UBC Faculty Research and Publications

The cell surface mucin podocalyxin regulates collective breast tumor budding Graves, Marcia L; Cipollone, Jane A; Austin, Pamela; Bell, Erin M; Nielsen, Julie S; Gilks, C. B; McNagny, Kelly M; Roskelley, Calvin D Jan 22, 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-13058_2015_Article_670.pdf [ 3.26MB ]
JSON: 52383-1.0361797.json
JSON-LD: 52383-1.0361797-ld.json
RDF/XML (Pretty): 52383-1.0361797-rdf.xml
RDF/JSON: 52383-1.0361797-rdf.json
Turtle: 52383-1.0361797-turtle.txt
N-Triples: 52383-1.0361797-rdf-ntriples.txt
Original Record: 52383-1.0361797-source.json
Full Text

Full Text

RESEARCH ARTICLE Open AccessThe cell surface mucin podocalyxinregulates collective breast tumor buddingMarcia L. Graves1, Jane A. Cipollone1, Pamela Austin1, Erin M. Bell1, Julie S. Nielsen2, C. Blake Gilks3,Kelly M. McNagny4 and Calvin D. Roskelley1*AbstractBackground: Overexpression of the transmembrane sialomucin podocalyxin, which is known to play a role inlumen formation during polarized epithelial morphogenesis, is an independent indicator of poor prognosis in anumber of epithelial cancers, including those that arise in the breast. Therefore, we set out to determine ifpodocalyxin plays a functional role in breast tumor progression.Methods: MCF-7 breast cancer cells, which express little endogenous podocalyxin, were stably transfected withwild type podocalyxin for forced overexpression. 4T1 mammary tumor cells, which express considerable endogenouspodocalyxin, were retrovirally transduced with a short hairpin ribonucleic acid (shRNA) targeting podocalyxin for stableknockdown. In vitro, the effects of podocalyxin on collective cellular migration and invasion were assessed intwo-dimensional monolayer and three-dimensional basement membrane/collagen gel culture, respectively.In vivo, local invasion was assessed after orthotopic transplantation in immunocompromised mice.Results: Forced overexpression of podocalyxin caused cohesive clusters of epithelial MCF-7 breast tumor cellsto bud off from the primary tumor and collectively invade the stroma of the mouse mammary gland in vivo.This budding was not associated with any obvious changes in histoarchitecture, matrix deposition or proliferation inthe primary tumour. In vitro, podocalyxin overexpression induced a collective migration of MCF-7 tumor cellsin two-dimensional (2-D) monolayer culture that was dependent on the activity of the actin scaffolding protein ezrin, acytoplasmic binding partner of podocalyxin. In three-dimensional (3-D) culture, podocalyxin overexpression induced acollective budding and invasion that was dependent on actomyosin contractility. Interestingly, the collectively invasivecell aggregates often contained expanded microlumens that were also observed in vivo. Conversely, when endogenouspodocalyxin was removed from highly metastatic, but cohesive, 4T1 mammary tumor cells there was adecrease in collective invasion in three-dimensional culture.Conclusions: Podocalyxin is a tumor cell-intrinsic regulator of experimental collective tumor cell invasion andtumor budding.BackgroundIt is often proposed that metastatic carcinoma progres-sion begins when single cells which have undergone anepithelial to mesenchymal transition (EMT) break awayfrom the primary tumor mass and begin to invade thesurrounding tissue stroma [1]. However, when the clin-ical tumor–stroma interface is examined in detail, thepresence and extent of small cohesive clusters of inva-sive cells, often denoted as “tumor buds”, correlates withmetastatic progression and poor prognosis in a numberof solid tumor types [2–8], including those of the breast[6–8]. The cells within these tumor buds, which invadecollectively, maintain at least a modicum of their originalepithelial phenotype and they often continue to form celljunctions with their neighbors [9]. While collectivetumor cell invasion has been extensively modeled experi-mentally [10], clinically-relevant drivers of the processare only now beginning to be identified [11, 12].Podocalyxin is a transmembrane sialomucin that isnormally localized to the apical surface of a variety ofepithelia, including the luminal cells of breast ducts andlobules [13]. The physiologic function of podocalyxin* Correspondence: roskelly@mail.ubc.ca1Department of Cellular and Physiological Sciences, University of BritishColumbia, 2350 Health Sciences Mall, Vancouver, BC V6T 1Z3, CanadaFull list of author information is available at the end of the article© 2016 Graves et al. Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, andreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link tothe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.Graves et al. Breast Cancer Research  (2016) 18:11 DOI 10.1186/s13058-015-0670-4was initially characterized in the developing kidney wherethe anti-adhesive properties of the mucinous extracellulardomain act coordinately with the actin cytoskeleton-associated cytoplasmic domain to physically separate ap-ical membrane processes between neighboring glomerularepithelial cells during the formation of the primary urinaryfilter [14, 15]. Podocalyxin also plays a more general rolein directing the morphogenesis of tubular and glandularepithelia by facilitating the formation of nonadhesive ap-ical membrane domains during lumen formation [16–18].Interestingly, Mostov and colleagues recently demon-strated that podocalyxin can also cause normal epithelialcells to become collectively invasive in three-dimensional(3-D) culture when polarity cues are actively disruptedsuch that apical membrane domains and lumens donot form efficiently [19].Podocalyxin is highly expressed in a number of humantumors, most of which are epithelially derived, with vary-ing degrees of prognostic significance [20–33]. In the caseof invasive breast cancer, podocalyxin overexpression is anindependent marker of poor outcome [13]. At the cellularlevel, podocalyxin overexpression expands apical mem-brane domains on the free surface of epithelial breastcancer cells that are maintained in monolayer culturewhere it alters the subcellular localization of two associ-ated actin-binding scaffolding proteins, NHERF-1 andezrin [34], both of which have been implicated in breasttumor progression when they are mistargeted [35–37].Importantly, Casey and colleagues demonstrated that apodocalyxin–ezrin complex increases breast tumor mo-tility [38]. These findings, coupled with the recent ob-servation that podocalyxin overexpression is positivelycorrelated with lymphovascular invasion (LVI) in breastcancer [31], which itself is well correlated with tumorbudding [8], led us to ask whether podocalyxin over-expression plays a functional role in collective breasttumor cell invasion.We found that podocalyxin overexpression caused epi-thelial MCF-7 breast tumor cell-derived xenotransplantsto generate cohesive micronodules that budded from theprimary tumor and collectively invaded the stroma ofthe mammary fat pad of immunocompromised mice. Intwo-dimensional (2-D) monolayer culture, podocalyxinoverexpression induced a collective MCF-7 cell migrationthat was ezrin dependent. In addition, in 3-D culture, podo-calyxin overexpression caused cohesive MCF-7 tumor cellclusters to expand their poorly formed internal microlu-mens and induced a collective invasion that was dependenton actomyosin contractility. Conversely, the stable down-regulation of podocalyxin attenuated the collective invasionof epithelial, but highly metastatic, 4T1 mammary tumorcells. Therefore, podocalyxin is a clinically-relevant markerof breast cancer progression that induces collectivetumor cell motility and invasion as well as experimentaltumor budding when it is highly expressed in epithelialbreast tumor cells.MethodsCell linesHuman MCF-7 breast cancer cells were purchasedfrom ATCC (Manassas, VA, USA). These cells were cohe-sive, epithelial, and estrogen dependent for growth in vivo.Mouse 4T1 mammary tumor cells were kindly providedby Dr Fred Miller (Wayne State University, Detroit,MI, USA).Podocalyxin overexpression and downregulationFor overexpression studies, MCF-7 cells were transfectedwith a pIRES2-based vector (Clonetech, Mountain View,CA, USA) without or with the full-length murine podo-calyxin cDNA [34]. Pooled, genetically selected, stablecell populations were subjected to two independent roundsof fluorescent-activated cell sorting (FACS) after indirectimmunofluorescent labeling of murine podocalyxin usinga species-specific antibody (anti-mouse PCLP-1; MBL,Nagoya, Japan).For downregulation studies, three different short hairpinRNA (shRNA) sequences designed for targeted knock-down of murine podocalyxin were identified using PSIOligomaker v1.5 freeware (http://web.mit.edu/jacks-lab/protocols/pSico.html) and were cloned into the pLL3.7lentiviral vector. Lentiviral particles were generatedafter transfecting 293T cells with the pLL3.7 vector andpackaging plasmids (pVSVg, pRSV-Rev, pMDLgag/pol)followed by transduction into subconfluent 4T1 mousemammary tumor cells. Infected 4T1 cells were then sub-jected to bulk FACS for green fluorescent protein expres-sion. Two of the resulting cell populations demonstratedreduced podocalyxin expression, one of which was usedfor functional analysis (4T1-Podo-KD 3603; see Fig. 7).To assess steady-state protein levels, cells were lysedin RIPA buffer (150 mM NaCl, 50 mM Tris pH 7.4,5 mM ethylenediamine tetraacetic acid, 1.0 % NP-40,0.5 % sodium deoxycholate, and 0.1 % sodium dodecylsulfate) with protease and phosphatase inhibitors. Equalamounts of protein were then separated by SDS-PAGE,transferred onto PVDF membranes and probed with ananti-mouse podocalyxin primary antibody (1 μg/ml; R&DSystems, Minneapolis, MN, USA), and primary antibodiesagainst E-cadherin (0.1 μg/ml; BD Biosciences, Mississauga,ON, Canada), ZO-1 (0.25 μg/ml; Invitrogen-ZymedLaboratories, South San Francisco, CA, USA), and epithe-lial cytokeratins (clones AE1/AE3; Dako, Troy, MI, USA).Primary antibody binding was visualized with the appro-priate species-specific horseradish peroxidase-conjugatedsecond antibody (Jackson Immunoresearch, West Grove,PA, USA) followed by chemoluminescence.Graves et al. Breast Cancer Research  (2016) 18:11 Page 2 of 17Orthotopic breast tumor xenografts, quantification, andhistological analysis17β-estradiol tablets (60-day release; IRA, Sarasota, FL,USA) were implanted subcutaneously into the cervicalscapular space of 12-week-old female Rag 2 M mice(Taconic, Hudson, NY, USA) and 1.0 × 106 MCF-7 cellsmixed 2:1 with Matrigel were innoculated into the rightabdominal (#4) mammary fat pad. After 6 weeks, the micewere sacrificed and the entire mammary fat pad togetherwith embedded tumor material was excised. Final tumorvolumes were calculated using the formula:Final tumor volume ¼ 0:52 length mmð Þð Þ  width mmð Þð Þ height mmð Þð Þ:Approval for the xenograft study was obtained fromthe Animal Care Committee of the University of BritishColumbia.Formalin-fixed and paraffin-embedded mammary glands(control, n = 8; podocalyxin, n = 7) were serial sectioned intheir entirety. Every 10th section was deparaffinized andstained with hematoxylin and eosin (H & E) as well asMasson’s trichrome using standard procedures. Stainedsections were qualitatively assessed for tumor morphology,tumor budding/collective microinvasion at tumor/stromalborders, invasion of tumor cells into existing normalmammary ducts, and infiltration into the surroundingvasculature and local lymph node.Microinvasion was quantified by counting the numberof tumor buds that extended into the stroma surround-ing the main primary tumor. The assessed sections wereselected based on the relative location of the inguinallymph node (20 H & E-stained sections, each at least50–100 μm apart, on either side of the lymph node wereanalyzed; i.e., 40 total sections per tumor through an ap-proximate tissue depth of 2–4 mm) and the average totalnumber of small nodules/buds and the ratio of the quan-tity of small nodules/buds per tumor volume are shown.Deparaffinized tissue sections were antigen-retrievedin heated citrate buffer, blocked, and incubated withprimary antibodies against E-cadherin (0.1 μg/ml; BDBiosciences), multiple epithelial cytokeratins using abroad-spectrum antibody (1:200; Dako), murine podoca-lyxin (1 μg/ml; R&D Systems), Ki67 (1:100; Abcam Inc.,Cambridge, MA, change to: 'USA'), and ezrin (1 μg/ml; CellSignaling Technology Inc., Danvers, MA, USA). Antibodybinding was visualized using a horseradish peroxidase-labeled polymer (EnVision™ + System; Dako), devel-oped with Nova Red™ (Vector Laboratories, Burlingame,CA, USA), and counterstained with Mayer’s hematoxylin.2-D migration assaysConfluent MCF-7 cell monolayers were serum-starvedovernight, scratched with a p-200 micropipette tip, andtreated with 100 ng/ml epidermal growth factor (EGF)for 16 hours. For live imaging, phase micrographs weretaken with a Nikon TMS phase Microscope equippedwith a Nikon Coolpix digital camera (Nikon, Mississauga,ON, Canada) to capture the migratory infilling of thescratch/wound. Wound areas were quantified using Cell-Sens Software (Olympus, Richmond Hill, ON, Canada).In some experiments the scratched monolayers weretreated with either dimethylsulfoxide (DMSO; vehiclecontrol) or the ezrin inhibitor NSC668394 (Millipore,Etobicoke, ON, Canada [39]) diluted to 10 μM in DMSO.3-D culture assay, quantification, and image analysisMCF-7 cells in low serum media (Dulbecco’s modifiedEagle’s medium/F12 with 1 % fetal bovine serum) wereplated overnight onto reconstituted basement membranegels (Matrigel™; BD Biosciences). Spheroidal cell aggregateswere then overlaid with neutralized collagen I (2.5 mg/ml;BD Biosciences) and maintained in low serum media sup-plemented with EGF (100 ng/ml; Sigma, St Louis, MO,USA) for an additional 4 days. The myosin II inhibitorblebbistatin (10 mg/ml; Sigma) was added to the mediaduring extracellular matrix (ECM) overlay (day 0) andagain on day 2. On day 4 the tumor spheroids/elongatingaggregates were fixed for immunocytochemistry and con-focal microscopic analysis as described in the next section.To quantify spheroid/aggregate shape, the longest andshortest axes (length and width respectively) were mea-sured using CellSens Software (Olympus) and expressed asan elongation index (length/width).For time-lapse imaging, collagen I-overlaid MCF-7 cell3-D spheroids were maintained in a stage-top incubator(Chamlide TC™; Live Cell Instrument, Seoul, Korea), andinvasion was imaged live over a 24-hour period using aLeica inverted DMI 4000 phase microscope equippedwith a DFC345 FX CCD camera and AF6000 imagingsoftware (Leica Microsystems, Concord, ON, Canada).The 3-D culture of 4T1 cells was performed as alreadydescribed except that the cells were maintained in theabsence of serum and EGF, given that these metastaticcells are completely growth factor-independent formotility. Collagen I-overlaid 4T1 cultures were imaged livefor 24 hours as described, and cultures were then fixedand immunostained for confocal analysis (described inthe following).Immunostaining and confocal microscopyCultured cells were fixed in 4 % paraformaldehydefollowed by 0.25 % triton incubation. Samples were thenrehydrated in phosphate-buffered saline, blocked with1 % bovine serum albumin/10 % normal goat serum,and incubated with primary antibodies for 1 hour atroom temperature that included: rat anti-mouse podoca-lyxin (1 μg/ml; R&D Systems), mouse anti-E-cadherinGraves et al. Breast Cancer Research  (2016) 18:11 Page 3 of 17(0.25 μg/ml; BD Biosciences), rabbit anti-ZO-1 (2.5 μg/ml;Invitrogen-Zymed Laboratories), mouse anti-cytokeratins(clones AE1/AE3, 1:1000; Dako), rabbit anti-ezrin (1 μg/ml;Cell Signaling Technology Inc.), and mouse anti-Muc1(1:500; gift from Dr John Stingl, Cambridge ResearchInstitute, Cambridge, UK)). Cells were then incubated withfluorescently conjugated species-specific secondary anti-bodies (Invitrogen-Molecular Probes, Burlington, ON,Canada), and nuclei were counterstained with 4′,6-diami-dino-2-phenylindole (Sigma), mounted in glycerol con-taining the anti-fade agent diazabicyclo[2.2.2]octane(Sigma), and imaged using an Olympus FV1000 confocalmicroscope (Olympus) followed by processing usingFV1000 Fluoview (Olympus) and Adobe Photoshop v12.0software (Adobe, San Jose, CA, USA).ResultsPodocalyxin overexpression induces locally invasivecollective tumor budding in vivoHuman MCF-7 breast tumor cells are epithelial, formwell-differentiated non-invasive tumors in vivo, and ex-press little endogenous podocalyxin [13, 38, 40]. Thus,we generated stable MCF-7 transfectants overexpressingmouse podocalyxin (MCF-7-podo) which can be identifiedunambiguously in human cells using a species-specificantibody [34]. MCF-7-podo cells did not exhibit a dif-ference in steady-state levels of the epithelial markersE-cadherin, ZO-1, or cytokeratins compared with emptyvector-transfected MCF-7-control cells (Fig. 1a). Likethe controls, MCF-7-podo cells also continued to formE-cadherin-containing adherens junctions in both 2-Dmonolayer (Fig. 3) and 3-D spheroid/aggregate (Fig. 5)culture. Therefore, podocalyxin-overexpressing MCF-7cells remained cohesive and epithelial.Human