IJC International Journal of Cancer
Breast carcinoma cells modulate the chemoattractive activity of human bone marrow-derived mesenchymal stromal cells by interfering with CXCL12 Manja Wobus1, Catrin List1, Tobias Dittrich1, Abhishek Dhawan1, Regina Duryagina1, Laleh S. Arabanian1, Karin Kast2, Pauline Wimberger2, Maik Stiehler3, Lorenz C. Hofbauer4,5, Franz Jakob6, Gerhard Ehninger1, Konstantinos Anastassiadis7 €user1 and Martin Bornha 1
Cancer Cell Biology
Division of Hematology, Oncology and Stem Cell Transplantation, Department of Medicine I, University Hospital Carl Gustav Carus, Technische Universi€at Dresden, Germany 2 Department of Obstetrics and Gynaecology, University Hospital Carl Gustav Carus, Dresden, Germany 3 University Centre for Orthopedics & Reconstructive Surgery, University Hospital Carl Gustav Carus and Centre for Translational Bone, Joint and Soft Tissue Research, Technische Universit€at Dresden, Germany 4 Division of Endocrinology, Diabetes and Bone Diseases, Technische Universit€ at Dresden, Germany 5 Department of Medicine III, Technische Universit€at Dresden, Germany 6 €rzburg, Germany Orthopedic Center for Muscoskeletal Research, University of Wu 7 Stem Cell Engineering, Biotechnology Center, Technische Universit€at Dresden, Germany
We investigated whether breast tumor cells can modulate the function of mesenchymal stromal cells (MSCs) with a special emphasis on their chemoattractive activity towards hematopoietic stem and progenitor cells (HSPCs). Primary MSCs as well as a MSC line (SCP-1) were cocultured with primary breast cancer cells, MCF-7, MDA-MB231 breast carcinoma or MCF-10A non-malignant breast epithelial cells or their conditioned medium. In addition, the frequency of circulating clonogenic hematopoietic progenitors was determined in 78 patients with breast cancer and compared with healthy controls. Gene expression analysis of SCP-1 cells cultured with MCF-7 medium revealed CXCL12 (SDF-1) as one of the most significantly downregulated genes. Supernatant from both MCF-7 and MDA-MB231 reduced the CXCL12 promoter activity in SCP-1 cells to 77% and 47%, respectively. Moreover, the CXCL12 mRNA and protein levels were significantly reduced. As functional consequence of lower CXCL12 levels, we detected a decreased transwell migration of HSPCs towards MSC/tumor cell cocultures or conditioned medium. The specificity of this effect was confirmed by blocking studies with the CXCR4 antagonist AMD3100. Downregulation of SP1 and increased miR-23a levels in MSCs after contact with tumor cell medium as well as enhanced TGFb1 expression were identified as potential molecular regulators of CXCL12 activity in MSCs. Moreover, we observed a significantly higher frequency of circulating colony-forming hematopoietic progenitors in patients with breast cancer compared with healthy controls. Our in vitro results propose a potential new mechanism by which disseminated tumor cells in the bone marrow may interfere with hematopoiesis by modulating CXCL12 in protected niches.
Disseminated tumor cells can be detected in the bone marrow at the time of the initial diagnosis in patients with breast cancer.1 Besides the hematogenic spread to visceral organs, breast cancer cells preferentially spread to bone. Apart from Key words: mesenchymal stromal cells, breast carcinoma, CXCL12, hematopoiesis Conflicts of interest: No potential conflicts of interest were disclosed. Grant sponsor: DFG Collaborative Research Center SFB 655 “Cells into tissues”; Grant sponsor: DFG Research Group-1586 “SKELMET” DOI: 10.1002/ijc.28960 History: Received 16 Dec 2013; Accepted 16 Apr 2014; Online 8 May 2014 Correspondence to: Dr. Manja Wobus, Medical Clinic and Polyclinic I, University Hospital Carl Gustav Carus, Fetscherstr. 74, D-1307 Dresden, Germany, Tel.: 149-351-458-4765, Fax: 149-351458-7398, E-mail: [email protected]
C 2014 UICC Int. J. Cancer: 136, 44–54 (2015) V
case-based surgical strategies, bone metastases are incurable and therefore better treatments need to be developed. Metastasis is an inefficient, multi-step process in which bonemarrow derived cells may be involved early throughout the seeding process.2 In the bone marrow, hematopoietic stem and progenitor cells (HSPCs) reside in an area that is defined as the stem cell niche. Cellular and biochemical definition of the HSC niche in bone marrow has been an active area of investigation. Several molecules expressed and secreted by perivascular cells and osteoblasts3–6 play critical roles in niche formation and selectivity. The bone microenvironment alters gene expression of breast cancer cells. In particular, cancer cells exhibit an increased production of cytokines and other active factors that act on cells on adjacent bone, making the respective microenvironment vulnerable for cancer colonization. Thus, the specific molecular interactions between breast cancer cells and the bone marrow microenvironment drive the
45
Wobus et al.
What’s new? Disseminated breast tumor cells tend to metastasize to the bones, possibly using similar pathways as hematopoietic stem and progenitor cells (HSPCs) to colonize bone marrow—such as the CXCL12/CXCR4 chemotactic pathway. This study reports a novel mechanism by which breast cancer cells interfere with CXCL12 production by bone marrow-derived mesenchymal stromal cells in vitro. Such effect reduces the stromal support for CD341 HSPCs and involves regulation by SP1 transcription factors, miR-23a, and TGFb1. The authors also found a higher frequency of circulating colony-forming hematopoietic progenitors in patients with breast cancer, compared to healthy controls, that may result from this interference.
