The tumor-promoting potential of CCL5 has been proposed but remains poorly understood. We demonstrate here that an autocrine CCL5–CCR5 axis is a major regulator of immunosuppressive myeloid cells (IMC) of both monocytic and granulocytic lineages. The absence of the autocrine CCL5 abrogated the generation of granulocytic myeloid-derived suppressor cells and tumor-associated macrophages. In parallel, enhanced maturation of intratumoral neutrophils and macrophages occurred in spite of tumor-derived CCL5. The refractory nature of ccl5-null myeloid precursors to tumor-derived CCL5 was attributable to their persistent lack of membrane-bound CCR5. The changes in the ccl5-null myeloid compartment subsequently resulted in increased tumor-infiltrating cytotoxic CD8+ T cells and decreased regulatory T cells in tumor-draining lymph nodes. An analysis of human triple-negative breast cancer specimens demonstrated an inverse correlation between "immune CCR5" levels and the maturation status of tumor-infiltrating neutrophils as well as 5-year-survival rates. Targeting the host CCL5 in bone marrow via nanoparticle-delivered expression silencing, in combination with the CCR5 inhibitor Maraviroc, resulted in strong reductions of IMC and robust antitumor immunities. Our study suggests that the myeloid CCL5–CCR5 axis is an excellent target for cancer immunotherapy. Cancer Res; 77(11); 2857–68. ©2017 AACR.
Altered myelopoiesis in which immature immunosuppressive myeloid cells are generated and expanded has been long associated with tumor growth. To date, immunosuppressive myeloid cells (IMC) have been documented to include myeloid-derived suppressor cells (MDSC), tumor-associated neutrophils (TAN), tumor-associated dendritic cells (TADC) and tumor-associated macrophages (TAM; refs. 1–4). MDSCs are further divided into two subsets based on expression levels of Ly6C and Ly6G: monocytic (M-MDSCs, Ly6Chigh/Ly6G−) and granulocytic (G-MDSCs, Ly6Clow/−/Ly6Ghigh; ref. 5). TANs and G-MDSCs are ambiguously defined by same markers (CD11b+/Ly6G+). The protumor TAN (namely N2, as opposed to antitumor N1) has been considered to be G-MDSCs. On the other hand, M-MDSCs (also referred as CCR2+/Ly6C+ inflammatory monocytes) are able to differentiate to TADCs and TAMs (6). Importantly, it was recently demonstrated that TAMs can originate directly from bone marrow, rather than being alternatively activated from mature macrophages (2, 7).
Ample evidence indicates that the mechanisms used by these IMCs include, but are not limited to, production of reactive oxygen species (ROS), nitric oxide (NO), IL10 and TGFβ, nitration of T-cell–specific chemokines and T-cell receptors (TCR), ineffective costimulations, induction of T regulatory cells (Treg), and exhaustion of CD8+ T cells (4, 5, 7, 8).
The gene encoding CC chemokine ligand 5 (CCL5) is amplified in human breast cancers (9). Antagonism of CCL5-CCR5 signaling of cancer cells has been hypothesized to prevent metastasis (10). We previously observed the generation of aberrant MDSCs in 4T1 tumor-bearing ccl5−/− mice (11). However, important questions remain unanswered, such as whether the aberrant ccl5−/− MDSCs are functionally deprived, the mechanism by which myeloid CCL5 regulates MDSCs, and how to specifically target myeloid CCL5. We revisited the subject and identified the autocrine CCL5–CCR5 axis as a profound player in myeloid compartment. It essentially to determine the immunosuppressive phenotypes of all major IMCs, including granulocytic G-MDSCs / TANs and monocytic TAMs. An immunohistochemistry analysis of triple-negative breast cancer (TNBC) specimens showed that this CCL5–CCR5 axis also regulated IMCs in patients. Because of the off-target activation of tumor-promoting myeloid cells by Fc-portion of an antibody (12), it is necessary to develop non-antibody tools to specifically inhibit the autocrine CCL5–CCR5 signaling in bone marrow. We explored the efficacy of porous silicon-based, bone marrow CCL5-targeting nanoparticles, and its synergistic effects with FDA-approved CCR5 inhibitor, Maraviroc. Our studies point to a great potential of an immunotherapy targeting myeloid CCL5–CCR5 axis.
Materials and Methods
WT BALB/c, WT C57BL/6 mice and ccl5−/− mice on C57BL/6 background were purchased from The Jackson Laboratory. ccl5−/− mice on BALB/c background were generated as previously described (11). All mice were maintained in a pathogen-free facility. All animal protocols are proved by the Research Animal Resource Center at Weill Cornell Medicine.
4T1 cell line (CRL-2539) was obtained from the ATCC in 2012. Cells were maintained RPMI1640 supplemented with 10% FBS, 2 μmol/L glutamine and 100 U/mL penicillin and 100 μg/mL streptomycin (referred as complete media hereafter). The cell line was most recently authenticated in March of 2017 by Genetica DNA Laboratories. The authentication test involved generating STR DNA profiles of the 4T1 cell line for 15 independent human genetic sites and amelogenin (the sex identity locus), and confirmed the lack of human cell contamination.
