Background

The sina qua non of inflammatory breast cancer (IBC) is numerous lymphovascular tumor emboli within the breast and especially within overlying dermal lymphatics where they are thought responsible for the clinical signs of IBC including rubor, calor, dolor and swelling [1,2,3]. All molecular subtypes of IBC exhibit this unifying phenotype whose mechanism remains unknown [4].

Previous studies by us and others using human xenograft models of inflammatory breast cancer and related cell lines have shed some light on the mechanisms responsible for the genesis of tumor clumps and derived emboli in the metastatic cascade of IBC [5,6,7,8,9,10]. In these studies membrane E-cadherin is an important adhesion molecule and its overexpression responsible for the tumor clumping that occurs [10,11,12]. Although E-cadherin generally is also considered a tumor suppressor gene / protein in tumor invasion and metastasis whose loss contributes to epithelial-mesenchymal transition (EMT) [13,14,15], in the setting of the tumor cell clump or the lymphovascular embolus of IBC, E-cadherin acts more like a dominant oncogene. [16,17,18]

However, these previous observations do not explain the plethora of emboli that occur in IBC, especially within dermal lymphatics. Certainly E-cadherin overexpression while contributing to the tumor cell clump and the tumor cell embolus can not alone account for the plethora of lymphovascular emboli observed in IBC.

For these reasons we sought explanations that might be responsible for this unique phenotype of IBC. We had collected a number of cases of human breast cancer of all types including IBC and non-IBC that were available to study through TMA digital algorithmic image analysis for detailed human observational studies. We also had pooled microdissected emboli available from the IBC cases. We had also established two contrasting xenograft models of IBC, Mary-X and Karen-X, to use in experimental studies to explain the unique phenotype of florid lymphovascular emboli observed in IBC.

Methods

Institutional approvals and initial establishment of the xenografts

Collection and use of human tissues from patients with breast cancer including non-IBC as well as IBC cases, completely anonymized, was approved by The Ohio State University Cancer Institutional Review Board (IRB) under protocol 2006C0042.

Both Mary-X and Karen-X were derived from patients with a biopsy proven diagnosis of IBC in the 1990’s and made into transplantable xenografts. Patient consents were obtained and studies were conducted under the UCLA’s Human Subject Protection Committee and the Chancellor’s Animal Research Committee (Certification 95–127-11). Mary-X was derived from a biopsy of primary IBC whereas Karen-X was derived from a biopsy of a hepatic metastases. Mary-X exhibited triple (ER, PR and Her-2/neu) negativity whereas Karen-X was weakly estrogen receptor (ER) positive. Both xenografts have been phenotypically stable for over 25 years of passage.

Both patient-derived xenografts (PDX) initially were directly implanted into the 4th mammary fat pad of both nude as well as Scid mice. When tumors emerged, they were both serially transplanted as well as placed in suspension culture. Both Mary-X and Karen-X spontaneously formed spheroids. The spheroids of Mary-X were tight and highly dense whereas the spheroids of Karen-X appeared as loose aggregates. Neither line formed monolayers on either plastic or fibronectin, vitronectin or collagen coated plastic. 103 spheroids (100µ diameter) with each spheroid consisting of approximately 103 cells or a total of 106 cells were also injected into the 4th mammary fat pad. In the case of both serially transplants or injected spheroids, tumors emerged which illustrated a stable phenotype and a characteristic histopathology. Over the years the sites of injection in the mice were varied to include the other mammary fat pads and non-orthotopic sites. Irrespective of the site of injection, both the phenotypes, histopathology and biology of Mary-X and Karen-X remained constant with passage. In previous studies with Mary-X, sorted signal cells, eg. stem cells, were also obtained on the basis of several stem cell markers and reinjected with the maintenance of a stable phenotype [19, 20]. Cells lacking stem cell markers were not tumorigenic.

Continuing animal studies were approved by The Ohio State University’s Institutional Animal Care and Use Committee (IACUC), protocol 2007A0218 and by The Ohio State University’s Institutional Biosafety Committee, protocol 2007R0057. Additional animal studies were approved by the University of Nevada’s School of Medicine and the Nevada Cancer Institute’s IACUC, protocols 00439 and 00440 when the corresponding author of this study was affiliated with these previous institutions. Final animal studies were conducted under an Interinstitutional Agreement between the California University of Science and Medicine and Anticancer, Inc. using the latter’s IACUC protocol D16-00503 and Office of Animal Welfare (OLAW) A3873-01.

