Bone Marrow-Derived Mesenchymal Stromal Cells Promote Survival and Drug Resistance in Tumor Cells

Abstract

Bone marrow mesenchymal stromal cells (BMMSC) have anti-tumorigenic activities. Here we hypothesized that circulating BMMSC are incorporated into tumors and protect tumor cells from therapy-induced apoptosis. Adherent cells harvested from murine bone marrow and expressing phenotypic and functional characteristics of BMMSC were tested for their anti-tumor activity against murine 4T1 mammary adenocarcinoma and LL/2 Lewis lung carcinoma cells. BMMSC but not NIH3T3 or murine skin fibroblasts stimulated the expansion of 4T1 cells in 3D co-cultures, and conditioned medium from these cells increased the viability of 4T1 and LL/2 cells in 2D cultures. 4T1 cells exposed to BMMSC conditioned medium exhibited a 2-fold reduction in apoptosis under low serum concentrations (0.5 to 1%). Furthermore, exposure of 4T1 and LL/2 cells to BMMSC conditioned medium increased their viability in the presence of paclitaxel or doxorubicin at therapeutic concentrations. This effect was accompanied by reductions in caspase-3 activity and Annexin V expression. When co-injected with 4T1 cells in the mammary fat pad of mice subsequently treated with doxorubicin, BMMSC (and not fibroblasts) also inhibited drug-induced apoptosis in tumor cells by 44 percent. We demonstrated that BMMSC were attracted by 4T1 and LL/2 cells but not by NIH3T3 cells in vitro and that when injected intravenously in 4T1 tumor bearing mice, these cells (and not NIH 3T3) were specifically detected in tumors within 12 to 18 days where they preferentially localized at the invasive front. Overall, our data identify BMMSC as an important mediator of tumor cell survival and treatment resistance in primary tumors.

Introduction

The process of tumorigenesis, previously thought to stem primarily from genetic changes within transformed cells, is now known to also depend on extracellular signals from non-cancerous cells present in the tumor microenvironment (TME) (1). Among the normal cells that contribute to the TME are bone marrow-derived myeloid cells whose function and contribution to neoplastic growth has been best understood (2). Recently, there has been increased evidence that bone marrow-derived mesenchymal cells also contribute to tumorigenesis and metastasis, but their role has been less well characterized (3, 4). Within the bone marrow microenvironment, BMMSC and osteoblasts provide a supportive niche for homing of hematopoietic stem cells (HSC) and tumor cells which promotes quiescence and survival (5). These cells also serve to generate the various connective tissue lineages (adipocyte, osteoblast, chondrocyte, myocyte, neuron, fibroblast) found in the marrow and in distal organs, and are actively recruited to sites of injury, inflammation and neoplastic transformation where it has been suggested they give rise to fibroblasts (6).

The kinetics of BMMSC recruitment from the marrow niche, as well as their distribution in primary tumors, are still poorly characterized. BMMSC display a chemotactic response to chemokines, growth factors, and extracellular matrix (ECM) proteases produced by tumor cells, which mimics the activity of inflammatory HSC and is mechanistically indistinguishable from the generalized wound healing response (7). Though BMMSC exhibit potent effects even in small numbers, studies of tumor-recruited BMMSC suggest that their migration is inefficient and difficult to quantify in vivo (8).

However, once recruited to tumor sites BMMSC differentiate into myofibroblasts (9) as well as tumor-associated fibroblasts (TAF), which produce mitogenic and angiogenic factors and display potent ECM remodeling capabilities (10). Cytokines secreted by BMMSC are also known to modulate immune responses within the TME, creating immunosuppressive effects which drive tumor progression (11). Concordantly, introduction of BMMSC into tumor bearing mice by intravenous injection or co-injection shows a net positive effect on tumor growth in a majority of studies (12, 13). However, anti-tumorigenic effects, driven by increased caspase-3 and PARP-1 cleavage, have also been reported (14). Most published work on the MSC-tumor interaction has focused on proliferative, angiogenic and immunoregulatory effects. Previous studies conducted in our laboratory have identified a pro-survival effect of human BMMSC on metastatic human neuroblastoma cells in the bone marrow microenvironment that promotes drug resistance (15, 16). This observation provides the basis for our present examination of a novel role of these mesenchymal cells and their derivatives within primary tumors, rather than the bone marrow.

We hypothesized that circulating BMMSC are incorporated into primary tumor sites and protect tumor cells from spontaneous and therapy-induced apoptosis via the production of soluble factors, similar to the role of native BMMSC in promoting metastatic tumor cell survival in the bone marrow microenvironment.

Material and Methods

Cells

The murine cell lines 4T1 mammary carcinoma, LL/2 Lewis lung carcinoma and NIH3T3 fibroblasts were purchased from ATCC (American Type Culture Collection), which uses short terminal repeat (STR) profiling for characterization. All cells were passaged for less than 6 months after resuscitation. Cells were cultured in DMEM (Dulbecco’s Modified Eagle Medium) or RPMI-1640 (4T1 cells) containing 10% fetal calf serum (FCS) and supplemented with 1% penicillin-streptomycin. Normal murine fibroblasts were obtained from skin samples from 6–8 week-old Balb/cJ mice (Jackson Laboratories). Four mm² fragments were placed in a 6 cm culture dish (3 sections per dish) and covered with 100 μL DMEM containing 10% FCS. Skin fragments were removed from the culture dish when adherent colonies of growing cells could be identified. These colonies of fibroblast cells were allowed to expand to 70% confluence before being harvested by trypsinization and transferred to 10 cm culture dishes for routine passaging. Murine BMMSC were obtained from 6–8 week-old Balb/cJ mice using a protocol adapted from Kirshner, et al. (17). In brief, animals were sacrificed and femurs, iliac bones and tibias were used. The extremities of each bone were removed with scissors and the marrow cavity was flushed a first time with phosphate buffered saline (PBS) using a 27-gauge needle to remove non-adherent hematopoietic cells, and the material was discarded. Each bone was then flushed a second time with PBS, using the needle to simultaneously scrape the endosteal surface of the marrow cavity. Flow through material from the second flush was collected, subjected to hemolysis and spun to remove debris. Pooled bone marrow cells from each animal were plated in BMMSC culture medium (Iscove’s Modified Dulbecco Medium IMDM) containing 15% FCS and 15% denatured horse serum (DHS) supplemented with hydrocortisone (10⁻⁶ mol/L), 2-mercaptoethanol (10⁻⁴ mol/L), and 1% penicillin-streptomycin in one well on a 12-well culture dish coated with collagen I and fibronectin, both at 5 μg/cm². Bone marrow cells were cultured for 24 hours and non-adherent cells removed by gentle washes two times. Adherent cells were grown for 3–6 weeks to allow the development of a confluent stromal monolayer and then harvested for routine passaging in regular plastic tissue culture dishes.

