Carcinoma associated fibroblasts (CAFs) promote breast cancer motility by suppressing mammalian Diaphanous-related formin-2 (mDia2)

The tumor microenvironment (TME) promotes tumor cell invasion and metastasis. An important step in the shift to a pro-cancerous microenvironment is the transformation of normal stromal fibroblasts to carcinoma-associated fibroblasts (CAFs). CAFs are present in a majority of solid tumors and can directly promote tumor cell motility via cytokine, chemokine and growth factor secretion into the TME. The exact effects that the TME has upon cytoskeletal regulation in motile tumor cells remain enigmatic. The conserved formin family of cytoskeleton regulating proteins plays an essential role in the assembly and/or bundling of unbranched actin filaments. Mammalian Diaphanous-related formin 2 (mDia2/DIAPH3/Drf3/Dia) assembles a dynamic F-actin cytoskeleton that underlies tumor cell migration and invasion. We therefore sought to understand whether CAF-derived chemokines impact breast tumor cell motility through modification of the formin-assembled F-actin cytoskeleton. In MDA-MB-231 cells, conditioned media (CM) from WS19T CAFs, a human breast tumor-adjacent CAF line, significantly and robustly increased wound closure and invasion relative to normal human mammary fibroblast (HMF)-CM. WS19T-CM also promoted proteasome-mediated mDia2 degradation in MDA-MB-231 cells relative to control HMF-CM and WS21T CAF-CM, a breast CAF cell line that failed to promote robust MDA-MB-231 migration. Cytokine array analysis of CM identified up-regulated secreted factors in WS19T relative to control WS21T CM. We identified CXCL12 as a CM factor influencing loss of mDia2 protein while increasing MDA-MB-231 cell migration. Our data suggest a mechanism whereby CAFs promote tumor cell migration and invasion through CXCL12 secretion to regulate the mDia2-directed cytoskeleton in breast tumor cells.


Introduction
Approximately 90% of cancer-related deaths are due to advanced metastatic disease [1]. In metastatic breast cancer, invasive primary tumor cells can migrate to regional lymph nodes en PLOS  signaling, which in turn modulates cancer cell morphology and invasion [38]. mDia1 was shown to be important for lamellae and filopodia formation following EGF stimulation in MTln3 breast adenocarcinoma cells [39]. mDia1-3 were shown to be important for invadopodia formation and subsequent matrix degradation [40]. mDia2, which is encoded by DIAPH3, increased invasive cell egress from epithelial ovarian cancer spheroids [41]. Functional inhibition of mDia2 through association with its negative regulator, Dia-interacting protein (DIP), caused non-apoptotic blebbing, a hallmark of amoeboid motility in breast tumor cells [42]. Conversely, mDia2 activation using small molecule agonists inhibited glioblastoma invasion and migration both in vitro and ex vivo [43]. Thus, the role of mDia proteins within different tumor microenvironments is likely complex and dictated by specific environmental cues.
In this study, we sought to understand how CAF-soluble factors affect the mDia-directed Factin cytoskeleton in MDA-MB-231 human breast adenocarcinoma cells. Here we demonstrated conditioned media (CM) from WS19T breast tumor-adjacent CAFs significantly increases MDA-MB-231 breast tumor cell migration and invasion, and is correlated with significant loss of mDia2 protein expression through a proteasomal-dependent mechanism. DIAPH3 expression was not diminished in response to CAF-CM treatment. Finally, we determined by membrane-based cytokine array that stromal-secreted CXCL12 is a significantly upregulated component of CAF-CM that underlies mDia2 loss in MDA-MB-231 cells and the resultant increase in cell migration.
Lactacystin (Santa Cruz Biotechnology) was used at 10 μM in dH 2 O with 16h treatment. Vehicle treatments were equal volumes. SMIFH2 in DMSO (EMD Biochemicals; Tocris Bioscience, Avonmouth) was used at 10-40 μM with 16h for wound closure assays and 8-72h for western blot analysis. Cycloheximide in dH 2 O was obtained from Sigma-Aldrich and used at 10 μg/ml for 1-24h. CXCL12/SDF-1α in dH 2 O (R&D Systems) was used at 15-100 ng/ml for 16h for wound closure assays and 8-72h for western blot analysis.