Breast tumor models and tumor measurement
4T1 and PyMT breast tumor model were established as described previously (13, 14). In our studies, 5 × 104 4T1 cells and 1 × 106 primaryPyMT cells/Matrigel (BD Biosciences) mixture were subcutaneously injected into the mammary pad of adult female BALB/c mice and C57BL/6 mice, respectively. Tumors were measured every other day using an electronic caliper. Tumor volume was calculated using the equation (length x width2)/2. The mean value of tumor volumes of each group was used to plot tumor growth curves.
Bone marrow-MDSCs (CD11b+/Gr-1+) were sorted as described previously (11). To sort bone marrow Ly6C+ M-MDSCs, the sorted CD11b+/Gr-1+ were briefly treated with multisort stop reagent, and further negatively selected against Ly6G via a LD column, followed by positive selection against Ly6C via a LS column. To sort tumor-infiltrating MDSCs, tumor was minced and digested with tissue dissociation buffer [0.25% collagenase IV (384 unit/mg; Worthington), 0.2% Dipase II (Roche) and 0.01% DNase I (Sigma) in HBSS] with periodic vortexing for 1 hour in 37°C water bath. Digested tissues were mashed through 70-μm filters, layered on a 20% and 80% Percoll gradient (GE), and centrifuged at 2800 rpm for 20 minutes without brake. Cells at the interface were collected and negatively selected by anti-B220, -CD4 and -CD8 microbeads via a LD columns, followed by a positive selection by anti-CD11b microbeads via LS columns to obtain CD11b+/Gr-1+ intratumoral MDSCs. To further separate intratumoral Ly6C+ from Ly6G+, CD11b+/Gr-1+ cells were treated with multisort stop reagent, followed by anti-Ly6G microbeads via a LS column. The Ly6G-depleted population was further selected via Ly6C microbeads via second LS column. All the sorted cells were ≥ 85% viable. The purities of bone marrow-MDSCs and tumor-infiltrating MDSCs are approximately 95% and approximately 81%, respectively (verified by flow cytometry). All microbeads and columns were from Miltenyi Biotechnology.
Flow cytometric analysis
The erythrocyte-depleted single-cell suspension from bone marrow, inguinal lymph nodes and tumor (as described in “Cell sorting”) were incubated with anti-FcγR antibody, followed by incubation with fluorescently labeled antibodies. Ly6G-PE, Ly6C-PerCP/Cy5.5, CD11b-FITC, CD4-PE/Cy7, and MHCII- PE/Cy7 were purchased from Biolegend. Gr-1-PE, Foxp3-FITC, Granzyme B-PE, MHCII-PE, CD11b-PerCP/Cy5.5, LAMP2-eF660, and CD8-AF647 were purchased from eBioscience. PD-1-APC, CCR5-APC and CCR3-PE were purchased from Myltenyi Biotec. Purified anti-CCL5 (Peprotech), secondary APC-conjugated goat anti-rabbit (Columbia Bioscience), iNOS-FITC (BD Biosciences), and CCR2-APC (R&D), CCR1 (R&D) were also acquired from different commercial sources. Data acquisition was performed using FACScan (Becton Dickinson) and analyzed via FlowJo (Tree Star, Inc).
Evaluation of secretion of NO and ROS
WT or ccl5−/− bone marrow-MDSCs were cultured in complete media supplemented with GM-CSF (20 ng/mL), IL6 (20 ng/mL), and 30% 4T1 supernatant (referred as MDSC media hereafter). Detection of NO (red) and ROS (green) was performed according to instruction manual of ROS/RNS Detection Kit (Enzo Life Science).
A total of 1.5 × 105 WT or ccl5−/− bone marrow-MDSCs were seeded in 24-well plate with 0.5 mL MDSC media. ELISA on TGFβ (R&D) and IL10 (Biolegend) were performed as instructed by manufacturers. To stimulate IL10 secretion, 0.5μg/mL LPS was added to the culture 24 hours before supernatant collection.
Evaluation of the morphologies of tumor-infiltrating Ly6G+ cells
Tumor-infiltrating Ly6G+ cells sorted as described above were further sorted against Ly6G/Ly6C markers using Becton-Dickinson Vantage cell sorter. Sorted populations (P1-P3) loaded onto slides using the Cytospin (Thermo Scientific). Attached cells were fixed and dried, followed by hematoxylin and eosin (H&E) staining. Cell morphologies was evaluated under light microscopy (Olympus BX51, 60× oil).
RNA Isolation, RT-qPCR, and RNA sequencing
Total RNA was extracted from various sorted cell populations using the RNeasy Plus Mini Kit (Qiagen). cDNA synthesis was performed using SuperScript VILO cDNA Sythesis Kit (Invitrogen Life Technology). RT-qPCR was performed on ABI PRISM 7900HT (Applied Biosystem). Second generation of RNA sequencing was performed by GenoIMCs Core Facilities at Weill Cornell Medicine. Primer information was listed in Supplement.
Tumor-infiltrating Ly6Clow/−/Ly6Ghigh, Ly6Chigh/Ly6Ghigh and Ly6Cint/Ly6Gint cells sorted as described in “Evaluation of the Morphologies of Tumor-infiltrating Ly6G+ Cells” were cultured in MDSC media. After 2-hour initial culture, sorted Ly6G+ cells were cocultured with sonicated fluorescence-labeled (494/518) Escherichia coli K-12 bioparticles (V6694, Life Technologies) for another 2 to 4 hours in dark. Phagocytosis abilities were evaluated according to the manufacturer's instruction. Bone marrow–derived mature macrophages and 4T1 cells served as a positive control and a negative control, respectively.
