale bars, 0.05 mm in all images.) (C) The identity of cells showing activated MAPK was assessed by overlaying images of phospho-MAPK-positive cells(immunocytochemistry) and SMA (immunofluorescence). Most cells with activated MAPK colocalize with SMA, indicative of myoepithelial cells (whitearrowheads), whereas others lie adjacent to or outside of the SMA reactivity (pink arrowheads), suggesting they may be nonepithelial cells, epithelial precursors,or myoepithelial cell fragments from adjacent sections. (Scale bars, 0.05 mm in all images.) (D) The addition of rSED1 to primary epithelial cell cultures leads toa transient activation of MAPK. It is unknown why SED1 leads to activation of both MAPK isoforms in vitro, whereas the 44-kDa isoform is more sensitive to SED1in vivo (A). (E) Hypothetical model to account for SED1 function during mammary gland morphogenesis. SED1 is secreted by myoepithelial and/or basally fromluminal epithelial cells. The RGD motif within the second EGF repeat anchors SED1 to myoepithelial cells by the �v�5/3 integrin, whereas adhesion to luminalepithelial cells is thought to be mediated by intercalation of the discoidin/C domains into the phospholipid bilayers (5, 14). Ligation of �v�5/3 integrin inducesMAPK activation in myoepithelial cells, leading to proliferation of the epithelial compartment, duct elongation, and branching.

Ensslin and Shur PNAS � February 20, 2007 � vol. 104 � no. 8 � 2719

DEV

ELO

PMEN

TAL

BIO

LOG

Y

Dow

nloa

ded

by g

uest

on

July

14,

202

0

by using Student’s t test. Primary mammary gland organoidsfrom 5- to 6-week-old mice were prepared (21) and cultured in3D collagen gels or on Matrigel substrates in defined medium[DMEM/F12 containing Nutridoma NS (Roche, Indianapolis,IN), ITS (Sigma, St. Louis, MO), and EGF (20 ng/ml; BDBiosciences, Franklin Lakes, NJ)].

Histology and Immunocytochemistry. Paraffin-embedded sectionswere rehydrated, subjected to microwave antigen retrieval, andstained for SED1, as described (10). SMA was detected by usingmouse monoclonal anti-SMA (1:200; Sigma), followed by anti-mouse Alexa Fluor 594 (Molecular Probes, Eugene, OR). DNAwas stained with DAPI.

To assess cell proliferation, 5-week-old mice were injectedwith BrdU (100 �g/g body weight; Sigma) 4 h before death.Incorporated BrdU was detected with a BrdU staining kit(Zymed, Carlsbad, CA). Total and BrdU-positive cells werecounted in random microscope fields. Significance was deter-mined by using Student’s t test. TUNEL staining was conductedon sections after microwave antigen retrieval, and apoptotic cellswere detected with the In Situ Cell Death Detection kit (Roche).Ten-micrometer cryostat sections were used for immunolocal-ization of �v integrin (1:100; Chemicon, Hampshire, U.K.),MAPK (1:100; Upstate Biotechnology, Lake Placid, NY), andp42/44 MAPK (1:100; Cell Signaling Technology, Beverly, MA),and antibodies were detected by using HRP-conjugated second-ary antibodies (1:1,000; Santa Cruz Biotechnology, Santa Cruz,CA). The percentage of cells with activated (i.e., phospho)MAPK was determined by counting the total and DAB-positivecells in 19 (SED1-null) or 20 (WT) random sections. Significancewas determined by using Student’s t test.

In Situ Hybridization. Digoxigenin (DIG)-labeled RNA probeswere generated from a 160-bp DNA fragment of SED1 cDNA,spanning exons 2 and 3, cloned into Bluescript vector KS�.Sense and antisense RNA was produced by using T7 or T3polymerase (Roche). Mammary gland sections were permeabil-ized with Proteinase K (10 �g/ml, 30 min, 25°C). Hybridizationwas carried out overnight at 41°C, and unbound probe wasremoved by washing in graded dilutions of SSC. DIG-labeledRNA was detected with anti-DIG alkaline phosphatase conju-

gate (1:200; Roche) and developed by using the BCIP/NBTstaining kit (Vector Laboratories, Burlingame, CA).

Cell Adhesion Assays. Organoids were treated with 0.2% collage-nase/2.5 units/ml dispase (Invitrogen, Carlsbad, CA). Single cellswere separated from undigested clumps and washed repeatedlywith DMEM/F12. Epithelial cells were plated on bicameral filterdishes (1.1 cm2), previously coated with 10 �g of rSED1, inDMEM/F12 containing 0.5 mM GRGDNP (Biomol, Plymouth,PA), GRADSP, or no peptide. After 20 min, cells and mediumwere replaced with fresh DMEM/F12 and cultured for 5 h toallow the attached cells to become adherent. The adherent cellswere fixed, washed, and photographed.