Reagents: Anti-mouse CD34-PE, CD45-PerCP, CD31-PE, VEGFR2-PE, and CD44-FITC were purchased from BD (Becton Dickinson). Anti-mouse E-Cadherin antibody was purchased from R&D Systems. Paclitaxel was purchased from Sigma, and doxorubicin from GensiaSicor.

Fluorescent cell labeling

BMMSC were labeled for detection by fluorescence microscopy using the PKH26 Red Fluorescent Cell Linker Kit (Sigma). Briefly, 10⁷ BMMSC were rinsed with PBS, loosely pelleted, and suspended in 1 mL PBS for staining with 8 μL red fluorescent dye. Cells were stained for 90 seconds, followed by treatment with 2 mL BMMSC culture medium and an additional 60 seconds of staining. Cells were pelleted and rinsed three times with culture medium, with a fresh tube used for each rinse. Cells were rinsed two times with PBS, plated on a 10 cm dish in culture medium, incubated for 24 hours at 37°C, and harvested for use.

Flow cytometry

BMMSC were harvested with PBS-based cell dissociation buffer (Gibco) and suspended in PBS containing 5% FCS (v/v) and 0.1% sodium azide (w/v). Analysis was performed using a FACS Caliber flow cytometer and data were analyzed using CellQuestPro software.

Osteocyte/adipocyte staining

For analysis of spontaneous differentiation, BMMSC were plated in 48-well dishes, allowed to expand for 14 days, and fixed in 4% paraformaldehyde (PFA) prior to staining for either alkaline phosphatase activity (AP), using a detection kit (Millipore), or for triglyceride deposits, using oil red O (ORO; Sigma). For directed osteocyte differentiation, BMMSC were treated with osteocyte differentiation medium (StemPro osteocyte/chondrocyte differentiation medium from Gibco) for 21 days, fixed in 4% PFA (w/v), and stained with alizarin red (Sigma) to detect calcium deposits. For directed adipocyte differentiation, BMMSC were plated in 12-well dishes treated with adipocyte maintenance medium (DMEM 15% FCS, 20 mM HEPES, 1 mM sodium pyruvate, 2 mM glutamase-1, and 1% penicillin-streptomycin) up to 100% confluency, switched to adipocyte induction medium for 2 days (150 mM insulin, 125 mM dexamethasone, 500 mM IBMX), fixed in 4% PFA, and stained with ORO.

BMMSC conditioned medium

Conditioned medium (CM) from BMMSC was obtained from BMMSC plated in 10 cm dishes and grown to a confluent monolayer. Cells were rinsed with PBS and treated with two changes of BMMSC culture medium containing 0.5% FCS (v/v) and 0.5% DHS (v/v) (1% total serum), 0.25% FCS (v/v) and 0.25% DHS (v/v) (0.5% total serum), or no serum for 30 minutes each. Medium was replaced with 10 mL culture medium containing 0.5% total serum, 1% total serum, or no serum. After 3–4 days exposure to BMMSC, the CM was collected and spun to remove cellular debris and stored at 4°C prior to testing. CM was also generated from NIH3T3 cells and normal murine fibroblasts by the same method.

Cell proliferation and apoptosis assays

For 2D proliferation assays, tumor cells were cultured in 6-well plates at 5×10⁴ per well in the presence of CM from MSC, NIH3T3, skin fibroblasts or IMDM, all containing 1% serum (v/v). Cells were harvested by trypsinization and counted in a hemocytometer. For 2D transwell proliferation assays, 4T1 tumor cells were cultured in 6-well plates at 5×10⁴ per well and exposed to 5×10⁴ MSC, NIH3T3 cells or normal fibroblasts plated in 0.4 μm pore size transwell inserts, or to IMDM containing 1% serum (v/v). Cells were harvested by trypsinization and counted in a hemocytometer. For 3D transwell proliferation assays, 4T1 cells were suspended in 4.5 mg/mL high growth factor Matrigel (BD Biosciences) and plated on 24-well plates at 1×10⁴ per well. After polymerization, 0.4 μm pore size transwell inserts containing 1×10⁴ BMMSC, NIH3T3 cells, or normal fibroblasts in serum-free medium were added to the wells. The total number of tumorspheres (>50 μm diameter) in each well was counted after 21 days. For 3D proliferation assays, 4T1 cells suspended in 4.5 mg/mL high growth factor Matrigel were plated on 24-well plates at 1×10⁴ per well. After polymerization, serum-free conditioned medium from BMMSC, NIH3T3 cells or normal skin fibroblasts was added to the wells. The total number of tumorspheres (>50 μm diameter) in each well was counted after 21 days. For apoptosis assays, cells were cultured in 6-well plates at 2.5×10⁴ (4T1) or 1×10⁵ (LL/2) per well in the presence of CM or control medium containing 1% serum (v/v). Cells were harvested by trypsinization, rinsed twice with PBS, and suspended in Annexin V binding buffer. Annexin V and PI staining were performed as per the manufacturer’s instructions (Annexin V-FITC apoptosis detection kit I, BD Biosciences). Apoptotic cells were defined as Annexin V-FITC positive. Caspase-3, caspase-9, and caspase-8 activity was determined using the appropriate ApoTarget colorimetric protease assay kit (Life Technologies) on aliquots containing 50 μg protein.