MCF-7-control and MCF-7-podo cells bothformed palpable estrogen-dependent tumors after theywere orthotopically transplanted into the mammary fatpads of immunocompromised mice. After 6 weeks therewas a trend towards slightly larger tumors in theMCF-7-podo group that was not statistically significant(Fig. 1b; p = 0.175; n = 8 for each condition). The sametrend towards a slight increase in tumor size in thepodocalyxin overexpression condition was also observedin subcutaneously grown tumors (Additional file 1:Figure S1A). Histopathologically, the central portions ofMCF-7-control and MCF-7-podo orthotopic tumorswere indistinguishable; both were characterized by denselypacked nests of cohesive tumor cells that were surroundedby an aniline blue-positive collagenous matrix and smallareas of tissue necrosis (Fig. 1c, top panel). Furthermore,there were no distinguishable differences in cell size(Fig. 1c, bottom panel) or proliferation based on Ki67staining (Fig. 1d, left panel) in the central portions ofthe tumors in either condition.While MCF-7-control tumors formed well-defined outerborders, the outer edges of MCF-7-podo tumors werehighly irregular when observed at low power (Additionalfile 1: Figure S2). Associated with the latter was a signifi-cantly greater number of cohesive micronodules, or buds,that surrounded the primary MCF-7-podo tumor masses(Fig. 1c, right panel, arrows; quantified in Fig. 1e, f). WhileKi67 staining indicated an increase in proliferation at thetumor/stromal interface compared with the tumor centerin both conditions, there was no obvious difference in Ki67positivity in this region between the MCF-7-control andMCF-7-podo tumors at that interface (Fig. 1d, right panels).Through a careful analysis of serial sections we deter-mined that some of the invasive buds located at the edge ofMCF-7-podo tumors were completely disconnected fromthe primary mass (Fig. 2a, b, arrows) while others remainedconnected to protrusions that emanated directly from theprimary mass (Fig. 2a, b, arrowheads). The MCF-7-podotumor micronodules were made up of densely-packed, co-hesive tumor cells that, like the primary tumor, continuedto express epithelial cytokeratins (Fig. 2a) and E-cadherin(Fig. 2b), some of which was membranous (Fig. 2b, lowerpanels, higher magnification insets). In addition, thesemicronodules often contained small podocalyxin-linedmicrolumens (Fig. 2c, arrows). Ezrin, which is an actin scaf-folding protein that complexes with the cytoplasmic tail ofpodocalyxin [34], also lined the invasive microlumenswithin the invasive nodules at the tumor/stromal interfacein the MCF-7-podo condition (Fig. 2d, right panel, arrows).While ezrin was also localized to the plasma membrane oftumor cells in both conditions, we observed very little evi-dence of ezrin-lined microlumens in the MCF-7-controlcondition (Fig. 2d, left panel). We concluded that podoca-lyxin overexpression facilitates the collective budding ofinvasive epithelial MCF-7 breast tumor cell micronodulesinto the mammary stroma in vivo and that this is associ-ated with microlumen formation.Podocalyxin overexpression induces collective migrationin 2-D monolayer culturePodocalyxin overexpression is known to increase theserum-induced motility of individual MCF-7 cells in 2-Dmonolayer culture [38]. However, when MCF-7-podocell monolayers were maintained at confluency theyremained cohesive and, while there was an expansion ofthe apical domain, the cells continued to form E-cadherin-containing adherens and ZO-1-containing tightjunctions and they still expressed epithelial cytokeratins(Fig. 3a). When confluent monolayers were scratchedand stimulated with EGF, MCF-7-podo cells at thescratch edge migrated collectively into the wound morerapidly than did MCF-7-control cells (Fig. 3b). It is un-likely that proliferation contributed significantly to theincreased wound filling by the MCF-7-podo cells as theGraves et al. Breast Cancer Research  (2016) 18:11 Page 4 of 17Fig. 1 (See legend on next page.)Graves et al. Breast Cancer Research  (2016) 18:11 Page 5 of 17increased collective migration was observable within thefirst 16 hours after the monolayers were scratched. Inaddition, there was no significant difference in the prolif-eration of MCF-7-control and MCF-7-podo cells inmonolayer culture (Additional file 1: Figure S1B).MCF-7-podo cells retained their E-cadherin and actin-containing adherens junctions during collective migra-tion (Fig. 3c, d). Interestingly, the f-actin-containing an-terior lamellipodia were more prominent at the leadingedges of the migrating MCF-7-podo cell sheets thanwere those formed by MCF-7-control cell sheets (Fig. 3b,lower panel, note phase dark lamellipodia in the MCF-7-podo cells at the wound edge; Fig. 3d, note increased f-actin in the lamellipodia of MCF-7-podo cells at thewound edge). Interestingly, podocalyxin was concen-trated apically just behind these anterior lamellipodiawhere it colocalized with a less prominent pool ofnonlamellar apical f-actin (Fig. 3d, bottom panel, z axis,arrow). This suggests that the demonstrated abilityof podocalyxin to segregate membrane domains inan actin cytoskeleton-dependent manner [19, 34] may playa role in its ability to stimulate collective tumor cellmigration.Podocalyxin interacts with the actin cytoskeleton viaezrin which binds to its cytoplasmic domain [34] andthe separate interaction of ezrin with actin requires it tobe phosphorylated in its “ERM” domain. When wetreated MCF-7-podo cells with a pharmacological inhibi-tor of this phosphorylation, NSC668394 [39], there wasa significant loss of the small punctate accumulations ofpodocalxyin and pERM at the cell surface (Additionalfile 1: Figure S3) which we have previously shown to beassociated with microvilli in the apical domain of MCF-7-podo cell monolayers [34]. Importantly, treatmentwith NSC668394 also decreased the collective migrationand the enhanced wound edge lamellipodia formation ofscratched MCF-7-podo cell monolayers (Fig. 4).Podocalyxin overexpression induces collective epithelialinvasion and a bud-like phenotype in 3-D cultureMCF-7 cells cluster together to form non-invasive aggre-gates when they interact with a reconstituted basementmembrane ECM (i.e., Matrigel) in 3-D culture [41].Thus, we pre-clustered MCF-7 cell populations on Matrigeland then overlaid them with collagen type I because wenoted an accumulation of stromal collagen in the MCF-7cell-derived tumors that developed orthotopically withinthe mammary fat pads in vivo (see Fig. 1). In addition,other investigators have shown that the presence of stromalcollagen facilitates collective breast tumor cell invasion[11, 12]. Under these conditions, the MCF-7-controlcell aggregates gradually increased in size over a 4-dayperiod. Importantly, the control cell aggregates remainedrelatively spherical throughout (Fig. 5a, left panel). The lat-ter characteristic was quantified by determining the elong-ation index (longest length/shortest width of the cellclusters where a perfect sphere has an index of 1.00)which was 1.32 ± 0.04 at the end of the experiment(Fig. 5b). In contrast, while they also started out as smallspheroidal aggregates, over the 4-day culture period manyof the MCF-7-podo cell clusters elongated and pushed outinto the matrix as cohesive multicellular extensions withblunt-ended tips (Fig. 5a, right panel) As a result, themean elongation index of the MCF-7-podo cell clusterswas significantly larger than the controls (2.89 ± 0.1,p <0.001 vs. controls; Fig. 5b). The dynamic natureof this podocalyxin-mediated increase in collectivetumor cell invasion into the ECM was observable by livevideo phase microscopy (compare Additional file 2:Movie S1 for MCF-7-control cell clusters with Additionalfile 2: Movie S2 for MCF-7-podo cell clusters).MCF-7-control and MCF-7-podo cell clusters bothmaintained their E-cadherin-containing adherens junc-tions in 3-D culture (Fig. 5c). However, unlike nor-mal mammary epithelial cells expressing podocalyxin(See figure on previous page.)Fig. 1 Podocalyxin overexpression induces collective tumor invasion in vivo. a Western blot analyses of whole cell lysates show that totalexpression of E-cadherin, the tight junction protein ZO-1, and epithelial cytokeratins were unaffected by stable podocalyxin overexpression inMCF-7-podo cells compared with MCF-7-control cells. b Podocalyxin overexpression did not significantly affect overall estrogen-dependent MCF-7cell-derived tumor growth after orthotopic xenotransplantation (n = 8 for each condition; tumors were excised after 6 weeks; p = 0.175, two-tailed,unpaired Student’s