Background
Chemo-resistance remains a significant challenge in acute myeloid leukemia (AML) and is one of the critical prognostic elements. Evidence has suggested that the interaction between leukemia cells and the BM stroma could affect the chemo-sensitivity of the tumor cells [
1]. As an important element of the tumor microenvironment, fibroblasts with an activated phenotype, referred to as ‘activated myofibroblasts’ or ‘cancer-associated fibroblasts’ (CAFs), have been reported to play a critical role in the chemo-resistance of solid tumors [
2]. However, CAFs still appear not to be a concern in leukemia because BM biopsies are not routinely performed in AML patients. Therefore, whether CAFs exist in the BM microenvironment and whether they are involved in resistance to chemotherapy still remain unclear.
CAFs display phenotypes that are similar to those of myofibroblasts derived from quiescent fibroblasts, which are activated when they interact with carcinoma cells during tumorigenesis [
3]. These powerful cells are immunologically defined primarily by the expression of α-smooth-muscle actin (α-SMA), fibroblast-activated protein (FAP), and fibroblast-specific protein 1 (FSP1) [
4‐
6]. Datas have revealed that CAF cells are involved in the migration, growth and chemo-resistance of tumor cells in several solid tumors, including human prostate, pancreatic, colorectal carcinoma, and in hematologic malignancies, such as multiple myeloma, and the function mainly by secreting a variety of cytokines [
7‐
9]. In the BM from mouse models of acute lymphoblastic leukemia (ALL), we have previously determined a protective niche, which is dynamically transient between Nestin
+ and α-SMA
+ cells,causing the production of reticular fibers in response to chemotherapy. CAF-like cells and fibers may provide chemoresistance for leukemia cells in vivo. Meanwhile, leukemia cells contribute to the protective niche formation by secreting chemotactic factors and growth factors, in particular, growth differentiation factor 15 (GDF15) [
10]. GDF15, a divergent member of the TGF-β superfamily, is highly expressed only in the placenta and in macrophage cells under physiological conditions [
11]. However, it is clinically associated with disease progression in numerous tumors [
12‐
14]. Recently, it has been demonstrated that CAFs could be a novel source of GDF15 in human prostate cancer and multiple myeloma [
7,
15]. We propose that CAFs within the BM of AML could also counter the chemo-sensitivity of leukemia cells by producing GDF15.
In the current study, excessive reticular fibers in the BM led to a poor prognosis in primary AML patients, making it important to investigate the functional roles of the fibroblast cells. After demonstrating the presence of CAFs in the marrow microenvironment of primary AML and their ability to protect leukemia cells from chemotherapy agents in vitro, we assessed whether GDF15 contributed to the CAF-mediated chemo-protection of AML cells either by using a neutralizing anti-GDF15 antibody or by knocking down GDF15 (siGDF15) in the CAFs as well as by investigating the distribution of GDF15 in the cytoplasm of CAFs within BM sections from AML patients. These findings were used to investigate the functional roles of CAF-derived GDF15 in chemo-resistance of leukemia and whether CAF cells, by producing GDF15, could be a survival and chemo-protective element for leukemia cells.
Methods
Primary AML samples and normal controls
Primary AML patients except acute promyelocytic leukemia (
n = 63) were prospectively enrolled in our study after providing written informed consent. Cases of newly diagnosed lymphoma patients without bone marrow involvement were selected as the normal control group (
n = 59). There were no statistical differences in age and sex between these two groups (data were not shown). The studied patients with AML received the induction/reinduction chemotherapy protocols and all the BM trephine biopsy specimens and BM aspirates were obtained from Shanghai Jiaotong University-affiliated hospitals (Shanghai, China). The patients were diagnosed according to the FAB classification. Complete remission (CR) and relapse were diagnosed according to Cheson et al [
16]. And the refractory AML cases were defined in accordance with Schmid et al [
17]. The AML patient characteristics are presented in Table
1. The pathologists who examined the BM samples were not participant in this study and were innocent about the group situation.