C 2014 UICC Int. J. Cancer: 136, 44–54 (2015) V
Material and Methods Cells and cell culture
The hTert immortalised MSC line SCP-1 was kindly provided by M. Schieker (Munich)15 and was cultured in MEM-alpha medium (aMEM, Invitrogen Life Technologies, Darmstadt, Germany) supplemented with 10% fetal calf serum (FCS; Biochrom, Berlin, Germany). MCF-10A normal breast epithelial cells were obtained from ATCC (LGC Standards GmbH, Wesel, Germany) and cultured in DMEM/F12 (Gibco Life Technologies, Darmstadt, Germany) supplemented with 5% horse serum (Invitrogen), 100 mg/ml EGF (Peprotech, Hamburg, Germany), 1 mg/ml hydrocortisone, 10 mg/ml insulin and 1 mg/ml cholera toxin (all from Sigma-Aldrich, Steinheim, Germany). MCF-7 and MDA-MB231 breast carcinoma cells were obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ, Braunschweig, Germany) and were maintained in DMEM GlutaMax (Invitrogen) supplemented with 10% FCS. Primary breast cancer cells were isolated from two biopsies of invasive lobular tumors. The tissue samples were cut into pieces and digested with Liberase (Roche Diagnostics, Mannheim, Germany) overnight. After resuspension and filtration through a 70 mm mesh, cells were washed two times with PBS/2% FCS and seeded in aMEM containing 10% FCS and 1% penicillin/streptomycin. Non-adherent cells were removed after 24 hr. Human primary MSCs were isolated from bone marrow aspirates derived from healthy donors after informed consent and approval by the local ethics committee and cultured as described previously.16 For isolation of CD341 HSPCs, mobilized peripheral blood (PB) was collected from healthy donors after a 5 day treatment with 7.5 mg/kg granulocyte colony-stimulating factor (G-CSF). HSPCs were purified using CD34 antibody conjugated magnetic beads according to the manufacturer’s instructions (Miltenyi Biotec). The purities were > 95% as assessed by flow cytometry and the vitality as measured by trypan blue exclusion was > 96%. Patient samples
PB samples of patients with breast cancer (n 5 78) and an age-matched healthy control group (n 5 22) were collected after informed written consent and histologically proven cancer diagnosis. The characteristics of both cohorts are provided in the Supporting Information. The study had been approved by the local institutional review board. Colonyforming units were quantified after 14 days of standardized
Cancer Cell Biology
subsequent formation of visceral and skeletal metastases.2,7 Several characteristics of breast cancer cells allow them to preferentially home to the bone marrow, for example, chemokine receptors, integrins, matrix metalloproteinases and several other tumor-secreted factors. In the context of hematopoietic stem cell homing, the CXC chemokine CXCL12 (SDF-1) and its receptors CXCR4 and CXCR7 are critical molecular determinants and perhaps the best characterized molecules for this regulation.8,9 Osteoblasts and perivascular stromal cells appear to be a major source of CXCL12, where synthesis is regulated by several inflammatory stimuli, DNA-damaging agents and hormones.9 Moreover, expression of CXCL12 in nestin-positive mesenchymal stromal cells (MSCs) is regulated in a circadian rhythm.10 CXCL12 expression can also be modulated by local production of Transforming Growth Factor beta1 (TGFb1) and this may play an important role in regulating the balance between proliferation and differentiation of HSPCs.11 Only recently, Pillai et al. have suggested, that CXCL12 can also be regulated by microRNAs (miRNA) in human MSCs.12 Metastatic breast carcinoma cells may use similar pathways as HSPCs to colonize the bone marrow and thus compete for their natural habitats.13 In prostate cancer, it was demonstrated that CXCL12 signaling through CXCR4 may trigger the dissemination of tumor cells by activating avb3 integrins and CD164 expression.14 CXCR4 signaling participates in both metastasis and the angiogenesis process of prostate cancer by downregulating the expression and secretion of the glycolytic enzyme phosphoglycerate kinase 1 and angiostatin.14 By means of cell-cell interaction networks, we have already shown that paracrine signaling between those tumor cells and MSCs may interfere with hematopoiesis (unpublished observation). In this study, we tested the hypothesis whether tumor cell induced changes of MSC characteristics have an impact on the HSPC support. We demonstrate a downregulation of MSC derived CXCL12 which correlates with decreased HSPC migratory potential. Further relevant findings of possible upstream regulatory mechanisms were the downregulation of SP1 transcription factors and increased miR-23a levels in MSCs after contact with tumor cell medium as well as an enhanced TGFb1 expression. The in vitro data are supported by our finding that breast cancer patients display elevated levels of circulating progenitor cells and a significantly higher number of colony forming units of the granulocytic/monocytic lineage (CFU-GM) compared with the age-matched healthy control group.
46
Modulation of MSCs by breast carcinoma cells
a pore size of 5 mm (Corning Costar Corporation, Cambridge, MA). We further investigated the dependence of migration against test media on the CXCL12/CXCR4 axis by initially adding 1 mmol/l of the bicyclam AMD3100, a highly specific CXCR4 antagonist (Sigma-Aldrich). CXCL12 ELISA
To measure CXCL12 secreted by MSCs either cultured with tumor cells directly or with conditioned medium, an antibody sandwich ELISA (R&D Systems, Wiesbaden, Germany) was used following the manufacturer’s instructions. Gene expression analysis
Cancer Cell Biology
SCP-1 cells were incubated with MCF-7 conditioned or normal culture medium for 24, 48 and 72 hr. Afterwards, cells were lyzed in Trizol reagent and shipped on dry ice to Miltenyi Biotec for one color Agilent Whole Human Genome Oligo Microarrays. Samples were pooled from three independent experiments. Figure 1. Gene expression analysis of SCP-1 cells treated with MCF-7 conditioned medium. Scatter plot of signal intensities of all spots. As an example, the data of 1 out of 3 independent array experiments are shown (experiment no. 1). The signal intensities of each feature represented by a dot are shown in double logarithmic scale. X-axis: control-log signal intensity; y-axis: sample-log signal intensity. Red diagonal lines define the areas of 2-fold differential signal intensities. Blue cross: unchanged genes, n 5 36,576. Red cross: significantly upregulated genes, n 5 2,060 (p-value < 0.01). Green cross: significantly downregulated genes, n 5 2,353 (p-value < 0.01). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
culture of PB mononuclear cells in semisolid medium containing recombinant growth factors.
Statistical analysis
All results of in vitro experiments are presented as means 6 standard error of the mean. Data were analyzed using two-way ANOVA for each time point or condition with a Bonferroni post test using GraphPad Prism 5.0. Statistical significance was established as p < 0.05 (*), p < 0.01 (**) or p < 0.001 (***). Univariate and multivariate analyses were used for analyzing the influence of tumor stage, age and body mass index on the frequency of circulating progenitor cells in patients. Experiments were done in triplicate and all results were expressed as the mean 6 SEM. A detailed description of all additional experimental procedures can be found in the Supporting Information.
Coculture models
To evaluate the expression of cytokines in direct coculture, MSCs were seeded at 5 3 103 cells/cm2 and grown for 3 days to 95% confluency. Thereafter, a total media exchange was combined with addition of tumor cells at 2.5 3 103 cells/cm2. At determined time points, media were collected and stored at 280 C until evaluation of cytokine content. To discriminate the effects of carcinoma cells on MSCs that are strictly based on soluble factors from effects that are cellcontact-dependent, indirect coculture models with conditioned media were applied. For preparation of conditioned medium from tumor cells or MCF-10A cells, 2 3 106 cells were seeded in 20 ml of serum-free aMEM into a T-75 tissue culture flask. After 72 hr, conditioned media were collected and stored at 280 C until the day of use. After thawing and prior to use on MSCs, 10% FCS was added to both conditioned media and fresh aMEM, which served as a control. Transwell chemotaxis assay
The migration of HSPCs to selected media samples was evaluated using a polycarbonate membrane transwell system with
Results MCF-7 conditioned medium altered MSC gene expression profile
To test the effects of cancer cells on bone marrow stroma we cultured MSCs with conditioned medium from the breast cancer line MCF-7 and analyzed their expression profile by microarrays. The gene expression analysis of SCP-1 cells after 24, 48 and 72 hr revealed 2353 downregulated and 2060 upregulated genes (p < 0.01, Fig. 1) belonging to diverse cellular pathways, for example, proliferation, differentiation or cell cycle. Furthermore, the number of modulated genes increased from day 1 to day 3. The complete data sets have been deposited in NCBI’s Gene Expression Omnibus and are accessible through GEO Series accession number GSE49858 (http://www.ncbi.nlm.nih. gov/geo/query/acc.cgi?acc5GSE49858). Results for some of the differentially expressed genes and pathways are shown in Supporting Information Tables S1 and S2. One of the significantly downregulated genes at all three time points was CXCL12. Since CXCL12 is an C 2014 UICC Int. J. Cancer: 136, 44–54 (2015) V
47
Wobus et al.
