Retinoblastoma Inactivation Induces a Protumoral Microenvironment via Enhanced CCL2 Secretion

The tumor microenvironment (TME) in solid tumors consists of extracellular matrix as well as the associated stromal cells including immune cells, fibroblasts, and vascular networks. Bidirectional interactions between tumor cells and the TME enhance tumor progression at multiple levels: by supplying a variety of cytokines and growth factors, and nurturing cancer cells via the promotion of angiogenesis (1). Immune checkpoint blockade (ICB) accelerates the antitumor function of the TME through cytotoxic T-cell–mediated immunosurveillance, which is frequently attenuated by the aberrant expression of immune checkpoint molecules such as PD-L1 (2). Furthermore, the recruitment of immunosuppressive cells such as myeloid-derived suppressor cells (MDSC) and regulatory T cells (Treg) into the TME also leads to a poorly immunogenic tumor; thus the ratio of tumor-infiltrating CD8⁺ cytotoxic T cells to immunosuppressive cells has been shown to predict patient outcomes for cancer immunotherapies using ICB (1).

Recent advances in cancer immunotherapies including ICB have been dramatic; however, they still provide limited benefits for the majority of patients. For example, according to an early-phase I trial in patients with breast cancer, the effectiveness of ICB has been recognized particularly in triple-negative breast cancer (TNBC) due to its lack of targetable molecules such as estrogen receptor (ER) and human EGFR-2 (HER2), with an approximately 20% overall response rate (3). In addition to the expression of immune checkpoint molecules or the frequencies of tumor-infiltrating CD8⁺ cytotoxic T cells in tumor tissues, several groups recently demonstrated that genetic aberrations in cancer cells such as loss-of-function mutations in JAK1/2, APLNR, PTPN2, and PBRM1 are significantly correlated with the efficacy of cancer immunotherapies (4–9). In spite of such extensive efforts, it remains highly challenging to effectively determine which patient will respond to current immunotherapies because of the complexity of tumor heterogeneity in terms of genetic, epigenetic, and/or microenvironment levels in aggressive tumors.

Although cancer cell–intrinsic properties resulting from oncogenic mutations have been well characterized, it is becoming increasingly clear that oncogenic mutations impact angiogenesis and/or the recruitment and phenotype of immune cells in the TME via increasing secretion of cytokines and chemokines by tumor cells (10). The BRAF^V600E mutation in melanoma cells promotes the secretion of multiple cytokines including IL6 and VEGFα, contributing to the establishment of a protumoral microenvironment (11). The oncogenic KRAS mutation in pancreatic ductal adenocarcinoma induced GM-CSF secretion, which enhanced the recruitment of Gr1⁺CD11b⁺ MDSCs and subsequent suppression of CD8⁺ cytotoxic T-cell infiltration in the TME (12, 13). In addition, the loss of tumor suppressor genes such as STK11/LKB1 and PTEN affects T-cell infiltration into the TME by modulating the secretion of cytokines and chemokines (14, 15). We previously reported that retinoblastoma (RB) loss in Trp53-null sarcoma cells and ARF-deficient breast cancer cells increased the secretion of cytokines such as IL6 via enhanced mitochondrial superoxide (MS) production, which stimulates their self-renewal activity in a cell-autonomous manner (16, 17). However, the role of enhanced cytokine production following RB inactivation in the TME remains unclear.

In this study, to elucidate the significance of non–cell-autonomous RB function in tumor tissue, we employed a syngeneic and orthotopic murine soft-tissue sarcoma model and a mammary carcinogenesis model to analyze the TME of RB-deficient tumors. We found that enhanced CCL2 secretion following RB inactivation contributes to the establishment of a tumor-promoting microenvironment due to the recruitment of immunosuppressive cells such as Gr1⁺CD11b⁺ MDSCs and F4/80⁺ tumor-associated macrophages (TAM) into the TME.

