Il-6 signaling between ductal carcinoma in situ cells and carcinoma-associated fibroblasts mediates tumor cell growth and migration

Ductal carcinoma in situ (DCIS) is a non-obligate precursor lesion of invasive breast cancer in which approximately half the patients will progress to invasive cancer. Gaining a better understanding of DCIS progression may reduce overtreatment of patients. Expression of the pro-inflammatory cytokine interleukin-6 increases with pathological stage and grade, and is associated with poorer prognosis in breast cancer patients. Carcinoma associated fibroblasts (CAFs), which are present in the stroma of DCIS patients are known to secrete pro-inflammatory cytokines and promote tumor progression. We hypothesized that IL-6 paracrine signaling between DCIS cells and CAFs mediates DCIS proliferation and migration. To test this hypothesis, we utilized the mammary architecture and microenvironment engineering or MAME model to study the interactions between human breast CAFs and human DCIS cells in 3D over time. We specifically inhibited autocrine and paracrine IL-6 signaling to determine its contribution to early stage tumor progression. Here, DCIS cells formed multicellular structures that exhibited increased proliferation and migration when cultured with CAFs. Treatment with an IL-6 neutralizing antibody inhibited growth and migration of the multicellular structures. Moreover, selective knockdown of IL-6 in CAFs, but not in DCIS cells, abrogated the migratory phenotype. Our results suggest that paracrine IL-6 signaling between preinvasive DCIS cells and stromal CAFs represent an important factor in the initiation of DCIS progression to invasive breast carcinoma.


Background
Ductal carcinoma in situ (DCIS) of the breast is a preinvasive lesion and a risk factor for subsequent invasive ductal carcinoma (IDC) [1]. DCIS represents about 20 % of newly diagnosed breast cancers in the United States [2]. If left untreated, approximately half of DCIS tumors will progress to IDC while half will remain indolent [3,4].
Although there are many subtypes of DCIS, it is not currently possible to identify which will progress. This has led to aggressive treatments, specifically radiation with either lumpectomy or mastectomy [5].
Components of the tumor microenvironment are increasingly implicated in the progression of many cancers. Early morphological and physiological changes in breast epithelium are minimal, and compounding factors such as tumor-suppressive paracrine signaling from myoepithelial cells [6] or the extracellular matrix [7] may hide early indicators of ductal cell aberration. Such changes may include, but are not limited to: gene expression modulation, epigenetic alterations, and loss of genomic stability in both the epithelial and stromal compartments. In the tumor microenvironment, carcinoma-associated fibroblasts (CAFs) represent a fibroblast population or mixture of sub-populations that can promote tumor progression [8][9][10][11][12][13]. Although this mechanism is not fully understood, it is known that CAFs secrete numerous cytokines and growth factors [14].
Interleukin 6 (IL-6) is a pro-inflammatory cytokine shown to alter cell morphology, modulate cell migration and the epithelial to mesenchymal transition [15][16][17]. Many of these processes occur upon IL-6 activation of the transmembrane IL-6 receptor (IL-6R), which heterodimerizes with the ubiquitously expressed cell surface receptor glycoprotein 130 (gp130). Downstream activation of the Janus kinase/signal transducers and activators of transcription (JAK/STAT) pathway initiates IL-6 target gene transcription [18,19]. Alternatively, secreted IL-6 can bind to the soluble IL-6 receptor (sIL-6R), which then binds gp130 at the cell surface and initiates intracellular signaling. This form of IL-6 signaling has been coined "IL-6 trans-signaling" (IL-6TS) [20]. IL-6 has been linked to the upregulation of proteases such as cysteine cathepsins and matrix metalloproteinases that are known to play a role in cancer progression [21,22]. The IL-6 signaling axis is commonly upregulated in invasive cancers, suggesting that IL-6 may be an important mediator of events involved in tumor cell invasion [23][24][25].
The effects of IL-6 signaling inhibitors on breast cancer cell morphology and proliferation have been evaluated in monotypic cultures. Two studies have shown that inhibiting autocrine IL-6 signaling in either triple negative breast cancer cell lines [26], or the ER-positive MCF7 cell line [27] significantly inhibited cell growth. Additionally, Leslie et al. show that knockdown of IL-6 in an invasive variant of the MCF10A cells, MCF10A-H-RasV12, inhibited cell migration in a transwell assay, and inhibited growth in a xenograft mouse model [28]. Although these studies evaluated paracrine signaling, cells were treated with exogenous recombinant protein rather than co-culturing different cell types. Therefore the authors were unable to evaluate in 3D the dynamic cell:cell interactions between two separate human cell types or the cell:microenvironment interactions.
Our 3D mammary architecture and microenvironment engineering (MAME) culture model mimics in vivo architecture, providing a suitable setting to study cell:cell interactions and, notably, physiologically relevant cell:cell signaling over time. Additionally, 3D in vitro cell culture models better represent in vivo tumor drug response, which would facilitate efficacious therapy development at the preclinical stage [29]. Here we examine the role of IL-6 in progression of preinvasive breast DCIS to an invasive phenotype, and show how co-culture of DCIS cells with CAFs changes DCIS growth and invasive potential.