Tumor cell medium decreased CXCL12 promoter activity and mRNA expression levels
To gain insights into the molecular control of CXCL12 expression and its modulation by tumor cells, we investigated both CXCL12 promoter activity using SCP-1 reporter cell lines and mRNA expression in primary MSCs by real-time PCR. Conditioned media from the tumor cell lines or MCF-10A cells were incubated with CXCL12 reporter SCP-1 cells for 6, 12 and 24 hr. As shown in Figure 2a, MDA-MB231 conditioned medium caused a significant decrease in CXCL12 promoter activity to 47% at all three time points. Furthermore, incubation with MCF-7 conditioned medium resulted in lowered activities of the CXCL12 promoter to about 77%. In contrast, the supernatant from non-malignant MCF-10A cells had no significant influence on promoter activity (Fig. 2a). The mRNA expression levels of CXCL12 mirror these effects. Primary MSCs were incubated with conditioned tumor cell medium, medium from nonmalignant MCF-10A cells or with normal culture medium as a control for 24, 48 or 72 hr. At all three time points, MCF-7 medium caused a significant decrease of CXCL12 expression in MSCs to about 50% and MDA-MB231 medium had a similar effect on day 2 and 3 (Fig. 2b). In contrast, MCF-10A conditioned medium strongly increased CXCL12 mRNA expression at day 3 (Fig. 2b). These results suggest a CXCL12 regulation in MSCs by tumor cell conditioned medium at the transcriptional level. CXCL12 secretion by MSCs was decreased by tumor cells
Whether MSCs change the secretion of CXCL12 after contact to tumor cells or their conditioned medium, respectively, was investigated by ELISA at specified time points. CXCL12 is not present in supernatants from MCF-10A, MCF-7 or MDA-MB231 in single culture. Conditioned media used for indirect coculture as well as control medium did not contain significant amounts of CXCL12 (data not shown). In direct cocultures with MCF-7, CXCL12 levels decreased continuously when compared with control on day 1 (289.6 vs. 324.8 pg/ml), day 3 (204.9 vs. 253.2 pg/ml) and day 5 (193.0 vs. 339.3 pg/ml). After complete media exchange, CXCL12 levels remained unaffected in MCF-7 cocultures while control levels markedly increased, resulting in significantly reduced CXCL12 levels after media exchange. Although direct coculture with MDA-MB231 also led to reduced CXCL12 levels (284.5; 222.6; 275.7 pg/ml), the difference to control was only significant at day 8 (Fig. 2c). In indirect cocultures, both breast cancer cell lines involved significantly reduced CXCL12 contents in supernatants of MSCs. In contrast to the results of direct cocultures, MDA-MB231 cells reduced CXCL12 secretion more strongly at day 3 and 5. After complete media exchange to control C 2014 UICC Int. J. Cancer: 136, 44–54 (2015) V
medium in all three conditions, CXCL12 levels normalized to control levels. This suggested that the downregulation of CXCL12 in MSCs in indirect cocultures with breast cancer cells is partially reversible (Fig. 2d). To investigate whether the observed effects were tumor cell specific, MSCs were incubated with conditioned medium of MCF-10A non-carcinogenic breast epithelial cells. As shown in Figure 2e, the CXCL12 levels were not affected or even increased. These results suggest a cell contact-independent downregulation of CXCL12 by breast tumor cells. Cell culture supernatants from primary tumor cell cultures were collected after 1, 3 and 5 days of coculturing MSCs with tumor conditioned medium or with tumor cells, respectively, and analyzed for CXCL12 concentration. For both conditions we could detect a reduction of CXCL12 levels to about 80 and 78%, respectively (Fig. 2f). Tumor cell-primed MSCs displayed a reduced chemoattractive potential
The functional relevance of CXCL12 downregulation in MSCs in indirect coculture conditions was evaluated by assessing the HSPC migration capacity using a Boyden Chamber assay. As expected, the test media prepared for the lower chamber of migration assay system showed a significantly lower CXCL12 content in MSCs in indirect cocultures with both breast cancer cell lines (Fig. 3a). This is more pronounced in indirect cocultures with MDA-MB231 (22% of control) than with MCF-7 (54% of control). Tumor cell conditioned medium did not change the expression levels of CXCR4 or CXCR7 on HSPCs (not shown). When compared with control, the migration of HSPCs against supernatants of MSCs in indirect cocultures with MCF-7 and MDA-MB231 was significantly reduced to 46 and 54%, respectively (Fig. 3b). This is akin to the basal migration rate without chemoattractant. Blocking of CXCL12/CXCR4 signaling by 1 mmol/l of the CXCR4 antagonist AMD3100 resulted in comparable migration rates slightly above the basal level in control and reference. Although the downregulation of CXCL12 is more prominent in indirect co-cultures with MDA-MB231, the migration of HSPCs is reduced most significantly in indirect cocultures with MCF-7. However, blocking of CXCL12/CXCR4 signaling significantly reduced the migration against supernatants from indirect cocultures with MCF-7, while AMD3100 had no effect in the case of MDA-MB231. These data suggest that the reduced expression of CXCL12 in MSCs in indirect cocultures with breast cancer cells leads to a markedly reduced chemoattraction of HSPCs by MSCs. Tumor cell medium modified the expression of SP1 transcription factor and miR-23a
To unravel mechanisms which are involved in the CXCL12 downregulation in MSCs by tumor cells, we investigated different regulatory levels. First, a possible CXCL12 regulation at the transcriptional level by the SP1 transcription factor was
Cancer Cell Biology
important niche-derived factor playing a crucial role in the regulation of HSC fate,17 we investigated its regulation by tumor cells in more detail although it was not as strongly downregulated as other genes.
Modulation of MSCs by breast carcinoma cells
Cancer Cell Biology
48
Figure 2. CXCL12 promoter activity, mRNA and protein levels were downregulated by tumor cells or conditioned medium. (a) Conditioned medium from MDA-MB231 cells triggers downregulation of CXCL12 promoter activity. CXCL12 reporter SCP-1 cells were cultured with aMEM as control, conditioned aMEM from MCF-10A, MCF-7 or MDA-MB231 cells for 6, 12 and 24 hr. At indicated time points, culture supernatants were collected for GLuc and SEAP activity measurement. GLuc activity was normalized to SEAP activity and results are presented in RLU (n 5 3 for 6 and 12 hr, n 5 6 for 24 hr). Bars represent mean 6 SEM of all independent experiments; p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***). Abbreviations: GLuc, Gaussia luciferase; SEAP, secreted alkaline phosphatase; RLU, relative luciferase units. (b) Quantitative real-time PCR for CXCL12 revealed downregulation of mRNA levels after 24, 48 and 72 hr incubation with MCF-7 or MDA-MB231 conditioned aMEM, whereas no change or a significant increase was observed after incubation with MCF-10A conditioned aMEM. Results are represented as the median of fold change compared with control cultures in aMEM which were set to 1 of three independent experiments in duplicate, p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***). (c) CXCL12 content was determined by ELISA in supernatants of direct cocultures of primary MSCs with MCF-7 or MDA-MB231 cells at indicated time points. Full media exchange to control medium was applied after sample preparation at day 5 (dashed line). Bars represent mean 6 SEM of three independent experiments, p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***). (d) CXCL12 content was determined by ELISA in supernatants of indirect cocultures of primary MSCs with tumor cell conditioned aMEM at indicated time points. Full media exchange to control medium was applied after sample preparation at day 5 (dashed line). Bars represent mean 6 SEM of three independent experiments, p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***). (e) CXCL12 content was determined by ELISA in supernatants of indirect cocultures of primary MSCs with tumor cell as well as MCF-10A conditioned aMEM after two and five days. Bars represent mean 6 SEM of three independent experiments. (f) CXCL12 production as measured by ELISA was suppressed also after incubation with conditioned medium (Tm medium) from two primary breast cancer samples and after direct contact with these primary cells (MSC/Tm cells). MSC/MSC cocultures served as control. Bars represent mean 6 SEM of two independent experiments with four time points each; p < 0.01 (**).