Trp53 knockout mice (18) were obtained from RIKEN BioResource Center (#CDB0001K). Ccr2 knockout mice were obtained from Dr. W.A. Kuziel (PDL Bio Pharma Inc., Incline Village, NV; ref. 19). Ccl2 knockout mice (#004434) and MMTV-Cre mice (#003553) were purchased from the Jackson Laboratory. Rb^flox/flox mice were gifted from Dr. A. Berns (Netherlands Cancer Institute, Amsterdam). Wild-type C57BL/6 mice were purchased from Japan SLC. Female of MMTV-Cre; Rb^flox/flox mice for mammary carcinogenesis were never mated to male mice during analysis. Mouse experiments were conducted in accordance with a Kanazawa University Institutional Animal Care and Use Committee-approved protocol (AP-153426).

Minced pieces of soft-tissue sarcoma samples derived from p53-knockout mice were digested with 300 U/mL collagenase (#C-5138, Sigma-Aldrich), 100 U/mL hyaluronidase (#H3506-1G, Sigma-Aldrich) and 100 μg/mL DNase I (#DN25-100MG, Sigma-Aldrich) in α-modified Eagle's medium (αMEM) supplemented with 10% FBS (17). Primary cells from surgically removed human breast carcinomas were established in the laboratory of Dr. Noriko Gotoh (Kanazawa University; ref. 20), and maintained in HuMEC (#12752-010, Life Technologies). The institutional review boards of Kanazawa University (Ishikawa, Japan; #335) and the Institute of Medical Science, The University of Tokyo (Tokyo, Japan), approved this study. RAW 264.7 (RIKEN BRC, RCB0535), THP-1 (RIKEN BRC, RCB1189), MCF7 (RIKEN BRC, RCB1904), and MDA-MB-231 [ATCC (HTB-26)] were cultured in DMEM containing 10% FBS. Hs578t [ATCC (HTB-126)] were maintained in DMEM supplemented with 10% FBS and 10 μg/mL recombinant insulin. HCC1187 [ATCC (CRL-2322)] were maintained in RPMI1640 supplemented with 10% FBS. These cell lines were authenticated by RIKEN BRC and ATCC, and all experiments were performed before reaching 10 passages. Mycoplasma infection was regularly checked by PCR using the conditioned media derived from each cell line (#G238, abm).

MISSION TRC validated shRNA target sets for mouse Rb (TRCN0000042543 and TRCN0000042544), human RB (TRCN0000040163 and TRCN0000010419), and negative control (Scramble; SHC002) were purchased from Sigma-Aldrich. pQCXIH-PSM-RB7LP was purchased from Addgene (#37106), and RB7LP-lacking stop codon was amplified by PCR (16). PCR products were then cloned into pDONR223, and subcloned into pLX304. pCL20c-CMV-EGFP and pCL20c-CMV-EGFP-DN-c-Jun were obtained from Dr. Katsuji Yoshioka (Kanazawa University; ref. 21). Generation and infection of lentivirus were performed according to the manufacturer's instructions.

Cells suspended in 50 μL αMEM with 10% FBS were mixed with 50 μL Matrigel (#354234; Corning) and injected subcutaneously into wild-type or Ccr2 KO C57BL/6 male mice. Tumors were assessed 14–28 days after injection.

tRNA was isolated from cultured cells or tumor tissues by using TRIzol (#15596018, Life Technologies) according to the manufacturer's instructions. qPCR of tRNA was performed as described previously (17) using Taqman probes. Taqman probes: mouse Actb (Mm00607939_s1), mouse Rb (Mm00485586_m1), mouse Vegfα (Mm01281449_m1), mouse Perforin1 (Mm00812512_m1), mouse Ccl2 (Mm00441242_m1), mouse Ccr2 (Mm99999051_gH), mouse Il6 (Mm00446190_ m1), mouse Il1α (Mm00439620_m1), mouse Cxcl1 (Mm04207460_ m1), mouse Cxcl5 (Mm00436451_g1), mouse Ptgs2 (Mm00478374_m1), human ACTB (Hs99999903_m1), human RB (Hs01078066_m1), human CCL2 (Hs00234140_ m1), and human CCL5 (Hs00982282_m1). The relative level of gene expression was normalized using the level of Actb or ACTB.

