|Home | About | Journals | Submit | Contact Us | Français|
The immunosuppressive tumor microenvironment is a key obstacle to hinder a cancer immunotherapy. Myeloid-derived suppressor cells (MDSCs) have been considered as a major player in immunosuppression. In this study, we find that tumor-infiltrating MDSCs (tiMDSCs) are less immunosuppressive than tumor-associated macrophages (TAMs) in multiple murine orthotopic breast tumor models. Compared to TAMs, tiMDSCs produce higher levels of pro-inflammatory factors and lower levels of anti-inflammatory factors. Furthermore, tiMDSCs are preferentially located in hypoxic areas and are more pro-angiogenic than TAMs. Consistent with these functional disparities, a shift from tiMDSCs to TAMs is observed during the progression of breast cancer. Moreover, infiltration of tiMDSCs is also noted in distal colonization of breast cancer cells in the lung. Taken together, our findings indicate that tiMDSCs are more pro-angiogenic and promote tumor initiation, while TAMs are more immunosuppressive and facilitate tumor immune evasion. This study suggests that selectively targeting on TAMs could alleviate the immunosuppressive tumor microenvironment and potentiate cancer immunotherapy.
Inducing angiogenesis and evading host immune destruction are two hallmarks of cancer . Tumor-infiltrating myeloid cell, including myeloid-derived suppressor cell (MDSC), tumor-associated neutrophil (TAN), and tumor-associated macrophage (TAM) plays diverse roles in tumor progression [2–6].
TAM represents a dominant myeloid cell population within many kinds of cancer lesions, and its accumulation often correlates with poor prognosis [4, 7, 8]. TAM usually has an M2-like phenotype and expresses typical markers, such as CD11b and F4/80 in murine models. M2-like TAM promotes tumor progression through multiple mechanisms, including the support of tumor angiogenesis, the induction of tumor cell invasion and migration, the promotion of extracellular matrix remodeling, and the suppression of host anti-cancer immune responses [8–12].
TAN expresses two typical neutrophil markers, CD11b and Ly-6G, in murine tumor models. TAN is involved in tumor initiation and progression via several mechanisms, including pro-angiogenesis and immunosuppression [6, 13, 14]. In a mesothelioma AB12 tumor model, depletion of Ly-6G neutrophils in vivo induced CD8+ T cell activation, indicating the immunosuppression of TAN .
MDSC has been identified in cancer patients and tumor-bearing mice. MDSC is a heterogeneous myeloid cell population with ability to suppress T cell activation. In tumor-bearing mice, MDSC is CD11b+Gr1+ and accumulates in the bone marrow, the spleen, and peripheral blood [15–19].
Although the phenotypes and functions of MDSC in peripheral immune organs are well defined, what are the critical roles of MDSC in the tumor microenvironment, as well as its relationship with TAN and TAM, remains not fully understood [6, 20, 21]. In this study, we characterized the immunological and angiogenic properties of these tumor-infiltrating myeloid cells in breast tumor models. Our data showed that tumor-infiltrating MDSC (tiMDSC) was less immunosuppressive, while more angiogenic, than TAM. Thus, selectively targeting TAM, rather than tiMDSC, could recondition the immunosuppressive tumor microenvironment and improve the efficacy of cancer immunotherapy.