Immune function assay
Tumor-infiltrating MDSCs (Gr-1+/CD11b+) cells were sorted as described above. T cells isolated from splenocytes (B220 microbeads-mediated depletion, followed by CD4/CD8 microbeads-mediated selection) were stimulated with soluble CD3 (0.5 μg/mL) and CD28 (0.5 μg/mL) antibodies in the presence of tumor-infiltrating MDSCs at different ratios in MDSC media for 6 hours. Ten μmol/L Edu was added to the culture, and cells were allowed to proliferate for another 18 hours. Proliferation was evaluated on the basis of Edu incorporation according to the manufacturer's instruction (C10337, Life technologies).
Same number of sorted MDSCs from different groups were lysed with SDS PAGE sample buffer (containing 0.2 mol/L DTT, 2% SDS, 5% glycerol, 0.06 mol/L Tris·Cl pH:6.8 and 0.002% bromophenol blue). Lysates were boiled briefly and separated with 8% to 16% gradient SDS PAGE gel (Genescript) followed by standard transfer to nitrocellulose membranes and immunoblotting against various target proteins. Anti-Rb (C-15), anti-NOS2 (M-19), anti–p-JAK3 (Tyr980), anti-GAPDH (FL335), and anti–p-IĸB-α (B-9) were purchased from Santa Cruz Biotechnology.
MDSC adoptive transfer
Adoptive transfer of MDSCs was performed between mice bearing 4T1 tumor at same stage. A total of 5 × 106 bone marrow M-MDSCs (CD11b+/Ly6C+) sorted from ccl5−/− mice bearing 4T1 tumors were mixed with 25 nmol/L Qdot nanocrystal dye (Qtracker 525 Cell Labeling Kit, Life Technologies) and incubated at a humidified atmosphere of 5% CO2 at 37°C for 2 hours based on the manufacturer's guidance book. Successful loading of green fluorescence was confirmed with flow cytometer. Green fluorescent ccl5−/− bone marrow-MDSCs were injected intravenously into WT tumor-bearing recipient mice. Same adoptive transfer was performed from WT to WT as a control. Mice were sacrificed for endpoint analyses at 7th to 8th day after adoptive transfer.
All TNBC specimens were collected from Asian females with age ranging from 21 to 71 (detailed clinical profiles in Supplementary Table S1) at Xiangya Hospital, Central South University, China. All the clinical samples were surgically excised and immediately fixed with 10% neutral buffered formalin. All the patients signed an agreement of informed consent (IRB number: 201308381, Xiangya Hospital) for the use of their cancer samples and publication of the anonymized data.
Formalin-fixed paraffin-embedded TNBC sections from 128 patients were dewaxed and rehydrated in Xylene, 100% ethanol, 95% ethanol, 70% ethanol, 50% ethanol, and PBS sequentially. Acidic antigen retrieval and serum blocking were performed before incubation with anti-human CCR5 primary antibody (R&D), followed by applications of biotinylated secondary antibody, Streptavidin-HRP, AEC-chromogen (R&D CTS003) and hematoxylin (CCR5: red to brown; hematoxylin: blue). After initial microscopic scanning on tumor sections, 62 out of 128 patients are eligible for calculating statistical significance (tumor sections, which have more than 10 intratumoral neutrophils for evaluating the nuclear morphologies, were considered eligible). On the basis of the CCR5 levels of infiltrating immune cells, 62 patients were grouped into immune CCR5high (n = 29) and immune CCR5low (n = 33) groups. The corresponding H&E-stained sections (4 sections/patient) were employed to evaluate the maturation status of neutrophils. Neutrophils with cloverleaf-shaped nuclei (3–4 nuclear lobes) were considered mature neutrophils. Immature neutrophils are mixed population with horse-shoe-shaped or 2-lobed nuclei.
MSV nanoparticle synthesis.
Porous silicon–based multistage vector (MSV) nanoparticles were fabricated by electrochemical etching of silicon wafer, surface modified with 3-aminopropyltriethoxysilane (APTES), and conjugated with E-selectin thioaptamer (ESTA) as previously described (15). To prepare CCL5 siRNA polyplex, CCL5 siRNA oligos were mixed with PEG(5k)2–PEI(10k; PEG–PEI; nitrogen in cationic polymer: phosphorus in siRNA oligo ratio = 15:1) in 10 μmol/L HEPES buffer containing 5% glucose, and incubated at 20°C for 15 minutes. To load CCL5 siRNA polyplex into ESTA-MSV, 200 μL polyplex suspension containing 20 μg siRNA was mixed with 1 × 109 dry ESTA-MSV particles, and sonicated for 3 minutes on ice. The siRNA loading efficiency was 87.3% ± 0.9%. Tumor-bearing mice (PyMT and 4T1) were intravenously injected with 100 μL nanoparticles once a week.
Maraviroc was purchased from Sigma (PZ002). 4T1 tumor-bearing mice were treated with Maraviroc at 8 mg/kg (i.p.) every day after tumor inoculation.