Drug resistance assay

Tumor cells were plated in opaque white 96-well culture plates at 1×10⁴ per well and treated with paclitaxel or doxorubicin for 72 hours (LL/2) or 96 hours (4T1) in the presence of 1% serum (LL/2) or 0.5% serum (4T1) CM or BMMSC conditioned medium. CellTiter-Glo reagent (Promega) was then added to each well at a 1:1 ratio with the cell medium, and luminescent reporter activity was measured via GloMax Multi Detection System (Promega), using Instinct software (Promega). For determination of caspase-3/7 activity, tumor cells were treated with paclitaxel or doxorubicin for 24 hours in the presence of 1% serum (LL/2) or 0.5% serum (4T1) CM or BMMSC culture medium. Caspase-Glo 3/7 reagent (Promega) was then added to each well at a 1:1 ratio with the cell medium, and luminescent reporter activity was measured as above described. For apoptosis induction, 4T1 cells were cultured in 6-well plates at 3×10⁵ per well and treated with 0.5 μmol/L paclitaxel or doxorubicin for 16–24 hours in the presence of 0.5% serum (v/v) CM or 0.5% serum (v/v) control medium. Cells were harvested by trypsinization, rinsed twice with PBS, and suspended in Annexin V binding buffer. Annexin V and PI staining were performed as per the manufacturer’s instructions (Annexin V-FITC apoptosis detection kit I, BD Biosciences).

Transwell migration assay

For transwell migration studies, 8×10³ BMMSC were plated in 24-well, 8.0 μm pore size transwell inserts and exposed to serum-free tumor cell CM or control medium added to the lower chamber of the transwell dish. After migration, transwell inserts were removed and the upper side of the filter was swabbed to remove non-migratory cells. Transwell filters were then fixed/stained with the Hema3 Stat Pack (Protocol) as per the manufacturer’s instructions. Filters were cut free and mounted on slides in a drop of immersion oil for microscopic examination.

BMMSC iron labeling

BMMSC were grown in 10 cm dishes to form a low-density monolayer (1–2×10⁶ cells per dish). Cells were rinsed with PBS and treated with 6.5 mL serum-free IMDM supplemented with 2U/mL heparin (Sigma). Within 15 seconds, 60 μg/mL protamine sulfate (APP Pharmaceuticals) was added to medium, with gentle agitation of the dish. Within another 15 seconds, 100 μg/mL Feraheme (AMAG Pharmaceuticals) was added to medium, with simultaneous agitation of the dish. Cells were incubated for 2 hours at 37°C and 5% CO2, followed by addition of 6.5 mL BMMSC culture medium. After 24 hours, cells were rinsed with PBS, followed by a second rinse with PBS containing 10U/mL heparin, followed by a third rinse with PBS. Cells were harvested by trypsinization for use in vivo.

Animal experiments

Studies in mice were performed under a protocol approved by the Institutional Animal Care and Use Committee at Children’s Hospital Los Angeles (Protocol 41-11). For tumor recruitment studies, Balb/cJ mice were injected s.c. with 6×10⁵ 4T1 cells in the left flank. On day 12 of tumor growth, mice received ~2×10⁶ PKH26-labeled BMMSC via tail vein. Forty-eight hours after BMMSC injection, mice were sacrificed and tumors extracted for cryopreservation. For early-stage tumor recruitment studies, Balb/cJ mice were injected s.c. with 5×10⁵ 4T1 cells in the left flank. Four days after tumor cell injection, mice received 3×10⁵ iron-loaded or control BMMSC via tail vein. Six days after BMMSC injection, mice were sacrificed and tumors extracted for formalin fixation and paraffin embedding. For in vivo/ex vivo bioluminescence tracking studies, Balb/cJ mice were injected s.c. with 2×10⁶ 4T1 cells in the left flank. On day 2 after injection, mice received ~2×10⁶ luciferase-positive BMMSC or luciferase-positive NIH3T3 cells by retro-orbital injection. Bioluminescent signal data was collected from mice at regular intervals by Xenogen imaging (Caliper), performed 15 minutes after i.p. injection of luciferin (1.5 mg/mouse) starting at 30 minutes after BMMSC/NIH3T3 implantation. On day 18 after BMMSC/NIH3T3 injection, mice were sacrificed and tumors and secondary organs extracted. Approximately 100 mg of tissue from each organ was suspended in lysis buffer and homogenized. Additionally, total bone marrow was collected from the left femur by flushing the marrow cavity with 1 mL lysis buffer. Flow through was collected and homogenized by vortexing. Tissue/bone marrow lysates were transferred to 96-well plates at 100 μL/well and treated with 5 μL/well luciferin at 2 mg/mL, and luminescent reporter activity was measured via GloMax Multi Detection System (Promega) using Instinct software (Promega). For drug resistance studies, Balb/cJ mice were injected with 1×10⁶ 4T1 cells or 1×10⁶ 4T1 cells plus 2×10⁵ normal mouse fibroblasts (5:1 ratio) in the right 4^th mammary fat pad, along with 1×10⁶ 4T1 cells plus 2×10⁵ BMMSC (5:1 ratio) in the contralateral fat pad. On days 3 and 4 after tumor cell injection, mice received doxorubicin (5 mg/kg) by i.p. injection. On day 5 after tumor cell injection, mice were sacrificed and tumors extracted for cryopreservation.