ATCC patent deposits and cell identification

Mary-X and its in vitro derived spheroids were deposited in the American Type Culture Collection (ATCC) cell repository (Manassas, VA) as PTA-2737 and PTA-27376, respectively, and recently verified and re-verified to be both novel and human in origin (STRA4993). Likewise Karen-X and its in vitro derived spheroidal aggregates were deposited in the ATCC cell repository as PTA-126676 and PTA-126677 respectfully and recently verified to be both novel and human in origin (STRA4980). Mary-X is a well-known transplantable model of inflammatory breast cancer and has been used in numerous previous studies [19,20,21]. Karen-X has not previously been reported, although it has been ATCC deposited.

Observational studies

Cases of IBC

2000 cases of non-IBC and 100 cases of IBC were randomly selected from a database and The Ohio State University’s Information Warehouse and anonymized.

TMA construction

Multiple 2-mm tissue cores of tumor from each paraffin-embedded donor block (average of 10 cores / block) were arrayed into recipient TMA blocks, each of which contained 100 cores. All retrieved cases were processed in this manner.

Histological and immunohistochemistry studies

Primary antibodies used included rat anti-human Ki-67 (DakoCytomation, Carpinteria, CA), D2-40 anti-podoplanin (clone D2-40, Dako, catalog number M3619) and anti-CD31 (rabbit polyclonal, Spring Bioscience, catalog number E11114). The appropriate secondary antibodies were linked to an identifying chromogen.

Imaging analysis and quantitation of Ki-67 index, mitotic count, nuclear size (grade) and lymphovascular tumor emboli (embolic density, shape and size)

Our specific TMA algorithms carried out virtual alignment, image processing, and the application of the epithelial recognition algorithms (ERAs) and specific recognition algorithms (SRAs), the latter based upon immunocytochemical studies that quantitated nuclear Ki-67 immunoreactivity [22,23,24,25]. Additional specific morphological algorithms included a mitosis algorithm [26], a nuclear grade (size) algorithm [27], and epithelial clustering algorithms quantitating embolic density and measuring embolic shape (elongation ratio: short/long axis) and size (perimeter) utilizing ImageJ software [28]. Image acquisition by either the Aperio ScanScope T2 System (Aperio, Vista, CA) or the iSCAN System (BioImagene, Inc, Cupertino, CA.) produced equivalent results. The imaging apparatus captured sequential fields of view (FOV) from each core to cover the core in its entirety. Modes and means ± standard deviations (SD) were calculated for all these measurements. All experiments were performed in quintuplicate.

Experimental studies

Continued xenograft studies

Athymic (nude) mice as well as Scids on BALB/c backgrounds, 4 week old females, purchased from Anticancer, Inc. (San Diego, CA) were derived from their respective breeding colonies. Mary-X and Karen-X were grown as transplantable xenografts. At selected intervals ranging up to 4 months, the mice were euthanized with 25 mg/ml ketamine, 2 mg/ml xylazine and acepromazine and subjected to necropsy. Routine histological and immunocytochemical studies were carried out on excised tumors. Antibodies used included biotinylated E-cadherin rabbit mAb (24E10) (Santa Cruz Biotechnology, Santa Cruz, CA) which recognized both full length E-cadherin (E-cad/FL) and its E-cadherin/N-terminal fragment1 (E-cad/NTF1), followed by a labeled Streptavidin–Biotin Complex (Santa Cruz Biotechnology, Santa Cruz, CA). DAB chromogen was used to develop the brown color.

In vitro studies

Both Mary-X and Karen-X were placed in culture and 2 mm thick slices of each tumor were minced in RPMI media with 10% fetal calf serum giving rise to liberated loose aggregates in suspension culture. These aggregates were then passed through cell strainers whose pore sizes ranged from 40 to 500 µm. The morphology of the aggregates were observed over the next 24–72 h by phase contrast microscopy using the CytoSMART Lux2 (CytoSmart Technologies B.V. The Netherlands).