IHC was performed on paraffin-embedded subcutaneous tumor sections. After deparaffinizing tissue blocks, antigen retrieval was performed by 0.01% trypsin at 37°C for 10 minutes (CD31, Gr-1, and F4/80), 10 mmol/L Tris-1 mmol/L EDTA buffer (pH 9.0) at 90°C for 10 minutes (CD3, CD4, CD8, and Foxp3), or 10 mmol/L citrate at 100°C for 10 minutes (PCNA and CCL2). To block nonspecific signal, tissue sections were incubated for 10 minutes at room temperature using PBS containing 5% goat serum, 1% BSA, and 0.1% TritonX-100. Serial tissue sections were stained with following antibodies: CD31 (#557355, BD Pharmingen), Gr-1 (#550291, BD Pharmingen), F4/80 (#MCA497G, Bio-Rad), CD3 (#MCA1477, Bio-Rad), CD4 (ab183685, Abcam), CD8a (#14-0195, Thermo Fisher Scientific), Foxp3 (#623801, BioLegend), PCNA (#13110, Cell Signaling Technology), and CCL2 (#ab25124, Abcam). Sections were visualized with ZEISS microscope equipped with ZEISS AxioCam HRc camera and Axio Vision 4.1. Obtained digital images were analyzed by Photoshop to determine the proportion of immune-positive cells.

Tumors were surgically removed from transplanted mice. The tumors were cut into smaller fragments, and digested in 5-mL DMEM medium containing 3,045 units collagenase, 1,050 units hyaluronidase, and 200–400 units deoxyribonuclease at 37°C for 1 hour. Then 1 × 10⁶ cells were stained by CD11b-APC (#553312, BD Pharmingen), Gr-1-FITC (#11-5931, eBioscience), F4/80-PE (#12-4801, eBioscience), CD45-PerCP-Cy5.5 (#45-0451, eBioscience), and CCR2-PE (#150609, BioLegend) antibodies. MS level was determined by MitoSOX Red (#M36008, Life Technologies). A total of 1 × 10⁶ cells suspended in 500 μL PBS containing 3% FBS were analyzed by FACSCanto ll (BD Biosciences).

The tRNA was extracted using the TRIzol reagent (#15596018, Life Technologies). From 15 μg of tRNA, a RNA-sequence library was constructed using the mRNA-sequence Sample Preparation Kit, according to the manufacturer's instructions (Illumina). A total of 36 base-pair single-end-read RNA-sequence tags were generated using a HiSeq 2000 sequencer, according to the standard protocol. The RNA-sequence tags were mapped to the mouse genomic sequence (mm9 from the UCSC Genome Browser) using the ELAND program (Illumina). Unmapped or redundantly mapped sequences were removed from the dataset, and only uniquely mapped sequences without any mismatches were used for the analyses (22). The raw data of RNA sequence are available in DNA Data Bank of Japan (DDBJ; DRA002911).

Blood samples were collected directly from the heart of anesthetized Ccl2 KO mice 17 days after subcutaneous tumor injection, incubated for 24 hours at 4°C, and then centrifuged at 1,200 rcf for 30 minutes at 4°C to separate the serum. Conditioned medium was collected from 53KOSTS and MCF7 cells transduced with shRNA. Mouse Ccl2 ELISA (#88-7391, eBioscience) and human CCL2 (#DCP00, R&D Systems) and human CCL5 ELISA (#DRN00B, R&D Systems) were performed according to the manufacturer's instructions.

The following reagents were used: anti-mouse Ccl2 antibody (#554440, BD Biosciences), Armenian Hamster-IgG (#400916, BioLegend), carmine alum (#07070, STEMCELL Technologies), N-acetylcysteine (#A7250, Sigma-Aldrich), Trolox (#202-17891, Wako), mitoquinone (#10-1363, Focus Biomolecures), carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP; #C2920, Sigma-Aldrich), and SP600125 (#8177, Cell Signaling Technology).