t test). c Representative images of trichrome-stained tumor sections (upper panels) show that MCF-7-control and MCF-7-podoxenografts both contained densely packed tumor cells at the center of the primary tumor nodule. Both tumor types also had small areas of visiblenecrosis and developed a collagenous stroma (light blue staining) that surrounded nests of cohesive tumor cells. Representative images of thetumor/host stromal interfaces show that, compared with controls, MCF-7-podo cells more extensively protruded into the surrounding stroma atthe edge of tumor (arrowhead). These tumors also formed multiple small cohesive tumor foci, or micronodular bud-like structures (arrows), thatappeared to be separate from the primary tumor nodule (labeled P; local lymph node is labeled LN). Higher power H & E images (lower panels)indicate that there was no a difference in tumor cell density within either the centers or edges of the lesions at the tumor/stromal interface. Scale= 100 μm. d Tumors were immunostained with the proliferation marker Ki-67 (brown). While there was an overall increase in Ki-67-positive cells atthe tumor/stromal interface, there was no discernible difference in Ki-67 frequency between MCF-7-control and MCF-7-podo cells either at thecenter of the tumors (left panel) or at the tumor/stromal interface (right panel). Scale = 100 μm. e, f The number of observable micronodules wasquantified either per tumor e or per tumor volume f. Note that, in both cases, there was a statistically significant increase in the MCF-7-podotumors compared with the control tumors (two-tailed, unpaired Student’s t test)Graves et al. Breast Cancer Research  (2016) 18:11 Page 6 of 17Fig. 2 Podocalyxin overexpression induces cohesive epithelial invasion of the stroma and regional lymph node. a, b Serial sections (approximately 100 μmapart) immunostained for cytokeratin a and E-cadherin b indicated that the stromal invasion of MCF-7-podo-derived tumor micronodules was cohesiveand epithelial. Notably, some cohesive micronodules were extensions originating from the main primary tumor (arrowheads), while others were completelydisconnected from it (arrows) a, b. Shown in the lower panels are magnified views of the outlined areas within the upper images, with further magnifiedinsets in the lower panels of b showing the presence of some membranous E-cadherin. Scale = 200 μm. c Representative invasive microno-dules were serially sectioned and immunostained for epithelial cytokeratins, E-cadherin, and podocalyxin. Note that a considerable portionof the podocalyxin localized to small microlumenal membrane surfaces within the tumor micronodules (arrows). Scale = 50 μm. d MCF-7-control andMCF-7-podo tumor sections were immunostained for ezrin and the tumor/stromal interface was imaged. Note that while ezrin was often localized tocell membranes in both conditions, it also lined microlumenal structures in the MCF-7-podo tumors (arrows). Scale = 50 μmGraves et al. Breast Cancer Research  (2016) 18:11 Page 7 of 17Fig. 3 Podocalyxin stimulates collective breast tumor cell migration in 2-D monolayer culture. a Confocal XZ vertical images show that, as expected,podocalyxin alters the architecture of the apical membrane surface of MCF-7-podo cells maintained in 2-D culture. As a result it causes the cells toassume different shapes within the monolayers, which are much more uniform in the controls. The basolateral localization of E-cadherin is relativelyunaffected in the podocalyxin-expressing cells and tight junctions are still present apically as indicated by discrete puncta of localized ZO-1, althoughthe location of the latter varies within the vertical plane given the change in cell architecture. Podocalyxin does not cause a loss of epithelial keratinfilaments but it does disrupt the uniformity of their localization at the apical surface. Scale = 10 μm. b Serum-starved MCF-7-control and MCF-7-podocell monolayers attached to a rigid collagen I-coated substratum (0.25 μg/cm2) were subjected to a wound assay under growth factor-stimulatedconditions (EGF 100 ng/ml). The ability of the cells to close the wound after 16 hours was monitored by phase microscopy, and photomicrographs ofthe same wound area after that period are shown (upper panel, 0 hours after wounding; bottom panels, 16 hours after wounding at low and highpower). There was little difference in wound closure in non-EGF-stimulated conditions (data not shown), but the podocalyxin-expressing cells wereable to close the wound more readily in response to EGF treatment compared with controls (graph, mean ± SD, unpaired Student’s t test, *p >0.05).Data shown are from one of three representative experiments. Scale = 50 μm. c MCF-7 cells were subjected to wounding as in b and after 16 hoursthey were fixed and immunostained for podocalyxin (red) and E-cadherin (green). Projections of confocal stacks at the leading edge of the woundshow that podocalyxin-expressing cells continued to form adherens junctions. Scale = 10 μm. d MCF-7 cells were subjected to woundingas in b, and after 16 hours they were fixed and immunostained for podocalyxin (red) and f-actin (green). Projections of x/y-axis confocalstacks show that MCF-7-podo cells have enhanced, f-actin-rich lamellipodia at their leading migratory edges compared with MCF-7-control cells (upperthree panels). X–Z-axis images (sliced along the white line shown in the x/y merged image) show that podocalyxin is polarized to the free, apical surfacemembrane (arrow), but does not extend completely into the enhanced f-actin rich lamellipodia (lower panels). Scale = 10 μmGraves et al. Breast Cancer Research  (2016) 18:11 Page 8 of 17(Additional file 1: Figure S4), MCF-7-podo cell aggregatesdid not fully polarize to form a single large, central lumen.Instead, the elongated MCF-7-podo cell clusters oftenformed multiple small podocalyxin-lined microlumens(Fig. 5c, right panel). Interestingly, the cells arrangedaround some of these microlumens formed small bud-likestructures that were readily apparent when we generated3-D reconstructions of the cell aggregates (Fig. 5c, lowerright panel, arrows).As already noted, we previously demonstrated thatpodocalyxin recruits the actin-scaffolding protein ezrinto the free apical surface of MCF-7 cells maintained asmonolayers in 2-D culture [34]. Herein we found thatezrin co-localized with podocalyxin around the internalFig. 4 The ezrin inhibitor NSC668394 decreases collective migration and leading lamellipodia formation of podocalyxin-overexpressing cells.a Serum-starved MCF-7-control and MCF-7-podo cell monolayers were cultured and scratched as described in Fig. 3 in the presence of DMSO(vehicle control) or the ezrin inhibitor NSC668394, and they were imaged by phase microscopy after 16 hours. Note that NSC668394 significantlydecreased the collective migration of the MCF-7-podo cells into the wound as quantified using the migration index described in Fig. 3 (mean ± SD,unpaired Student’s t test, *p >0.05). b MCF-7-control and MCF-7-podo cells were subjected to wounding in the absence and presence of NSC668394for 16 hours. The cells were then fixed and immunostained for podocalyxin (red) and f-actin (green). Projections of x/y confocal stacks (upper panels)and X–Z plane images (lower panels) indicated that NSC668394 decreased the formation of f-actin-containing leading lamellipodia in the MCF-7-podocells. Scale = 10 μm. DMSO dimethylsulfoxideGraves et al. Breast Cancer Research  (2016) 18:11 Page 9 of 17microlumens that formed in the collectively invasiveMCF-7-podo cell aggregates in 3-D culture (Fig. 6a,upper panels). While the control cell clusters containedclosed internal apical membrane domains that weremarked by the mucin Muc1 (Fig. 6a, lower left panel),podocalyxin expression caused these internal domains toexpand and open up as microlumens (Fig. 6a, lowerright panel, arrows) that preferentially recruited f-actinFig. 