Confocal single and double label immunofluorescence experiments on Mary-X and Karen-X

Mary-X spheroids and Karen-X loose aggregates were subjected to single label immunofluorescence studies using Alexa Fluor 488-conjugated 24E10 (#3199) (Cell Signaling Technology, Inc.) which recognized both E-cad/FL as well as E-cad/NTF1. In order to immobilize the spheroids and loose aggregates, glass-bottom dishes were coated with Cell-TEK adhesive. The adherent spheroids were then fixed with 4% paraformaldehyde, after permeabilizing with TX-100 and blocking with normal goat serum. After washing with PBS 4–5 times, each for 10 min, the spheroids and aggregates were incubated with 24E10. The dishes were finally mounted with Vectorshield mounting medium with DAPI (#H-1200) (Vector Laboratories) and viewed with a Olympus Fluoview-1000 confocal scanning system.

Mary-X and Karen-X were also subjected to double label immunofluorescence studies. Double label immunofluorescence experiments were carried out using the following combinations of antibodies: Alexa Fluor 488-conjugated 24E10 which recognized both E-cad/NTF1 as well as E-cad/FL and goat polyclonal antibody to mouse podoplanin or CD31 (R&D Systems, Inc.) which recognized murine lymphatics or blood vessels followed by Alexa Fluor 594-conjugated donkey anti-goat (#A11058) (Invitrogen, Inc.). Tumor emboli were recognized by E-cadherin positivity surrounded by circumferential podoplanin (lymphatics) or CD31 positivity.

Isolation of Circulating Tumor Cells (CTCs) using EpCAM microbeads

CTCs were isolated from 10 mice each harboring the Mary-X and Karen-X xenografts. Blood from mice harboring at least 2 cm in diameter tumors were obtained by cardiac puncture and CTCs isolated with Epithelial Cell Adhesion Molecule (EpCAM) microbeads [29].

Laser capture microdissection

Frozen Sects. (8 µm) of Mary-X and human IBC cases were obtained, fixed in 70% ethanol, stained with hematoxylin and progressively dehydrated. Tumor emboli and non-embolic solid areas of Mary-X and human IBC cases were microdissected using a Pixcell II Laser Capture Microdissection 788 Laboratory System (Arcturus, Inc., Mountain View, CA) and stored at − 80 °C. 100 microdissected emboli /case from 10 cases were pooled together in individual aliquots. A total of 20 aliquots were prepared and stored for future studies.

Preparation of protein lysates and western blot analysis

The xenografts and laser-captured microdissected and pooled frozen IBC emboli were lysed using ice-cold RIPA lysis buffer (Pierce Biotechnology, Inc, Rockford, IL). For western blot analysis, boiled protein was loaded onto a 4–12% precast gradient gel (Invitrogen, Inc.), transferred to nitrocellulose membranes (Bio-rad, Hercules, CA) and incubated with either E-cadherin rabbit mAb (24E10) or rabbit anti-human ectodomain E-cadherin (H108) (Santa Cruz Biotechnology, Santa Cruz, CA) followed by anti-rabbit IgG, HRP-linked antibody (Cell Signaling Technology Inc., Danvers, MA). Rabbit mAb (13E5) was used to a housekeeping protein, ACTB. Bound antibodies were detected by a chemiluminescent detection system (West Femto) (Pierce Biotechnology, Inc.) according to the manufacturer's instructions. The relative intensity of the bands was determined by densitometric analysis.

Synthesis, labelling and use of SC and SB decapeptides

The calpain cleavage and binding sites on full length E-cadherin were used as a template to design mimicking decapeptides to block calpain cleavage and binding. Two decapeptides: the site of cleavage (SC) peptide: DARPEVTRND and the site of binding (SB) peptide: GGEEDQDFDL were purchased and used in concentration ranges of 50 nM to 50 µM (ThermoFisher Scientific, Inc., Waltham, MA). The same decapeptides were also obtained labelled, respectively, with Alexa Fluor 488 and Alexa Fluor 594 at their NH2-terminal ends and used in the same concentration ranges. Both the labelled and unlabelled decapeptides were studied in vitro for their abilities to penetrate into the spheroidal aggregates of both Mary-X and Karen-X, transgress the cell membrane, block E-cad/NTF1 generation and affect spheroidgenesis. Specifically dose response effects of the decapeptides on E-cad/NTF1 generation and effects on both spheroid disadherence and budding were observed and recorded. Decapeptides with random amino acid sequences in the same concentration ranges of 50 nM–50 µM served as negative controls in both the studies of E-cad/NTF1 generation as well as spheroidgenesis.