The Student t test was employed to calculate statistical significance for difference between groups. A P value of <0.05 was considered statistically significant (*, P < 0.05; **, P < 0.01; ***, P < 0.001). Kaplan–Meier estimate was used to analyze the 5-year-survival rates of TNBC patients.
Enhanced maturation of intratumoral neutrophils in the absence of host CCL5
4T1 tumor microenvironment (TME) is a complex ecology that favors inflammation. CCL5 is one of the elevated chemokines in 4T1 TME (16). Therefore, 4T1-bearing ccl5−/− mice represent an excellent model to study CCL5-deficient myeloid cells in CCL5-sufficient milieu. Compared with WT mice, ccl5−/− mice have approximately 50% rate of potent rejection (Supplementary Table S2) and much smaller total burden of engrafted 4T1 tumor (Fig. 1A). We inspected the H&E-stained tumor sections from WT and ccl5−/− mice. Despite comparable levels of tumor-derived CCL5 between WT and ccl5−/− hosts (Supplementary Fig. S1), marked morphology differences of intratumoral Ly6G+ cells were discerned. On tumor sections from WT, polymorphonuclear populations with horse-shoe-shaped nuclei were observed; in comparison, cells with hypersegmented / cloverleaf-shaped nuclei were observed on sections from ccl5−/− (Fig. 1B). The nuclear morphologies of ccl5−/− polymorphonuclear cells signify a more mature form of neutrophils (17), and have been associated with anticancer neutrophils, namely N1 (1).
It is generally accepted that myeloid-derived suppressors are a heterogeneous population deviated from full maturation because of tumor-elicited inflammation (18). We postulated that such a deviation was dependent on host CCL5. We first sorted the Ly6G+ populations from 4T1 tumors via magnetic beads. Flow-cytometric analysis showed that the tumor-infiltrating Ly6G+ population spawned from WT bone marrow consisted mostly of G-MDSCs (Ly6Clow/−/Ly6Ghigh, Population 1). Two distinct Ly6G+ populations in tumor carried by ccl5−/− mice were observed: Ly6Cint/Ly6Gint (Population 2) and Ly6Chigh/Ly6Ghigh (Population 3; Fig. 1C). These populations referred as P1, P2 and P3 hereafter were further FACS-sorted and examined for cell morphologies using H&E staining. Consistently, the tumor-infiltrating P1 derived from WT bone marrow are polymorphonuclear with nuclei curved into a horse-shoe shape, whereas the intratumoral P2 derived from ccl5−/− bone marrow showed hypersegmented/cloverleaf-shaped nuclei. The cells of P3 showed ring-shaped nuclear morphology (Fig. 1D). The side-scatter parameters (SSC) of these populations agreed with the microscopic observations: the cells of P2 had the highest SSCs (Fig. 1E), confirming the highly complex nuclear morphology of P2. In addition to the nuclear morphology, mature neutrophils exhibit significantly higher expression of lysosomal proteins and stronger phagocytosis activity than tumor-induced G-MDSCs (19). To verify that the P2 were mature neutrophils, we compared LAMP2 expression and phagocytosis abilities between WT P1 and ccl5−/− P2. The cells of ccl5−/− P2 demonstrated higher LAPM2 expression and stronger phagocytosis abilities than WT P1 (Fig. 1F and G).
Lastly, we examined the costimulatory molecule CD86 on sorted tumor-infiltrating Ly6G+ myeloid cells. Unlike CD80, CD86 is a costimulatory molecule usually associated with immunogenic antigen-presenting cells (APC; ref. 20), but not with immunosuppressive functioning of MDSCs (21, 22). The tumor-infiltrating Ly6G+/ MHCII+ cells from WT host expressed low levels of surface CD86. This costimulatory molecule was found to be positive on almost all MHCII+/Ly6G+ cells in tumor carried by ccl5−/− mice (Fig. 1H), suggesting that CCL5-deficient MHCII+/Ly6G+ cells more closely resemble mature APCs (23), and could possibly activate T-cell responses.
Taken together, our findings indicate that host CCL5, but not tumor-derived CCL5, plays an important role in impeding the maturation of neutrophils under tumorigenesis-associated inflammation; hence, facilitating the generation of immature Ly6G+ myeloid cells, which are “immunosuppressive” in nature.
Arrested differentiation of ccl5−/− bone marrow precursors into TAMs
It has been recently demonstrated that TAMs are different from alternatively activated macrophages (M2), and can directly derive from bone marrow Ly6C+/CCR2+ inflammatory monocytes. In contrast to nonproliferative CD11bhigh/MHCIIhigh mammary tissue macrophages (MTM), TAMs exhibit CD11blow/MHCIIhigh and CD11c+; they also proliferate upon differentiation (7). Proliferation is a feature ascribed to immature myeloid precursors, but not terminally differentiated macrophages (24). Being positive for CD11c also blurs the identity of TAMs between a dendritic cell and a macrophage. We hypothesized that TAMs are a heterogeneous monocytic population prevented from full maturation in tumor, and might also be regulated by the host CCL5 signaling.