C 2014 UICC Int. J. Cancer: 136, 44–54 (2015) V
Figure 3. HSPC migration potential to MSC/tumor cell cocultures is impaired in a CXCL12/CXCR4-dependent manner. (a) Concentration of CXCL12 in supernatants of primary MSCs after 4 days in control conditions (“CTRL”) and indirect coculture with MCF-7 (“MCF”) or MDAMB231 (“MDA”). Data is represented as mean 6 SEM of four independent experiments, p < 0.01 (**); p < 0.001 (***). (b) Migration of HSPCs towards the above mentioned supernatants, assayed with blocking of CXCR4 by AMD3100 (AMD1) and without blocking (AMD2). Recombinant CXCL12 (rSDF) served as internal reference. Basal migration without chemoattractant is indicated with a dashed line. Data is represented as mean 6 SEM of four independent experiments, p < 0.05 (*); p < 0.01(**) when compared with CTRL. p < 0.05 (†); p < 0.01 (††) when compared with AMD1.
investigated. Primary MSCs were treated with 50 mM mithramycin A which binds GC-rich DNA sequences and thereby inhibits the activity of SP transcription factors.18 We detected a dramatically decreased CXCL12 secretion after 24 hr in all investigated MSC samples suggesting a regulation by Sp transcription factors also in human MSCs (Fig. 4a). Total cell numbers were not influenced by mithramycin (not shown). To test whether tumor cell conditioned medium can influence SP1 in MSCs, we determined the mRNA levels by realtime PCR and protein expression by Western blot. Whereas tumor cell medium caused a significant decrease of SP1 mRNA expression after 1 and 2 days, MCF-10A medium had almost no impact (Fig. 4b). At day 3, the effects were again slightly reversed. The SP1 protein levels in MSCs after incubation with tumor cell conditioned medium were determined by Western blot analysis. As demonstrated in Figure 4c, medium of both tumor cell lines caused a clear decline of the signal intensity whereas MCF-10A medium had no influence. Only recently, in another study of our group miR-23a was identified to regulate CXCL12 expression in MSCs.19 Therefore, we quantified the miR-23a levels in MSCs after incubation with tumor cell conditioned medium. Interestingly, MCF-7 medium caused a significantly upregulation of miR23a after one and two days (Fig. 4d). Additionally, MDAMB231 conditioned medium slightly increased the miR-23a expression levels, whereas MCF-10A medium had no significant impact suggesting a tumor cell specific effect.
Increased TGFb1 expression was involved in the CXCL12 regulation in MSCs
Our microarray data revealed a strong upregulation of the TGFb1 receptor pathway in SCP-1 cells after incubation with C 2014 UICC Int. J. Cancer: 136, 44–54 (2015) V
MCF-7 conditioned medium. Therefore, we tested a possible modulation of TGFb1 in MSCs after contact to tumor cell medium and an influence on CXCL12 expression. A downregulation of CXCL12 secretion by TGFb1 in primary MSCs was demonstrated by addition of the recombinant protein. A blocking anti-TGFb1 antibody was able to reverse this effect at least in part (Fig. 5a). Interestingly, the addition of this blocking anti-TGFb1 antibody to tumor cell conditioned medium in which the MSC were incubated for 24 hr was able to completely abolish the effect of decreased CXCL12 secretion (Fig. 5b). Next, we tested whether tumor cells can modulate the TGFb1 expression in MSCs. The TGFb1 mRNA levels were increased in MSCs by conditioned medium of MCF-7 and MDA-MB231 cells at all three time points, whereas MCF10A conditioned medium resulted in decreased expression (Fig. 5c). The TGFb1 protein in MSCs was studied by immunofluorescence staining. Again, a clear increase in TGFb1 expression was observed when MSCs were cultured with conditioned medium of MCF-7 or MDA-MB231 cells compared to cultures in normal medium (Fig. 5d).
Patients with breast cancer have higher CFU-GM numbers
As a reduced production of CXCL12 may lead to a perturbed balance between circulating and resident HSPCs, we expected a higher frequency of clonogenic HSPCs in the PB of patients with breast cancer. Indeed, as shown in Figure 6, a significantly higher number of CFU-GM (1363) compared with age-matched healthy control group (5 6 2; p < 0.05) was detected. Moreover, the stage of tumor disease did not affect the number of CFU-GM or circulating progenitor cells (data not shown). This effect could be confirmed when body-mass
Cancer Cell Biology
49
Wobus et al.
50
index was included in a multivariate model as confounding variable, but not the age.