Histology

Prussian blue staining was carried out on formalin-fixed, paraffin-embedded (FFPE) sections (6 μm) of 4T1 tumors from mice injected with iron-loaded and control BMMSC. Sections were deparaffinized, rehydrated, and treated with a solution of potassium hexacyanoferrate (II) trihydrate (Sigma) in HCl for 20 minutes. Slides were counterstained with nuclear fast red (Sigma), dehydrated, cleared in xylene, and mounted in Cytoseal. Frozen sections (6 μm) of 4T1 tumors from mice injected with PKH26-labeled BMMSC were mounted in Vectashield medium for fluorescence with DAPI (Vector Laboratories) and examined for red fluorescent signal. Frozen sections of orthotopic 4T1 tumors from mice co-injected with BMMSC or normal mouse skin fibroblasts were stained for E-Cadherin-positive cells and for apoptotic cells using the Cell Death Detection Kit (Roche), as previously reported (18).

Statistical analysis

Comparisons between two groups were performed by unpaired Student t test with unequal variances assumed, while multiple group comparisons were performed by 2-way ANOVA or linear regression analysis. T tests and ANOVA were performed using Excel (Microsoft), and linear regression analysis was performed using Stata 11 software (StataCorp). Two sided p values are reported.

Results

BMMSC characterization

Murine MSC harvested from the bone marrow and expanded in vitro were tested for their stem cell potential by a combination of surface marker staining and functional studies (Table 1; Fig. 1). FACS analysis of BMMSC revealed positive staining for CD44, a surface marker that is highly enriched on mesenchymal cells at both early (<10) and late (10–20) passages (19) (Table 1). The cells were negative for the endothelial markers CD31 and VEGFR2 at early and late passages, along with the myeloid marker CD45. Weak positive staining for the hematopoietic marker CD34 was observed at early passages, potentially due to the presence of some myeloid cells that were not eliminated in the washing procedure. Late passage BMMSC were negative for CD34, which parallels the loss of contaminating cell types over subsequent passages (Fig. 1A). The tested cells thus displayed characteristic BMMSC markers by cell surface phenotype.

Table 1. Characterization of BMMSC surface marker expression.

FACS analysis of BMMSC surface markers in cultures of primary murine stromal cells at early (≤10) and late (>10) passages. Data are displayed as the percentage of cells positive for indicated markers.

Marker	Early Passage (≤10)	Late Passage (>10)
CD44	99.93%	99.88%
CD34	10.83%	1.40%
CD45	1.12%	0.44%
CD31	0.52%	0.68%
VEGFR2	0.66%	0.79%

Figure 1.

Primary murine bone marrow stromal isolates display mesenchymal cell markers. A, FACS analysis of BMMSC surface markers. Representative histograms for early passages (<10) shown (open histogram represents the autofluorescent control). B, spontaneous osteoblast/adipocyte differentiation of BMMSC in culture. Upper panel: light micrographs of early passage BMMSC cultured for 2–3 weeks and stained for alkaline phosphatase (AP) activity (osteoblasts) or lipids by Oil Red O (ORO, adipocytes). Scale bar = 250 μm (left) and 100 μm (right). Lower panel: percent of BMMSC with positive staining for AP and ORO over passages in culture. The data represent the mean ±SD of positive cells counted in five 40x fields. C, mature osteoblast differentiation of BMMSC cultured 3 weeks in osteogenesis medium vs. control medium and stained with Alizarin red. D, early adipocyte differentiation of BMMSC cultured for 2 weeks in maintenance medium or switched to adipogenic medium after 1 week, and stained for the presence of lipids with ORO (Scale bar = 100 μm).

To establish whether these cells also retained their pluripotent function, they were cultured for an extended period and tested for spontaneous generation of both adipocytes and osteoblasts. Analysis of a single BMMSC culture over multiple passages showed relatively stable levels of adipogenesis (2–8%), while early osteoblastic differentiation potential (around 2%) decreased at around passage 8 (Fig. 1B). Late passage BMMSC were able to generate mature osteocytes, characterized by alizarin red-positive calcium deposits, when cultured in osteoblast induction medium (Fig. 1C), and cultures of BMMSC treated with adipogenesis induction medium showed an increased percentage of ORO-positive cells versus controls, which generated smaller numbers of more mature adipocytes (Fig. 1D). The observed spontaneous and inducible differentiation of osteoblasts and adipocytes in these experiments confirmed the identity of the tested cells as BMMSC.

BMMSC enhance tumor cell expansion

In order to determine the effect of BMMSC on tumor cell expansion, 4T1 murine mammary adenocarcinoma cells were first grown in 3D cultures of Matrigel in the presence of BMMSC, NIH3T3 cells, or normal mouse fibroblasts plated in transwells. This experiment revealed that BMMSC stimulated the formation of a number (46 ± 12 per well) of tumorspheres whereas 4T1 cells failed to form tumorspheres in the presence of NIH3T3, fibroblasts, or controls (p < 0.02) (Fig. 2A). This stimulatory effect on tumorspheres required the presence (although not the contact) of BMMSC because Matrigel-embedded 4T1 cells grown in the presence of serum-free CM from BMMSC, NIH3T3, fibroblasts or serum-free control medium did not form any discernible tumorspheres (data not shown). Using 2D cultures of 4T1 grown in the presence of 1% serum CM from BMMSC, NIH3T3 and fibroblasts, we then observed a 1.5-fold increase in cell numbers in cultures exposed to BMMSC CM but not to CM from NIH3T3 or fibroblasts (p = <0.001 by ANOVA) (Fig. 2B). A similar increase in cell number in the presence of BMMSC but not NIH3T3 or skin fibroblasts was observed when 4T1 cells were co-cultured in transwells (p = <0.001 by ANOVA) (Fig. 2B). Similarly, LL/2 cells exposed to 1% serum CM from BMMSC showed a significantly higher (2 fold) increase in the number of viable cells (Supplementary Fig. S1). The data thus indicated that BMMSC increased the viability of tumor cells by a mechanism that did not require direct cell-cell contact and was primarily mediated by soluble factors present in the CM. To determine if this increase in tumor cell expansion stemmed from enhanced proliferative potential vs. changes in spontaneous apoptosis, 4T1 cells cultured in the presence or absence of BMMSC CM were examined for proliferation by cell cycle analysis. The data revealed that 4T1 cells grown in the presence of 1% serum CM from BMMSC showed no significant increase in the percentage of cells in S-phase or G2/M-phase at 2 days, but a significant decrease in the percentage of cells in S-phase and G2/M-phase at 4 days, suggesting in fact an inhibitory effect on proliferation (Supplementary Fig. S2A). Using BrdU labeling, we did not find any significant difference in the percentage of cells in G2 phase in the presence of BMMSC CM versus regular medium (Supplementary Fig. S2B). These data thus suggested that the increase in viability observed when tumor cells were cultured in the presence of BMMSC CM was not due to a positive effect on proliferation.