Western blotting
Whole cell lysates were collected with lysis buffer (0.5M Tris-HCl, pH 6.8, glycerol, 10% SDS (wt/vol), 0.1% bromophenol blue (wt/vol) supplemented with 0.1M diothiothreitol (DTT)) and SDS-PAGE was performed to resolve proteins. Proteins were transferred to PVDF membranes using a BioRad Trans-Blot turbo transfer system. Western blots were exposed using Clarity Western ECL (BioRad) and Alpha Innotech imaging system (Azure Biosystems). Densitometry was performed using ImageJ software.

Conditioned media preparation
Human mammary, NIH-3T3, WS19T, and WS21T fibroblasts were plated in DMEM growth media, in T-75 culture flasks, and grown to confluence. Media were removed 5 days post-confluence, centrifuged at 1000 rpm to remove cellular debris, and stored at -20˚C until use.

Wound healing assays
Confluent MDA-MB-231 breast cancer cells in 6-well culture plates were scratched with a sterile pipette tip, followed by 16h incubation at 37˚C with 5% CO 2 . Immediately following scratching, media were changed to either DMEM, HMF-CM, WS19T-CM or WS21T-CM. Image acquisition occurred at introduction of the scratch (0h) and 16h post scratching. "Wounds" were measured using MetaMorph Image Analysis software to determine percent wound closure. Each condition was performed in triplicate within a single experiment and with a minimum of three experimental repeats.

Spheroid formation
Spheroids were generated using centrifugation and poly-HEMA coated, low attachment plates as previously described [48]. Coated wells in a 96-well culture plate were seeded with 4,000 cells suspended in DMEM with 10% FBS supplemented with 2.5% of 15μl/ml matrigel (BD Biosciences). Cells were pelleted at 1,000xg for 2m. Spheroids were grown for 72h prior to the start of all assays.

Spheroid invasion assay
Collagen-1 (BD Biosciences) was used at a concentration of 2mg/ml and prepared as previously described (modified from [49]). 8-well chamber slides (LabTek) were coated with a thin layer of the diluted collagen, spheroids were added in 15μl of media and a thin collagen layer was added on top of the initial collagen layer and spheroid. Slides were incubated at 37˚C for 45m to allow for polymerization prior to adding media. Serum-free media (SFM), full growth media (DMEM containing 10% FBS), HMF-CM, or WS19T-CM was used per 8-well slide. Spheroids were imaged upon embedding and 24, 48h, and 72h post-embedding. Media were refreshed every 24h. MetaMorph image analysis software was used to determine the area of each spheroid by drawing a region of interest (ROI) that encompassed at least 90% of the invasion edges. Change in area was used as a measure of invasion. A single experiment included measurements from at least 8 wells of each media type and the experiment was repeated three times.
and an Applied Biosystems 7500 PCR system. Analysis was performed on SDS software as previously described [50]. The average Ct values for tested (DIAPH3) and housekeeping genes (GAPDH and cyclophilin B) were calculated from the individual Ct values generated from the PCR reaction. The average Ct values for the experimental housekeeping genes were subtracted from the tested experimental values for the ΔCt experimental. The average Ct values for the control housekeeping genes were subtracted from the control experimental values for the ΔCt control. The ΔCt control was then subtracted from the ΔCt experimental to produce the ΔΔCt.

Proliferation assay
MTT (Biosynth International) cell viability assay was performed following the manufacturer's specifications. Metabolic activity was assessed at 16, 24, 48, and 72h. Within a single experiment each condition and time point included nine measurements and the experiment was repeated thrice.

Cytokine antibody array
Custom cytokine antibody arrays were from RayBiotech (S1 Table). The assay was performed following the manufacturer's instructions. Briefly, treated membranes were exposed using the Alpha Innotech chemilluminescence imaging system (Azure Biosystems). Quantification was performed using ImageJ software and analysis was performed with the RayBiotech analysis software. Background measurements were subtracted and values were normalized to the corresponding target incubated with HMF-CM. WS19T and WS21T-CM were screened three times with each target spotted in duplicate per membrane (S2 Table). Control HMF-CM was screened twice with each target spotted in duplicate per membrane.