We first examined the precursor of TAMs in bone marrow, and further traced the Ly6C+ cells in tumors. We noted the presence of extravasating inflammatory monocytes (CCR2+/CX3CR1low; ref. 25) in WT bone marrow, but not in ccl5−/− bone marrow, in response to tumor growth (Fig. 2A). Approximately 76% of intratumoral Ly6C+ cells from WT were CCR2+ with upregulation of F4/80, indicating an in situ transition from infiltrating monocytes to TAMs (7). In contrast, the equivalent intratumoral Ly6C+ cells from ccl5−/− bone marrow were CCR2int/− with significantly less upregulation of F4/80 in the presence of tumor-derived CCL5 (Fig. 2B). We further observed that TAMs (CD11blow/MHCIIhigh) dominated numerically over MTMs (CD11bhigh/MHCIIhigh) with a ratio of 2.5:1 in 4T1 tumors from WT hosts. Such a preponderance of TAMs over MTMs was not seen in tumors carried by ccl5−/− (Fig. 2C), suggesting that the differentiation to TAMs is possibly dependent on host CCL5. To accurately enumerate TAMs, tumor-infiltrating cells positive for Vcam1, CD11c and MHCII were counted using flow cytometer (7). We observed significantly less TAMs in ccl5−/− mice when compared with WT counterparts (P < 0.01; Fig. 2D). In line with the proliferative feature of TAMs, the CD11b+/MHCII+-gated intratumoral cells from WT showed approximately 70% Edu incorporation; in contrast, only approximately 15% Edu incorporation was observed in the same cell population from ccl5−/− (Fig. 2E).
To determine whether the failed differentiation from inflammatory monocytes to TAMs was a consequence of the absence of host CCL5 in bone marrow, adoptive transfer of bone marrow Ly6C+ cells from tumor-bearing ccl5−/− to WT counterparts was performed. CD11b+/Ly6C+ cells sorted from ccl5−/− bone marrow were loaded with Qdot nanocrystal dye (green fluorescence 485/525) before intravenous transfer to tumor-bearing WT mice. As a control, WT CD11b+/Ly6C+ cells loaded with same dye were also transferred (intravenous) to tumor-bearing WT mice. TAMs (CD11blow/MHCIIhigh) and MTMs (CD11bhigh/MHCIIhigh) in recipient mice were gated and further analyzed against green fluorescence. As shown in Fig. 2F, the TAM population in mice that received ccl5−/− bone marrow Ly6C+s was largely dim for green fluorescence; whereas the MTM population showed detectable emissions. These results indicate that the majority of TAMs were derived from endogenous WT myeloid precursors (non-fluorescent) rather than the adoptively transferred CD11b+/Ly6C+ cells from ccl5−/− bone marrow (green fluorescent); in the same proinflammatory milieu, myeloid precursors derived from ccl5−/− bone marrow inclined to differentiate into non-proliferative MTMs, which more closely resemble mature macrophages. Such a preferential differentiation into MTMs was not seen when donor Ly6C+s were from WT bone marrow. Collectively, these data reveal host CCL5 in bone marrow as a critical player in the TAM differentiation during tumorigenesis.
Altered monocytic and granulocytic subtypes of GR-1+ cells in the absence of CCL5
Normal monocytes/macrophages and neutrophils share a common myeloid progenitor. As described above, both monocyte- and granulocyte-lineages are inclined toward more mature forms in a TME lacking host CCL5, but not tumor-derived CCL5. These observations prompted us to investigate their common bone marrow precursors (Gr-1+ population) that comprises both immature granulocytes (Ly6G+) and TAM precursors (Ly6C+; ref. 6).
We first sought to determine the cellular source of CCL5 in myeloid compartment by screening bone marrow CD34+, CD90+, and Gr-1+/CD11b+ (MDSCs) cells for intracellular CCL5 expression. The analysis revealed that in both 4T1 and MMTV-PyMT (PyMT) mammary tumor models, bone marrow Gr-1+/CD11b+ themselves secreted CCL5 after tumor inoculation (Supplementary Fig. S2; Fig. 3A; Supplementary Fig. S3).
When compared with MDSCs in tumor-bearing ccl5−/− mice, MDSCs in naïve ccl5−/− exhibited the same atypical Ly6Chigh/Ly6Ghigh phenotype, but at a lower abundance. These observations suggest that the absence of autocrine CCL5, not 4T1 tumor-derived factors, caused the accumulation of the atypical Ly6Chigh/Ly6Ghigh population with a deficiency in the generation of the G-MDSC subset (Fig. 3B). Of note, Youn and colleagues (26) suggested that a certain proportion of M-MDSCs (Ly6Chigh/Ly6G−) acquired phenotypic and morphological features of G-MDSCs (Ly6Clow/−/Ly6Ghigh) upon epigenetic silencing of the retinoblastoma gene (Rb1). We cultured M-MDSCs sorted from WT or ccl5−/− bone marrow in vitro (27). Approximately 15% WT M-MDSCs gained the phenotype of G-MDSCs by downregulating Ly6C and expressing Ly6G after 4 days in culture; this typical G-MDSC population was nearly not detectable in ccl5−/− M-MDSC culture. Instead, approximately 35% ccl5−/− M-MDSCs became the atypical Ly6Chigh/Ly6Ghigh cells (Fig. 3C). We next examined the mRNA and protein levels of Rb1 in WT and ccl5−/− bone marrow MDSCs from mice with or without tumor. A failure in epigenetic silencing of Rb1 in ccl5−/− MDSCs was revealed in response to tumor growth (Fig. 3D). These observations demonstrate that autocrine CCL5 is a key regulator of Rb1 activation and in the development of the G-MDSC subset. Detailed demonstrations of MDSCs' developmental alterations were included in Supplement (Supplementary Fig. S4).