5 Podocalyxin stimulates collective breast tumor cell invasion in 3-D spheroid culture. MCF-7 cells were forced to form aggregates in 3-Dculture by first plating them overnight on Matrigel. The aggregates were then overlaid with type I collagen in the presence of EGF for the timesindicated. a The MCF-7-control cell aggregates grew considerably in size between day 1 and day 4 and remained relatively spherical. The MCF-7-podocell aggregates also grew in size but, in addition, they became elongated as they protruded and collectively pushed into the matrix. Livephase microscopy; scale bar = 100 μm. b The collective invasion of the aggregates was quantified after 4 days in 3-D culture by determiningtheir “elongation index”, which consisted of the longest length to width ratio for each aggregate. ***p <0.001, unpaired Student’s t test. c Cellaggregates maintained for 4 days in 3-D culture were fixed and coimmunostained for podocalyxin (red) and E-cadherin (green). Single x/y-axisconfocal images (top panels; scale = 40 μm) and 3-D reconstructed images (bottom panels) are shown. The reconstructions show that podocalyxinpromotes the cohesive elongation of the aggregates without disrupting E-cadherin localization at cell–cell adhesions. Note also that podocalyxinprimarily localized at multiple microlumina within cohesive tumor cell bud-like structures that bulged out from the elongated cell aggregates (arrows).In the lower panels the white outlined box indicates the rotation of the reconstructed confocal stack; the long axis of each box is 211 μm in lengthGraves et al. Breast Cancer Research  (2016) 18:11 Page 10 of 17Fig. 6 Podocalyxin-mediated collective invasion is actomyosin contractility dependent. a MCF-7-control and MCF-7-podo cell aggregates overlaidwith collagen were fixed and immunostained for exogenous mouse podocalyxin (red) and either the actin cytoskeletal linker ezrin (green, upperpanel) or the apical membrane marker Muc1 (green, lower panel). Single x/y-axis confocal images show that MCF-7-control cells localize ezrin tobasolateral surfaces and small membranous pockets within the spheroid interior. Apical Muc1 localized to small, closed internal domains in theMCF-control cell aggregates. In contrast, MCF-7-podo cells predominantly localized ezrin and Muc1 to expanded microluminal surfaces within theelongated invasive aggregates. Scale = 20 μm. b MCF-7-control and MCF-7-podo cell aggregates were fixed and stained for exogenous mousepodocalyxin (red) and f-actin (green). Single x/y-axis confocal micrographs (top panel) and 3-D projected images (bottom panel) show an enrichment off-actin along the expanded microlumenal surfaces within the MCF-7-podo cell aggregates where it colocalized with podocalyxin. Also, the outer cellsurfaces of the podocalyxin aggregates exhibited prominent f-actin-rich, but podocalyxin-spare, membrane projections that probed the surroundingECM (arrows). Scale bar = 40 μm. c, d MCF-7 cell aggregates were maintained in the presence of either DMSO (vehicle control) or themyosin II inhibitor blebbistatin (10 μM) for 4 days. Phase contrast images show that blebbistatin significantly blocked MCF-7-podo cellsfrom collectively invading the matrix by elongation c. This is quantified in d. ***p <0.001, *p <0.05, Student’s t test; scale bar = 100 μm.DMSO dimethylsulfoxideGraves et al. Breast Cancer Research  (2016) 18:11 Page 11 of 17to them (Fig. 6b). Interestingly, 3-D reconstructions re-vealed that the outer surface of MCF-7-podo cell clus-ters, but not MCF-7-control cell clusters, often formedf-actin-rich protrusions that were actually podocalyxin-negative (i.e., similar to the situation in the enhancedlamellipodia-like structures in 2-D monolayers) whichappeared to probe the surrounding collagen matrix(Fig. 6b, bottom right panel, arrows). Very dynamicperipheral protrusions were also observed at the edgeof the collectively invading MCF-7-podo cell aggre-gates by live video microscopy (Additional file 2: MovieS2). Thus, contractile actin-mediated protrusion for-mation may contribute to the collective invasion of theMCF-7-podo cells, a tentative conclusion that was sup-ported by the finding that blebbistatin, a pharmaco-logic inhibitor of actomyosin-dependent contractility,greatly diminished their elongation/collective invasion(Fig. 6c, d).Podocalyxin depletion inhibits collective invasion in 3-DcultureThe highly metastatic mouse mammary tumor 4T1cell line expresses significant amounts of endogenouspodocalyxin compared with genetically similar mammarytumor cell lines that are either weakly metastatic (66 cl4)or nonmetastatic (67NR; Fig. 7a). Despite their high meta-static potential, 4T1 cells form epithelial E-cadherin-containing adherens junctions [42]. We thus reasoned thatthese cells were good candidates to be collectively invasivein 3-D culture, which turned out to be the case. Specific-ally, when we preclustered 4T1-control cells and overlaidthem with collagen type 1, the clusters sent out collective,bud-like protrusions into the matrix that containedpolarity-disrupted aggregations of podocalyxin, some ofwhich appeared to be associated with microlumenal struc-tures (Fig. 7c, left panels; Additional file 3: Movie S3). Incontrast, when we stably knocked down podocalyxin,the resulting 4T1-podo KD 3603 cells (Fig. 7b) did notproductively invade the matrix as cohesive cell cohorts.Instead, a small proportion of 4T1-podo KD cells movedinto the gel singly in an elongated fashion (Fig. 7c, rightpanels, arrowheads; Additional file 3: Movie S4).DiscussionPodocalyxin was initially described as the major glom-erular anion in the kidney that plays a critical role in theformation of the small gaps between the individual pro-cesses of podocytes that become the slit diaphragmswhich are critical for plasma filtration during the initialphase of urine formation [14]. This was confirmed inpodocalyxin knockout mice where the glomerular epi-thelium remains patent and no filtration occurs [15].Podocalyxin was later shown to be gp135, which hadlong been used as a marker of the apical membranedomain of MDCK kidney epithelial cells in 2-D mono-layer culture [16]. Functional experiments in MDCKcells demonstrated that podocalyxin contributes to theformation of a preapical membrane domain during theformation of a single central lumen within polarizedMDCK cell spheroids maintained in 3-D collagen gels[16]. Interestingly, when apical–basal polarity is activelydisrupted by blocking integrin function, podocalyxin isnot trafficked properly, a single central lumen does notform efficiently, and the otherwise normal MDCK epi-thelial cell spheroids start to move collectively throughthe collagen matrix [19].When we force-expressed podocalyxin in normalmammary epithelial cells and placed the cells in 3-D cul-ture, they clustered together and, similar to MDCK cells,formed polarized spheroids with a large, single lumenthat was centrally located (Additional file 1: Figure S3).In contrast, epithelial MCF-7 breast tumor cells have anintrinsic polarity defect such that control MCF-7 spher-oids did not generate a single central lumen. Instead, asdemonstrated by Muc1 staining, the control MCF-7 cellclusters formed multiple small internal apical membranedomains. Forced podocalyxin overexpression in MCF-7cells served to facilitate the expansion and separation ofthese domains during microlumen formation withinelongating MCF-7-podo cell clusters that began to pro-trude collectively into the ECM. Therefore, the intrinsicpolarity defect in the MCF-7 spheroids, which has beenshown to be mediated, at least in part, by aberrant integ-rin signaling [41], may have contributed to the ability ofpodocalyxin to initiate their collective invasion in a fash-ion analogous to that which occurred in integrin-blocked MDCK cell spheroids [19].In MDCK cells, the movement of podocalyxin to ap-ical and luminal membrane domains that occurs duringthe normal polarization process is dependent on theinteraction of its cytoplasmic domain with the actinscaffolding proteins ezrin, NHERF-1, and NHERF-2 [19].We showed here that ezrin colocalized with podocalyxinat the microlumenal membranes of collectively invasiveMCF-7 cell aggregates, and we have shown previouslythat NHERF-1 also colocalizes with podocalyxin in thesecells in small punctate microvillus-like structures whenthey are maintained in 2-D monolayer culture [34].Mostov and colleagues have suggested that an inappro-priate targeting of podocalyxin and these associated scaf-folding proteins contributes to the formation of leadinglamella-like structures that drive collective MDCK cellinvasion under polarity-disrupted conditions in 3-D cul-ture [19]. We found that force-expressed apical podoca-lyxin localized directly adjacent to expanded actin-richlamella at the free edge of collective migrating MCF-7monolayers in 2-D culture. Importantly, when we treatedthese cells with a pharmacologic inhibitor of ezrinGraves et al. Breast Cancer Research  (2016) 18:11 Page 12 of 17activation, we observed a disruption of ezrin and podoca-lyxin localization in the apical domain of these cells and adecrease in collective migration which broadly agrees withthe findings of Mostov and colleagues who used normalepithelial cells where polarity was actively disrupted in3-D culture to initiate an ezrin-dependent podocalyxin-mediated collective invasion [19].We also found that podocalyxin overexpression stimu-lated the formation of actin-rich lamellipodia-likeprotrusions that emanated from the collectively invad-ing MCF-7-podo cell aggregates in 3-D culture. Thus,podocalyxin likely acts in subtly different ways frompodoplanin, which is another small mucin that initiatescollective tumor cell motility by inducing the formation offilopodial rather than lamella-like structures [43]. Regard-less, the ability of polarity-disrupted podocalyxin to initi-ate the formation of cellular processes that protrude intothe matrix is likely critical for facilitating collectiveFig. 