Target sequences for CRISPR interference were designed using the sgRNA designer (http://portals.broadinstitute.org/gpp/public/analysis-tools/sgrna-design), and then cloned into pCRISPRv2. A nontargeting sgRNA from the Gecko library v2 was used as a scramble sgRNA.

Scramble sgRNA: ACGTGTAAGGCGAACGCCTT

Mouse Ccl2 sgRNA: TGTCACCAAGCTCAAGAGAG

tRNA was extracted using the RNeasy Mini Kit (#74106, Qiagen) according to the manufacturer's instructions. The quality of the tRNA samples was assessed using the RNA 6000 Nano Lab Chip Kit (Bio analyzer 2100, Agilent Technologies). The microarray analysis was performed with a Human Gene Expression 4 × 44K v2 Microarray (#26652, Agilent Technologies). The fluorescence intensity was measured by the G2505C Microarray Scanner (Agilent Technologies). Three samples were analyzed per group. Data were analyzed by Gene Spring 12.6.1 - GX - PA (Agilent Technologies) and R version 3.1.1. The raw data of microarray analysis are available in Gene Expression Omnibus database (GSE64525).

Whole-cell lysates were prepared as described previously (23). Immunoblotting was performed as described previously (23) using following antibodies: Phospho-Rb (#9308, Cell Signaling Technology), total RB (#554136, BD Biosciences), Phospho-ACC (#3661, Cell Signaling Technology), total ACC (#3676, Cell Signaling Technology), Phospho-AMPK (#2535, Cell Signaling Technology), total AMPK (#5832, Cell Signaling Technology), α-tubulin (#CP06-100UG, EMD Millipore), and β-Actin (#3700, Cell Signaling Technology).

Conditioned medium was collected from MCF7 transduced with shRNA cultured in serum-free RPMI1640 for 48 hours. A total of 1 × 10⁵ THP-1 cells suspended in serum-free RPMI1640 (WAKO) were loaded into the upper chamber of the 24-well–type Microchemotaxis Chamber (#CLS3421, Sigma-Aldrich), and 600 μL conditioned medium derived from MCF7 cells was added to the lower chamber. After 24-hour incubation at 37°C, the number of THP-1 cells migrated into the lower chamber was analyzed by BZ analysis software on BZ-9000 (Keyence) and Photoshop.

Immunofluorescence was performed on paraffin-embedded subcutaneous tumor sections. After deparaffinizing tissue blocks, antigen retrieval was performed by boiling the sections in pH6 10 mmol/L citrate buffer for 10 minutes. To block nonspecific signal, tissue sections were incubated for 10 minutes at room temperature using PBS containing 5% goat serum, 1% BSA, and 0.1% TritonX-100. Serial tissue sections were stained with following antibodies: F4/80 (#MCA497G, Bio-Rad), and CK18 (#GTX105624, GeneTex). Put the sections in 4 °C overnight. The next day, we labeled the F4/80 antibody with Alexa Fluor 633 (#A-21094, Thermo Fisher Scientific), labeled the CK18 antibody with Alexa Fluor 488 (#A-11034, Thermo Fisher Scientific), and then we mount the sections with Antifade Mounting Medium with DAPI (#H-1200, VECTOR). Sections were visualized with Leica TCS SP8 microscope and LAS X 1.8. Obtained digital images were analyzed by photoshop to determine the proportion of immune-positive cells.

Statistical significance was assessed using unpaired two-tailed Student t test, or one-way ANOVA followed by Tukey post hoc test. P values less than 0.05 were considered significant. Asterisks used to indicate significance correspond with: *, P < 0.05; **, P < 0.01. Columns represent means ± SD. In one-way ANOVA followed by post hoc tests, we showed asterisks only in pairs of our interest. GraphPad Prism7 was used for all statistical analysis, data processing, and presentation.