Figure 2.

BMMSC enhance tumor cell growth by suppressing spontaneous apoptosis. A, Matrigel-embedded 4T1 cells cultured for 21 days in the presence or absence of BMMSC, NIH3T3 cells, or normal mouse skin fibroblasts in a transwell chamber were examined for the formation of tumorspheres. The data represent the mean ±SD number of 4T1 tumorspheres in quadruplicate wells from one of two experiments showing similar results. Inset: optical light micrographs of 4T1 spheroids in the central field of representative control and co-culture wells. Scale bar = 500 μm. B, 4T1 cells were cultured in 2D for 4 days in the presence or absence of transwells containing BMMSC, NIH3T3, or fibroblasts (left panel) or in the presence or absence of CM from BMMSC, NIH3T3, or fibroblasts (right panel). The data represent the mean ±SD of viable cells at indicated times from triplicate wells. C, 4T1 cells were cultured in 2D for 5 days in the presence or absence of CM from BMMSC or NIH3T3 cells. Top panel: representative analysis of apoptosis by flow cytometry at day 5. Lower panel: the data represent the mean ±SD of Annexin V-FITC positive cells at indicated times in triplicate wells from one of two experiments showing similar results. D, 4T1 cells were cultured for 5 days in the presence or absence of CM from BMMSC and examined for caspase activity at the indicated times. The data represent the mean ±SD of caspase-3 (top), caspase-9 (middle), and caspase-8 (bottom) activity at day 3, 4, and 5 from triplicate wells from one of two experiments showing similar results.

BMMSC protect tumor cells from spontaneous apoptosis

We therefore explored whether BMMSC could affect tumor cell survival. 4T1 grown in 1% serum CM or control medium for 3–5 days were stained with Annexin V-FITC and PI and examined by flow cytometry (Fig. 2C). We observed significant reductions in the percentage of Annexin V(+) cells in cultures exposed to BMMSC CM for 3 (16.8% CM vs. 24.8% control; p = 0.012), 4 (21.8% CM vs. 46.1% control; p = 0.002), and 5 days (7.4% CM vs. 19.1% control; p = 0.0004). This effect was specific to BMMSC since we did not observe a statistically significant decrease in spontaneous apoptosis when cells were cultured in 1% serum CM from NIH3T3 cells (16.4% NIH3T3 CM vs. 19.1% control; p=0.21). Experiments utilizing LL/2 cells revealed a similar reduction in the percentage of Annexin V(+) cells in the CM-treated samples versus controls that reached statistical significance by day 5 (51.8% CM vs. 63.6% control; p = 0.033 (Supplementary Fig. S3A). Consistent with a protective effect of BMMSC CM on apoptosis, an analysis of cell lysates from 4T1 cells indicated a statistically significant decrease in the level of caspase-3 activity in cells exposed to CM from BMMSC for 5 days, compared to controls (44.5% of control) (Fig. 2D). Similar data were observed with LL/2 cells (Supplementary Fig. S3B). The data thus suggest that the observed increase in viability in tumor cells cultured in the presence of BMMSC was primarily the result of a protective effect against apoptosis induced by serum starving. In order to determine if this protective effect stemmed from a suppression of intrinsic or extrinsic apoptotic signaling, we examined the effect of CM from BMMSC on caspase-9 and -8 activity. These data revealed a statistically significant decrease in caspase-9 activity (44.2% of control) and a non-significant inhibition of caspase-8 activity (25% of control) in 4T1 cells cultured in the presence of CM from BMMSC at day 5, consistent with a primary protective effect on intrinsic apoptosis (Fig. 2D). The data thus indicated that CM of BMMSC increased the viability of tumor cells primarily by having a protective effect on intrinsic apoptosis.

BMMSC protect tumor cells from drug-induced apoptosis

We then asked whether BMMSC would also protect tumor cells from drug-induced apoptosis. For these experiments, 4T1 and LL/2 cells were exposed to apoptosis-inducing drugs in the presence or absence of CM containing 1% total serum (LL/2) or 0.5% total serum (4T1). 4T1 cells cultured in the presence of BMMSC-derived CM and treated with paclitaxel or doxorubicin (concentrations ranging from 0.00001 to 100 μmol/L) showed a significantly greater survival than control cells cultured in regular medium at therapeutic concentrations (Fig. 3A), as confirmed by linear regression analysis (p ≤ 0.001). Similarly, LL/2 cells treated with paclitaxel or doxorubicin at therapeutic concentrations showed a significantly greater survival when cultured in the presence of BMMSC-derived CM than in the presence of regular medium (Fig. 3B) (p ≤ 0.016). Consistent with a protective effect of BMMSC on drug-induced apoptosis, we observed a significantly lower level of caspase-3/7 activity in 4T1 cells cultured in the presence of BMMSC-derived CM than controls when exposed to paclitaxel (0.1 to 10 μmol/L) or doxorubicin (5 to 10 μmol/L) (Fig. 4A) (p ≤ 0.02). LL/2 cells treated with BMMSC-derived CM also showed significantly lower levels of caspase-3/7 activity than controls in the presence of paclitaxel (5 μmol/L) or doxorubicin (1 to 10 μmol/L) (Fig. 4B) (p ≤ 0.019). The protective effect of BMMSC-derived CM on drug-induced apoptosis was confirmed by flow cytometry analysis of 4T1 cells grown in 0.5% serum CM or control medium for 16–24 hours in the presence of 0.5 μmol/L paclitaxel or doxorubicin and stained with Annexin V-FITC and PI (Fig. 4C). This analysis indicated a significant reduction in the percentage of Annexin V(+) cells in paclitaxel-treated cultures exposed to BMMSC CM for 16 hours (11.8% CM vs. 29.7% control), although this difference was no longer significant after 24 hours of paclitaxel treatment (23.3% CM vs. 31.1% control). Doxorubicin-treated 4T1 cells exposed to BMMSC-derived CM also showed a significant reduction in Annexin V(+) cells at both 16 hours (3.6% CM vs. 12.2% control) and 24 hours (6.4% CM vs. 17.1% control). The data thus demonstrated that BMMSC had a pro-survival effect on tumor cells that not only protected them from serum starvation-induced apoptosis but also from drug-induced apoptosis.