Profound functional defects of the arrested Ly6Chigh/Ly6Ghigh cells
To confirm that the subtype-switching-arrested ccl5−/− MDSCs are functionally inactive, we analyzed differences between sorted Ly6Clow/−/Ly6Ghigh (WT) and Ly6Chigh/Ly6Ghigh (ccl5−/−) cells via next-generation RNA-sequencing. The gene-expression profiling demonstrated that Ly6Clow/−/Ly6Ghigh bone marrow MDSCs from WT bearing 4T1 tumor turned off expressions of genes related to red blood cell biosynthesis and survival, but upregulated expression of proinflammatory [e.g., Saa3, Prok2, Ptges2 (COX-2), Lipg (endothelial lipase), CXCR1, VEGFα, IL6Ra, CD14 and TLR1] and immunosuppressive (IL4Rα, IL13Ra1, and NOS2) molecules. In contrast, Ly6Chigh/Ly6Ghigh MDSCs from ccl5−/− bone marrow failed to upregulate these genes to a large extent in response to tumor growth, suggesting functional defects of ccl5−/− MDSCs (Fig. 4A). Some of the key findings from the sequencing profiles were corroborated by flow cytometry, RT-qPCR and etc. (Fig. 4B and C; Supplementary Fig. S5). When compared with WT, significantly lower expression of ROS, NO, IL10, and TGFβ by Ly6Chigh/Ly6Ghigh (ccl5−/−) were observed (Fig. 4D and E).
We further examined tumor-infiltrating CD8+ T cells and splenetic T-cell counts (CD4+/CD3+ and CD8+/CD3+ cells) in WT versus ccl5−/− mice bearing 4T1 tumor (4 weeks), which could largely be affected by nitration of T-cell–specific chemokines (28) and by TGFβ-suppressed T-cell proliferation/NO-induced T-cell apoptosis, respectively (29–32). Indeed, ccl5−/− mice showed both enhanced tumor infiltration of CD8+ T cells and higher splenetic T cells counts when compared with WT peers (Fig. 4F and G).
Together, these observations confirm that the lack of autocrine CCL5 results in aberrant Ly6Chigh/Ly6Ghigh MDSCs with multitiered deficiency of immunosuppressive functions in bone marrow.
Unaltered phenotypic and functional defects of ccl5−/− myeloid-suppressor cells in tumor
To determine whether exposure of ccl5−/− myeloid-suppressor cells to tumor-derived CCL5 could “correct” their phenotypic features (Ly6Chigh/Ly6Ghigh) and functional defects, we analyzed the tumor-infiltrating Gr-1+ myeloid cells. The intratumoral Gr-1+/CD11b+ originated from ccl5−/− bone marrow remained Ly6Chigh/Ly6Ghigh with unsilenced Rb1 expression (Fig. 5A and B). When compared with Ly6Clow/−/Ly6Ghigh cells, the predominant MDSC subset in tumor carried by WT, a consistently lower expression of NOS2 was observed in tumor-infiltrating Ly6Chigh/Ly6Ghigh cells (Fig. 5C). Furthermore, tumor-infiltrating Gr-1+/CD11b+ cells spawn from ccl5−/− bone marrow showed significantly weakened ability to inhibit anti-CD3/CD28–induced T-cell proliferation than the WT counterparts (Fig. 5D). Together, CCL5 in TME appeared incapable of compensating the deficiency of host CCL5 in bone marrow.
IMCs have all been demonstrated to induce “exhaustion” in effector T cells and expansion of regulatory T cells (Tregs; refs. 1, 33, 34). To further demonstrate that ccl5−/− myeloid cells are able to activate productive T-cell responses in CCL5-sufficient peripheries, we analyzed the expression of granzyme B (GzmB) and PD-1, markers for cytolytic activities and an inhibitory coreceptor of T cells, respectively, by tumor-infiltrating CD8+ T cells. In WT, GzmB− cells made up most of tumor-infiltrating CD8+ T cells, of which approximately 55% expressed PD-1. On the contrary, approximately 95% tumor-infiltrating CD8+ T cells in ccl5−/− mice were PD-1−, of which approximately 35% were also GzmB+ (Fig. 5E). Consistent with this observation, CD4+/Foxp3+ Tregs in the inguinal lymph nodes draining the 4T1 tumor were 37 times more abundant in WT than ccl5−/− (Fig. 5F).