To assess whether RB status affects tumor progression in a non–cell-autonomous manner, we established a syngeneic and orthotropic murine soft-tissue sarcoma model to analyze the TME of RB-deficient tumors (Fig. 1A). First, we depleted Rb in a Trp53 knockout (KO) C57BL/6 mouse–derived sarcoma cell line named 53KOSTS (Trp53 knockout soft-tissue sarcoma) cells (Supplementary Fig. S1A), which we established previously (16, 17). As we expected, Rb depletion in 53KOSTS cells caused an acceleration of cell growth in vitro (Supplementary Fig. S1B) and an increase in tumor size following orthotopic engraftment in wild-type C57BL/6 mice. Interestingly, tumor tissues derived from Rb-depleted 53KOSTS cells appeared to be more vascularized than the control tissues (Fig. 1B). To validate this observation, we performed endothelial cell marker CD31 staining to confirm higher angiogenesis in Rb-depleted tumors (Fig. 1C). Consistent with higher CD31 expression in Rb-depleted tumors and a previous report describing induction of Vegfα expression following Rb-inactivation (24), Vegfα expression was 3-fold higher in tumor tissues derived from Rb-depleted 53KOSTS cells, even though Rb depletion only weakly induced Vegfα expression in 53KOSTS cells themselves in vitro (Fig. 1D and E). On the basis of these findings, we expected that stromal cells that formed the TME might also be stimulated to express Vegfα upon Rb depletion in 53KOSTS cells due to cell–cell interactions including humoral factors. Indeed, we observed that treatment with conditioned medium (CM) derived from Rb-depleted 53KOSTS cells significantly increased expression of Vegfα, as well as that of two other activation markers, IL6 and Il1α, in an RAW264.7 mouse macrophage cell line (Fig. 1F), implying that Rb depletion in tumor cells might affect the TME via secreted factors.

Given these changes in the TME, we next characterized infiltration of immune cells into tumor tissues derived from Rb-depleted 53KOSTS cells. Interestingly, Rb depletion significantly promoted the infiltration of Gr1⁺CD11b⁺ MDSCs, which are known to be very potent suppressors of cytotoxic T-cell immunity (Fig. 1G and H). In addition, Rb depletion slightly enhanced the infiltration of F4/80⁺ TAMs in the TME (Fig. 1I). Moreover, consistent with a higher number of CD3⁺ pan-T cells in the TME (Fig. 1J), we observed a significantly higher number of CD4⁺ effector or Foxp3⁺ Tregs in the TME of Rb-depleted tumors, although the infiltration of CD8⁺ cytotoxic T cells was slightly lower in Rb-depleted tumors (Fig. 1K; Supplementary Fig. S1C). We confirmed lower expression of perforin, the cytolytic granule effector molecule of CD8⁺ T cells, in Rb-depleted tumors by qRT-PCR (Fig. 1L). Taken together, Rb depletion not only enhances angiogenesis, but also contributes to the establishment of a protumoral microenvironment by recruiting immunosuppressive cells such as MDSCs into the TME.

We next systematically assessed RB-regulated factors that could induce these changes in the TME by performing RNA sequencing of Rb-depleted 53KOSTS cells. This uncovered multiple chemokine genes including Ccl2 that were upregulated following Rb depletion in these cells (Fig. 2A). Furthermore, pathway analysis using Kyoto Encyclopedia of Genes and Genomes (KEGG) revealed that cytokine–cytokine receptor interactions and chemokine signaling pathways were significantly upregulated in Rb-depleted 53KOSTS cells (Table 1). We then validated the upregulation of Ccl2 and other chemokines such as Cxcl1 and Cxcl5 by qRT-PCR ((Fig. 2B; Supplementary Fig. S1D). In addition, we detected enhanced Ccl2 secretion following Rb depletion in 53KOSTS cells using ELISA (Fig. 2C). Despite higher secretion levels of Ccl2 from Rb-depleted 53KOSTS cells, they themselves seemed not to receive Ccl2 because of the lack of Ccr2 expression, which encodes the main receptor for Ccl2 (Fig. 2D). These data suggested that enhanced Ccl2 secretion from Rb-depleted 53KOSTS cells mainly involved stromal cells, but not tumor cells in vivo, and might contribute to the remodeling of the TME in a non–cell-autonomous manner. Consistent with this idea, in addition to Ccl2 expression, tumor tissues derived from Rb-depleted 53KOSTS cells showed significantly high Il1α expression (Fig. 2E), which was induced in the macrophage cell line by treatment with CM derived from Rb-depleted 53KOSTS cells (Fig. 1F).