Figure 3.

BMMSC enhance survival in drug-treated tumor cells. A, cell viability of 4T1 cells (1×10⁴) cultured in the presence or absence of 0.5% serum BMMSC-derived CM and treated with paclitaxel (upper panel) or doxorubicin (lower panel) for 96 hours. The data represent the mean ±SD percentage of surviving cells for indicated drug concentrations in triplicate wells from one of three experiments showing similar results. B, cell viability of LL/2 cells (1×10⁴) cultured and treated as indicated in A for 72 hours. The data represent the mean ±SD percentage of surviving cells for indicated drug concentrations in triplicate wells from one of two experiments showing similar results.

Figure 4.

BMMSC CM protects tumor cells from drug-induced apoptosis. A, 4T1 cells (1×10⁴) cultured in the presence or absence of 0.5% serum BMMSC CM and treated with paclitaxel (upper panel) or doxorubicin (lower panel) at indicated concentrations for 24 hours were examined for caspase-3/7 activity. The data represent the mean ±SD fold-change from 4T1 cells cultured in the presence of regular medium (controls) in triplicate wells. They are representative of two experiments showing similar results. B, LL/2 cells (1×10⁴) cultured as indicated in A were examined for caspase 3/7 activity. The data represent the mean ±SD fold-change from controls in triplicate wells from one of two experiments showing similar results. C, 4T1 cells cultured as indicated in A were examined for Annexin V expression by flow cytometry after 16 and 24 hours. Left panel: the data represent the mean ±SD of Annexin V-FITC positive cells in triplicate wells. Right panel: representative analysis by flow cytometry at 16 hours.

BMMSC suppress drug-induced apoptosis in vivo

We then asked whether exogenous BMMSC would protect tumors from chemotherapy-induced apoptosis in vivo. To examine this possibility, mice injected with 4T1 cells in the presence or absence of BMMSC or normal mouse fibroblasts (ratio tumor cell:stromal cell 5:1) were treated with doxorubicin on days 3 and 4 and tumors were analyzed on day 5 for the presence of apoptotic, E-cadherin-positive cells (Fig. 5). This analysis revealed a significantly reduced number of TUNEL/E-cadherin-positive cells in sections of 4T1 tumors co-injected with BMMSC, when compared to 4T1 tumors injected alone (p = 0.003). Additionally, the number of TUNEL/E-cadherin-positive cells in tumors co-injected with BMMSC was also significantly lower than in tumors co-injected with normal mouse fibroblasts (p = 0.02). The data thus indicate that in vivo BMMSC protected tumor cells from drug-induced apoptosis, and this effect was specific to BMMSC and not normal fibroblasts.

Figure 5.

BMMSC suppress apoptosis in drug-treated tumors in vivo. 4T1 cells (1×10⁶) were injected in the mammary fat pad of Balb/cJ mice in the absence or presence of 2×10⁵ BMMSC or normal mouse fibroblasts. Three days post-injection of tumor cells, mice were treated with doxorubicin for two consecutive days, and tumors were harvested on day 5 and examined for the presence of apoptotic E-cadherin-positive cells. A, fluorescent micrographs of representative tumor sections from control and tumors co-injected with BMMSC or fibroblasts, stained for the presence of E-cadherin and apoptotic nuclei. Top: TUNEL, middle: E-cadherin, and bottom: overlay with DAPI. Arrowheads indicate the presence of apoptotic nuclei. Scale bar = 50 μm. B, the data represent the mean ±SD of TUNEL-positive, E-cadherin-positive cells counted in a total of 5 32x fields in each of 5 tumors.

BMMSC migrate toward tumor-derived products in vitro

We then investigated whether BMMSC would be attracted by primary tumors. We initially explored this possibility by testing the effect of tumor cell-derived CM on BMMSC migration in vitro in a transwell migration assay. This experiment indicated a robust migration of BMMSC upon exposure to serum-free 4T1 and LL/2 CM for 2 days (Fig. 6A). Further experiments utilizing serum-free 4T1 and LL/2 CM also demonstrated significant BMMSC recruitment after only 3 hours of exposure to tumor-derived products. This effect seemed in part tumor specific since 3 hours of exposure to serum-free NIH3T3-derived CM resulted in a migratory response that was only 28% of the response observed in the presence of 4T1-derived CM. These experiments demonstrated that BMMSC were preferentially recruited by tumor cells.

Figure 6.

BMMSC exhibit a chemotactic response to tumor cells in vitro and in vivo. A, BMMSC (8×10⁴ per well) were cultured for 3 hours or 48 hours in a transwell culture well in the presence or absence of serum-free CM from 4T1, LL/2, and NIH3T3 cells. Left panel: the data represent the mean ±SD number of migrated BMMSC from triplicate filters from one of two separate experiments showing similar results. Right panel: light micrographs of the bottom side of representative transwell filters from 48 hour assays. Scale bar = 100 μm. B, fluorescence analysis of frozen sections (6 μm) of 4T1 tumors and indicated organs harvested on day 14 post injection of tumor cells (6×10⁵ injected s.c. in the left flank) and day 2 post-injection of PKH26-labeled BMMSC (2×10⁶ via tail vein). Cy3/DAPI overlay for tumor, liver, lung, and kidney are shown. Scale bar = 50 μm. C, histological analysis of iron deposits by Prussian blue staining on sections from FFPE 4T1 tumors (6 μm) harvested on day 10 of tumor growth (5×10⁵ injected s.c. in the left flank) and 6 days post-injection of iron-labeled BMMSC (3×10⁵ via tail vein). Scale bar = 250 μm (left) and 30 μm (right). T = tumor; N = normal tissue.