Cell lines
MCF10A human breast non-transformed epithelial cells, MCF10.DCIS and SUM102 human breast DCIS cells, and WS-12Ti human breast carcinoma-associated fibroblasts were provided by Dr. Bonnie F. Sloane. All primary fibroblasts were derived from human breast tissue. CAF40T were derived from biopsy tissue diagnosed as invasive carcinoma. NAF98 were derived from benign tissue. Both CAF40T and NAF98 fibroblast cell lines were provided by Dr. Simon W. Hayward. These fibroblasts were immortalized in Dr. Sloane's lab and are designated CAF40TKi and NAF98i, respectively. The FB-NF, NAF-FB, and FB-CAF primary fibroblasts were derived from patient biopsies diagnosed as: benign (FB-NF, NAF-FB), or invasive carcinoma with accompanying DCIS (FB-CAF) and provided by Dr. Fariba Behbod. The FB-NF fibroblasts were immortalized in Dr. Sloane's lab and designated FB-NF-i. The FB-CAF and NAF-FB fibroblasts were not immortalized. All patient derived cells were received de-identified and therefore are exempt from IRB oversight.

Cell culture
In this study we utilized non-tumor forming MCF10A human breast epithelial cells [30] and the human DCIS cell lines MCF10.DCIS and SUM102, which were maintained as previously described [31]. See supplemental methods for more detailed information (Additional file 1). All 3D MAME cultures were performed using Cultrex (3433-005-01, Trevigen, Gaithersburg, MD) similar to previously described [32]. Briefly, cell culture dishes were coated with 100 % Cultrex. Cells were added on top of solidified Cultrex and allowed to adhere for 30-45 min before being overlaid with 2 % Cultrex in phenol red-free DMEM F12 media containing 2.5 % fetal bovine serum (Additional file 2: Figure S1). In co-culture experiments, fibroblasts were added first and allowed to adhere before adding tumor cells. Once tumor cells had adhered to the matrix, an overlay of 2 % Cultrex was added. Media were changed every 4 days.

Measurement of multicellular structures
Differential interference contrast (DIC) images of three random fields at 20X magnification were used to measure multicellular structures. Three individuals of whom two were study-blinded measured the diameter and perimeter of structures and the number and length of interconnections between structures. All three data sets were used in the quantification. This analysis was performed using Zen imaging software (Zeiss, Thornwood, NY). Volume measurements were obtained using 3D fluorescent images and quantified using Volocity software (Perkin-Elmer, Waltham, Mass).