In the peripheral immune organs, such as lymph nodes and spleen, MDSC is considered to be a major immune suppressor [2, 15, 22]. Our previous study showed that low dose anti-VEGFR2 treatment improved cancer vaccine therapy, even though tiMDSC was increased . These results lead us to hypothesize that tiMDSC is not the major immune suppressor within the tumor microenvironment. To get more insights into the phenotypes of tumor-infiltrating myeloid cell populations, we established representative murine breast cancer models: spontaneously arising autochthonous mammary carcinoma (MMTV-PyVT) and orthotopic implanted breast cancers (EO771 and MCaP0008). MMTV-PyVT is a widely used murine breast cancer model that mirrors the progression of breast cancer in humans [24, 25]. In MMTV-PyVT breast tumor tissue, two major tumor-infiltrating myeloid cell populations were identified: CD45+CD11b+Gr1hiF4/80− (Gr1+F4/80−, tiMDSC) and CD45+CD11b+Gr1−F4/80+ (Gr1−F4/80+, TAM) (Figure (Figure1A1A and Supplementary Figure 1). In EO771 and MCaP0008 tumors, there were three major myeloid cell populations: CD45+CD11b+Gr1hiF4/80− (tiMDSC), CD45+CD11b+Gr1int/lowF4/80int/low, and CD45+CD11b+Gr1−F4/80+ (TAM) (Figure 1B–1C and Supplementary Figures 2–3,). In all breast tumor models tested here, CD11b+Gr1hiF4/80− (tiMDSC) cells were also Ly6G+Ly6Clow, an equivalent phenotype to that observed in TAN. Giemsa staining also indicated that CD11b+Gr1hiF4/80− (tiMDSC) cells had typical characteristics of neutrophil (Figure (Figure1D).1D). Most CD45+CD11b+Gr1int/lowF4/80int/low cells were Ly6G−Ly6C+, suggesting that they are monocytic myeloid cells (Figure (Figure1C).1C). In the breast cancer models evaluated here, the majority of TAMs were Gr1−Ly6G−, but some of them were Ly6C+ (Figure (Figure1).1). In EO771 cancer models, myeloid cell populations displayed very different patterns compared to the other two models tested in this study. CD45+CD11b+Gr1int/lowF4/80int/low cells were a big population, and most of them were Ly6G−Ly6C+. In addition, many TAMs also expressed Ly−6C in EO771 tumor (Figure (Figure1C).1C). Together, these data suggest that tiMDSC has a similar phenotype to TAN (CD11b+Gr1+Ly6G+Ly6ClowF4/80−). TiMDSC and TAM comprise two distinct tumor-infiltrating myeloid cell populations in established breast tumors.
As tumor-infiltrating myeloid cells have been suggested to play critical roles in the immunosuppression [18, 26], but the distinct roles of tiMDSC and TAM in tumor immune evasion remain not very clear. Thus, we compared the capability of TAM and tiMDSC to suppress T cell function by co-culturing them with splenocytes at the same ratio. In both MMTV-PyVT and MCaP0008 breast cancer models, tiMDSC (Gr1+F4/80−) inhibited T cell proliferation induced by anti-CD3/CD28 monoclonal antibodies; however, TAM (Gr1−F4/80+) was much more potent than tiMDSC in suppressing T cell proliferation (Figure (Figure2A).2A). The data showed that TAM is more immunosuppressive than tiMDSC in breast tumors.
To reveal molecular mediators involved in immune suppression, we isolated tiMDSC and TAM from MCaP0008 cancer tissues using CD11b-microbead enrichment followed by flow sorting. Then, we analyzed the cytokine/chemokine profiles of tiMDSC and TAM. TiMDSC had significantly higher levels of pro-inflammatory factors, such as IL12α, IL-1β, CXCL9 and CXCL10, compared to TAM (Figure (Figure2B).2B). Conversely, tiMDSC had significantly lower levels of anti-inflammatory factors, including IL10, Arg1, CCL17, and CCL22, compared to TAM (Figure (Figure3).3). In EO771 tumor model, we also analyzed the gene expression profiles in tiMDSC, TAM and CD45+CD11b+Gr1int/lowF4/80int/low cells. Again, tiMDSC expressed higher levels of IL12α, IL1β, CXCL11, while TAM had higher levels of IL10, Arg1, CCL17 and CCL22 (Figure (Figure4).4). The gene expression profile of CD45+CD11b+Gr1int/lowF4/80low was in between those measured for tiMDSC and TAM. Together, these data suggest that TAM is more immunosuppressive than tiMDSC, consistent with functional data (Figure (Figure22).
TAM is usually considered to accumulate in the hypoxic/necrotic areas of tumor tissue [27, 28], while the distribution of tiMDSC in tumor tissue remains unclear. In order to evaluate the distribution of tiMDSC and TAM in vivo, we adapted an intravital Hoechst 33342 staining technique [23, 29, 30] to label myeloid cells as Hoechst 33342 positive (Ho+, proximal to perfused tumor vessels) and negative (Ho−, hypoxic/necrotic area). Then, we analyzed their distribution using flow cytometry (Figure (Figure5A).5A). In both MMTV-PyVT and MCaP0008 tumor models, Ho+tiMDSCs were approximately 25% while Ho+TAMs were about 50% (Figure 5B–5C). From the patterns of Hoechst fluorescence intensity histograms, both Ho+tiMDSCs and Ho+TAMs were found to be evenly distributed, indicating that both tiMDSCs and TAMs were evenly distributed around the perfused tumor vessels (Figure (Figure5B).5B). Notably, more tiMDSCs (about 75%) were located in non-perfused areas (hypoxic/necrotic areas), compared to TAMs (about 50%) (Figure (Figure5C).5C). Consistently, immunohistochemistry data also showed that tiMDSCs were more located in distant to perfused vessels, while TAMs were relatively even distributed throughout the tumor tissue (Figure (Figure5D).5D). Thus, our data suggest that tiMDSC is more frequently located in hypoxic/necrotic areas.