Discussion
Cancer Cell Biology
Disseminated tumor cells can be detected in various malignancies and tend to home to the bone marrow. Several of these neoplasms have a striking tendency to subsequently metastasize to bone. The chemokine CXCL12 is expressed by
Modulation of MSCs by breast carcinoma cells
bone marrow stromal cells as well as osteoblast precursors and plays key roles in the migration, adhesion and activation of HSPCs and leukemic cells.20,21 It was hypothesized that tumor cells apply the CXCL12/CXCR4 pathway as homing mechanism to a similar extent as hematopoietic cells.22 Here, we describe a novel mechanism by which breast cancer cells interfere with CXCL12 production of bone marrow-derived MSCs in vitro. This effect reduces the stromal support for CD341 HSPCs and involves regulation by SP1 transcription factors, miR23a as well as by TGFb1. A significantly higher frequency of circulating colony-forming hematopoietic progenitors in patients with breast cancer compared to healthy controls may also result from this interference. Breast tumor cells or their conditioned medium caused a downregulation of CXCL12 mRNA and protein levels in MSCs. In contrast, MCF-10A non-tumorigenic breast epithelial cells did not influence the CXCL12 expression in MSCs or even increased it. The effect of tumor cell mediated downregulation of MSC-derived CXCL12 secretion is cell contact independent and partially reversible. Different kinetic properties of the coculture models, especially with respect to the concentration of paracrine signals, could explain for the different magnitudes of decreased CXCL12 levels. While they were reduced to 80% in direct coculture during a period of 5 days, respective CXCL12 levels were reduced to
Breast carcinoma cells modulate the chemoattractive activity of human bone marrow-derived mesenchymal stromal cells by interfering with CXCL12 Manja Wobus1, Catrin List1, Tobias Dittrich1, Abhishek Dhawan1, Regina Duryagina1, Laleh S. Arabanian1, Karin Kast2, Pauline Wimberger2, Maik Stiehler3, Lorenz C. Hofbauer4,5, Franz Jakob6, Gerhard Ehninger1, Konstantinos Anastassiadis7 €user1 and Martin Bornha 1
Cancer Cell Biology
Division of Hematology, Oncology and Stem Cell Transplantation, Department of Medicine I, University Hospital Carl Gustav Carus, Technische Universi€at Dresden, Germany 2 Department of Obstetrics and Gynaecology, University Hospital Carl Gustav Carus, Dresden, Germany 3 University Centre for Orthopedics & Reconstructive Surgery, University Hospital Carl Gustav Carus and Centre for Translational Bone, Joint and Soft Tissue Research, Technische Universit€at Dresden, Germany 4 Division of Endocrinology, Diabetes and Bone Diseases, Technische Universit€ at Dresden, Germany 5 Department of Medicine III, Technische Universit€at Dresden, Germany 6 €rzburg, Germany Orthopedic Center for Muscoskeletal Research, University of Wu 7 Stem Cell Engineering, Biotechnology Center, Technische Universit€at Dresden, Germany
We investigated whether breast tumor cells can modulate the function of mesenchymal stromal cells (MSCs) with a special emphasis on their chemoattractive activity towards hematopoietic stem and progenitor cells (HSPCs). Primary MSCs as well as a MSC line (SCP-1) were cocultured with primary breast cancer cells, MCF-7, MDA-MB231 breast carcinoma or MCF-10A non-malignant breast epithelial cells or their conditioned medium. In addition, the frequency of circulating clonogenic hematopoietic progenitors was determined in 78 patients with breast cancer and compared with healthy controls. Gene expression analysis of SCP-1 cells cultured with MCF-7 medium revealed CXCL12 (SDF-1) as one of the most significantly downregulated genes. Supernatant from both MCF-7 and MDA-MB231 reduced the CXCL12 promoter activity in SCP-1 cells to 77% and 47%, respectively. Moreover, the CXCL12 mRNA and protein levels were significantly reduced. As functional consequence of lower CXCL12 levels, we detected a decreased transwell migration of HSPCs towards MSC/tumor cell cocultures or conditioned medium. The specificity of this effect was confirmed by blocking studies with the CXCR4 antagonist AMD3100. Downregulation of SP1 and increased miR-23a levels in MSCs after contact with tumor cell medium as well as enhanced TGFb1 expression were identified as potential molecular regulators of CXCL12 activity in MSCs. Moreover, we observed a significantly higher frequency of circulating colony-forming hematopoietic progenitors in patients with breast cancer compared with healthy controls. Our in vitro results propose a potential new mechanism by which disseminated tumor cells in the bone marrow may interfere with hematopoiesis by modulating CXCL12 in protected niches.
Disseminated tumor cells can be detected in the bone marrow at the time of the initial diagnosis in patients with breast cancer.1 Besides the hematogenic spread to visceral organs, breast cancer cells preferentially spread to bone. Apart from Key words: mesenchymal stromal cells, breast carcinoma, CXCL12, hematopoiesis Conflicts of interest: No potential conflicts of interest were disclosed. Grant sponsor: DFG Collaborative Research Center SFB 655 “Cells into tissues”; Grant sponsor: DFG Research Group-1586 “SKELMET” DOI: 10.1002/ijc.28960 History: Received 16 Dec 2013; Accepted 16 Apr 2014; Online 8 May 2014 Correspondence to: Dr. Manja Wobus, Medical Clinic and Polyclinic I, University Hospital Carl Gustav Carus, Fetscherstr. 74, D-1307 Dresden, Germany, Tel.: 149-351-458-4765, Fax: 149-351458-7398, E-mail: [email protected]
C 2014 UICC Int. J. Cancer: 136, 44–54 (2015) V
case-based surgical strategies, bone metastases are incurable and therefore better treatments need to be developed. Metastasis is an inefficient, multi-step process in which bonemarrow derived cells may be involved early throughout the seeding process.2 In the bone marrow, hematopoietic stem and progenitor cells (HSPCs) reside in an area that is defined as the stem cell niche. Cellular and biochemical definition of the HSC niche in bone marrow has been an active area of investigation. Several molecules expressed and secreted by perivascular cells and osteoblasts3–6 play critical roles in niche formation and selectivity. The bone microenvironment alters gene expression of breast cancer cells. In particular, cancer cells exhibit an increased production of cytokines and other active factors that act on cells on adjacent bone, making the respective microenvironment vulnerable for cancer colonization. Thus, the specific molecular interactions between breast cancer cells and the bone marrow microenvironment drive the
45
Wobus et al.
What’s new? Disseminated breast tumor cells tend to metastasize to the bones, possibly using similar pathways as hematopoietic stem and progenitor cells (HSPCs) to colonize bone marrow—such as the CXCL12/CXCR4 chemotactic pathway. This study reports a novel mechanism by which breast cancer cells interfere with CXCL12 production by bone marrow-derived mesenchymal stromal cells in vitro. Such effect reduces the stromal support for CD341 HSPCs and involves regulation by SP1 transcription factors, miR-23a, and TGFb1. The authors also found a higher frequency of circulating colony-forming hematopoietic progenitors in patients with breast cancer, compared to healthy controls, that may result from this interference.