Gene expression
RNA was isolated from cells grown in either 2D monolayer or 3D MAME cultures. For 2D culture, cells were washed in phosphate buffered saline and harvested using 0.5 % Trypsin/EDTA (Life Technologies, Foster City, CA) and pelleted. The cell pellets were resuspended in TRIzol® Reagent (Life Technologies, Foster City, CA) for RNA extraction. All qRT-PCR reactions were performed using Taqman Assays (Life Technologies, Foster City, CA). See supplemental list of Taqman Assays (Additional file 3: Table S1).

ELISA
ELISA kits (human IL6R-ab46029, human IL6-ab46044, and human GAPDH-ab119627) were purchased from Abcam® (San Francisco, CA). Aliquots of lysates were collected for ELISA assays and measurement of total DNA.

Immunofluorescence
Nuclei were labeled with Hoechst (33342, Thermo Scientific) or EDU (Life Technologies, Foster City, CA). Polyclonal antibodies to human IL-6 (AF-206-NA, R&D Systems, Minneapolis, MN) were used at a concentration of 1 μg/ml. Mono-specific antibodies to human cathepsin B have been previously isolated and characterized [33]. Cathepsin B immunostaining was performed as previously described with the exception that 1 % Tween 20 replaced the 0.01 % saponin [34]. For some studies, CAF40TKi were pre-labeled, prior to seeding in MAME co-cultures, utilizing CellTrace CFSE (carboxyfluorescein diacetate succinimidyl ester; Life Technologies, Foster City, CA) according to the manufacturer's protocol.

Drug treatments
For treatment of MAME cultures with IL-6 nAb (R&D Systems, AF-206-NA), we added 1 μg/ml of IL-6 nAb in the 2 % Cultrex overlay to 3D cultures on the first day of culture and refreshed with IL-6 nAb and 2 % overlay every 4 days. The antibody concentration was selected based on preliminary studies in which we determined the lowest concentration needed to inhibit growth of tumor structures. Oxymatrine (Sigma, St. Louis, MO) at 1-mg/ml (3.7 mM) was added 24 hours after cell seeding and was replaced with fresh drug every 4 days. The oxymatrine concentration was determined empirically based on the concentration at which proliferation was inhibited to 50 % of control. The protease inhibitors CA074Me and E64d (Sigma, St. Louis, MO) were used at a concentration of 10 μM [35].

Live cell proteolysis assay
Dye-quenched collagen IV (DQ-collagen IV, Life Technologies, Foster City, CA) was used admixed in Cultrex as previously described [32]. MAME cultures in optical glass bottom cell culture dishes were imaged live for a period of 10 to 60 min under 5 % CO 2 at 37°C.

Confocal microscopy
Confocal microscopy was performed on either a Zeiss LSM 510 or LSM 780 upright confocal microscope (Zeiss, Thornwood, NY). All cell cultures used for imaging were seeded on 40 mm optical glass bottom culture dishes at a density of 45 cells/mm 2 (~5000 cells/dish).

Statistics
Data were statistically analyzed using Student's t-test on GraphPad Prism 6.0 (GraphPad Software Inc., La Jolla, CA).

Ethics statement
All human subject materials and experiments in this study have been reviewed by the Wayne State University Institutional Review Board and deemed exempt according to the definition codified in the common rule at 45 CFR 46.102(d)(f ).