Macrophage present in hypoxic areas displays altered gene expression with a pro-angiogenic phenotype [4, 27, 28]. Since tiMDSC is primarily present in non-perfused tumor areas, we proposed that tiMDSC and TAM possess differential pro-angiogenic capability. In flow sorted tiMDSCs and TAMs, tiMDSCs had significantly higher levels of SDF1α, MMP9, VEGFa, and PlGF, compared to TAMs (Figure (Figure6A).6A). Furthermore, in MMTV-PyVT breast tumors, tiMDSCs were more potent than TAMs in their ability to induce tube formation (Figure (Figure6B).6B). These data suggest that, in cell base, tiMDSC is more angiogenic than TAM.
The differentially immunosuppressive and angiogenic capabilities of tiMDSC and TAM (Figures (Figures22 and and6)6) indicated that they may play different roles in tumor initiation and progression. Thus, we analyzed tiMDSC and TAM at two time points after MCaP0008 breast cancer cell inoculation. Interestingly, approximately 27.6% of the CD11b+ cells (13.0% in total viable cells) and 2.6% of total CD11b+ cells (0.3% in total viable cells) were tiMDSCs on day 7 and 14 after tumor cell inoculation, respectively (Figure (Figure7).7). The absolute number of tiMDSC on day 7 and 14 was 4475 and 752 per tumor, and CD11b+ cells on day 7 and 14 was 16822 and 17812 per tumor. Conversely, 55.2% of CD11b+ cells (26.5% in total viable cells) were TAMs on day 7 post-inoculation, and that proportion increased to 82.0% (6.6% in total viable cells) on day 14 day after inoculation (Figure (Figure7).7). The absolute number of TAM on day 7 and 14 was 9640 and 14913 per tumor, respectively. Although the percentages of tiMDSC and TAM decreased on day 14 relative to day 7, the ratios of TAM/tiMDSC were dramatically increased on day 14 (TAM/tiMDSC = 20) compared to day 7 (TAM/ tiMDSC = 2), suggesting a shift of tiMDSC to TAM. These data show that the initiation and progression of MCaP0008 breast tumors are accompanied by a shift of tiMDSC to TAM, which is consistent with their distinct angiogenic and immunosuppressive activities of tiMDSC and TAM.
Breast cancer is the most common cancer in women. More than 90% of breast cancer-related deaths are due to metastasis of breast cancer cells into vital organs, especially the lung . The colonization of metastatic cancer cells in distant organs is a key step for cancer progression. TiMDSC is preferentially accumulated during the early stage of breast cancer and possesses strong pro-angiogenic activity (Figures (Figures66 and and7).7). Thus, we hypothesized that neutrophil facilitates lung colonization. To test this hypothesis, we analyzed the quantity of neutrophil in the lungs after we removed 4T1 primary tumors. Interestingly, the amount of neutrophil in the lungs dramatically decreased two days after primary tumor removal. On day 7, the levels of neutrophil were comparable to normal levels in the lung, and then began to increase (Figure (Figure8A).8A). We also examined lung metastases at different time points. After 4T1 primary tumor removal, there was no visible lung metastasis on day 2, and some of lungs had several tiny lung metastases on day 7. It is striking that there were dozens of lung metastases on day 14 and day 21. These data suggested that lung colonization by breast cancer cells was associated with an increase of neutrophil in the lung. We also compared tiMDSC present in different sizes of lung metastases. The proportion of tiMDSC in lung metastases was much higher than in that observed in primary breast tumors. In addition, the proportion of tiMDSC in small lung metastases was higher then larger lung metastases (Figure (Figure8B).8B). These data suggest that lung colonization by breast cancer cells is associated with the accumulation of neutrophil.