We further explored the impacted signaling pathways downstream of CCL5 receptors (CCR1, CCR3, and CCR5) in Gr-1+ myeloid cells due to the absence of autocrine CCL5. The RNAseq profiling of CD11b+/Gr-1+ bone marrowcells revealed different mRNA expression of CCR5, but not CCR1 and CCR3, between WT and ccl5−/− tumor-bearing mice (Fig. 5G). Flow cytometric analysis confirmed this expression pattern of CCL5 receptors (Fig. 5H and I). We further observed that the defective expression of CCR5 was conserved from bone marrow-residing to tumor-infiltrating ccl5−/− Gr-1+ cells (Fig. 5J). The persistent lack of membrane-bound CCR5 on may explain why supplementing tumor-derived CCL5 in trans could rescue neither phenotypic nor functional defects of ccl5−/−Gr-1+ cells. To verify the deficiency of CCL5-CCR5 signaling in both bone marrow and tumor-infiltrating Ly6Chigh/Ly6Ghigh cells, its downstream targets such as JAK/STAT (35, 36) and NFκB (37, 38) were inspected by immunoblotting. Indeed, the elevated phosphorylation of JAK3 and degradation of phosphorylated IκBα were observed in sorted CD11b+/Gr-1+s from tumor-bearing WT mice, but not in the ccl5−/− counterparts (Fig. 5K). The dampened JAK/STAT and NFκB signaling transductions of ccl5−/− MDSCs, in turn, explain their loss of function phenotype, because these pathways are known to mediate the immunosuppressive function of MDSCs (39).
Collectively, these findings indicate that (i) the autocrine CCL5 is required for the initial expression of CCR5 on developing Gr-1+ cells in bone marrow; (ii) the functional defects of Ly6Chigh/Ly6Ghigh myeloid cells are long-lasting due to the persistent lack of CCR5.
Therapeutic potential of the targeting myeloid CCL5–CCR5 axis in breast cancer
Inspired by our findings in mouse models, we further investigated the autocrine CCL5–CCR5 axis in patients with TNBCs. As we showed previously, CCR5 expression on myeloid cells was dependent on the autocrine CCL5 in bone marrow. In addition, tumor-infiltrating immune cells have a typical morphology of small and compact basophilic nuclei, thus they can be reliably differentiated from cancer cells on H&E-stained specimens (40). We, therefore, used the CCR5 expression in the infiltrating immune cells, not cancer cells, as an index of myeloid CCL5 in TNBC patients.
Immunohistochemical analysis demonstrated that “immune CCR5” (Fig. 6A) inversely correlated with the maturation status of neutrophils in tumor (Fig. 6B and C), as well as 5-year-survival rates of these patients (Fig. 6D; Supplementary Table S1).
On the basis of the mouse and human data, we reasoned that targeting CCL5 in bone marrow would have great therapeutic efficacy against breast cancer. Because CD11b+/Gr-1+s make up approximately 85% of the bone marrow cells in tumor-bearing mice, we designed bone marrow–targeting nanoparticles (Fig. 7A), which comprised two main moieties: (i) the biodegradable mesoporous silicon nanoparticles (41) incorporated with liposomal CCL5 siRNA; (ii) surface-conjugated affinity ligand (thioaptamer) for E-selectin expressed on bone marrow endothelium (15). The efficiencies of bone marrow–targeting and siRNA-releasing were verified experimentally before animal treatments (Fig. 7B).
CCL5-siRNA-loaded MSV nanoparticles were delivered to 4T1 and PyMT tumor-bearing mice via tail vein injection once a week. This regimen was chosen due to the sustained release of siRNA liposomes from MSV (42). Compared with PBS (mock) and control nanoparticle-treated mice, significantly reduced 4T1 tumor growth in mice receiving CCL5-targeting nanoparticles was observed after the second treatment (Fig. 7C). After three nanoparticle treatments, reductions of tumor burden (4T1: ∼65%; PyMT: ∼30%–40%) were evident (Fig. 7D; Supplementary Fig. S6A). Consistently, the enlargement of spleen, a sign for extramedullary hematopoiesis in which G-MDSCs significantly increase, reduced dramatically in 4T1-bearing mice receiving nano-treatment (Fig. 7E). The efficacy of nanoparticle-treatment was further corroborated by examining the mRNA levels of CCL5, CCR5, and NOS2 of bone marrow-MDSCs (Supplementary Fig. S6B). CCL5-targeting nano-treatment in bone marrow considerably reprogramed immunosuppressive myeloid cells, as evidenced by arrested M- to G-subset switching (Fig. 7F), decreased NOS2 expression of MDSCs, and crippled TAM development (Fig. 7G). The reprogrammed IMCs further led to reinvigorated antitumor immunities, exemplified by enhanced CD8+ T-cell infiltration (Fig. 7H) and approximately 9-fold increase of GzmB+/PD-1− /CD8+ T cells in tumor (Fig. 7I). We further proposed that the residual tumor growth is the consequence of incomplete inhibition of CCL5 in bone marrow, and the blockade of CCR5 by Maraviroc in periphery could synergistically improve the therapeutic efficacy. As expected, combination of CCL5 nanoparticles and Maraviroc led to vigorous reductions of 4T1 tumors (Fig. 7J and 7K). 2 out of 8 mice rejected tumors at day 16 and day 20 upon the synergistic treatments. Unsurprisingly, further enhanced antitumor immunities were observed (Fig. 7L).