Human breast DCIS cells express pro-inflammatory cytokines
Increased levels of pro-inflammatory cytokines, including IL-6, in tumors and serum of breast cancer patients correlate with poor prognosis [26,[36][37][38]. Immunohistochemistry on breast tissue from female patients, age 30 to 81 with an average age of 49.5, confirmed IL-6 protein expression in 65 % of patient samples diagnosed with DCIS ( Fig. 1a and b, cf. 1c and d).
We utilized a 3D MAME culture model (Additional file 2: Figure S1), to study the role of IL-6 signaling between human breast DCIS cells and human breast CAFs. The MAME model allows for co-culture of multiple cell types in the context of a three dimensional microenvironment. Our model is advantageous over commercially available 3D platforms as it can be utilized for live cell microscopy, and imaging live cell functional assays (such as proteolysis) [32]. MAME cultures are tractable and can be altered to grow multiple cell types at various ratios or at various relative positions within the matrix. Additionally, the reduced density overlay allows for realtime collection and analysis of cytokine secretion without disrupting the longitudinal growth of the culture.
Here we examined the expression of IL-6 and the associated pro-inflammatory genes interleukin 1β (IL1β) and nuclear factor kappa light chain enhancer of activated B cells 1 (NFκB1) in human breast MCF10.DCIS and SUM102 cell lines. Our data show higher IL-6 mRNA levels in both MCF10.DCIS and SUM102 cell lines, as compared to the MCF10A breast epithelial cell line (Fig. 1e). Levels of IL-6 protein in lysates of the three cell lines were below the level of detection. When we measured secreted IL-6 in conditioned media from MCF10.DCIS or SUM102 cultures, we found that both DCIS cell lines secreted 10 to 30-fold more IL-6 than the MCF10A cells (Fig. 1f ).

Blocking IL-6 autocrine signaling inhibits MCF10.DCIS growth
To test whether blocking IL-6 autocrine signaling affected DCIS cell growth, we treated MCF10.DCIS MAME cultures for 8 days with either an IL-6 neutralizing antibody (nAb) or an equivalent concentration of . f Secretion of IL-6 protein from DCIS cell lines and non-transformed MCF10A cells as determined by ELISA. *P < 0.05, Student's t-test; mean ± SD a species and isotype matched antibody (Fig. 2). IL-6 nAb treatment resulted in a reduction in diameter of the multicellular 3D structures (Fig. 2b, cf. 2a, quantified in 2d). This was reversible as replacement of media on day 8, with media lacking IL-6 nAb, resulted in a significant increase in diameter of the MCF10.DCIS structures after an additional 48 h (Fig. 2c, cf. 2b, quantified in 2d).
To determine the effects of IL-6 nAb on MCF10.DCIS structures, we examined the expression of a panel of candidate genes that have been associated with tumor growth and invasion [39][40][41][42][43]. The expression of IL-6 was upregulated 2-fold in the treated cultures. The expression of TWIST1, vimentin, and Fibroblast Growth Factor 2 was downregulated greater than 2-fold, while minimal changes were observed in the expression of E-cadherin, N-cadherin, and NFκB1 (Fig. 2e). To test whether pharmacological suppression of IL-6 could reproduce IL-6 nAb mediated growth inhibition, we treated cells with oxymatrine, a naturally occurring inhibitor of IL-6 gene expression. Oxymatrine has been shown to prevent nuclear translocation of NFκB-p65 thereby inhibiting transcriptional activation of its target genes, which include IL-6 [44]. Oxymatrine treatment was able to replicate the growth inhibitory effects observed with IL-6 nAb (Additional file 4: Figure S2B, cf. S2A, quantified in S2C). Neither oxymatrine nor IL-6 nAb treatment resulted in marked cell death as cytotoxicity assays showed no difference in cell viability after 48-hour drug treatment (Additional file 4: Figure S2D).
Carcinoma-associated fibroblasts express IL-6 and promote DCIS cell proliferation and motility CAFs represent a population or group of populations of stromal cells that can promote tumor cell growth [14,[45][46][47]. The mechanism of supported tumor growth is likely through stromal-epithelial paracrine signaling. Therefore, we next evaluated human breast CAFs to determine their contribution of IL-6 in the tumor microenvironment. Additionally, we examined the role that CAFs play in MCF10.DCIS cell proliferation and motility in the 3D MAME model.
We examined the expression of IL-6 mRNA in normal human fibroblasts and CAFs grown in 3D. Here we found that CAFs exhibited elevated expression of IL-6 mRNA compared to normal fibroblasts (Fig. 3a). Protein levels of IL-6 in FB-NF-i normal fibroblast lysates were near the lower limit of detection and undetectable in NAF-FB or NAF98i lysates. IL-6 levels in CAF40TKi lysates were significantly higher than in FB-NF-i lysates (Fig. 3b). Levels of IL-6 in CAF-conditioned media were higher than in normal fibroblast-conditioned media (Fig. 3c).
We next co-cultured MCF10.DCIS cells with CAFs in 3D, at a seeding ratio of five tumor cells to one CAF [32], to evaluate the effect of CAF-secreted cytokines on DCIS cell proliferation and related changes in morphology of multicellular structures. We found that in MCF10.DCIS:C AF40TKi co-cultures there was an increase in the average diameter and volume of the multicellular structures and a prominent formation of branch-like interconnections between the structures (Fig. 3e, cf. 3d, quantified in Additional file 5: Figure S3A). Fluorescent imaging of MCF10.DCIS:CAF40TKi 3D co-cultures revealed that the branch-like multicellular connections between structures were primarily composed of tumor cells, yet contained some CAFs (Additional file 5: Figure S3B and Additional file 6: Video S1).
We also observed that CAFs induced an increased rate of proliferation in MCF10.DCIS cells. Using a thymidine analog to evaluate the rate of DNA synthesis, we observed that co-cultures had a consistently higher rate of DNA synthesis than CAFs alone or DCIS cells alone (Additional file 7: Figure S4A-D). CAF and DCIS co-culture using a slower growing DCIS cell line, i.e., SUM102, also resulted in changes in multicellular structure formation. We observed the presence of multiple invasive processes in SUM102:CAF co-cultures that were completely absent in cultures of the SUM102 cells alone (Additional file 8: Figure  S5C and D, arrow, cf. S5A and B).