7 Stable knockdown of podocalyxin in highly invasive 4T1 murine breast tumor cells reduces collective invasion. a Whole cell lysates from syngeneicmurine breast tumor cell lines with different capacities to metastasize were probed for endogenous podocalyxin expression. Endogenous podocalyxin wasmost highly expressed in 4T1 cells, which are highly metastatic compared with the nonmetastatic 67NR and weakly metastatic 66 cl4 cell lines[42]. b 4T1 cells stably transfected with either control vector (4T1-control) or vectors containing an shRNA sequence targeting the 3′ untranslated region ofthe murine podocalyxin transcript were analyzed for endogenous podocalyxin expression by western blotting. Stable 4T1 cell populations, each expressinga different short hairpin sequence, show a moderate (podo KD-3293) and near complete knockdown of endogenous podocalyxin (podo KD-3603),respectively. c 4T1-control or 4T1-podo KD 3603 tumor cells were aggregated on Matrigel overnight, overlaid with collagen type I, and maintained in3-D culture for 4 days. They were then imaged, live, by phase microscopy (upper panels) or fixed and immunostained for endogenous podocalyxin(red) and the tight junction protein ZO-1 (green) followed by confocal imaging (lower panels). The 4T1-control cells invaded as cohesive multicellularstrands with scattered small pockets of poorly polarized podocalyxin throughout. In contrast, aggregates from podocalyxin knockdown cells did noteffectively invade as cohesive multicellular strands; they did, however, occasionally release single cells into the matrix (arrowheads). Scale = 50 μmGraves et al. Breast Cancer Research  (2016) 18:11 Page 13 of 17invasion because such processes are prominent in “leader”[11] and “trailblazer” [12] cells that drive collective inva-sion by heterogeneous breast cancer cell populations.Additionally, as is the case with leader and trailblazercell-driven collective invasion, MCF-7-podo cells onlygenerated these processes and became collectively in-vasive when stromal collagen (type I) was added to the3-D culture system (i.e., they did not form processes andthey were not invasive in Matrigel-only 3-D cultures;data not shown).Podocalyxin has been shown to activate the actinnucleator cortactin in a NHERF-1-dependent manner[44]. This activation may also contribute to podocalyxin’sability to initiate leading lamellipodial protrusions underconditions that favor collective cell motility. Addition-ally, podocalyxin-mediated alterations in NHERF proteinlocalization may act to augment growth factor-mediatedsignaling that contributes to the collective motilephenotype given the ability of NHERF proteins tointeract with multiple hormone and growth factor recep-tors [45]. Indeed, we found that podocalyxin overexpres-sion increased EGF-dependent mitogen-activated proteinkinase and phosphoinositide 3-kinase-dependent signalingin MCF-7 cells (Additional file 1: Figure S5). We are nowmanipulating the interactions of NHERF-1 (MCF-7 cellsdo not express appreciable amounts of NHERF-2) withthe cytoplasmic tail of podocalyxin to determinewhich of these interactions influences EGF-dependentsignaling, lamellipodial protrusion, and/or collective mo-tility in the presence or absence of ezrin. Finally, the heav-ily glycosylated extracellular domain of podocalyxin couldact to augment a glycocalyx-mediated stimulation ofintegrin-dependent signaling to further facilitate epithelialtumor cell motility [46].A polarity protein that is linked to the cytoskeleton,podocalyxin may modulate subcellular asymmetries incytoskeletal contractility that have also been shown tocontribute to experimental collective cell motility [10].In accordance with this, it has been demonstrated thatpodocalyxin augments contractile RhoA activation at thefree surface of kidney epithelial cells in an ezrin-dependent manner [47, 48] and we have shown thatpodocalyxin-induced collective invasion is dependent onacto-myosin contractility. Sahai and colleagues havedemonstrated that the Par3 and Par6 polarity proteinsgenerate contractile asymmetries in collectively migrat-ing squamous carcinoma cells through their interactionwith the collagen-binding Discoidin Domain Receptor 1(DDR1) [49]. Interestingly, the expression of DDR1 helpsto clinically distinguish invasive ductal breast carcinomas,which are often cohesive and epithelial, from invasivelobular breast carcinoma which often generate mesenchy-mal cells that move through the stromal matrix in a“single file” arrangement [50]. It is intriguing to considerthe possibility that podocalyxin and DDR1 may bemechanistically linked given that their expression iscoordinately suppressed by miR-199b-5p, a microRNAwhose loss leads to the elevation of both DDR1 andpodocalyxin in acute myeloid leukemia [51].Tumor budding, which is regarded to be an importantcomponent of local tumor cell dissemination, has signifi-cant prognostic significance in a number of solid tumortypes, including breast cancer [6–8]. By the strictest def-inition, tumor buds are small (i.e., less than five cells[9]). Interestingly, however, larger invasive cell clustersand micronodules can have as great, or greater, prognos-tic significance. The latter structures, which can consistof dozens of cells, have been classified as “poorly differ-entiated clusters” (PDCs) in invasive breast cancer giventhat their polarity is often disrupted such that they donot contain single, well-defined central lumens [8]. Basedon morphology, size, and lack of a single central lumen, itwould appear that the collectively invasive cell clustersthat we observed in MCF-7-podo cell xenotransplantsmore closely resemble clinical PDCs. Whether or notincreases in proliferation rates play a role in this buddingis not clear clinically [7–9], and while proliferatingcells were present in the budding clusters of MCF-7-podo cells in the xenotransplants, any proliferativedifferences with control tumors were, at best, minimal.Thus, it is unlikely that proliferation is a major driverof podocalyxin-mediated collective invasion in vivo.Additionally, differences in proliferation rates were not afactor in the increased collective motility of the MCF-7-podo cells in culture.Like smaller tumor buds, the presence of PDCs is wellcorrelated with LVI which is another prognostic indica-tor of progression in invasive breast cancer [8]. Initially,we found that podocalyxin overexpression is also a prog-nostic indicator of poor outcome, overall, in invasivebreast cancer [13]. This was later confirmed by Andrulisand colleagues, who demonstrated that podocalyxinoverexpression is also positively correlated with LVIamongst an entire invasive breast cancer cohort [31].Interestingly, however, Andrulis’ group also found thatpodocalyxin overexpression was negatively correlatedwith outcome within the LVI breast cancer subco-hort itself. This suggests to us that podocalyxin maycontribute to an initial local dissemination via collect-ive tumor invasion but that a subsequent retention ofpodocalyxin could suppress a later more aggressivespread by helping to prevent, for example, a switchfrom a collective to a more mesenchymal mode of inva-sion. Given this, it will be interesting to determinewhether endogenous podocalyxin contributes to thetightly regulated collective invasion that helps to drivebranching morphogenesis during normal mammarygland development [52].Graves et al. Breast Cancer Research  (2016) 18:11 Page 14 of 17There is one report that podocalyxin facilitates a trans-forming growth factor beta-mediated mesenchymal trans-formation of A549 lung carcinoma cells [53]. However, wefound that podocalyxin overexpression did not preventcell–cell junction formation or downregulate epithelialmarker expression in MCF-7 cells. We also force-expressed podocalyxin in epithelial ovarian carcinomacells [30] and saw no evidence that this leads to theemergence of