Human CXCL12/SDF1-α ELISA
A human CXCL12/SDF1-α ELISA kit was obtained from Sigma-Aldrich. The assay was performed following manufacturer's instructions. The provided standard was performed in triplicate. HMF-CM, WS19T-CM, and WS21T-CM were screened in three independent CM sample collections, collected as previously described. Each collection included parallel HMF-CM, WS19T-CM, and WS21T-CM sample. HMF-CM samples were in duplicate, while WS19T-CM and WS21T-CM were assessed in triplicate. Assays were read at 450nm absorbance using a SpectraMax plate reader and results were plotted with SigmaPlot software.

Statistical analysis
One-tail Student's t-tests were used with a 95% confidence interval with p < 0.05 interpreted as statistically significant. Standard deviations are shown on histograms.

WS19T CAF-derived CM drives increases MDA-MB-231 breast tumor cell motility in vitro
Given the notion that CAF-secreted factors promote invasion and metastasis, we hypothesized that mammary tumor-derived CAFs may influence tumor cell motility through modifying the actin cytoskeleton. We first assessed if factors present in WS19T CAF-conditioned CM altered tumor cell motility in wound healing assays. WS19T CAFs are a tumor-adjacent patientderived mammary carcinoma associated fibroblast cell line [44]. To assess the effects of WS19T-CM conditioned for various lengths of time upon MDA-MB-231 cell motility, wound closure assays were performed using WS19T-CM collected 1-5d post confluency. Confluent cell monolayers were "scratched" (T0) and allowed to fill in the "wound" for 16h. MDA-MB-231 cell wound closure increased in a time -dependent manner in CM collected at 3, 4, and 5d post-confluency, with greatest wound closure in 5d CM (S1 Fig

WS19T-CM reduces MDA-MB-231 cell proliferation
To validate that increased wound closure throughout 16h was not due to increased cell proliferation, we analyzed the effects of WS19T-CM on MDA-MB-231 proliferation using MTT and cell counting assays. Interestingly, WS19T-CM reduced MDA-MB-231 metabolic activity ( Fig  2A) by 55-85% through 72h of WS19T-CM incubation relative to MDA-MB-231 cells in DMEM. We then manually counted cells incubated in DMEM growth media or WS19T-CM. We plated 50,000 cells and assessed growth ever 24h through 72h. MDA-MB-231 cells in the presence of WS19T-CM still proliferated, but did so at a modest yet significantly decreased rate (~2,678 cells/h) compared to the corresponding DMEM-treated cells (~3,125 cells/h) ( Fig  2B). Therefore, increased MDA-MB-231 cell proliferation does not account for increased wound closure in response to WS19T-CM.

WS19T-CM increases MDA-MB-231 cell invasion in a 3D collagen matrix
We next assessed whether WS19T-CM affected MDA-MB-231 cell invasion. MDA-MB-231 spheroids were formed for 72h, embedded in 2mg/ml Type-1 collagen gels, and allowed to invade for an additional 72h [48] (Fig 3A). Embedded spheroids incubated with serum-free medium (SFM), control HMF-derived CM, and DMEM growth media showed moderate invasion through 72h (Fig 3A and 3B). Spheroids embedded with WS19T-CM showed significantly increased invasion compared to controls at corresponding culture times through 72h invasion. Thus, CAF-CM increases MDA-MB-231 motility in both 2D and 3D environments.

WS19T-CM mediated mDia2 protein expression loss is recoverable
Kinetic washout assays were performed to evaluate if mDia2 loss in response to WS19T-CM was recoverable (Fig 5). MDA-MB-231 cells were first incubated in either DMEM growth media or WS19T-CM for 8h. The respective media were then washed out and replaced with DMEM growth media. mDia2 protein was re-expressed as early as 1h after the CM-washout (dark grey vs. light grey bars, Fig 5B). Levels approached DMEM control levels (white bars) by 2h post washout. mDia1 protein levels remained relatively unchanged with time (Fig 5A and  5B) in the presence of WS19T-CM and throughout the washout.