DCIS cells migrate with CAFs at the invasive front
CAFs express and secrete a number of proteases, which enhance their ability to migrate and remodel extracellular matrices [48][49][50][51]. In CAF:MCF10.DCIS co-cultures, we evaluated cell:cell interactions and motility. Here we found that MCF10.DCIS spheroids formed attachments to CAFs and remained associated with them throughout an 8-day culture. On day 1 (24 h after seeding) of a coculture we observed many single DCIS cells and a few CSFE labeled CAFs (Fig. 4a). After 3 days in culture DCIS cells had formed small spheroids that were in contact with one or more CAFs (Fig. 4b). On day 5 we observed CAFs at the invasive edges of DCIS tumor spheroids. A representative image shows a CAF (Fig. 4c, arrow) in contact with a DCIS spheroid. A time-lapse video shows the CAF leading the DCIS spheroid (Additional file 9: Video S2). High magnification imaging shows heterocellular contact between a tumor structure and a single CAF (Additional file 10: Figure S6). We did not observe these invasive characteristics in DCIS cell grown alone, as the tumor structures tended to roll in the extracellular matrix (Additional file 11: Video S3). By day 7 CAFs were seen at the invasive edge of most tumor structures (Fig. 4d).
Using confocal microscopy we examined the 3D spatial organization of the DCIS cells in relation to CAFs. A top view of a 3D reconstruction shows an MCF10.DCIS structure (only nuclei labeled) branching out toward CFSE labeled CAFs (pseudo-colored white) (Fig. 4e, Additional file 12: Video S4). The core of this structure (dashed circle) formed shortly after seeding and grew in size before the two branching outgrowths formed. These outgrowths are reminiscent of the "strand" multicellular migration pattern previously described [52]. A side view of the same structure shows that DCIS cells migrated downward from the core structure toward the CAFs (Fig. 4f). IL-6 protein expression in these co-cultures was detected using immunofluorescent staining (Additional file 13: Figure S7). A CFSE-labeled CAF showed high expression of IL-6, whereas the leading edge of the DCIS multicellular structure showed a gradient of IL-6 that was strongest near the attachment to the CAF (Additional file 13: Figure S7A, arrow). We also showed that MCF10.DCIS cells that do not migrate to serumfree media (Additional file 13: Figure S7D) migrated toward CAF-conditioned serum-free media (Additional file 13: Figure S7E, S7F).