Statistical analysis

For Ki-67 index, mitotic count and nuclear grade (size) and measurements of embolic density, shape (elongation ratio) and size, both modes and means ± SD values were determined. All experiments were performed in quintuplicate and representative results depicted showing %cases, modes or means ± standard deviations. All stated or calculated differences imply differences of statistical significance, assessed by the two tailed students t test as well as analysis of variance (ANOVA).

Results

Observational studies

In the human TMA studies, compared to other tumor features like nuclear grade (size), mitosis and Ki-67 immunoreactivity which showed that IBC is only modestly more proliferative with larger nuclei than non-IBC, what really distinguished IBC was the markedly greater number of tumor emboli and distinctly smaller emboli (Fig. 1). IBC cases compared to non-IBC cases showed a modestly greater Ki-67 mean index (p = 0.05) (Fig. 1A), a modesty higher mean mitotic count (p = 0.05) (Fig. 1B) and a modestly increased mean nuclear grade (p = 0.05) (Fig. 1C) but the distribution of these tumor features showed considerable overlap. In contrast, mean embolic density in IBC dwarfed non-IBC by a factor of 100-fold (p = 0.0001) (Fig. 1D). Furthermore although the overall mode and mean shape of the emboli in IBC v non-IBC did not differ (p = 0.5) (Fig. 1E), the modal and mean size of the IBC emboli were significantly smaller (p = 0.01) (Fig. 1F).

Fig. 1
figure 1figure 1

Image and algorithmic analyses of multiple histological and immunocytochemical parameters in IBC v non-IBC human cases. A Ki-67 immunoreactivity (left), specific algorithmic recognition (middle) and Ki-67 index quantitation (right) in IBC v non-IBC are illustrated. B Mitosis histology (left), specific algorithmic recognition (middle) and mitotic count determination (right) in IBC v non-IBC are illustrated. C Nuclear histology (left), specific algorithmic quantitation (middle) and nuclear grade (size) distribution (right) in IBC v non-IBC are illustrated: yellow, < 20 µm; green, 20–30 µm; red, 31–50 µm. D Tumor embolic histology (left) and density determinations (right) in IBC v non-IBC are illustrated. E Tumor embolic histology (left) and algorithmic imaging shape determinations (right) in IBC v non-IBC are illustrated. F Tumor embolic imaging (left) and algorithmic perimeter determinations (right) in IBC v non-IBC are illustrated. Scale bars are provided. For each of these parameters, the graph depicts %cases, modal values (plotted) and calculated means ± SD. Differences of significance are depicted. All experiments were performed in quintuplicate

Full size image

Experimental studies

Initial xenograft studies

Mary-X grew equally well in both athymic (nude) mice as well as Scids as a transplantable xenograft (Fig. 2A). Mary-X exhibited a stable phenotype for over 25 years. Mary-X grew in a nodular fashion with islands of tumor cells embedded within a murine fibrovascular matrix (Fig. 2B). The nodules of Mary-X expressed strong membrane E-cadherin immunoreactivity (Fig. 2D). Mary-X gave rise to large numbers of CTCs (Fig. 2F: inset) which measured out at 10–15 cells / 100 µl blood as well as distal metastases including pulmonary metastases (Fig. 2F). These metastases also strongly expressed E-cadherin.

Fig. 2
figure 2

Xenograft Properties. A The growth of Mary-X and Karen-X are depicted. Each time point displays the mean ± standard deviation of 10 mice. Routine histology of Mary-X (B) and Karen-X (C) shows similar nodular islands of tumor cells. E-cadherin immunocytochemistry shows more intense and more circumferential membrane immunoreactivity in Mary-X (D) compared to Karen-X (E). Mary-X gives rise to E-cadherin positive pulmonary metastases (F) as well as CTCs (F: inset) whereas Karen-X lacks both (G, G: inset). Scale bars are provided.