an overt mesenchymal phenotype on itsown. However, we have not fully investigated the possibil-ity that podocalyxin upregulates selected mesenchymalmarkers within otherwise epithelial tumor cells whichhave been shown to be a part of the collectively invasivebreast tumor budding phenotype [9, 12]. Interestingly,when we knocked down endogenous podocalyxin inhighly metastatic, but still epithelial, 4T1 mammary tumorcells, while there was a clear decrease in collective inva-sion there was also an increase in the movement of singleelongated cells into the matrix. Thus, it is possible that alate-stage loss of podocalyxin may stimulate a shift to-wards a single cell mode of motility which could be associ-ated with the emergence of a mesenchymal phenotype.Under some conditions, such a mode shift could poten-tially lead to the emergence of a more aggressive pheno-type, a notion that is supported by the clinical data ofAndrulis and colleagues already described [31]. Whetheror not the loss of podocalyxin can initiate a switch be-tween collective and mesenchymal modes of tumor cellinvasion is something we are actively investigating.In addition to invasive breast cancer, podocalyxin over-expression is also correlated with poor outcome in otherepithelial-derived cancers where alterations in polarityfigure prominently in tumor formation and progression.These include renal cell, colorectal, high-grade serousovarian, and bladder carcinomas [28–30, 32, 33]. In thelatter case, increased podocalyxin levels are also predictiveof outcome [33]. Given these findings, we developed anti-bodies against the podocalyxin extracellular domain in aneffort to block its tumor progression promoting function.Our initial studies indicate that this is feasible in preclin-ical transplantation experiments using metastatic breastand mammary tumor cell lines that overexpress podo-calyxin [54]. Therefore, podocalyxin overexpression mayrepresent a novel therapeutic target in metastatic carcin-oma progression.ConclusionsPodocalyxin overexpression causes epithelial tumor cellsto become collectively motile in culture and to initiatecollectively invasive tumor budding in orthotopic trans-plants in vivo. Therefore, in addition to being prognos-tically significant for breast cancer outcome, podocalyxinoverexpression may functionally contribute to breastcancer progression.Additional filesAdditional file 1: Is Figure S1 showing podocalyxin has little effect onsubcutaneous tumor size a or proliferation in monolayer culture b,Figure S2 showing podocalyxin overexpression promotes local invasionof MCF-7 tumor cell xenografts,. Figure S3 showing that the ezrin inhibitorNSC668394 disrupts apical podocalyxin localization in monolayer culture,Figure S4 showing normal mammary epithelial cells continue to form spheresand form single, polarized lumens in 3-D culture, and Figure S5 showingpodocalyxin expression increases EGF-mediated signaling. (ZIP 1056 kb)Additional file 2: Is Movie S1 showing control MCF-7 cells present littlecollective invasion 3-D culture, and Movie S2 showing podo-overexpressingcells send out very dynamic processes and show considerable collectiveinvasion in 3-D culture. (ZIP 2262 kb)Additional file 3: Is Movie S3 showing collective invasion of control 4T1cells that express considerable endogenous podocalyxin in 3-D culture, andMovie S4 showing decreased collective invasion of 4T1 cells in 3-D culturewhen endogenous podocalyxin is stably knocked down. (ZIP 1243 kb)Abbreviations2-D: Two dimensional; 3-D: Three dimensional; DDR1: Discoidin DomainReceptor 1; DMSO: Dimethylsulfoxide; ECM: Extracellular matrix; EGF: Epidermalgrowth factor; EMT: Epithelial to mesenchymal transition; FACS: Fluorescent-activated cell sorting; H & E: Hematoxylin and eosin; LVI: Lymphovascularinvasion; PDC: Poorly differentiated cluster; shRNA: Short hairpin RNA.Competing interestsMLG, KMM, and CDR hold a patent to utilize podocalyxin as a prognosticindicator of breast and ovarian cancer outcome. The remaining authorsdeclare that they have no competing interests.Authors’ contributionsMLG designed the study, carried out the in vivo and 3-D culture assays,analyzed the data, and helped draft the manuscript. JAC helped design thestudy, characterized the cell lines where podocalyxin was manipulated, andhelped draft the manuscript. PA carried out the ezrin inhibition assays andhelped revise the manuscript critically for important intellectual content. EMBcarried out videomicroscopic assays and helped revise the manuscript criticallyfor important intellectual content. JSN helped design the study, helped carryout the podocalyxin overexpression and in vivo experiments, and helped draftthe manuscript. CBG assessed the in vivo phenotypes and helped develop thecollective invasion hypothesis for podocalyxin which was critical to the draftingof the manuscript and revising it critically. KMM helped design the study,analyze data, and draft the manuscript. CDR conceived of the study,helped analyze the data, and helped draft the manuscript. All authorsread and approved the final manuscript.AcknowledgementsThe authors thank Eileen Harder for technical contributions to theimmunostaining and polarity assessments of podocalyxin-overexpressingcells. This work was supported by a grant from the Canadian Institutesof Health Research (MOP-125992) to KMM and CDR.Author details1Department of Cellular and Physiological Sciences, University of BritishColumbia, 2350 Health Sciences Mall, Vancouver, BC V6T 1Z3, Canada.2Deeley Cancer Research Center, 2410 Lee Avenue, Victoria V8R 6V5BC,Canada. 3Genetic Pathology Evaluation Center, Vancouver General Hospital,2660 Oak Street, Vancouver V6H 3Z6BC, Canada. 4Biomedical ResearchCenter, University of British Columbia, Vancouver, BC V6T 1Z3, Canada.Received: 29 June 2015 Accepted: 31 December 2015References1. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell.2011;144(5):646–74.2. Mitrovic B, Schaeffer DF, Ridell RH, Kirsh R. Tumor budding in colorectalcarcinoma: time to take notice. Mod Pathol. 2012;25:1315–25.Graves et al. Breast Cancer Research  (2016) 18:11 Page 15 of 173. Ohike N, Coban I, Kim GE, Basturk O, Tajiri T, Krasinskas A, et al. Tumorbudding as a strong prognostic indicator in invasive ampullary carcinomas.Am J Surg Pathol. 2010;34(10):1417–24.4. Karamitopoulou E, Zlobec I, Born D, Kondi-Pafiti A, Lykoudis P, Mellou A,et al. Tumour budding is a strong and independent prognostic factor inpancreatic cancer. Eur J Cancer. 2013;49(5):1032–9.5. Taira T, Ishii G, Nagai K, Yoh K, Takahashi Y, Matsumura Y, et al.Characterization of the immunophenotype of the tumor budding and itsprognostic implications in squamous cell carcinoma of the lung. LungCancer. 2012;76(3):423–30.6. Man Y-G. Tumor cell budding from focally disrupted tumor capsules: acommon pathway for all breast cancer subtyped-derived invasion? J Cancer.2010;1:32–7.7. Liang F, Cao W, Wang Y, Li L, Zhang G, Wang Z. The prognosticvalue of tumor budding in invasive breast cancer. Pathol Res Pract.2013;209(5):269–75.8. Sun Y, Liang F, Cao W, Wang K, He J, Wang H, et al. Prognostic value ofpoorly differentiated clusters in invasive breast cancer. World J SurgOncology. 2014;12:310–8.9. Bronsert P, Enderle-Ammour K, Bader M, Timme S, Kuehs M, Csanadi A, et al.Cancer cell invasion and EMT marker expression: a three-dimensional studyof the human cancer–host interface. J Pathol. 2014;234:410–22.10. Friedl P, Locker J, Sahai E, Segall JE. Classifying collective cancer cellinvasion. Nat Cell Biol. 2012;14(8):777–83.11. Cheung KJ, Gabrielson E, Werb Z, Ewald EJ. Collective invasion in breastcancer requires a conserved basal epithelial program. Cell. 2013;155:1639–51.12. Westcott JM, Prechtl AM, Maine EA, Dang TT, Esparza MA, Sun H, et al.An epigenetically distinct cancer cell subpopulation promotes collectiveinvasion. J Clin Invest. 2015;125(5):1927–43.13. Somasiri A, Nielsen JS, Makretsov N, McCoy ML, Prentice L, Gilks CB, et al.Overexpression of the anti-adhesin podocalyxin is an independent predictorof breast cancer progression. Cancer Res. 2004;64(15):5068–73.14. Kerjaschki D, Sharkey DJ, Farquhar MG. Identification and characterization ofpodocalyxin–the major sialoprotein of the renal glomerular epithelial cell.J Cell Biol. 1984;98(4):1591–6.15. Doyonnas R, Kershaw DB, Duhme C, Merkens H, Chelliah S, Graf T, et al.Anuria, omphalocele, and perinatal lethality in mice lacking the CD34-relatedprotein podocalyxin. J Exp Med. 2001;194(1):13–27.16. Meder D, Shevchenko A, Simons K, Fullekrug J. Gp135/podocalyxin andNHERF-2 participate in the formation of a preapical domain duringpolarization of MDCK cells. J Cell Biol. 2005;168(2):303–13.17. Ferrari A, Veligodskiy A, Berge U, Lucas MS, Kroschewski R. ROCK-mediatedcontractility, tight junctions and channels contribute to the conversion of apreapical patch into apical surface during isochoric lumen initiation. J CellSci. 