C 2014 UICC Int. J. Cancer: 136, 44–54 (2015) V
Material and Methods Cells and cell culture
The hTert immortalised MSC line SCP-1 was kindly provided by M. Schieker (Munich)15 and was cultured in MEM-alpha medium (aMEM, Invitrogen Life Technologies, Darmstadt, Germany) supplemented with 10% fetal calf serum (FCS; Biochrom, Berlin, Germany). MCF-10A normal breast epithelial cells were obtained from ATCC (LGC Standards GmbH, Wesel, Germany) and cultured in DMEM/F12 (Gibco Life Technologies, Darmstadt, Germany) supplemented with 5% horse serum (Invitrogen), 100 mg/ml EGF (Peprotech, Hamburg, Germany), 1 mg/ml hydrocortisone, 10 mg/ml insulin and 1 mg/ml cholera toxin (all from Sigma-Aldrich, Steinheim, Germany). MCF-7 and MDA-MB231 breast carcinoma cells were obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ, Braunschweig, Germany) and were maintained in DMEM GlutaMax (Invitrogen) supplemented with 10% FCS. Primary breast cancer cells were isolated from two biopsies of invasive lobular tumors. The tissue samples were cut into pieces and digested with Liberase (Roche Diagnostics, Mannheim, Germany) overnight. After resuspension and filtration through a 70 mm mesh, cells were washed two times with PBS/2% FCS and seeded in aMEM containing 10% FCS and 1% penicillin/streptomycin. Non-adherent cells were removed after 24 hr. Human primary MSCs were isolated from bone marrow aspirates derived from healthy donors after informed consent and approval by the local ethics committee and cultured as described previously.16 For isolation of CD341 HSPCs, mobilized peripheral blood (PB) was collected from healthy donors after a 5 day treatment with 7.5 mg/kg granulocyte colony-stimulating factor (G-CSF). HSPCs were purified using CD34 antibody conjugated magnetic beads according to the manufacturer’s instructions (Miltenyi Biotec). The purities were > 95% as assessed by flow cytometry and the vitality as measured by trypan blue exclusion was > 96%. Patient samples
PB samples of patients with breast cancer (n 5 78) and an age-matched healthy control group (n 5 22) were collected after informed written consent and histologically proven cancer diagnosis. The characteristics of both cohorts are provided in the Supporting Information. The study had been approved by the local institutional review board. Colonyforming units were quantified after 14 days of standardized
Cancer Cell Biology
subsequent formation of visceral and skeletal metastases.2,7 Several characteristics of breast cancer cells allow them to preferentially home to the bone marrow, for example, chemokine receptors, integrins, matrix metalloproteinases and several other tumor-secreted factors. In the context of hematopoietic stem cell homing, the CXC chemokine CXCL12 (SDF-1) and its receptors CXCR4 and CXCR7 are critical molecular determinants and perhaps the best characterized molecules for this regulation.8,9 Osteoblasts and perivascular stromal cells appear to be a major source of CXCL12, where synthesis is regulated by several inflammatory stimuli, DNA-damaging agents and hormones.9 Moreover, expression of CXCL12 in nestin-positive mesenchymal stromal cells (MSCs) is regulated in a circadian rhythm.10 CXCL12 expression can also be modulated by local production of Transforming Growth Factor beta1 (TGFb1) and this may play an important role in regulating the balance between proliferation and differentiation of HSPCs.11 Only recently, Pillai et al. have suggested, that CXCL12 can also be regulated by microRNAs (miRNA) in human MSCs.12 Metastatic breast carcinoma cells may use similar pathways as HSPCs to colonize the bone marrow and thus compete for their natural habitats.13 In prostate cancer, it was demonstrated that CXCL12 signaling through CXCR4 may trigger the dissemination of tumor cells by activating avb3 integrins and CD164 expression.14 CXCR4 signaling participates in both metastasis and the angiogenesis process of prostate cancer by downregulating the expression and secretion of the glycolytic enzyme phosphoglycerate kinase 1 and angiostatin.14 By means of cell-cell interaction networks, we have already shown that paracrine signaling between those tumor cells and MSCs may interfere with hematopoiesis (unpublished observation). In this study, we tested the hypothesis whether tumor cell induced changes of MSC characteristics have an impact on the HSPC support. We demonstrate a downregulation of MSC derived CXCL12 which correlates with decreased HSPC migratory potential. Further relevant findings of possible upstream regulatory mechanisms were the downregulation of SP1 transcription factors and increased miR-23a levels in MSCs after contact with tumor cell medium as well as an enhanced TGFb1 expression. The in vitro data are supported by our finding that breast cancer patients display elevated levels of circulating progenitor cells and a significantly higher number of colony forming units of the granulocytic/monocytic lineage (CFU-GM) compared with the age-matched healthy control group.
46
Modulation of MSCs by breast carcinoma cells
a pore size of 5 mm (Corning Costar Corporation, Cambridge, MA). We further investigated the dependence of migration against test media on the CXCL12/CXCR4 axis by initially adding 1 mmol/l of the bicyclam AMD3100, a highly specific CXCR4 antagonist (Sigma-Aldrich). CXCL12 ELISA
To measure CXCL12 secreted by MSCs either cultured with tumor cells directly or with conditioned medium, an antibody sandwich ELISA (R&D Systems, Wiesbaden, Germany) was used following the manufacturer’s instructions. Gene expression analysis
Cancer Cell Biology
SCP-1 cells were incubated with MCF-7 conditioned or normal culture medium for 24, 48 and 72 hr. Afterwards, cells were lyzed in Trizol reagent and shipped on dry ice to Miltenyi Biotec for one color Agilent Whole Human Genome Oligo Microarrays. Samples were pooled from three independent experiments. Figure 1. Gene expression analysis of SCP-1 cells treated with MCF-7 conditioned medium. Scatter plot of signal intensities of all spots. As an example, the data of 1 out of 3 independent array experiments are shown (experiment no. 1). The signal intensities of each feature represented by a dot are shown in double logarithmic scale. X-axis: control-log signal intensity; y-axis: sample-log signal intensity. Red diagonal lines define the areas of 2-fold differential signal intensities. Blue cross: unchanged genes, n 5 36,576. Red cross: significantly upregulated genes, n 5 2,060 (p-value < 0.01). Green cross: significantly downregulated genes, n 5 2,353 (p-value < 0.01). [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]
culture of PB mononuclear cells in semisolid medium containing recombinant growth factors.
Statistical analysis
All results of in vitro experiments are presented as means 6 standard error of the mean. Data were analyzed using two-way ANOVA for each time point or condition with a Bonferroni post test using GraphPad Prism 5.0. Statistical significance was established as p < 0.05 (*), p < 0.01 (**) or p < 0.001 (***). Univariate and multivariate analyses were used for analyzing the influence of tumor stage, age and body mass index on the frequency of circulating progenitor cells in patients. Experiments were done in triplicate and all results were expressed as the mean 6 SEM. A detailed description of all additional experimental procedures can be found in the Supporting Information.
Coculture models
To evaluate the expression of cytokines in direct coculture, MSCs were seeded at 5 3 103 cells/cm2 and grown for 3 days to 95% confluency. Thereafter, a total media exchange was combined with addition of tumor cells at 2.5 3 103 cells/cm2. At determined time points, media were collected and stored at 280 C until evaluation of cytokine content. To discriminate the effects of carcinoma cells on MSCs that are strictly based on soluble factors from effects that are cellcontact-dependent, indirect coculture models with conditioned media were applied. For preparation of conditioned medium from tumor cells or MCF-10A cells, 2 3 106 cells were seeded in 20 ml of serum-free aMEM into a T-75 tissue culture flask. After 72 hr, conditioned media were collected and stored at 280 C until the day of use. After thawing and prior to use on MSCs, 10% FCS was added to both conditioned media and fresh aMEM, which served as a control. Transwell chemotaxis assay
The migration of HSPCs to selected media samples was evaluated using a polycarbonate membrane transwell system with
Results MCF-7 conditioned medium altered MSC gene expression profile
To test the effects of cancer cells on bone marrow stroma we cultured MSCs with conditioned medium from the breast cancer line MCF-7 and analyzed their expression profile by microarrays. The gene expression analysis of SCP-1 cells after 24, 48 and 72 hr revealed 2353 downregulated and 2060 upregulated genes (p < 0.01, Fig. 1) belonging to diverse cellular pathways, for example, proliferation, differentiation or cell cycle. Furthermore, the number of modulated genes increased from day 1 to day 3. The complete data sets have been deposited in NCBI’s Gene Expression Omnibus and are accessible through GEO Series accession number GSE49858 (http://www.ncbi.nlm.nih. gov/geo/query/acc.cgi?acc5GSE49858). Results for some of the differentially expressed genes and pathways are shown in Supporting Information Tables S1 and S2. One of the significantly downregulated genes at all three time points was CXCL12. Since CXCL12 is an C 2014 UICC Int. J. Cancer: 136, 44–54 (2015) V
47
Wobus et al.