Full size image

Karen-X, in contrast, grew only in Scids as a transplantable xenograft (Fig. 2). Karen-X also exhibited a stable phenotype for over 25 years. Karen-X exhibited a significantly longer latency than Mary-X but once a tumor emerged exhibited a similar rate of growth (Fig. 2A). Karen-X was similar to Mary-X histologically (Fig. 2C). Karen-X also expressed E-cadherin membrane immunoreactivity which overall was less intense and less circumferential (Fig. 2E) than that expressed in Mary-X. In contrast to Mary-X, Karen-X did not generate CTCs (Fig. 2G: inset) or metastases (Fig. 2G).

In vitro studies

Mary-X and Karen-X exhibited markedly different in vitro properties when grown in suspension culture (Fig. 3). When their respective xenografts were excised and minced in culture, both initially gave rise to loose tumor aggregates. After 24–48 h, the Mary-X aggregates tightened to form high density spheroids (Fig. 3A) whereas the Karen-X aggregates remained loosely associated and did not tighten (Fig. 3G). After an ensuing 24 h–96 h period, the Mary-X spheroids exhibited both partial budding as well as complete budding. We counted over 1000 spheroids over a 96 h period and observed partial budding in 15 ± 5% and complete budding in 5 ± 3%. Partial budding was defined as budding that had not yet completely separated and complete budding was defined as full separation with the appearance of daughter spheroids. When observed with time-lapsed photography, the budding of Mary-X spheroids was rapid and quite dynamic, occurring over minutes (Supplement 1). In this time-lapsed photography, one parent spheroid divided into two and the two divided into four, illustrating a geometric progression. Still images from the time-lapsed video demonstrate this unique budding of Mary-X spheroids (Fig. 3B-3F). Mary-X budding gave rise to daughter spheroids which were smaller in size (Fig. 4A) which nevertheless expressed strong E-cadherin immunoreactivity (Fig. 4B). In contrast, Karen-X spheroids did not bud over time in suspension culture (Supplement 2). Still images of the Karen-X spheroids confirmed the lack of budding and their persistence as loose aggregates (Fig. 3G). Neither Mary-X nor Karen-X ever attached or grew as monolayers even when grown on fibronectin or vitronectin-coated dishes (data not shown). Furthermore, both Mary-X and Karen-X spheroids grew only minimally in suspension culture and then underwent a growth arrest. Initially, Ki-67 immunoreactivity in both Mary-X and Karen-X spheroids was high but then decreased over several weeks in culture. In contrast both the Mary-X and the Karen-X xenografts grew fairly rapidly with a relatively high Ki-67 index (> 50%). Yet despite the growth latency in vitro, both Mary-X and Karen-X spheroids were fully tumorigenic when reinjected into mice, even after 6 months in culture.

Fig. 3
figure 3

In Vitro Clump Properties. Time-lapsed phase contrast microscopy depicts Mary-X spheroids growing in suspension culture as tight aggregates which completely bud into daughter spheroids (Supplement 1). Still images of this dynamic budding are depicted (A–F). In contrast, Karen-X spheroids remain as loose aggregates which do not bud (Supplement 2). Still images confirm this absence of budding (G, H)

Full size image
Fig. 4
figure 4

Complete Budding both In Vitro and In Vivo. Complete budding of Mary-X with phase contrast (A) and single label E-cadherin immunofluorescence (B). E-cadherin immunoreactivity remains intense even at the points of the budding (B). DAPI was used as a nuclear counterstain. White arrows indicate budding points. Time-lapse photography confirms the dynamics of the budding process (Supplement 1). Double label immunofluorescent studies demonstrate tumor emboli exhibiting green immunofluorescence within podoplanin-positive lymphovascular channels exhibiting red immunofluorescence with DAPI used as a nuclear counterstain. Dramatic complete embolic budding is also observed in vivo (C). E-cadherin immunoreactivity remains intense even at the points of the budding. White arrows indicate budding points. As a result of this geometric budding, large numbers of tumor emboli are propagated (D). Scale bars are provided

Full size image

Additional xenograft studies

Double label immunofluorescence studies of the excised Mary-X xenograft showed florid tumor embolic lymphovascular budding (Fig. 4C). In contrast the Karen-X xenograft showed no lymphovascular invasion and no budding (data not shown). The parental tumor embolic budding of Mary-X gave rise to a dramatic geometric increase in the number of smaller tumor emboli within lymphovascular channels (Fig. 4C, D). Both parental and daughter tumor emboli retained strong membrane E-cadherin immunofluorescence (Fig. 4C, D). Non-embolic tumoral areas of Mary-X showed no evidence of budding.