Cancer immunotherapy and anti-angiogenic therapy are two key cancer treatment modalities [32, 33]. Tumor-infiltrating myeloid cells often compromise their efficacy [34, 35]. Myeloid cells are a heterogeneous cell population within the tumor immune microenvironment. They play critical roles in tumor angiogenesis and tumor immune evasion [34, 35], but the distinct characteristics of individual populations remain not fully defined. Here we showed that tiMDSC and TAM possess distinct angiogenic and immunologic properties in breast cancer models. In the established breast cancers, TAM is the most abundant immune cell population with potent immunosuppression. Therefore, polarization of TAM away from immunosuppression, or depletion of TAM could enhance cancer immunotherapy and/or improve survival [12, 23, 25, 36–38]. Conversely, tiMDSC is pro-angiogenic and preferentially accumulates during the early stage of breast cancer initiation or metastasis [14, 39]. Therefore, selectively targeting on distinct myeloid cell population based on breast cancer development stage could achieve better therapeutic outcome.
Tumor-infiltrating myeloid cells include multiple cell populations, including tiMDSC, TAM, TAN, granulocytic immature myeloid cell, and monocytic immature myeloid cell [5, 25, 35, 40]. Accumulating evidence suggests that there are phenotypic and functional overlaps between these populations. MDSC is initially identified in the peripheral immune organs. According to data on surface markers and their functions, MDSC within the tumor microenvironment may include neutrophil, monocyte and CD45+CD11b+Gr1int/lowF4/80low/+ cell. Polymorphonuclear CD11b+Ly6G+Ly6Clow cells and monocytic CD11b+Ly6G−Ly6Chi cells are two major populations in the spleen of tumor-bearing mice [9, 35, 41]. In MMTV-PyVT, tumor-infiltrating myeloid cells have two distinct populations based on the expression of Gr1 and F4/80: CD11b+Gr1hiF4/80− and CD11b+Gr1−F4/80+. CD11b+Gr1hiF4/80− cells, which are usually considered to be MDSC, are also Ly6G+Ly6Clow, thus CD11b+Gr1hiF4/80− MDSC phenotypically overlaps with TAN (CD11b+Ly6G+Ly6ClowF4/80−) and granulocytic immature myeloid cell [5, 6, 25]. Since MDSC is generally considered to be a heterogeneous cell population, we suggest referring tumor-infiltrating CD11b+Ly6G+Ly6ClowF4/80− (tiMDSC) cell as TAN. Some investigators consider all tumor-infiltrating CD11b+ cells as MDSC, which includes both TAN and TAM [42, 43]. However, TAM is typically Gr1−F4/80+ and should not be considered as MDSC [4, 44]. Monocytic CD45+CD11b+Ly6G−Ly6C+F4/80int/low cell population varies in different tumors and at different tumor stages, and it also changes by certain treatments [45, 46].
TAM is a multifunctional cell population. During tumor initiation, TAM forms an inflammatory environment to potentiate oncogenic mutations, but at the tumor progression stage, TAM secretes growth factors and cytokines to facilitate angiogenesis and suppress cytotoxic T cell responses. However, after progression to malignancy, TAM produces proteases and chemokines to induce tumor metastasis [37, 38, 47, 48]. Among myeloid cells present in breast cancers, TAM is very immunosuppressive. Interestingly, we found that tiMDSC was predominately located in less perfused and necrotic areas and expressed higher levels of pro-angiogenic factors. In addition, tiMDSC is more prevalent at early stages of cancer initiation or during metastatic cancer cell colonization. These properties suggest that tiMDSC is pro-angiogenic and pro-metastatic. This is consistent with previous reports that Gr1+ myeloid cells promote tumor angiogenesis and metastasis [14, 39, 49–51]. Gr1+ myeloid cells have been shown to be involved in the formation of lung metastatic niches .
Different myeloid cell populations possess various properties and may play different roles in tumor initiation and progression [9, 52, 53]. In general, the proportion of myeloid cells increases when tumor grows bigger in either peripheral immune organs or in the tumor microenvironment. Meantime, the composition of myeloid cell populations also changes at the various stages of tumor. We observed a higher ratio of tiMDSC/TAM during tumor initiation, compared to established tumors. TiMDSC is a heterogeneous cell population and can differentiate into other myeloid cell types. The shift of TAN to TAM could be due to the differentiation of tiMDSC to TAM. The hypoxic tumor microenvironment promotes rapid differentiation of tiMDSC into TAM [26, 54, 55]. In addition, the short lifespan of neutrophil may also contribute to the decreased ratio of tiMDSC/TAM .