Tumor cell medium decreased CXCL12 promoter activity and mRNA expression levels
To gain insights into the molecular control of CXCL12 expression and its modulation by tumor cells, we investigated both CXCL12 promoter activity using SCP-1 reporter cell lines and mRNA expression in primary MSCs by real-time PCR. Conditioned media from the tumor cell lines or MCF-10A cells were incubated with CXCL12 reporter SCP-1 cells for 6, 12 and 24 hr. As shown in Figure 2a, MDA-MB231 conditioned medium caused a significant decrease in CXCL12 promoter activity to 47% at all three time points. Furthermore, incubation with MCF-7 conditioned medium resulted in lowered activities of the CXCL12 promoter to about 77%. In contrast, the supernatant from non-malignant MCF-10A cells had no significant influence on promoter activity (Fig. 2a). The mRNA expression levels of CXCL12 mirror these effects. Primary MSCs were incubated with conditioned tumor cell medium, medium from nonmalignant MCF-10A cells or with normal culture medium as a control for 24, 48 or 72 hr. At all three time points, MCF-7 medium caused a significant decrease of CXCL12 expression in MSCs to about 50% and MDA-MB231 medium had a similar effect on day 2 and 3 (Fig. 2b). In contrast, MCF-10A conditioned medium strongly increased CXCL12 mRNA expression at day 3 (Fig. 2b). These results suggest a CXCL12 regulation in MSCs by tumor cell conditioned medium at the transcriptional level. CXCL12 secretion by MSCs was decreased by tumor cells
Whether MSCs change the secretion of CXCL12 after contact to tumor cells or their conditioned medium, respectively, was investigated by ELISA at specified time points. CXCL12 is not present in supernatants from MCF-10A, MCF-7 or MDA-MB231 in single culture. Conditioned media used for indirect coculture as well as control medium did not contain significant amounts of CXCL12 (data not shown). In direct cocultures with MCF-7, CXCL12 levels decreased continuously when compared with control on day 1 (289.6 vs. 324.8 pg/ml), day 3 (204.9 vs. 253.2 pg/ml) and day 5 (193.0 vs. 339.3 pg/ml). After complete media exchange, CXCL12 levels remained unaffected in MCF-7 cocultures while control levels markedly increased, resulting in significantly reduced CXCL12 levels after media exchange. Although direct coculture with MDA-MB231 also led to reduced CXCL12 levels (284.5; 222.6; 275.7 pg/ml), the difference to control was only significant at day 8 (Fig. 2c). In indirect cocultures, both breast cancer cell lines involved significantly reduced CXCL12 contents in supernatants of MSCs. In contrast to the results of direct cocultures, MDA-MB231 cells reduced CXCL12 secretion more strongly at day 3 and 5. After complete media exchange to control C 2014 UICC Int. J. Cancer: 136, 44–54 (2015) V
medium in all three conditions, CXCL12 levels normalized to control levels. This suggested that the downregulation of CXCL12 in MSCs in indirect cocultures with breast cancer cells is partially reversible (Fig. 2d). To investigate whether the observed effects were tumor cell specific, MSCs were incubated with conditioned medium of MCF-10A non-carcinogenic breast epithelial cells. As shown in Figure 2e, the CXCL12 levels were not affected or even increased. These results suggest a cell contact-independent downregulation of CXCL12 by breast tumor cells. Cell culture supernatants from primary tumor cell cultures were collected after 1, 3 and 5 days of coculturing MSCs with tumor conditioned medium or with tumor cells, respectively, and analyzed for CXCL12 concentration. For both conditions we could detect a reduction of CXCL12 levels to about 80 and 78%, respectively (Fig. 2f). Tumor cell-primed MSCs displayed a reduced chemoattractive potential
The functional relevance of CXCL12 downregulation in MSCs in indirect coculture conditions was evaluated by assessing the HSPC migration capacity using a Boyden Chamber assay. As expected, the test media prepared for the lower chamber of migration assay system showed a significantly lower CXCL12 content in MSCs in indirect cocultures with both breast cancer cell lines (Fig. 3a). This is more pronounced in indirect cocultures with MDA-MB231 (22% of control) than with MCF-7 (54% of control). Tumor cell conditioned medium did not change the expression levels of CXCR4 or CXCR7 on HSPCs (not shown). When compared with control, the migration of HSPCs against supernatants of MSCs in indirect cocultures with MCF-7 and MDA-MB231 was significantly reduced to 46 and 54%, respectively (Fig. 3b). This is akin to the basal migration rate without chemoattractant. Blocking of CXCL12/CXCR4 signaling by 1 mmol/l of the CXCR4 antagonist AMD3100 resulted in comparable migration rates slightly above the basal level in control and reference. Although the downregulation of CXCL12 is more prominent in indirect co-cultures with MDA-MB231, the migration of HSPCs is reduced most significantly in indirect cocultures with MCF-7. However, blocking of CXCL12/CXCR4 signaling significantly reduced the migration against supernatants from indirect cocultures with MCF-7, while AMD3100 had no effect in the case of MDA-MB231. These data suggest that the reduced expression of CXCL12 in MSCs in indirect cocultures with breast cancer cells leads to a markedly reduced chemoattraction of HSPCs by MSCs. Tumor cell medium modified the expression of SP1 transcription factor and miR-23a
To unravel mechanisms which are involved in the CXCL12 downregulation in MSCs by tumor cells, we investigated different regulatory levels. First, a possible CXCL12 regulation at the transcriptional level by the SP1 transcription factor was
Cancer Cell Biology
important niche-derived factor playing a crucial role in the regulation of HSC fate,17 we investigated its regulation by tumor cells in more detail although it was not as strongly downregulated as other genes.
Modulation of MSCs by breast carcinoma cells
Cancer Cell Biology
48
Figure 2. CXCL12 promoter activity, mRNA and protein levels were downregulated by tumor cells or conditioned medium. (a) Conditioned medium from MDA-MB231 cells triggers downregulation of CXCL12 promoter activity. CXCL12 reporter SCP-1 cells were cultured with aMEM as control, conditioned aMEM from MCF-10A, MCF-7 or MDA-MB231 cells for 6, 12 and 24 hr. At indicated time points, culture supernatants were collected for GLuc and SEAP activity measurement. GLuc activity was normalized to SEAP activity and results are presented in RLU (n 5 3 for 6 and 12 hr, n 5 6 for 24 hr). Bars represent mean 6 SEM of all independent experiments; p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***). Abbreviations: GLuc, Gaussia luciferase; SEAP, secreted alkaline phosphatase; RLU, relative luciferase units. (b) Quantitative real-time PCR for CXCL12 revealed downregulation of mRNA levels after 24, 48 and 72 hr incubation with MCF-7 or MDA-MB231 conditioned aMEM, whereas no change or a significant increase was observed after incubation with MCF-10A conditioned aMEM. Results are represented as the median of fold change compared with control cultures in aMEM which were set to 1 of three independent experiments in duplicate, p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***). (c) CXCL12 content was determined by ELISA in supernatants of direct cocultures of primary MSCs with MCF-7 or MDA-MB231 cells at indicated time points. Full media exchange to control medium was applied after sample preparation at day 5 (dashed line). Bars represent mean 6 SEM of three independent experiments, p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***). (d) CXCL12 content was determined by ELISA in supernatants of indirect cocultures of primary MSCs with tumor cell conditioned aMEM at indicated time points. Full media exchange to control medium was applied after sample preparation at day 5 (dashed line). Bars represent mean 6 SEM of three independent experiments, p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***). (e) CXCL12 content was determined by ELISA in supernatants of indirect cocultures of primary MSCs with tumor cell as well as MCF-10A conditioned aMEM after two and five days. Bars represent mean 6 SEM of three independent experiments. (f) CXCL12 production as measured by ELISA was suppressed also after incubation with conditioned medium (Tm medium) from two primary breast cancer samples and after direct contact with these primary cells (MSC/Tm cells). MSC/MSC cocultures served as control. Bars represent mean 6 SEM of two independent experiments with four time points each; p < 0.01 (**).