Studies of Mary-X and Karen-X E-cadherin proteolysis

Our previous Mary-X studies had demonstrated the transcriptome equivalence of xenograft-generated spheroids with the lymphovascular emboli in mice with both structures also demonstrating E-cadherin overexpression and specific proteolytic processing with calpain and other proteases, producing a number of specific E-cadherin fragments including Ecad/NTF1-4 and Ecad/CTF1-4 [21, 30]. Western blots with either E-cadherin rabbit mAb (24E10) or rabbit anti-human ectodomain E-cadherin (H108) in the present study confirmed the high levels of both E-cad/FL as well as E-cad/NTF1 in the Mary-X spheroids but absent E-cad/NTF1 in the Karen-X aggregates (Fig. 5A). Laser capture microdissected pooled lymphovascular tumor emboli from patients (100 emboli / case from 10 cases with IBC) (Fig. 5B) showed similarly high levels of both E-cad/FL as well as E-cad/NTF1 (Fig. 5C).

Fig. 5
figure 5

Expression of E-cad/FL and E-cad/NTF1. Western blot using H108 reveals both E-cad/FL and E-cad/NTF1 in Mary-X but essential absence of E-cad/NTF1 in Karen-X (A). Laser capture microdissection of tumor emboli (B) from cases of IBC also confirm the presence of E-cad/NTF1 by Western blot (C)

Full size image

Because calpains, as intracellular proteases, have many different actions on many different targets, blocking calpains with calpastatin would not be expected to solely block only the cleavage of E-cadherin and the generation of its E-cad/NTF1 fragment. So in the present study, we used specific oligopeptides (decapeptides) that mimicked, respectively, either calpain’s site of cleavage (SC) or site of binding (SB) on the full length E-cadherin molecule. Both decapeptides by Western blot significantly inhibited the generation of E-cad/NTF1 in a dose-dependent manner confirmed by densitometric analysis (Fig. 6A). No effects on the generation of E-cad/NTF1 were observed in the Karen-X spheroids likely because Karen-X spheroids did not generate E-cad/NTF1 de novo (Fig. 6B).

Fig. 6
figure 6figure 6

Effects of SC and SB decapeptides. Dose response of SC and SB decapeptides on E-cad/NTF1 generation in Mary-X spheroids (A). Each decapeptide singly and in combination was effective at blocking calpain generation of E-cad/NTF1 illustrated by Western blots. In Karen-X spheroids, since there was no or negligible E-cad/NTF1, these decapeptides exerted no effects (B). In vitro delivery and action of fluorescently labelled decapeptides on tumor clumps are depicted (C). SC peptide labelled with Alexa Fluor 488 (green fluorescence) (top row, left panel) and the SB peptide labelled with Alexa Fluor 594 (red fluorescence) (top row, center panel) both singly and in combination (yellow fluorescence) (top row, right panel), penetrated the cell membranes (bottom row, left and right panels) and caused increased disadherence and decreased budding (D). Effects of both decapeptides were synergistic. Scale bars are provided

Full size image

Using the same SC and SB decapeptides as before but now labelled, respectively, with Alexa Fluor 488 and Alexa Fluor 594, the decapeptides were administered to the Mary-X spheroids. Each peptide singly and in combination penetrated the spheroids (Fig. 6C). The effects on the Mary-X spheroids were significant in terms of both increased disadherence (p < 0.05), decreased partial and complete budding (p < 0.05) (Fig. 6D). The effects of the SB peptide were greater than the inhibitory effects of the SC peptide on E-cad/NTF1 generation (p < 0.05) but both peptides promoted disadherence, and decreased budding equally. The effects of the two peptides in combination were synergistic on inhibition of E-cad/NTF1 generation (p < 0.05), increased disadherence and decreased budding (p < 0.05). There were no measurable effects of these decapeptides on the loose aggregates of Karen-X (data not shown). Decapeptides with random amino acid sequences which served as negative controls in both the E-cad/NFT1 inhibition experiments as well as in the studies of spheroidgenesis and budding exerted no effects.