WS19T-CM does not affect mDia2 mRNA levels
To determine if mDia2 expression loss is at the level of the mDia2 mRNA, qRT-PCR was performed with primers recognizing DIAPH3 or control genes encoding cyclophilin B (PPIB) or GAPDH. MDA-MB-231 control and DIAPH3 mRNA levels were assessed after 8-72h of WS19T-CM treatment (Fig 6), paralleling mDia2 protein expression (Fig 4). DIAPH3 levels remained statistically unchanged when normalized to PPIB or GAPDH mRNA, compared to the corresponding DMEM treatment time points (Fig 6A and 6B). Thus, WS19T-CM regulates mDia2 protein expression and not DIAPH3 mRNA levels. mDia functional inhibition does not affect mDia2 protein levels or cell motility mDia2 functional inhibition using the small molecule inhibitor of FH2 domain, SMIFH2, resulted in loss of mDia2 protein in U2OS cells within 2-16h, and within 5h in HEK 293T, A375, and MDA-MB-231 cells through an unidentified non-proteasomal mechanism of protein degradation [51]. SMIFH2 functionally inhibits the FH2 domain of mDia formins and prevents F-actin nucleation, decreases formin affinity for the barbed end of F-actin, and reduces F-actin elongation [57]. We assessed whether mDia functional inhibition decreased mDia2 protein expression. We first validated the functionality of the SMIFH2 used in these studies by assessing induction of non-apoptotic plasma membrane blebs in MDA-MB-231 cells, a functional consequence we previously observed [42]. Indeed, 10μM SMIFH2 induced robust blebbing, confirming functionality of individual lots of SMIFH2 (data not shown). MDA-MB-231 cells treated with 40μM SMIFH2 for 8-72h revealed no loss of mDia2 protein expression compared to MDA-MB-231 cells in DMEM and DMSO-treated cells (Fig 7A and  7B). In our system, cell motility was unaffected upon mDia2 functional suppression, with no significant difference in percent wound closure between DMEM-and SMIFH2-treated cells (Fig 7C) after 16h.

WS19T-CM reduces mDia2 protein expression through a proteasomedependent mechanism
We next sought to understand mechanisms whereby mDia2 protein expression is lost in MDA-MB-231 cells in response to WS19T-CM. We first utilized cycloheximide to determine the half-life of mDia2 in culture and compare to WS19T-CM-affected mDia2 expression kinetics. MDA-MB-231 cells treated with 10μg/mL of cycloheximide yielded an mDia2 half-life of 5.6h (Fig 8A and 8B). When MDA-MB-231 cells were treated with WS19T-CM for the same time course, mDia2 half-life decreases to 3.9h. This shortened mDia2 half-life in the presence  of WS19T-CM supports the notion of a mechanism other than the normal turnover of mDia2 protein is at play in response to WS19T-CM in MDA-MB-231 cells. mDia2 protein expression during the cell cycle is tightly regulated by ubiquitination and subsequent degradation [28]. mDia2 is expressed in S-and G2/M phase with a significant drop following progression into G0/G1 phase. This marked drop is due to poly-ubiquitination followed by degradation. We next assessed whether proteasomal-mediated degradation of mDia2 is initiated by CAF-CM factors. MDA-MB-231 cells were treated with the proteasome inhibitor lactacystin for 16h in the presence of DMEM or WS19T-CM and mDia2 protein expression was evaluated. Neither vehicle-treated nor proteasome inhibitor-treated MDA-MB-231 cells showed loss of mDia2 protein expression when cultured in DMEM (Fig 8C and 8D). Proteasome inhibition in the presence of WS19T-CM restored mDia2 protein expression to that of DMEM and vehicle control levels. Proteasome inhibition also inhibited CM-mediated motility (via wound closure) to levels comparable to MDA-MB-231 cells cultured in DMEM (Fig 8E). Thus, intracellular sequestration does not appear to be a mechanism impacting mDia2 protein expression and/or recovery in our system.