CAFs and MCF10.DCIS cells utilize cathepsin B to degrade matrix
IL-6 has previously been shown to upregulate cathepsin B [21], a protease associated with breast cancer progression. Therefore, we examined the expression and localization of cathepsin B in relation to active proteolysis in our 3D MAME co-cultures. Our active proteolysis assays were performed during live cell imaging [33]. Fluorescent imaging of live cells in MAME co-culture revealed regions of proteolytic activity (green) around several multicellular structures (Fig. 5a). The cells were then fixed for cathepsin B immunofluorescent staining. Visualizing the same structures revealed cathepsin B expression at the interface between the multicellular structures and the matrix (Fig. 5b). Fluorescent overlay shows co-localization between DQ-collagen IV degradation and areas with high cathepsin B expression (Fig. 5c); however, matrix degradation also occurred in areas that did not stain for cathepsin B (arrow). Upon visualization using light microscopy, the cells in these areas had morphological features consistent with CAFs, which would be consistent with current knowledge of fibroblast migration and their secretion of many proteases in addition to cathepsin B [53,54].
We then used protease inhibitors to determine the contribution of all cysteine proteases and/or cathepsin B and L, to the overall proteolysis in the MAME culture. We treated cultures with E64d, a cell permeable pan-cysteine protease inhibitor, and found a 30 -40 % reduction of matrix degradation (Fig. 5d). We also treated cultures with the cell permeable cathepsin B/L inhibitor, CA074Me, and found a 60 -70 % reduction in matrix degradation (Fig. 5e). CA074Me has previously been shown to be a more efficacious inhibitor against cathepsins B and L than is E64d [35,55]. We did not observe complete inhibition of proteolysis with the cysteine protease inhibitors, a finding in accord with previous findings that DCIS cells and CAFs secrete many non-cysteine proteases [31,56,57].

Blocking IL-6 inhibits CAF-stimulated effects on human breast DCIS cells
We have shown that CAFs promote the proliferation and migration of human breast DCIS cells. To determine c Fluorescent overlay of dDQ-Col.IV (green) and cathepsin B immunostaining (red). Scale bar, 200 μm. Quantification of dDQ-Col.IV fluorescent intensity in MAME co-cultures treated with cell permeable inhibitors E64d (pan-cysteine protease inhibitor) (d) or the cathepsin B selective inhibitor CA074-Me (e). *P < 0.05, Student's t-test; mean ± SD whether CAF-stimulated migration and interaction with MCF10.DCIS cells could be inhibited by blocking IL-6, we co-cultured MCF10.DCIS cells with two human breast carcinoma-associated fibroblast lines WS-12Ti or CAF40 TKi, in the presence of IL-6 nAb or an isotype control antibody. IL-6 nAb reduced the size of MCF10.DCIS:CAF multicellular structures when grown with either CAF cell line (Fig. 6a-e). Multicellular structure volume measurements confirmed a significant reduction when treated with IL-6 nAb (Fig. 6f-h).
To confirm the role of IL-6 as a key player in the formation of the multicellular structure interconnections, we utilized shRNA targeting IL-6 expression. In CAF40T Ki and MCF10.DCIS cells, we achieved greater than 50 % reduction in secreted IL-6 (Additional file 14: Figure S8A). When we co-cultured CAF40TKi-shRNA control fibroblasts with MCF10.DCIS cells, we saw a phenotype similar to non-shRNA transduced cultures (Additional file 14: Figure S8B, cf. 3E). Knocking down CAF40TKi IL-6 expression in co-culture resulted in the formation of multicellular structures with uniform borders and few invasive processes (Additional file 14: Figure S8C). Co-culture with non-shRNA transduced CAF40TKi fibroblasts and shRNA-IL-6 transduced MCF10.DCIS cells showed greater MCF10.DCIS:CAF 40TKi interaction and multicellular structure branching (Additional file 14: Figure S8D). IL-6 signaling is propagated through either direct cell membrane receptor signaling or soluble receptor transsignaling (TS). In the DCIS cell lines, we found that IL-6R expression was low and sometimes undetectable by qRT-PCR. Additionally, we detected very low levels of sIL-6R secreted from MCF10.DCIS cells and none from SUM102 cells. In contrast, CAFs had a higher level of IL-6R gene expression and high secretion of sIL-6R as verified by ELISA. These findings suggest that IL-6TS is a likely mechanism of IL-6 action in these DCIS cells [58] (Additional file 15: Figure S9).