Discussion

IBC is the deadliest type of human breast cancer that presents with florid tumor emboli especially involving overlying dermal lymphatics [1,2,3,4,5]. These lymphovascular tumor emboli also reoccur locally, making skin-sparing and breast-conserving surgery not a usual option. [31,32,33,34,35]

In our human TMA studies, our detailed image analysis with artificial intelligence algorithms demonstrated what really set IBC apart was the markedly greater number of tumor emboli and distinctly smaller emboli whose numbers exhibited dramatic geometric or exponential differences. That novel observation together with our in vitro / murine in vivo observations supported our hypothesis that IBC emboli bud into daughter and smaller emboli in patients. These emboli likely were the source of numerous CTCs and metastases [36, 37] although more recent studies have questioned whether the density of lymphovascular emboli necessarily correlate with numbers of CTCs [38]. The budding of Mary-X in vitro was dynamic, dramatic and fairly rapid (occurring in minutes).

We used only two xenografts, Mary-X and Karen-X, xenografts we originated and established in immunodeficient mice because although there have a number of other aggressive triple negative breast cancer (TNBC) lines and xenografts reported in the literature and studied, we are not aware of any published studies that show that they spontaneously form spheroids in suspension culture or bud or form tumors as xenografts that also bud into daughter emboli. In our study we are attempting to take advantage of the fact that one of our PDX models, Mary-X still metastasizes, forms lymphovascular emboli through budding whereas our other model, Karen-X, neither metastasizes, forms lymphovascular emboli nor buds, arguing that these three properties are causally related. We do plan in future studies to study both Mary-X and Karen-X in mice whose immune system has been restored and humanized. [39, 40]

In a previous study [21] we reported the calpain-mediated cleavage in Mary-X and the generation of E-cad/NTF1. In that study and in our present work, we note that calpain cleaves E-cadherin at amino acid site: 782 and the size fragments expected from that cleavage are 100 kDa for the NH2-terminal end and 10 kDa for the COOH terminal end. We used two different E-cadherin antibodies, H108 and 24E10 that recognized amino acid sites: 695–701 and 778–781, respectively, to distinguish the different cleavage products produced by the different proteases including calpain (Fig. 7A).

Fig. 7
figure 7

Generation of E-cadherin and its fragments by different proteases. A Schematic (adapted by permission from Macmillan Publishers Ltd,; The EMBO Journal 2000; 21: 1948–1956, ref 71) depicts cleavage sites of calpain, caspase-3, MMP and γ-secretase including its obligate PS1 binding domain (highlighted) which is also thought to be the site of calpain binding as well and 8 of the E-cadherin fragments (E-cad/NTF1-4; E-cad/CTF1-4) generated by these specific cellular proteases. EC1-5 denote the extracellular E-cadherin repeats. TM indicates transmembrane domain (also highlighted). The sites of binding of two anti-E-cadherin antibodies, H108 and 24E10 are also depicted. Schematic adapted from Fig. 3, ref 71. B Schematic of E-cad/NTF1 membrane mobility depicts the structure of E-cad/NTF1 bound to p120ctn but not to β- or α-catenin nor the actin cytoskeleton. Therefore while E-cad/FL is tethered to the membrane in traditional CAJs, E-cad/NTF1 becomes more mobile, redistributing itself along adjacent areas of the membrane, contributing to increased budding

Full size image

Our current work significantly advances this observation in several ways: it shows that the lymphovascular emboli of Mary-X exhibit complete budding both in vitro by time-lapsed photography and in murine studies with the generation of completely intact and smaller daughter emboli and that this budding can be inhibited in vitro by blocking calpain cleavage of E-cadherin and the generation of E-cad/NTF1 with decapeptides that mimic either the cleavage site or the binding site on E-cadherin in a dose-dependent manner.

Metastases is a complex process regulated by multiple determinants so just because calpain-mediated cleavage of E-cadherin into E-cad/NTF1 that contributes to a budding phenomenon is observed in IBC, it does not follow necessarily that reduced calpastatin which would contribute to increased calpain or increased calpain alone would be responsible for increased metastases and poor survival across the board if the rate-limiting steps of metastasis of other cancers including non-IBC were governed by other pathways such as adhesion or proliferation. Furthermore while certainly our studies suggest that E-cadherin/NTF1 could be a marker of poor prognosis and this should be studied, other markers characteristic of other pathways could well be confounding in other tumor types. For starters, we do plan to measure E-cad/NTF1 levels in lymphovascular emboli retrieved by laser capture microdissection in a cohort of just IBC cases and see whether the levels predict disease-free survival and overall survival.