WS19T-CM contains up-regulated, cancer-associated cytokines
To characterize factors present in WS19T-CM that underlie enhanced MDA-MB-231 cell motility and/or loss of mDia2 expression, we performed a cytokine array analysis. HMF-CM, WS19T-CM, and WS21T-CM were applied to cytokine arrays assessing 28 target proteins (S1 Table). HMF-CM was used as a control for baseline levels of these targets in a non-cancerous stromal population and as HMF-CM promoted neither MDA-MB-231 migration nor influenced mDia2 expression. Cytokine targets were chosen based on known CAF-secreted factors and tumor-promoting factors. Cytokines of interest were identified by up-regulation in WS19T-CM arrays relative to WS21T-CM arrays to identify factors with potential to differentially affect MDA-MB-231 cell motility and reduce mDia2 protein expression. TGFα, PDGF, IL-17, TNF-β MMP-13, and CXCL12 all showed dramatically increased expression in the WS19T-CM compared to WS21T-CM (Fig 9A and 9B and S2 Table).

CXCL12 is a key effector in WS19T-mediated MDA-MB-231 cell motility
We initially focused upon CXCL12 and its role in promoting MDA-MB-231 invasion and in regulating mDia2 expression. We first evaluated the physiological levels of CXCL12 within our CM panel via CXCL12/SDF1-α ELISA. We assessed CXCL12 levels in HMF-CM, WS19T-CM, and WS21T-CM. Within HMF-CM, WS21T-CM, and WS19T-CM, CXCL12 levels were measured at~10.5, 31.0, and 136 pg/mL, respectively (Fig 10A). We therefore observed the highest level of CXCL12 secretion in WS19T-CM-the cell line whose CM most dramatically contributed to migration, invasion, and loss of mDia2 in MDA-MB-231 cells. , and WS19T fibroblasts were plated and CM collected concurrently for each replicate. HMF-CM, WS19T-CM, and WS21T-CM from three independent collections were applied in triplicate to a CXCL12/SDF1α ELISA assay. The experiment was repeated three times. CXCL12 levels were averaged for each CM and compared to HMF-CM. Ã p<0.0008 HMF-CM relative to WS21T-CM, WS21T-CM relative to WS19T-CM, and HMF-CM relative to WS19T-CM. p<0.001. B. MDA-MB-231 cells were treated with 15, 25, and 100ng/ml CXCL12 and wound closure assays performed for 16h. The assay performed in triplicate and repeated three times. Ã p<0.001 C. MDA-MB-231 cells were treated with 100ng/ml CXCL12 for 8-72h and cell lysates were Western blotted. mDia2 expression was normalized to GAPDH and compared to the DMEM control. The experiment was repeated three times. Ã p<0.01 D. MDA-MB-231 cells were treated with WS19T-CM as indicated. Cells were pretreated (P) for 15m with AMD3100, and/or simultaneously and continuously (C) with AMD3100 and CM (CM) or DMEM (DM) for 16h. Cell lysates were Western blotted as indicated. https://doi.org/10.1371/journal.pone.0195278.g010 To evaluate the effects of CXCL12 upon MDA-MB-231 cell migration, we performed wound closure assays in response to purified CXCL12. We treated MDA-MB-231 cells for 16h with 15, 25, and 100ng/ml CXCL12. CXCL12 treatment resulted in significantly increased wound closure relative to MDA-MB-231 cells in DMEM (~60% closure vs. 45% closure, respectively) (Fig 10B). Modest yet significant increases in wound closure was observed with the lower CXCL12 concentrations as well.
Finally, we evaluated the effects of CXCL12 upon mDia2 expression in MDA-MB-231 cells. MDA-MB-231 cells treated with 100ng/ml CXCL12 for 8-72h had significantly reduced mDia2 protein expression relative to MDA-MB-231 cells in DMEM (Fig 10C). Pre-incubation (PCM) as well as continuous (CCM) treatment of MDA-MB-231 cells with AMD3100, an inhibitor blocking the CXCL12 receptor CXCR4, blocked WS19T-CM-mediated loss of mDia2 protein relative to DMEM growth media-treated cells, supporting a role, in part, for CXCL12 signaling and regulation of mDia2 protein expression in our system (Fig 10D).