2008;121(Pt 21):3649–63.18. Strilic B, Kucera T, Eglinger J, Hughes MR, McNagny KM, Tsukita S, et al. Themolecular basis of vascular lumen formation in the developing mouseaorta. Dev Cell. 2009;17(4):505–15.19. Bryant DM, Roignot J, Datta A, Overeem AW, Kim M, Yu W, et al. A molecularswitch for the orientation of epithelial cell polarization. Dev Cell. 2014;31:1–17.20. Kelley TW, Huntsman D, McNagny KM, Roskelley CD, Hsi ED. Podocalyxin:a marker of blasts in acute leukemia. Am J Clin Pathol. 2005;124(1):134–42.21. Casey G, Neville PJ, Liu X, Plummer SJ, Cicek MS, Krumroy LM, et al.Podocalyxin variants and risk of prostate cancer and tumor aggressiveness.Hum Mol Genet. 2006.22. Ney JT, Zhou H, Sipos B, Buttner R, Chen X, Kloppel G, et al. Podocalyxin-likeprotein 1 expression is useful to differentiate pancreatic ductal adenocarcinomasfrom adenocarcinomas of the biliary and gastrointestinal tracts. Hum Pathol.2007;38(2):359–64.23. Heukamp LC, Fischer HP, Schirmacher P, Chen X, Breuhahn K, Nicolay C,et al. Podocalyxin-like protein 1 expression in primary hepatic tumours andtumour-like lesions. Histopathology. 2006;49(3):242–7.24. Schopperle WM, Kershaw DB, DeWolf WC. Human embryonal carcinomatumor antigen, Gp200/GCTM-2, is podocalyxin. Biochem Biophys ResCommun. 2003;300(2):285–90.25. Yasuoka H, Tsujimoto M, Hirokawa M, Tori M, Nakahara M, Miyauchi A, et al.Podocalyxin expression in undifferentiated thyroid carcinomas. J Clin Pathol.2008;61(11):1228–9.26. Yasuoka H, Tsujimoto M, Inagaki M, Kodama R, Tsuji H, Iwahashi Y, et al.Clinicopathological significance of podocalyxin and phosphorylated ezrin inuterine endometrioid adenocarcinoma. J Clin Pathol. 2012;65(5):399–402.27. Hayatsu N, Kaneko MK, Mishima K, Nishikawa R, Matsutani M, Price JE, et al.Podocalyxin expression in malignant astrocytic tumors. Biochem BiophysRes Commun. 2008;374(2):394–8.28. Larsson A, Johansson ME, Wangefjord S, Gaber A, Nodin B, Kucharzewska P,et al. Overexpression of podocalyxin-like protein is an independent factor ofpoor prognosis in colorectal cancer. Br J Cancer. 2011;105(5):666–72.29. Larsson AH, Nodin B, Syk I, Palmquist I, Uhlen M, Eberhard J, et al.Podocalyxin-like protein expression in primary colorectal cancer andsynchronous lymph node metastases. Diagn Pathol. 2013;8(1):109.30. Cipollone JA, Graves ML, Kobel M, Kalloger SE, Poon T, Gilks CB, et al. The anti-adhesive mucin podocalyxin may help initiate the transperitoneal metastasis ofhigh grade serous ovarian carcinoma. Clin Exp Metastasis. 2012;29(3):239–52.31. Forse CL, Yilmaz YE, Pinnaduwage D, O’Malley FP, Mulligan AM, Bull SB, et al.Elevated expression of podocalyxin is associated with lymphatic invasion,basal-like phenotype, and clinical outcome in axillary lymph node-negativebreast cancer. Breast Cancer Res Treat. 2013;137(3):709–19.32. Hsu YH, Lin WL, Hou YT, Pu YS, Shun CT, Chen CL, et al. Podocalyxin EBP50ezrin molecular complex enhances the metastatic potential of renal cellcarcinoma through recruiting Rac1 guanine nucleotide exchange factorARHGEF7. Am J Pathol. 2010;176(6):3050–61.33. Boman K, Larsson AH, Segersten U, Kuteeva E, Johannesson H, Nodin B, et al.Membranous expression of podocalyxin-like protein is an independent factor ofpoor prognosis in urothelial bladder cancer. Br J Cancer. 2013;108(11):2321–8.34. Nielsen JS, Graves ML, Chelliah S, Vogl AW, Roskelley CD, McNagny KM.The CD34-related molecule podocalyxin is a potent inducer of microvillusformation. PLoS One. 2007;2(2):e237.35. Dai JL, Wang L, Sahin AA, Broemeling LD, Schutte M, Pan Y. NHERF (Na+/H+ exchanger regulatory factor) gene mutations in human breast cancer.Oncogene. 2004;23(53):8681–7.36. Elliott BE, Meens JA, SenGupta SK, Louvard D, Arpin M. The membranecytoskeletal crosslinker ezrin is required for metastasis of breast carcinomacells. Breast Cancer Res. 2005;7(3):R365–73.37. Sarrio D, Rodriguez-Pinilla SM, Dotor A, Calero F, Hardisson D, Palacios J.Abnormal ezrin localization is associated with clinicopathological features ininvasive breast carcinomas. Breast Cancer Res Treat. 2006;98(1):71–9.38. Sizemore S, Cicek M, Sizemore N, Ng KP, Casey G. Podocalyxin increases theaggressive phenotype of breast and prostate cancer cells in vitro throughits interaction with ezrin. Cancer Res. 2007;67(13):6183–91.39. Bulut G, Hong SH, Chen K, Beauchamp E, Rahim S, Kosturko GW, et al. Smallmolecule inhibitors of ezrin inhibit the invasive phenotype of osteosarcomacells. Oncogene. 2012;31:269–81.40. Schiemann S, Schwirzke M, Brunner N, Weidle UH. Molecular analysis of twomammary carcinoma cell lines at the transcriptional level as a modelsystem for progression of breast cancer. Clin Exp Metastasis. 1998;16:129–39.41. Wang F, Hansen RK, Radisky D, Yoneda T, Barcellos-Hoff MH, Petersen OW, et al.Phenotypic reversion or death of cancer cells by altering signaling pathways inthree-dimensional contexts. J Natl Cancer Inst. 2002;94(19):1494–503.42. Lou Y, Preobrazhenska O, Auf dem Keller U, Sutcliffe M, Barclay L, McDonaldPC, et al. Epithelial-mesenchymal transition (EMT) is not sufficient for spontaneousmurine breast cancer metastasis. Dev Dyn. 2008;237(10):2755–68.43. Wicki A, Lehembre F, Wick N, Hantusch B, Kerjaschki D, Christofori G.Tumor invasion in the absence of epithelial-mesenchymal transition:podoplanin-mediated remodeling of the actin cytoskeleton. Cancer Cell.2006;9(4):261–72.44. Lin CW, Sun MS, Lia MY, Chung CH, Chi YH, Chiou LT, et al. Podocalyxin-like1 promotes invadopodia formation and metastasis through activation ofRac1/Cdc42/cortactin signaling in breast cancer cells. Carcinogenesis.2014;35(11):2425–35.45. Weinman EJ, Hall RA, Friedman PA, Liu-Chen LY, Shenolikar S. Theassociation of NHERF adaptor proteins with G protein-coupled receptorsand receptor tyrosine kinases. Annu Rev Physiol. 2006;68:491–505.46. Paszek MJ, DuFort CC, Rossier O, Bainer R, Mouw JK, Godula K, et al. Thecancer glycocalyx mechanically primes integrin-mediated growth andsurvival. Nature. 2014;511(7509):319–25.47. Schmieder S, Nagai M, Orlando RA, Takeda T, Farquhar MG. Podocalyxinactivates RhoA and induces actin reorganization through NHERF1 and Ezrinin MDCK cells. J Am Soc Nephrol. 2004;15(9):2289–98.48. Fukasawa H, Obayashi H, Schmieder S, Lee J, Ghosh P, Farquhar MG.Phosphorylation of podocalyxin (Ser415) prevents RhoA and ezrin activationand disrupts its interaction with the actin cytoskeleton. Am J Pathol.2011;179(5):2254–65.Graves et al. Breast Cancer Research  (2016) 18:11 Page 16 of 1749. Hidalgo-Carcedo C, Hooper S, Chaudhry SI, Williamson P, Harrington K,Leitinger B, et al. Collective cell migration requires suppression of actomyosinat cell-cell contacts mediated by DDR1 and the cell polarity regulators Par3and Par6. Nat Cell Biol. 2011;13(1):49–58.50. Turashvili G, Bouchal J, Baumforth K, Wei W, Dziechciarkova M, Ehrmann J, et al.Novel markers for differentiation of lobular and ductal invasive breast carcinomasby laser microdissection and microarray analysis. BMC Cancer. 2007;7:55.51. Favreau AJ, Cross EL, Sathyanarayana P. miR-199b-5p directly targets PODXLand DDR1 and decreased levels of miR-199b-5p correlate with elevatedexpression of PODXL and DDR1 in acute myeloid leukemia. Am J Hematol.2012;87(4):442–6.52. Ewald AJ, Brenot A, Duong M, Chan BS, Werb Z. Collective epithelial migrationand cell rearrangements drive mammary branching morphogenesis. Dev Cell.2008;14(4):570–81.53. Meng X, Ezzati P, Wilkins JA. Requirement of podocalyxin in TGF-beta-inducedepithelial mesenchymal transition. PLoS One. 2011;6(4):e18715.54. Snyder KA, Hughes MR, Hedberg B, Brandon J, Hernaez DC, Bergqvist P, et al.Podocalyxin enhances breast tumor growth and metastasis and is a target formonoclonal antibody therapy. Breast Cancer Res. 2015;17(1):46.•  We accept pre-submission inquiries •  Our selector tool helps you to find the most relevant journal•  We provide round the clock customer support •  Convenient online submission•  Thorough peer review•  Inclusion in PubMed and all major indexing services •  Maximum visibility for your researchSubmit your manuscript atwww.biomedcentral.com/submitSubmit your next manuscript to BioMed Central and we will help you at every step:Graves et al. Breast Cancer Research  (2016) 18:11 Page 17 of 17


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items