C 2014 UICC Int. J. Cancer: 136, 44–54 (2015) V
Figure 3. HSPC migration potential to MSC/tumor cell cocultures is impaired in a CXCL12/CXCR4-dependent manner. (a) Concentration of CXCL12 in supernatants of primary MSCs after 4 days in control conditions (“CTRL”) and indirect coculture with MCF-7 (“MCF”) or MDAMB231 (“MDA”). Data is represented as mean 6 SEM of four independent experiments, p < 0.01 (**); p < 0.001 (***). (b) Migration of HSPCs towards the above mentioned supernatants, assayed with blocking of CXCR4 by AMD3100 (AMD1) and without blocking (AMD2). Recombinant CXCL12 (rSDF) served as internal reference. Basal migration without chemoattractant is indicated with a dashed line. Data is represented as mean 6 SEM of four independent experiments, p < 0.05 (*); p < 0.01(**) when compared with CTRL. p < 0.05 (†); p < 0.01 (††) when compared with AMD1.
investigated. Primary MSCs were treated with 50 mM mithramycin A which binds GC-rich DNA sequences and thereby inhibits the activity of SP transcription factors.18 We detected a dramatically decreased CXCL12 secretion after 24 hr in all investigated MSC samples suggesting a regulation by Sp transcription factors also in human MSCs (Fig. 4a). Total cell numbers were not influenced by mithramycin (not shown). To test whether tumor cell conditioned medium can influence SP1 in MSCs, we determined the mRNA levels by realtime PCR and protein expression by Western blot. Whereas tumor cell medium caused a significant decrease of SP1 mRNA expression after 1 and 2 days, MCF-10A medium had almost no impact (Fig. 4b). At day 3, the effects were again slightly reversed. The SP1 protein levels in MSCs after incubation with tumor cell conditioned medium were determined by Western blot analysis. As demonstrated in Figure 4c, medium of both tumor cell lines caused a clear decline of the signal intensity whereas MCF-10A medium had no influence. Only recently, in another study of our group miR-23a was identified to regulate CXCL12 expression in MSCs.19 Therefore, we quantified the miR-23a levels in MSCs after incubation with tumor cell conditioned medium. Interestingly, MCF-7 medium caused a significantly upregulation of miR23a after one and two days (Fig. 4d). Additionally, MDAMB231 conditioned medium slightly increased the miR-23a expression levels, whereas MCF-10A medium had no significant impact suggesting a tumor cell specific effect.
Increased TGFb1 expression was involved in the CXCL12 regulation in MSCs
Our microarray data revealed a strong upregulation of the TGFb1 receptor pathway in SCP-1 cells after incubation with C 2014 UICC Int. J. Cancer: 136, 44–54 (2015) V
MCF-7 conditioned medium. Therefore, we tested a possible modulation of TGFb1 in MSCs after contact to tumor cell medium and an influence on CXCL12 expression. A downregulation of CXCL12 secretion by TGFb1 in primary MSCs was demonstrated by addition of the recombinant protein. A blocking anti-TGFb1 antibody was able to reverse this effect at least in part (Fig. 5a). Interestingly, the addition of this blocking anti-TGFb1 antibody to tumor cell conditioned medium in which the MSC were incubated for 24 hr was able to completely abolish the effect of decreased CXCL12 secretion (Fig. 5b). Next, we tested whether tumor cells can modulate the TGFb1 expression in MSCs. The TGFb1 mRNA levels were increased in MSCs by conditioned medium of MCF-7 and MDA-MB231 cells at all three time points, whereas MCF10A conditioned medium resulted in decreased expression (Fig. 5c). The TGFb1 protein in MSCs was studied by immunofluorescence staining. Again, a clear increase in TGFb1 expression was observed when MSCs were cultured with conditioned medium of MCF-7 or MDA-MB231 cells compared to cultures in normal medium (Fig. 5d).
Patients with breast cancer have higher CFU-GM numbers
As a reduced production of CXCL12 may lead to a perturbed balance between circulating and resident HSPCs, we expected a higher frequency of clonogenic HSPCs in the PB of patients with breast cancer. Indeed, as shown in Figure 6, a significantly higher number of CFU-GM (1363) compared with age-matched healthy control group (5 6 2; p < 0.05) was detected. Moreover, the stage of tumor disease did not affect the number of CFU-GM or circulating progenitor cells (data not shown). This effect could be confirmed when body-mass
Cancer Cell Biology
49
Wobus et al.
50
index was included in a multivariate model as confounding variable, but not the age.
Discussion
Cancer Cell Biology
Disseminated tumor cells can be detected in various malignancies and tend to home to the bone marrow. Several of these neoplasms have a striking tendency to subsequently metastasize to bone. The chemokine CXCL12 is expressed by
Modulation of MSCs by breast carcinoma cells
bone marrow stromal cells as well as osteoblast precursors and plays key roles in the migration, adhesion and activation of HSPCs and leukemic cells.20,21 It was hypothesized that tumor cells apply the CXCL12/CXCR4 pathway as homing mechanism to a similar extent as hematopoietic cells.22 Here, we describe a novel mechanism by which breast cancer cells interfere with CXCL12 production of bone marrow-derived MSCs in vitro. This effect reduces the stromal support for CD341 HSPCs and involves regulation by SP1 transcription factors, miR23a as well as by TGFb1. A significantly higher frequency of circulating colony-forming hematopoietic progenitors in patients with breast cancer compared to healthy controls may also result from this interference. Breast tumor cells or their conditioned medium caused a downregulation of CXCL12 mRNA and protein levels in MSCs. In contrast, MCF-10A non-tumorigenic breast epithelial cells did not influence the CXCL12 expression in MSCs or even increased it. The effect of tumor cell mediated downregulation of MSC-derived CXCL12 secretion is cell contact independent and partially reversible. Different kinetic properties of the coculture models, especially with respect to the concentration of paracrine signals, could explain for the different magnitudes of decreased CXCL12 levels. While they were reduced to 80% in direct coculture during a period of 5 days, respective CXCL12 levels were reduced to