To determine the specific contribution of enhanced Ccl2 secretion to tumor progression in vivo, we next subcutaneously injected 53KOSTS cells into Ccl2 KO C57BL/6 mice to determine the concentration of tumor-derived Ccl2 in serum (Fig. 3A). Consistent with the results in vitro, Rb depletion elevated the concentration of tumor-derived Ccl2 in serum in vivo (Fig. 3B). Next, to elucidate the role of Ccl2 in vivo in tumor progression promoted by Rb depletion, we employed Ccr2 KO C57BL/6 mice to abolish the Ccl2–Ccr2 axis in the TME (Fig. 3C). Interestingly, Rb-depleted cells generated significantly smaller tumors in Ccr2 KO C57BL/6 mice than in wild-type C57BL/6 mice (Fig. 3D). To further study how the Ccl2–Ccr2 axis promoted tumor development, we then assessed angiogenesis and infiltration of immune cells such as MDSCs and T cells into the TME. Angiogenesis induced by Rb depletion in the TME was markedly suppressed in Ccr2 KO C57BL/6 mice. (Fig. 3E). In addition, infiltration of TAMs and MDSCs induced upon Rb depletion was clearly suppressed in Ccr2 KO C57BL/6 mice (Fig. 3E–G). Previously, several groups have shown that TAMs and MDSCs are recruited into the TME by CCL2 secreted from tumor cells and contribute to tumor angiogenesis by producing angiogenic factors such as VEGF (25, 26). Consistent with these reports, we observed that approximately 95% of Gr1⁺CD11b⁺ MDSCs in the TME express Ccr2 (Fig. 3H). However, the infiltration of T cells including Tregs did not decrease, suggesting that this was independent of the activation of the Ccl2–Ccr2 pathway (Supplementary Fig. S1E and S1F). Moreover, we demonstrated that CRISPR/Cas9–mediated depletion of Ccl2 significantly attenuated tumor growth, angiogenesis, and the infiltration of TAMs and MDSCs in Rb-depleted 53KOSTS cells in C57BL/6 mice (Fig. 3I–K; Supplementary Fig. S1G–S1I). Taken together, these findings suggest that tumor progression induced by Rb depletion, at least in part, depends on the elevated Ccl2 secretion and the subsequent activation of Ccr2-dependent angiogenesis or the recruitment of immunosuppressive cells into the TME.

To validate the role of the CCL2–CCR2 axis and confirm its relevance to human cancer, we focused on human breast cancer because the RB gene shows genetic alterations in approximately 10% of patients with breast cancer (27). Furthermore, we previously reported enhanced cytokine secretion following RB inactivation in breast cancer cells (16, 17). According to gene expression profiling using DNA microarray and subsequent pathway analysis in RB-depleted MCF7 cells (RB intact and ARF deficient), we confirmed that cytokine signaling is highly upregulated following RB depletion in MCF7 cells as observed in 53KOSTS cells (Fig. 4A; Supplementary Fig. S2A). Furthermore, among CC chemokine family members, the expression levels of CCL2 and CCL5 were specifically upregulated, although we did not observe Ccl5 upregulation in Rb-depleted 53KOSTS cells (Supplementary Table S1). We next examined CCL2 and CCL5 expression across a panel of RB-positive breast cancer cell lines. In most cell lines, CCL2 and CCL5 expression was significantly higher following RB depletion (Fig. 4B; Supplementary Fig. S2B). We further examined patient-derived primary breast cancer cells, in which RB depletion also upregulated both CCL2 and CCL5 (Fig. 4C). Moreover, we confirmed enhanced CCL2 and CCL5 secretion following RB depletion by ELISA (Fig. 4D). In contrast to RB depletion, overexpression of the constitutively active (nonphosphorylatable) form of RB (RB7LP; ref. 28) clearly decreased both CCL2 and CCL5 expression (Fig. 4E). Consistent with these findings, CCL2 and CCL5 expression showed a weak inverse correlation with RB expression according to gene expression profiling data from breast cancer cell lines found in the Cancer Cell Line Encyclopedia (CCLE) database (Fig. 4F; Supplementary Table S2). To functionally validate the role of the RB–CCL2 axis, we next examined CCL2-dependent migration of THP-1 cells using a transwell migration assay. Importantly, compared with the CM derived from control MCF7 cells, CM derived from RB-depleted MCF7 cells exhibited stronger chemoattractant activity for THP-1 cells, which was significantly antagonized by treatment with an anti–CCL2-neutralizing antibody (Fig. 4G). Taken together, these data suggest that RB inactivation might influence the TME of breast cancer through enhanced chemokine secretion including CCL2.