To quantify cell adhesion, single cells were prepared fromorganoids as above and cultured on plastic dishes for 3–4 days toincrease cell yield. Contaminating fibroblasts were removed byincubating 1–2 min in trypsin/EDTA (Invitrogen), followed by 5min in trypsin/EDTA to remove epithelial cells. Cells were washedin DMEM/F12/5% FBS and incubated 45 min to recover fromtrypsinization, the medium was replaced with DMEM/F12/1%BSA, and cells were incubated 10–30 min with or without inhibitors.Ten thousand cells in 100 �l of DMEM/F12/1% BSA were addedto 96-well plates previously coated with rSED1. After 10–20 min,unbound cells were aspirated, and the wells were washed with PBS.The adherent cells were fixed, rinsed with water, and stained with0.1% Crystal violet. Cells were washed in water, stain was releasedinto the supernatant with 100 �l of 10% acetic acid, and absorbancewas read at 595 nm.

MAPK Activation in Cultured Mammary Epithelial Cells. WT mam-mary epithelial cells were cultured in six-well plates containingDMEM/F12, 5% FBS, insulin transferin selenium, and 10 ng/mlEGF. At confluency, the medium was replaced with serum-freemedium, and cells were serum-starved overnight. rSED1 (50�g/ml) was added in serum-free medium, after which cells werelysed at the indicated times and subjected to immunoblotting forMAPK and phospho-MAPK.

We are grateful to Drs. Karen Schmeichel, Win Sale, and Victor Faundezfor helpful suggestions regarding the manuscript. This work was supportedby National Institutes of Health Grant HD23479 (to B.D.S.).

1. Bissell MJ, Rizki A, Mian IS (2003) Curr Opin Cell Biol 15:753–762.2. Stubbs JD, Lekutis C, Singer KL, Bui A, Yuzuki D, Srinivasan U, Parry G

(1990) Proc Natl Acad Sci USA 87:8417–8421.3. Larocca D, Peterson JA, Urrea R, Kuniyoshi J, Bistrain AM, Ceriani RL

(1991) Cancer Res 51:4994–4998.4. Couto JR, Taylor MR, Godwin SG, Ceriani RL, Peterson JA (1996) DNA Cell

Biol 15:281–286.5. Andersen MH, Berglund L, Rasmussen JT, Petersen TE (1997) Biochemistry

36:5441–5446.6. Couto JR, Blank EW, Peterson JA, Ceriani RL (1995) Cancer Res 55:1717–1722.7. Newburg DS, Peterson JA, Ruiz-Palacios GM, Matson DO, Morrow AL, Shults

J, Guerrero ML, Chaturvedi P, Newburg SO, Scallan CD, et al. (1998) Lancet351:1160–1164.

8. Atabai K, Fernandez R, Huang X, Ueki I, Kline A, Li Y, Sadatmansoori S,Smith-Steinhart C, Zhu W, Pytela R, et al. (2005) Mol Biol Cell 16:5528–5537.

9. Hanayama R, Nagata S (2005) Proc Natl Acad Sci USA 102:16886–16891.10. Ensslin MA, Shur BD (2003) Cell 114:405–417.11. Humphreys RC, Krajewska M, Krnacik S, Jaeger R, Weiher H, Krajewski S,

Reed JC, Rosen JM (1996) Development (Cambridge, UK) 122:4013–4022.12. Schmeichel KL, Bissell MJ (2003) J Cell Sci 116:2377–2388.13. Gouon-Evans V, Rothenberg ME, Pollard JW (2000) Development (Cambridge,

UK) 127:2269–2282.

14. Andersen MH, Graversen H, Fedosov SN, Petersen TE, Rasmussen JT (2000)Biochemistry 39:6200–6206.

15. Taddei I, Faraldo MM, Teuliere J, Deugnier MA, Thiery JP, Glukhova MA(2004) J Mammary Gland Biol Neoplasia 8:383–394.

16. Taylor MR, Couto JR, Scallan CD, Ceriani R, Peterson JA (1997) DNA CellBiol 16:861–869.

17. Penta K, Varner JA, Liaw L, Hidai C, Schatzman R, Quertermous T (1999)J Biol Chem 274:11101–11109.

18. Steffgen K, Dufraux K, Hathaway H (2002) Dev Biol 244:114–133.19. Weir ML, Muschler J (2004) J Mammary Gland Biol Neoplasia 8:409–419.20. Fata JE, Werb Z, Bissell MJ (2004) Breast Cancer Res 6:1–11.21. Simian M, Hirai Y, Navre M, Werb Z, Lochter A, Bissell MJ (2001)

Development (Cambridge, UK) 128:3117–3131.22. Radisky DC, Hirai Y, Bissell MJ (2003) Trends Cell Biol 13:426–434.23. Ensslin M, Vogel T, Calvete JJ, Thole HH, Schmidtke J, Matsuda T, Topfer-

Petersen E (1998) Biol Reprod 58:1057–1064.24. Hanayama R, Tanaka M, Miwa K, Shinohara A, Iwamatsu A, Nagata S (2002)

Nature 417:182–187.25. Hanayama R, Tanaka M, Miyasaka K, Aozasa K, Koike M, Uchiyama Y,

Nagata S (2004) Science 304:1147–1150.26. Wiseman BS, Werb Z (2002) Science 296:1046–1049.

2720 � www.pnas.org�cgi�doi�10.1073�pnas.0610296104 Ensslin and Shur

Dow

nloa

ded

by g

uest

on

July

14,

202

0