BMMSC are specifically localized to tumors and are present at the invasive front of developing tumors

We then asked the question whether when injected intravenously (i.v.) BMMSC would preferentially localized to tumors versus healthy organs in vivo. For these experiments, 4T1 tumor-bearing mice were injected i.v (into the orbital sinus) with luciferase-labeled BMMSC or NIH3T3 cells on day 2 after tumor cell injection. An analysis by bioluminescence imaging over time (Figure 7A) revealed that MSC rapidly but transiently localized to the lungs and that luciferase-positive signals that co-localized to the site of the tumor could be detected in all five MSC-injected mice on day 12 following retro-orbital implantation. These luciferase-positive signals persisted in two of the five MSC-injected mice on day 15 and day 18. Similar luciferase-positive signals at the site of the tumor implantation were not detected in mice injected with luciferase-expressing NIH3T3 cells. To confirm the presence of BMMSC in tumors and absence in normal organs, mice were sacrificed at day 18, and tumors and organs were subjected to ex vivo bioluminescence analysis (Fig. 7B). This analysis revealed the presence of detectable luminescent signals above background in tumor lysates from the two MSC-injected mice in which luciferase-positive signals were still detected in vivo (mouse 1 and 3), whereas none of the organ lysates (lung, liver, brain, marrow and kidney) showed luciferase-positive signal detectable above background. A similar experiment performed with luciferase-expressing NIH3T3 cells failed to reveal any bioluminescence activity in either tumor lysates or normal organ lysates. These experiments indicated that BMMSC, but not NIH3T3 cells, can preferentially and selectively home to developing tumors.

Figure 7.

BMMSC are specifically localized to tumors. Luciferase-expressing BMMSC or NIH3T3 cells (2×10⁶) were injected i.v. into Balb/cJ mice bearing 4T1 tumors (n= 5; 2×10⁶ cells injected s.c. in the left flank) on day 2 of tumor growth. A, serial bioluminescent imaging of BMMSC and NIH3T3-injected mice over time (at indicated days post BMMSC/NIH3T3 injection). B, quantification of luminescent signals from tumor and normal organ lysates obtained from mice sacrificed on day 18 post BMMSC/NIH3T3 injection. The data represent the luminescent signal (RLU) of 10 mg of tissue lysate, with SD for 2 replicates shown.

To obtain a confirmation of these experiments and a better insight into the localization of BMMSC in tumors, 4T1 tumor-bearing mice were also injected i.v. with PKH26-labeled BMMSC on day 12 after 4T1 tumor cell implantation. An analysis by fluorescent microscopy on frozen sections of primary tumor and organs, revealed the presence of isolated, red fluorescent cells in the lung, liver and kidney but multicellular aggregates near the margins of some tumor sections (Fig. 6B). To further examine the localization pattern of tumor-recruited BMMSC in vivo, 4T1 tumor-bearing mice were injected i.v. with iron nanoparticle-labeled BMMSC on day 4 after 4T1 implantation and examined 6 days after injection. An histological analysis of these tumor sections by Prussian blue staining revealed the presence of iron-containing cells primarily at the invasive front of the tumors (Fig. 6C) that was distinguishable from extracellular iron present in the tumor stroma at areas of hemorrhage and necrosis.

Discussion

Our studies identify a novel role for recruited BMMSC in primary tumors, namely a protective effect on drug-induced apoptosis by soluble BMMSC-produced factors, which is separate from their protective function in the bone marrow. BMMSC are reported to promote tumorigenesis through the production of a number of soluble factors, many of which operate synergistically. However, since these pro-tumorigenic pathways largely function through growth stimulation or immune modulation, our examination of apoptosis in the BMMSC-tumor crosstalk brings a new aspect to the interaction between tumor cells and MSC.

Our primary murine bone marrow stromal cells exhibit self renewal, adherence to tissue culture plastic, and lineage differentiation, similar to the multipotent mesenchymal cells originally identified by Friedenstein in bone marrow explants (20). However, it is demonstrated that these cells represent a true population of mesenchymal stem cells, rigorously defined by Caplan as the progenitor of all connective tissue lineages (21), or whether they are part of a subpopulation of multipotent cells alternately called stromal stem cells, multipotent stromal cells, mesenchymal stromal cells, or multipotent adult progenitor cells (22). Our examination of adipogenic and osteogenic differentiation was sufficient to confirm the mesenchymal origin of these BMMSC without conducting a full characterization of their ability to generate other lineages (chondrocyte, muscle, etc.). Additionally, the surface marker profile for these primary cells, which were positive for CD44 and negative for CD34, CD45, CD31, and VEGFR2, confirmed them to be of mesenchymal origin, as opposed to hematopoietic or endothelial origin, without a full characterization of their mesenchymal marker expression (19). As mentioned previously, the low level of CD34 found in early passage BMMSC may indicate the presence of non-adherent myeloid cells or adherent fibrocytes, which express hematopoietic markers but exhibit a mesenchymal-like phenotype and contractile properties (23). Weak CD34 expression may also represent a transient production of this marker by BMMSC (24). Given the phenotypic and functional characteristics of our primary cells, they are appropriately referred to as mesenchymal stromal cells (25).