Discussion
In this study, we discovered a role for stromal-derived CXCL12 in breast cancer migration and invasion potentially through proteasome-mediated loss of mDia2 protein expression. WS19T CAF-CM significantly increased motility and invasion of MDA-MB-231 cells relative to MDA-MB-231s in DMEM growth media, or control CM. We observed concomitant and specific loss of mDia2 protein expression with incubation with WS19T CAF-CM. The stromalderived factor CXCL12 was subsequently identified as a prominent factor enriched in WS19T CAF-CM that promoted, in part, both MDA-MB-231 cell migration, as well as loss of mDia2 protein.
We show a loss of mDia2 expression with increased cell migration and invasion in MDA-MB-231 cells in the presence of WS19T-CM. Indeed, we previously demonstrated that mDia2 expression and/or activity was involved in tumor cell motility. Sustained mDia2 activation using small molecule intramics (IMMs) reduced U87 and U251 glioblastoma invasion [43], while mDia2 functional inhibition promoted amoeboid motility in a MDA-MB-231 breast cancer model [42,47], and loss of mDia2 function and/or expression increased single cell dissemination in an ovarian cancer spheroid model [41]. Furthermore, DIAPH3 expression is reduced in invasive prostate cancer [53] and in breast and hepatocarcinoma cells [74]. Thus, tight regulation of mDia2 (and other mDia formins) is an important factor underlying cell motility in a variety of tumor models. Until now, little was known regarding what physiological drivers impact mDia2 expression and/or function. The loss of mDia2 expression in the presence of WS19T-CM highlights a physiologically relevant source of secreted factors targeting and downregulating mDia2, and subsequently promoting cancer cell dissemination and invasion. Interestingly, DIAPH3 expression was decreased in micro-dissected tumor adjacent stroma derived from invasive breast carcinoma [74][75][76][77]. Hence, tight regulation of DIAPH3, or mDia2 expression and/function as a mechanism to control cellular transformation and dissemination may not be employed solely by tumor cells, but by the TME cellular constituency as a whole. It must be noted that this is the first tumor cell model of mDia2 regulation in response to CAF-CM. It warrants expanding studies to additional cancer models where a role for mDia2 in tumor invasion and metastasis has been established, such as prostate, hepatocarcinoma and glioblastoma, to test the specificity of this mechanism of mDia2 dependent mode of motility regulation.
We examined the effects of the SMIFH2-mediated mDia suppression in our system to evaluate if mDia functional suppression impacted mDia protein stability and/or expression. Indeed, a previous report correlated SMIFH2-mediated functional suppression with loss of mDia2 expression in select cell lines [51]. In our system, unlike treatment with WS19T-CM, SMIFH2 treatment for 8-72h did not result in loss of mDia2 expression at specific time points, yet it did drive amoeboid conversions and non-apoptotic membrane blebbing (data not shown), as we previously demonstrated [42]. Within a continuous 16h treatment window, however, wound closure was not impacted with SMIFH2 treatment. Cell velocities, persistence, and total distance migrated were not measured, however. These results differ from a previous report, in which MDA-MB-231 cells treated with SMIFH2 lost mDia2 protein within 5h [51]. It is possible that due to the cyclical nature of SMIFH2 action [51], or lot-to-lot variability in the inhibitor, the kinetics of mDia2 expression upon SMIFH2 suppression are subtly altered in our system. Alternatively, functional suppression of mDia2 (via the SMIFH2 mechanism) is not a requisite for mDia2 protein loss. Whether functional suppression of mDia2 precedes protein loss in WS19T-CM-treated MDA-MB-231 is currently under investigation in our lab.
We sought to understand whether the loss of mDia2 was at the level of gene transcription or protein stability/degradation. RT-PCR indicated that levels of DIAPH3 transcripts are unchanged in the presence of WS19T-CM. We utilized cycloheximide to determine the half-life of mDia2 in culture and compare to WS19T-CM kinetics. The mDia2 half-life in MDA-MB-231 cells treated with CM is significantly shorter than that of cycloheximide-treated cells (~3.9 vs. 5.6h, respectively). Targeting the proteasome with lactacystin both restores mDia2 expression in the presence of CM and blocked the CM-induced increase in cell motility, pointing towards a proteasome-dependent mechanism for loss of mDia2. There is evidence for linkage between mDia2 and the proteasome in tightly regulating mDia2 expression. mDia2, ubiquitin, and the proteasome were shown to directly interact yet mDia2 did not undergo proteasomemediated degradation in a HEK 293T forced overexpression system [78]. Conversely, mDia2 was expressed through S-and G2/M phase of cell division in HeLa cells and was highly polyubiquitinated at the end of M phase, followed by substantial degradation and loss of mDia2 as cells enter G0/G1 phase [28]. Our results support these findings, and moreover, highlight physiological signals triggering mDia2 proteasomal degradation in breast tumor cells.