Discussion
A number of factors produced by CAFs have been shown to be involved in promoting malignant transformation in epithelial cells, these include TGFß [59][60][61] and CXCL12 (SDF-1) [59,61,62]. These factors are involved in eliciting a range of responses that are context-dependent and that benefit the tumor in various ways. The present study describes a novel addition to these known interactions and a new mechanism by which CAF can influence tumor progression. There is a direct correlation between serum IL-6 levels and poor prognosis in breast cancer patients [36,[63][64][65]. Studies have shown that CAFs [66], and various immune/inflammatory cells secrete pro-inflammatory cytokines including IL-6 and contribute to tumorigenicity [67][68][69][70][71][72][73].
In the current study, we show that IL-6 expression can be found in both the tumor and stromal compartments. In our IHC data we found approximately 65 % of patient samples had positive IL-6 staining; however, when we examined IL-6 expression in our DCIS tumor cell lines, we found the expression to be near the lower threshold of our assay. This discrepancy may be due to differences in gene expression between tissue and cells [74], differences in IL-6 expression with tumor grade/invasiveness [65], degree of "stemness" in cell lines vs. tissue [75], and differences between the assays and/or sample collection.
We next examined the interaction between human breast DCIS epithelial cells and human breast CAFs in the context of an in vitro 3D microenvironment. We hypothesized that CAFs promoted the migration of breast DCIS cells via paracrine signaling within the tumor microenvironment. Here we observed that DCIS cells migrated towards CAFs and upon attachment to CAFs, DCIS cells remained attached and migrated through the matrix following the lead of the CAFs. Similar findings of fibroblast-led migration have been reported for cocultures of invasive squamous carcinoma cells and CAFs [76]. It is also likely that CAFs migrated to DCIS cells, as some fibroblasts moved more easily and quickly through the matrix as seen in time-lapse videos; therefore it is possible that some fibroblasts migrated to and attached to DCIS cells thereupon leading migration of the DCIS cells.
An increase in DCIS cell proliferation and a change in multicellular structure morphology was observed in all of our co-culture experiments. DCIS multicellular structures showed invasive characteristics, having lost their uniform circular structure and had developed single or multiple protruding edges. An underlying mechanism for DCIS:CA F interaction, enhanced tumor cell proliferation, and migration was IL-6 signaling in the tumor microenvironment. We [31] as well as Sung et al. [77] have shown in 3D co-culture systems that paracrine HGF and its receptor cMet can drive the invasive potential of the MCF10.DCIS cells. We confirmed this in xenografts formed by orthotopic implantation of HGF-secreting fibroblasts and MCF10 .DCIS cells in SCID mice [31]. HGF and IL-6 have been shown to cooperatively enhance lung cancer cell invasion by upregulation of their corresponding cell surface receptors [78]. Also of interest is the stem cell-like properties of the proliferating DCIS cell population as Krishnamurthy et al. have shown that endothelial IL-6 enhances selfrenewal of cancer stem-like cells. Whether or not this is the case for CAF IL-6 in regard to proliferation of DCIS cells has yet to be determined [79].