Neutrophil is the first responder during inflammation. Indeed, CD11b+Gr1+ cells comprise about 75% of CD45+ cells infiltrated into inflammatory sites induced by OXd4 . Tumor has been considered as a wound that does not heal. At the early stage of tumorigenesis, MDSC (CD11b+Gr1+) cells are the major infiltrating immune cells, which is likely due to inflammation . During tumor development, the amount of TAM was increased and polarized to immunosuppression, and thus promoted tumor progression. TiMDSC appears to play a pivotal role in primary tumor initiation or metastatic cancer cell colonization via the induction of angiogenesis [13, 57], while TAM promotes tumor progression through the creation of an immunosuppressive tumor microenvironment. Thus, tiMDSC and TAM possess differential properties and fulfill distinct roles during different stages of cancer development and progression.
FVB/N, C57BL/6, Balb/c, and MMTV-PyVT mice were housed in pathogen free animal facilities. The EO771 breast tumor cell line was purchased from CH3 Biosystems (New York, USA). Dr. Peigen Huang at Massachusetts General Hospital generated the MCaP0008 breast tumor cell line . Both of the cell lines were cultured in DMEM medium (Gibco, USA) containing 10% fetal bovine serum (FBS, Gibco). 4T1 breast tumor cells were purchased from ATCC (USA) and cultured in RPMI-1640 medium (Gibco) containing 10% FBS. To obtain source tumor tissue, MCaP0008 breast tumor cells (1 × 106 cells) were injected orthotopically into the mammary fat pad of female FVB/N mice . When the tumor reached 8 mm in diameter, it was excised, and a small piece (about 1 mm3) of viable tumor tissue was orthotopically transplanted into new female FVB mouse. In some experiments, both MCaP0008 (1 × 106 cells) and EO771 (3 × 105 cells) tumor cells were orthotopically inoculated into the mammary fat pads of FVB/N and C57BL/6 mice, respectively. Animal procedures were carried out following the Public Health Service Policy on Humane Care of Laboratory Animals. All procedures were approved by the Institutional Laboratory Animal Care and Use Committee of Soochow University.
When tumors reached 6−8 mm in diameter, tumor-bearing mice were injected, via the tail vein, with 200 μl of 10 mg/kg Hoechst 33342 dye (Sigma-Aldrich, USA). After 5 minutes, mice were systemically perfused with PBS and the tumors were removed. Under these conditions, a reproducible perivascular tumor cell-labeling gradient was achieved, as described previously [29, 30]. Tumor tissues were cut into two pieces. One portion was fixed with 4% paraformaldehyde for histological analysis. The remaining piece was used for flow sorting or flow cytometric analysis. FITC-anti-Ly6G (BD Biosciences, USA) and APC-anti-F4/80 (eBioscience, USA) were used to stained neutrophil and macrophage in the frozen tumor tissues.
Tumors were harvested from MMTV-PyVT transgenic mice or MCaP0008 tumor-bearing mice. Tumor tissues were then minced and digested at 37°C for 45 min with DMEM containing collagenase type 1A (1.5 mg/ml), hyaluronidase (1.5 mg/ml), and DNase (20 U/ml). TiMDSCs and TAMs were enriched by CD11b-microbeads (Miltenyi, Germany). Enriched cells were then stained with PE-F4/80 (eBioscience), FITC-Ly6G, PE-Cy7-CD45, APC-Gr1, APC-Cy7-CD11b (BD Biosciences) and isolated by flow sorting. 50,000 tiMDSCs were subjected to cytospin preparation (Thermo Shandon), and were fixed and stained with a Wright Giemsa Staining according to the manufacturer's instructions (Sigma-Aldrich).
Mixed leukocyte reaction (MLR) was used to evaluate the immunosuppression of myeloid cell populations as previously described with minor modifications [9, 59]. Briefly, single cell suspensions of splenocytes were prepared from naïve mice. Splenocytes (2 × 105 cells) were co-cultured in triplicate with tiMDSCs or TAMs (5 × 104 cells) in a total volume of 200 ul containing anti-CD3/CD28 (1/5 μg/ml) for 24 hrs in U-bottom 96-well plates. Then the cultures were treated with 1 μCi of 3H-thymidine (GE Healthcare Life Sciences, USA) overnight. Cells were harvested using a cell harvester (Skatron Instruments), and 3H-thymidine uptaken was counted using a liquid scintillation counter. The background proliferation of splenocytes, tiMDSCs, or TAMs alone was subtracted to obtain the final proliferation value.