We noted dramatic upregulation of 6 factors by cytokine array analysis of WS19T-CM, relative to HMF-CM and WS21T-CM. We prioritized factors based upon known established links with cancer cell invasion, metastatic formation, cytoskeleton regulation, or mDia formins. From those criteria, CXCL12, PDGF [69][70][71], and TGF-α [72,73] were our top 3 candidates, with initial focus upon CXCL12. We revealed a novel finding that purified CXCL12 promoted mDia2 loss in MDA-MB-231 cells loss through engagement of its receptor CXCR4 (Fig 10), and promoted MDA-MB-231 cell motility. These results indicate an important intersection between CXCL12 signaling and regulation of the mDia2-directed cytoskeleton driving tumor motility.
CXCL12 ELISAs (Fig 10) revealed a CXCL12 average concentration within WS19T CAF-CM of >100 pg/ml. At this concentration, CXCL12 was, in part, sufficient to promote invasion and migration, as well as loss of mDia2 expression. How does this concentration compare to other studies measuring CXCL12 levels in invasive tumors? Serum CXCL12 levels in patients with esophageal cancer were measured at 1.27 ng/ml compared to 0.86 ng/ml in healthy controls [79], while gastric cancer tumors were approximately 3618 ng/ml compared to 1715 ng/ml in the non-cancer control group [80]. Such substantially disparate CXCL12 values could point to distinct functions for CXCL12 in different malignancies and pathologies, as well as additive/synergistic roles for other cytokines in promoting cancer phenotypes. In our studies using purified CXCL12 in monolayer culture, higher concentrations were needed to mimic the effects of WS19T-CM upon MDA-MB-231 migration and mDia2 suppression. Within the context of WS19T CM, CXCL12, likely in conjunction with other enriched cytokines and growth factors, may underlie a more complex system to fully regulate the actin cytoskeleton and cell motility. Within the TME, differences in temporally and spatially local CXCL12 concentrations may differentially influence mDia2 loss and/or increased motility. Collectively, our study does not indicate an exclusive role for CXCL12 in promoting both mDia2 loss and tumor migration. Rather it indicates an important and likely complementary role within a milieu of various cytokines. Future experiments will focus upon the additive or possible synergistic signaling nature of CXCL12 and other cytokines in our system, such as PDGF, in driving mDia2 loss while promoting motility.
In this study, we observed significant increases in breast cancer cell migration and invasion in response to CAF-conditioned media, which was accompanied by dramatic loss of mDia2 expression. We identified CXCL12 as an underlying factor within WS19T-CM that mediates, in part, these phenotypes. Previous work identified a role for exogenous CXCL12 in breast cancer motility and migration, yet the source of CXCL12 influencing tumor motility was uncertain. Here we identify for the first time a physiological source for CXCL12 in the TMEspecifically from tumor-adjacent carcinoma associated fibroblasts, and reveal a unique role for the TME in directly influencing tumor cell motility through mDia formin-dependent cytoskeletal regulation. This novel mechanism is a step towards understanding the role of CAF:tumor signaling in cancer progression and identifies potential therapeutic targets that could aid in blocking metastatic dissemination and improving patient prognosis.
Supporting information S1  MDA-MB-231 cells plated on glass coverslips were treated with the indicated media for 8h before fixation. Cells were immunostained with anti-mDia2 antibodies, phalloidin and DAPI. Percent nuclear mDia2 fluorescence was measured relative to plasma membrane/cytoplasmic mDia2 fluorescent signal with Metamorph software. At least 30 cells per condition were measured and the experiment was repeated three times. Scale bars = 25μm. (TIF)