Table 1
Patient demographics and clinical characteristics
Age (years) | | | 0.261 |
Median age | 53 | 60 | |
Range | 23-82 | 18-80 | |
Sex (Male/Female) | 18/7 | 27/11 | 0.935 |
FAB subtype | - | - | 0.709 |
M0 | 1 | 1 | |
M1 | 1 | 6 | |
M2 | 8 | 10 | |
M4 | 8 | 15 | |
M5 | 5 | 4 | |
M6 | 1 | 1 | |
M7 | 1 | 1 | |
Cell culture and regents
Mesenchymal stem cells (MSCs) were obtained from the BM of primary AML patients after informed consent. The marrow was diluted twice with phosphate buffered saline (PBS), and then isolated by Ficoll-Hypaque (Axis-Shield Diagnostics, Dundee, Scotland, UK) density-gradient centrifugation. Monocytes were collected by adherence to a plastic flask and incubated for 48h in MesenCult® medium (STEMCELL Technologies, Vancouver, BC, Canada). The phenotype and multipotential differentiation of MSC cells used in the study has been verified (Additional file
1: Figure S1–2). Then BM-MSCs were treated with recombinant human TGF-β1 (10ng/ml, R&D Systems, Minneapolis, MN, USA) to trigger CAF differentiation. Conditioned medium from the CAFs or MSCs was obtained from cells cultured in regular medium with 1 % FBS. The leukemia cell lines THP-1 and K562 (American Type Culture Collection, Manassas, VA, USA) were cultured in 1640 or low-glucose DMEM supplemented with 10 % fetal bovine serum (FBS) and penicillin-streptomycin (Gibco, Grand Island, NY, USA) at 37 °C in 5 % CO
2. Additionally, where appropriate, the leukemia cells were pre-incubated with the TGF-β1 inhibitor SB431542 (Sigma, St Louis, MO, USA) for 2 h at 37 °C in 5 % CO2.
Transwell co-culture of CAFs with leukemia cells
CAF cells (2 × 104 initial cell count) were cultured for 24 h in the upper side of a transwell chamber partitioned by a polycarbonate membrane (8.0um pore size, Corning Incorporated, Costar). And 2 × 105 THP-1and K562 cells were co-cultured in the lower part in 1640 or DMEM medium with 10 % FBS at 37 °C, 5 % CO2 for 48 h.
Evaluation of leukemic cell viability
THP-1 and K562 cells from suspension cultured, direct co-cultured or transwell co-cultured groups were treated with Ara-C (10uM, TCI Company, Tokyo chemical industry, Tokyo, Japan) for 48 h. Directly co-cultured THP-1/K562 cells were separated from the MSCs or CAFs monolayer by careful pipetting. After the leukaemic cells were collected, the monolayer was observed under the microscope (100×) to confirm that the monolayer was not damaged and that fewer than 10 leukaemic cells/vision field remained attached. Cell viability was evaluated by trypan blue exclusion in triplicate samples and quantified by a Cellometer Mini (Nexcelom, Lawrence, MA, USA).
Immunohistochemistry and the reticulin fiber density (RFD) assay
The BM tissue samples were used for immunohistochemistry and Gomori’s stain. The rabbit anti-FSP1/S100A4 monoclonal antibody (Millipore, Billerica, MA, USA), rabbit anti-FAP polyclonal antibody (Abnova, Taipei, Taiwan), goat anti-α-SMA polyclonal antibody (Abcam, Cambridge, MA, USA), rabbit anti-GDF15 monoclonal antibody (Cell Signaling, Danvers, MA, USA), mouse anti-Vimentin monoclonal antibody (Santa Cruz, Dallas, CA, USA), and mouse anti-CD68 monoclonal antibody (Abcam) incubations were performed according to the manufacturer’s instructions. Images of the slides stained for these markers were scanned at 40× magnification using the optical microscope (Olympus Co., Tokyo, Japan). The positive rate of each marker was quantified by counting the ratio of the positive cells in all nuclear cells. The reticulin fiber density (RFD) of the samples was assessed according to Norén-Nyström et al [
18]. All readings and estimations were performed in a blinded manner.
Flow cytometry
The Alexa Fluorescence488 (AF488)-conjugated monoclonal antibodies specific for α-SMA (R&D Systems, Minneapolis, MN, USA), a rabbit anti-FAP polyclonal antibody (Abcam, Cambridge, MA, USA) and a FITC-conjugated anti-rabbit polyclonal antibody were used (eBioscience, San Diego, CA, USA). The appropriate isotype-matched antibodies were used as controls (R&D Systems). For the apoptosis assay, THP-1 and K562 cells (5 × 105) were stained with Annexin-V and propidium iodide (PI) for apoptosis detection (BD, Franklin Lakes, New Jersey, USA) according to the recommended protocol on a Becton Dickinson Flow Cytometer. For cell cycle analyses, THP-1 cells were treated with RNase A and PI (Sigma, St Louis, MO, USA). The measurements were made using a flow cytometer (Beckman, Urbana, IL, USA).