The pro-angiogenic effect of myeloid cell populations was assessed by tube formation assay as descried previously with minor modifications [60, 61]. Briefly, human umbilical vein endothelial cells (HUVECs) were cultured in DMEM medium containing 10%FBS. Growth factor reduced matrigel matrix (CORNING) was thawed in a refrigerator (4°C) overnight. Matrigel matrix (100 μl/well) was added to the growth surface of 24-well plates and the coated plates were incubated at 37°C for 30 minutes to allow the gel to solidify. HUVECs (30,000 cells/well) were seeded onto the top of the gel and tiMDSCs or TAMs (60,000 cells/well) were added, and co-cultured with HUVECs in triplicate. Cells were incubated at 37°C with 5% CO2 for 12 hrs. The tube formation pictures were captured with a Box-Type Fluorescence Imaging Device (OLYMPUS). The numbers of tubes and branches were counted for at least three fields per well.
After intracardiac injection of PBS, breast cancer tissues, lung metastases or lungs were harvested, minced and digested at 37°C for 45 min with DMEM medium containing collagenase type 1A (1.5 mg/ml), hyaluronidase (1.5 mg/ml), and DNase (20 U/ml). The digestion mixtures were filtered through 70 μm cell strainers. Single-cell suspensions were incubated with rat anti-mouse CD16/CD32 mAb (BD Biosciences), and then stained, washed and re-suspended in cold buffer (1%BSA, 0.1% NaN3 in PBS). 7AAD reagent (eBioscience) was added to the stained tubes (5 μl/tube) just before running the flow analysis. Flow cytometry data were acquired on a Gallios flow cytometer (Beckman, USA), and data were analyzed with Kaluza software (version 1.3). The appropriate, fluorochrome-conjugated, isotype-matched, control IgGs were used in all experiments. The following monoclonal anti-mouse antibodies were used: CD45-PE-Cy7, CD45-PerCP, CD45-BV421, Gr1-PerCP-Cy5.5, Gr1-APC, Gr1-APC-Cy7, CD11b-APC-Cy7, CD11b-BV510, Ly-6G-FITC, Ly-6C-PE (BD Biosciences) and F4/80-PE, F4/80-FITC, F4/80-APC (eBioscience).
Total RNA was extracted from the following flow-sorted cells using a RNeasy Mini Kit (QIAGEN, USA): tiMDSC, CD45+CD11b+Gr1int/lowF4/80int/low, TAM. Full-length cDNAs were synthesized using an iScript™ cDNA Synthesis Kit (Bio-Rad, USA). Primers specific for β-actin, Arignase-1, IL10, CCL17, CCL22, MRC1, Nos2, IL12a, IL1β, TNFα, CXCL9, CXCL10, VEGF, PlGF, CXCL12 and MMP9 were provided in the Supplementary Table 1. qRT-PCR analysis was performed by using a Power SYBR® Green PCR Master Mix (Applied Biosystems) on a 7500 Fast Real-Time PCR System (Applied Biosystems) according to the manufacturer's instructions. The comparative threshold cycle method was used to calculate fold change in gene expression, which was normalized to β-actin as a reference gene.
Results were expressed as means ± SEM (Standard Error of Mean). Experimental differences were tested using Student's t-test (unpaired, two sided). Values were considered statistically significant when P < 0.05. Significant differences and the p-values are represented in figures by asterisks as follows: * < 0.05; ** < 0.01.
I would like to thank Dr. Peigen Huang for providing the MCaP0008 cell lines and Dr. Quansheng Zhou for providing HUVECs.
Authors’ contributionsZ.F., X.C., X.W., and Y.H. designed experiments, analyzed data and wrote the manuscript; Z.F., X.C., C.W., R.Y., and C.Z. performed experiments and analyzed data.
CONFLICTS OF INTEREST
The authors declare no conflicts of interest.
This work has been supported in parts by grants from the National Natural Science Foundation of China (Grant No: 81372245, 81673004, Y.H.), the fund of the Distinguished Professor of Jiangsu Province (SR21100114, Y.H.), the Collaborative Innovation Center of Hematology, and the Priority Academic Program Development of Jiangsu Higher Education Institutions.