IMCs include mainly a granulocyte lineage (G-MDSCs or TANs) and a monocyte lineage (TADCs and TAMs). During tumorigenesis, Ly6C+ myeloid cells further differentiate into both granulocyte-like-MDSCs/TANs (26, 43, 44) and monocyte-derived TAMs (6). In our study, the absence of autocrine CCL5 in evolving myeloid suppressors in bone marrow causes defective expression of CCR5. In TME, the persistent lack of membrane CCR5 allows the CCL5-deficient myeloid cells to largely neglect tumor-derived CCL5, thereby gaining the ability to advance toward more mature neutrophils and macrophages, not the otherwise immature myeloid suppressor cells. Genetic ablation of ccl5 gene or nanoparticle-mediated silencing of host CCL5 in bone marrow gives rise to not only the incomplete G-MDSC development, but also the suspended differentiation to CD11c+ TAMs (Cartoon illustrations in Supplementary Fig. S7). The concomitantly enhanced maturation of “pan-myeloids” in tumor, as well as the subsequent productive T-cell responses point to an important role of the autocrine CCL5–CCR5 axis in myeloid-originated suppression. However, the molecular mechanism by which CCL5 regulates CCR5 expression in MDSCs needs further investigations.
Our findings also highlight the importance of MDSC subtype-switching. Despite their low proliferation rates and relatively weak immunosuppressive activities on a per cell basis in comparison to M-MDSCs, G-MDSCs are the predominant form of MDSCs in animal tumor models and cancer patients (26, 45). The advantage underlying the transition from potent M-MDSCs to the less powerful and short-lived G-MDSCs as tumor progresses is not clear. Our whole exon RNA sequencing of 4T1 tumor-infiltrating Ly6G+ versus Ly6C+ MDSCs revealed that certain key immunosuppressive molecules such as S100A8/A9, NOS2, and PD-L1 were upregulated in Ly6G+ G-MDSCs, but not in Ly6C+ M-MDSCs (Supplementary Fig. S8). Consistent with their immune-activating phenotypes, ccl5−/− intratumoral MDSCs showed not only the reduced expression of S100A8/A9 and NOS2, but also the upregulation of immunostimulatory CD86 on Ly6G+ MDSCs (Supplementary Fig. S8).
We propose that immunosuppressive functions are likely derived from G-MDSCs, as evidenced by the direct demonstration of tumor-promoting neutrophils (1) and increased antitumor effect upon depletion of Ly6G+ cells, but not Ly6C+ cells (46–48). This notion also explains, at least in part, that the arrested Ly6Chigh/Ly6Ghigh MDSCs, which failed to develop into the typical G-MDSCs, have lost most of immunosuppressive activities. The observation that M-MDSCs are more immunosuppressive than G-MDSCs on a per cell basis could be the combined consequences of a higher proliferative capacity of M-MDSCs and the transition from M- to G-MDSCs, especially the Rb1low population (26). However, more research is needed to confirm the differential roles of MDSC subsets. Another limitation derives from the techniques (Percoll gradient centrifugation) employed to isolate MDSCs. The immunosuppressive cells with low densities may consist of other types of immune cells, which could be an independent project for the further studies.
Mounting evidence has put IMCs firmly on the map as a negative regulator of antitumor immunity. Various strategies are, therefore, sought to eliminate myeloid-originated immunosuppression. However, targeting a limited number of molecules of IMCs may not be effective, since multiple suppressive mechanisms with varying prevalence have been proposed to be operated by tumor at different stages. Our studies demonstrated that an autocrine CCL5–CCR5 axis is poised at the apex of the cascade directing bone marrow precursors toward IMCs of granulocyte- and monocyte-lineages during tumorigenesis. Our analysis of TNBC patients further suggests that bone marrow CCL5–CCR5 axis represent an excellent target in cancer immunotherapies.
Disclosure of Potential Conflicts of Interest
M. Ferrari has ownership interest (including patents) in Nanomedical Systems, Inc. No potential conflicts of interest were disclosed by the other authors.
Conception and design: Y. Ban, L. Chouchane, H. Shen, X. Ma
Development of methodology: Y. Ban, J. Mai, X. Li, L. Zhang, H. Shen
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Ban, J. Mai, M. Mitchell-Flack
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Ban, T. Zhang, L. Chouchane, X. Ma
Writing, review, and/or revision of the manuscript: Y. Ban, L. Chouchane, M. Ferrari, H. Shen, X. Ma
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Mai, X. Li
Study supervision: H. Shen, X. Ma
We thank Dr. Aihao Ding (Weill Cornell Medicine) and Dr. Ming Li (Memorial Sloan Kettering Cancer Center) for insightful discussions on TAM differentiation. Celecoxib was a generous gift from Dr. Andrew Dannenberg (Weill Cornell Medicine).
This work was supported by New York State Department of Public Health grantC028251 to X. Ma, Qatar National Research Foundation grantNPRP 7-136-3-031 to L. Chouchane and X. Ma, and NIH training grantT32 5T32AI007621-15 to Y. Ban, the Ernest Cockrell Jr. Distinguished Endowed Chair to M. Ferrari, and the NIH grants U54 CA210181 to M. Ferrari, and 1R01CA193880-01A1 to H. Shen. This publication was made possible by NPRP grant (7-136-3-031) from the Qatar National Research Fund (a member of Qatar Foundation). The findings achieved herein are solely the responsibility of the author[s].
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.