Quantitative reverse transcription PCR (qRT-PCR)
Total RNA was extracted using Trizol (Invitrogen, Paisley, UK), and the RNA was converted into cDNA using the Primer Script TM RT reagent Kit (Takara Bio Inc, Otsu, Shiga, Japan). All real-time PCR reactions were performed using an ABI 7500 real-time PCR system (Biosystems, Foster City, CA, USA) and the SYBR Premix Ex Taq reagent kit (Takara Bio Inc).
Western blotting and ELISA
MSC or CAF cells were harvested in a radioimmunoprecipitation assay buffer for western blotting analysis (Beyotime, Haimen, Jiangsu, China). Mouse anti-collagen I monoclonal antibody (Abcam, Cambridge, MA, USA), rabbit anti-collagen III polyclonal antibody (Abcam), rabbit anti-FAP polyclonal antibody (Abnova, Taipei, Taiwan), rabbit anti-α-SMA monoclonal antibody (Epitomics, Burlingame, CA,USA), rabbit anti-GDF15 polyclonal antibody (Bioworld, Louis Park, MN,USA) and mouse anti-β-actin monoclonal monoclonal antibody (Cell Signaling, Danvers, MA, USA) were used. Antigen detection was performed using a goat anti-rabbit or an anti-mouse IgG secondary antibody (Cell Signaling) conjugated to HRP and an ECL western blotting substrate (Pierce, Waltham, MA, USA). The BM aspirates were collected from newly diagnosed AML patients which were divided into CR group (n = 9) and refractory group (n = 8) and normal controls (n = 11). The ELISA analysis for GDF15 was performed according to the manufacturer’s instructions (R&D systems).
RNA interference
GDF15 siRNA and the control siRNA products were purchased from Gene Pharma Company (Suzhou, Jiangsu, China). Transfection of siRNA was performed using Lipofectamine2000 (Panc05.04) or Lipofectamine RNAiMax (Invitrogen, Paisley, UK) following the manufacturer’s protocol. The validation of target gene knockdown was performed using qRT-PCR using the delta-delta CT method.
Statistical analyses
A Mann-Whitney U test was applied to compare the RFD of the AML patients and the normal controls. The statistical analyses were performed using the two-sample two-tailed Student’s t-test. The survival curves were constructed using the Kaplan–Meier method, and the log-rank test was used to test the difference in RFS and OS between cases with different RFD values. RFS was defined as the time between diagnosis and the onset of the first relapse, and OS was defined as the time between diagnosis and the occurrence of death or when lost to follow-up. Independent prognostic parameters for OS and RFS were identified by univariable Cox regression analyses and a multivariable Cox regression model using backward elimination with an exclusion significance level of 5 %. The data are described as the mean ± standard deviation (SD), and p values ≤0.05 were considered significant. All statistical analyses were conducted by using the SPSS 18.0 software program (SPSS, Chicago, IL, USA).
Discussion
As it is known that the normal tumor microenvironment becomes ‘corrupted’ during tumor development, which is reflected by appearance of a large and heterogeneous category of cancer-associated fibroblasts (CAFs) [
28]. CAF cells are considered active modulators of the tumor microenvironment among many solid tumors [
29‐
31]. However, few studies have directly addressed the role of the CAFs in the BM of leukemia. In the present study, we demonstrate that functional CAFs are widespread in the BM of AML patients and could serve as a critical chemo-protective element for AML cells by producing an abundance of GDF15.
As we all know, AML cells interact both anatomically and functionally with the stroma within the BM microenvironment. These interactions have a critical role in the development, progression and relapse of AML. A recent study has suggested discoidin domain receptor 1 (DDR1), a class of collagen-activated receptor tyrosine kinase (RTK) was highly upregulated on bone marrow (BM)-derived CD33+ leukemic blasts of acute myeloid leukemia (AML) patients, suggesting that the remodeled collagen IV in BM microenvironment could modulate the migration and adhesion of myeloid leukemia cells during leukemogenesis [
32]. In this study, the RFD and the immunohistochemistry assay were used to detect the transformation of the stroma niche in BM during acute leukemogenesis. It has been reported that reticulin in BM is mainly composed of individual fibrils or small bunches of fibrils of type III collagen surrounding a core of type I collagen fibrils [
20]. Our experiments also confirmed the presence of the amounts of collagen I and collagen III in the marrow of AML (Additional file
1: Figures S3). This result is in agreement with those from previous studies, which have revealed that CAFs might produce and secrete various extracellular matrix proteins (i.e., collagen I, III, IV) in solid tumors [
33]. Data has been proposed that CAFs are responsible for collagen synthesis in the stroma of human hepatocellular carcinoma [
34]. But reports about the origin of the fiber is still rare, to our knowledge, Jean Y P et al. firstly developed the quantitative dissection of stromal cell-extracellular matrix interactions in living tissue by using multiphoton laser scanning microscopy and second harmonic generation (SHG) of fibrillar collagen. They showed that the CAF is necessary for fiber remodeling [
35]. Based on the support from the literatures, we considered that the increased fiber stroma in BM might imply the increased of CAFs during leukemogenesis. Positive staining for the CAF-specific markers α-SMA, FAP and FSP1 could also displayed that amount of the CAFs contributed to the abundant fibers in the BM. In addition, there existed sufficient TGF-β1 in BM microenvironment of leukemia to stimulate the formation of CAF niche (Additional file
1: Figure S4).