In our study we are distinguishing geometric budding of emboli from the phenomenon that triggers initial lymphovascular invasion in the first place, which, in itself remains unexplained. The recognition of the lymphovascular embolus as an entity unto itself has catalyzed larger observations that many cancers invade and metastasize as clumps or clusters [41,42,43,44,45,46,47,48]. Up until relatively recently, most scientists considered cancers to metastasize as single cells and their increased locomotion and degradation of the extracellular matrix with subsequent lymphovascular invasion as a manifestation of so called epithelial-mesenchymal transition (EMT) [13,14,15]. Models epitomizing this phenomenon in the breast included infiltrating lobular cancers where cells lose E-cadherin expression and infiltrate as single cells. [49, 50]

The study of tumor cell clumps or clusters became important not only in vivo but also in vitro because tumor clumps in suspension culture provided 3D models which were uniformly superior to 2D monolayer cultures of cancer cell lines. In vitro 3D models were thought to better recapitulate the 3D situation of in vivo cancer and to more faithfully recapitulate in vivo gene expression, provide better evaluation of physiological factors such as hypoxia and improved testing of anticancer pharmacological strategies. [51,52,53]

Most 3D models take the form of spheroids in suspension culture. The majority of these spheroids or multicellular aggregates (MTCs), however, need to be induced to grow as spheroids by growing them on agar or other repellent surface that interferes with the cancer cells natural proclivity to grow as monolayers. [54,55,56,57,58,59,60,61,62,63]

Some of these spheroids when injected into immunocompromised mice will develop into tumors but these tumors do not exhibit obvious lymphovascular invasion or tumor emboli formation. Furthermore no previous models have demonstrated the equivalence of in vitro spheroids with in vivo tumor emboli. In contrast, our in vitro spheroids of Mary-X spontaneously form and when re-injected into mice develop into tumors showing florid tumor emboli, which by Principal Component Analysis (PCA) not only express the transcriptome equivalence of Mary-X spheroids but also the equivalence of microdissected tumor emboli from cases of IBC [21, 30]. And the emboli of Mary-X and their in vitro equivalent, i.e., the spheroids are what undergo geometric budding.

Although budding per se of tumor cells has also been observed and studied in vivo in other studies, that budding applies to invasion of the primary cancer and not specific budding of the lymphovascular tumor embolus [64,65,66]. Budding per se from the lymphovascular tumor embolus has not previously been observed or studied.

It is interesting that the hypothesis of EMT still dominates the thinking of the metastatic process even when applied to metastasizing clumps and budding where E-cadherin overexpression is also observed [46,47,48]; [67,68,69]. Despite this observation, the evidence suggests that the mechanism of budding responsible for geometric embolic propagation in our model does not involve either complete or partial EMT. Our time-lapsed photography indicates that complete in vitro budding occurs in minutes and that both the bud as well as the parent spheroid or embolus overexpress E-cadherin even at the incident point of budding. It makes more sense that budding results not from EMT followed by MET but rather from pre-existing molecules like E-cad/NTF1. Certainly budding must involve additional steps some of which could be energy (ATP) dependent, hypoxia triggered, nutrient-related or density regulated. We plan to use our budding model to delineate these steps in future studies.

Conclusions

Our study therefore is the first study that observes and quantitates actual budding of tumor emboli in an IBC model both in vivo and in vitro which account for the markedly greater number of tumor emboli and distinctly smaller emboli whose numbers reflect a geometric or exponential process in IBC patients.

Blocking the calpain-mediated proteolysis of membrane E-cadherin by SC and SB mimicking decapeptides suggests that E-cad/NTF1 participates in both tight spheroidgenesis as well as budding. E-cad/NFT1 retains its p120ctn binding site but loses both the β-catenin and α-binding sites, facilitating its disassembly from traditional CAJs. E-cad/NFT1 can exhibit 360° mobility around the membrane and its vacillating levels would both initiate and subsequently consolidate the budding process (Fig. 7B). This insight could lead to therapies at inhibiting budding and embolic propagation, thereby decreasing skin recurrences in IBC [70].