Treatment of DCIS:CAF co-cultures with an IL-6 nAb abrogated the proliferation and migratory phenotype acquired by DCIS cells. This phenotype was primarily produced by the inhibition of CAF secreted IL-6 as shRNA knockdown of IL-6 in CAFs, but not in DCIS cells was able to replicate the phenotype. We also showed that MCF10.DCIS cells treated with IL-6 nAb had a down regulation of genes associated with EMT. A recent study suggests that tumor cell EMT is mediated through factors secreted from CAFs and that selective inhibition of TGFβ1 is sufficient to reverse EMT associated gene expression [80]. Other studies show cross talk between IL-6 and TGFβ signaling consistent with IL-6 and TGFβ both acting as drivers of EMT [81].
A Federal Drug Administration (FDA) approved humanized anti-IL-6R antibody, tocilizumab (Actemra), has shown promise in the treatment of inflammatory diseases particularly rheumatoid arthritis, Crohn's disease, and Castleman's disease. A major caveat of Actemra is that serum IL-6 levels are increased in patients after drug administration [82]. Since breast cancer patients with elevated serum IL-6 have poorer prognosis, Actemra may not be a practical therapy in these patients, however alternative approaches to reduce IL-6 signaling may prove efficacious.
Siltuximab, a monoclonal antibody against IL-6, is in clinical trials for therapies including; combinatorial treatment of metastatic renal cell carcinoma, multiple myeloma, and prostate cancer [83][84][85][86]. The FDA recently approved siltuximab for the treatment of multicentric Castleman's disease [87]. Studies have shown that siltuximab significantly inhibits the growth of nonsmall-cell lung cancer in primary xenografts [88], and ovarian cancer cell xenografts [89]. Therapeutic use of siltuximab for the treatment of breast cancer has not been fully evaluated, although preliminary studies with the estrogen receptor alpha positive MCF-7 breast cancer cell line suggests potential efficacy [90].
A limitation of the current study is that we did not have complete histories or demographic data to perform correlation studies to determine how IL-6 expression in DCIS relates to tumor prognosis. Another limitation is there is a limited number of commercially available DCIS cell lines. Here we used two cell lines that are commercially available. A third DCIS cell line that is commercially available is the SUM-225 DCIS cell line; however, this line comes from a metastatic recurrence so we chose not to use it. Another limitation is that our study only followed the progression and interaction of DCIS cells with CAFs, but not inflammatory cells associated with breast tumors. Additionally, we are investigating the probability of increasing the number of cell types grown together in our 3D MAME culture, for example: macrophages, adipocytes and lymphocytes.

Conclusion
In conclusion, we have shown that IL-6 paracrine signaling between DCIS cells and CAFs is a key mediator of early stage breast cancer cell proliferation and migration. Furthermore, IL-6 from CAFs facilitated the transition of DCIS cells from preinvasive to an invasive phenotype. This study highlights the necessity to explore paracrine signaling within the context of the tumor microenvironment and components therein.  Table S1. Taqman Gene Expression Assays. Taqman Assays (primers) were selected for the human gene targets. Each Taqman Assay was selected based on suppliers "best coverage" criteria. (C) Incorporation of EDU in a representative structure from a MCF10.DCIS: CAF40TKi co-culture. Note that similar sized structures were chosen to compare levels of EDU incorporation in a 24-h period. Scale bars, 100 μm. (D) DNA was extracted from MAME cultures to quantify EDU concentration. Data shows a significantly higher concentration of EDU incorporation in DNA of co-culture as compared to monotypic cultures. ***P < 0.001, ****P < 0.0001, Student's t-test; mean ± SD. (TIFF 344 kb) Additional file 8: Figure S5. Carcinoma-associated fibroblasts enhance invasive growth of SUM102 cells in MAME co-culture.