Resistance to anti-cancer therapies which is the major obstacle to a better prognosis of patients obtained increasingly attention in the modern biomedical research [
36]. Acting on cancer cells, CAFs have previously been reported to be involved in resistance to chemotherapy in many solid tumors. Our co-cultured experiments showed that these functional fibroblasts could also exhibit a chemo-protective effect on the leukemia cells. It is worth noting that our results did no find the similar protective effects of the stroma cells which were derived from the healthy donors and pretreated with TGF-β1(10ng/ml) for 14 d (Additional file
1: Figure S5). We suspected this may be due to the differences in BM-MSC themselves between the AML patients and the healthy ones. The mechanisms involved in the anti-chemotherapy effects of CAFs on AML cells are not clear. In this report, we detected that either knockdown of GDF15 in the CAFs or interference by neutralizing anti-GDF15 in the co-culture system decreased the mortality of the THP-1 cells, confirming that GDF15 has an essential function in the CAF-mediated chemotherapeutic resistance of the leukemic cells. When compared with MSC and the THP-1 cells, we detected an increased GDF15 concentration in the conditioned medium of the CAFs (Additional file
1: Figure S6). Therefore, we assumed that CAFs were likely to be the main contributor to the high concentrations of soluble GDF15, which was different form our previous publication that GDF15 derived from leukemia cells conferred chemoresistance in acute lymphoblastic leukemia, we suspect there existed originally differences in the lymphoid and myeloid leukemia cells. As we all know that GDF15, as an important component of the TGF-β superfamily, is notably increased in patients with prostate, colorectal, or pancreatic cancers and has been described as an important biomarker of poor clinical outcome [
37‐
39]. However, the GDF15 concentrations in the BM of AML patients are not well defined. In our experiments, the direct evidence for the CAFs contributing to high GDF15 expression is supported by the BM biopsies, which indicated that this factor was mainly expressed by the CAFs and colocalized with FAP by immunohistochemistry. GDF15 is present both on the membrane and in the cytoplasm of leukemic cells (Fig.
5c), suggesting that leukemia cells have an autocrine function, which is consistent with the results of ELISA analysis of the leukemia cell lines in vitro. Both the CAFs and the leukemia cells secreted GDF15, suggesting that GDF15 plays a critical role in the local cross-regulation between the leukemia cells and the CAFs within the BM in primary AML patients. However, the detection of CAF cells on BM biopsies were not routinely performed in AML patients, so if there resist a correlation between CAF density and AML survival still remains unclear. Based on this, it is necessary for us to construct the mouse transplantation models of AML to explore the correlation between CAF density and further to explore the potential mechanisms in vivo.
Conclusions
This study highlights the importance of the fiber stroma in leukemogenesis. The abundant reticulin fiber predicts a poor outcome of AML. Based on the experiments, we suggest that a portion of the CAF cells exist in the BM microenvironment of AML. And these functional cells affected the chemosensitivity of the leukemia cells by producing GDF15. Disrupting GDF15 via treatment with an anti-GDF15 antibody or si-GDF15 in the CAFs abrogated the protection of the leukemic cells by CAFs. To our knowledge, this is the first report of the molecular mechanisms of CAFs in AML. We suspect that there might maintain a chemo-protective niche in a pool of AML cells causes the relapse, which may need for further animal experiments to verification. In this regard, future studies from our laboratory will assess GDF15 as a novel target for therapeutic strategies in AML in vivo by using a mouse transplantation model of AML.
Acknowledgements
This study was supported by the National Natural Science Foundation of China (Grant No. 81570135), National Science and Technology Major Equipment Projects of China (Grant No.2013YQ03065109) and the Shanghai Committee of Science and Technology, China (Grant No.13D22293700).