Our current study adds to the body of literature on the effect of BMMSC on tumor cell growth by specifically pointing to an effect on survival rather than on proliferation. The growth enhancement effects of BMMSC reported typically involve a stimulation of tumor cell proliferation, as demonstrated for example in human and rat osteosarcoma (12, 13). Human colorectal carcinoma xenografts also show increased proliferation in the presence of BMMSC in multiple studies (26, 27). Additional growth stimulatory effects are reported in mammospheres of breast carcinoma cells grown in the presence of BMMSC (28), and in mammary carcinoma cells exposed to CCL5 produced by BMMSC (29). However, MSC have also been shown to have a growth-suppressive effect, reducing the proliferation of Kaposi’s sarcoma cells in vitro and in vivo (30) and limiting cell cycle progression in human hepatoma and ovarian carcinoma cells, glioma and neuroblastoma (31–33). These conflicting data on the effect of BMMSC on tumor cell proliferation may be due to differences in the types of MSC studies and also the cancer cells. While proliferation promoting activities have been primarily reported in cancer of epithelial origin, proliferation inhibitory activities have been shown in neural tumors. Polarization of BMMSC into an anti or pro-tumorigenic function, similar to the one observed in T cells and macrophages, may also explain the discrepancy in BMMSC functions (34).

The novelty of the data presented in this manuscript relates to the observation that the positive effect of BMMSC on tumor cell viability was not due to an effect on proliferation but rather on survival, an effect that has not been much explored so far. Several published studies have in fact reported a pro-apoptotic effect of BMMSC on tumor cells. For example, intravenous injection of MSC has been shown to increase PARP-1 and caspase-3 cleavage in mammary carcinoma xenografts (35), and murine hepatoma and lymphoma cells were reported to increase the production of caspase-3 and p21 proteins when exposed to BMMSC in vitro and in vivo (14). Additionally, direct injection of BMMSC in rat gliomas increased tumor cell expression of pro-apoptotic caspase-3 and Bax while downregulating anti-apoptotic Bcl-2 (36). The disparity between these findings and our own observation of BMMSC-mediated suppression of tumor cell apoptosis may stem from our use of serum depletion and drug-induced cell death models. It should be pointed out that these previous studies have examined apoptosis of tumor cells exposed to MSC under unstressed conditions, suggesting that MSC can increase baseline apoptotic signaling in tumors. An important difference in our studies is that we have examined the effect of BMMSC under stress conditions such as low serum and drug treatment. Our data clearly demonstrate that under these conditions BMMSC promote rather than inhibit survival, a role that is highly relevant to clinical chemotherapeutic interventions.

The use of co-injection of BMMSC with 4T1 tumor cells in the mammary fat pad represents a helpful model to examine the interaction between BMMSC and tumor cells in vivo, although it has the limitation that BMMSC were not recruited from the bone marrow by primary tumors. However similar co-injection models have been used to examine other aspects of BMMSC-tumor interaction in vivo. Karnoub et al. found that co-injection of human mammary carcinoma xenografts with BMMSC increased the metastatic spread of tumor cells (37), an effect linked to the production of CCL5 by BMMSC, while Suzuki et al. reported in Lewis lung carcinoma tumors co-injected with BMMSC an enhancement of tumor growth, driven by increased angiogenesis (38). These co-injected tumors, artificially enriched with BMMSC, developed stromal elements with functional properties similar to TAF, which are known descendants of BMMSC (9, 39–41). Our current study leaves open the question of whether BMMSC and BMMSC-derived TAF would play a similar protective role on apoptosis as they are recruited from the bone marrow.

Importantly, we clearly demonstrate that the protective effect of BMMSC on tumor cells is specific, as it was not observed with either NIH3T3 cells or normal mouse skin fibroblasts which were used as control in multiple experiments. It thus underlines a property of MSC that is not shared with other cells of mesenchymal nature but does not eliminate a similar property in non-mesenchymal cells such as endothelial cells or myeloid cells. In fact, a protective effect against chemotherapy-induced apoptosis has been reported for tumor-associated macrophages (42).

The observed protective effect of BMMSC on primary tumor cell survival was contact-independent, as has been found in previous studies of metastatic tumor cells in the bone marrow. Production of GDF15 by bone marrow-resident BMMSC protects multiple myeloma cells from melphalan, bortezomib and lenalidomide (43), and several laboratories including ours have identified a chemoprotective effect of BMMSC on tumor cells driven by Il-6, which increases in a STAT3-dependent manner, the expression of anti-apoptotic Bcl2, BclXl, survivin and XIAP, as well as multidrug resistance proteins MDR and MRP (44–47). The identity of the protective factor(s) involved in our observed suppression of tumor cell apoptosis is currently investigated by our laboratory and preliminary data points to soluble proteins. It should be noted that other signaling molecules besides proteins could also be involved in this protective effect, as fatty acids released by BMMSC in response to platinum-based therapies have been shown to protect murine colon carcinoma and Lewis lung carcinoma tumors from platinum-induced cytotoxicity (48).

Finally, our data indicate that our primary BMMSC displayed a migratory response to tumor cells in vitro and in vivo. Whereas our in vitro data demonstrate that tumor cells specifically attract MSC, our in vivo data demonstrate that circulating MSC preferentially localize at the site of a growing tumor over time. However the fate of these cells over time remains unclear as our data show that in 3 of the 5 mice, no luciferase signal could be detected either in vivo or ex-vivo at days 15 and 18. This aspect will require further investigation. The mechanism behind the chemoattraction of MSC by tumor cells is still being investigated, with preliminary studies eliminating SDF1/CXCR4 and CD44 as the molecules involved (data not shown). MCP-1 is another potential recruitment factor for BMMSC, as this chemokine has been shown to stimulate BMMSC migration into murine mammary tumors in vivo (49). Alternatively multiple factors may be involved since BMMSC have been shown to exhibit migration in response to a large variety of growth factors and cytokines (50, 51).

In summary, in this manuscript we provide data identifying BMMSC as an important mediator of tumor cell survival and drug resistance, and evidence that BMMSC could exert this function not only in the bone marrow as previously assumed but also in the primary tumor.