Previously, we found that RB inactivation induced the enhanced secretion of several cytokines such as IL6 in breast cancer cell lines through enhanced MS production (17). In brief, the transcription of mitochondria-related genes especially related to fatty acid oxidation (FAO) such as CPT1 is upregulated due to deregulation of E2F activity following RB inactivation. CPT1 provides a rate-limiting step in long-chain fatty acid oxidation. CPT1 controls transportation of long-chain fatty acids into mitochondria. Long-chain fatty acids transported into mitochondria are used for β-oxidation. Consistent with higher expression of FAO-related genes, RB-depleted cells showed an elevated oxygen consumption rate upon palmitate stimulation and also an elevated MS production (17). In the same previous work (17), we demonstrated that RB inactivation significantly decrease malonyl-CoA level. This finding was highly consistent with elevated FAO because malonyl-CoA strongly suppresses FAO through inhibition of CPT1. However, why malonyl-CoA level drops following RB inactivation was not fully cleared in the previous study. In this work, we discovered that RB loss increases the phosphorylation of AMP-activated protein kinase (AMPK) and one of its substrates acetyl-CoA carboxylase (ACC; Fig. 5A). ACC plays a crucial role in regulating FAO. The malonyl-CoA, which is generated by ACC, specifically inhibits the CPT1 activation. The phosphorylated AMPK phosphorylates and thus inactivates ACC. Therefore, upon AMPK phosphorylation, the level of malonyl-CoA drops, leading to increased activity of CPT1. The RB–AMPK–ACC axis may provide a possible pathway whereby RB loss increases FAO.

To further understand whether the induction of CCL2 and CCL5 following RB depletion depends on MS, we first analyzed the association between CCL2 or CCL5 expression and MS production in a variety of breast cancer cell lines, noting that CCL2, but not CCL5, showed a significant positive correlation with MS production (Fig. 5B). Moreover, consistent with our previous study (17), MS production was significantly upregulated following RB depletion in multiple breast cancer cell lines except MDA-MB-231 (Fig. 5C) in which CCL2 was not upregulated by RB depletion (Fig. 4B). To examine whether higher MS production in RB-depleted cells directly contributes to higher CCL2 production, we treated MCF7 cells with several antioxidants such as N-acetyl cysteine (NAC), Trolox, and mitochondria-targeted antioxidant MitoQ, and found that CCL2 production was, at least in part, dependent on MS (Fig. 5D). However, CCL5 induction following RB depletion was not antagonized upon treatment with antioxidants (Supplementary Fig. S2C). In addition, treatment with trifluoromethoxy carbonylcyanide phenylhydrazone (FCCP), a potent uncoupler of oxidative phosphorylation in mitochondria known to induce MS production, strongly induced CCL2, but not CCL5 expression, to the same degree as RB depletion in MCF7 cells (Fig. 5E; Supplementary Fig. S2D). Collectively, these data suggest that enhanced MS production following RB inactivation induces CCL2 expression. Finally, as we previously reported that enhanced cytokine secretion following RB inactivation was mediated by JNK activation, we verified the JNK dependency of CCL2 induction by treatment with JNK inhibitor SP600125 (Fig. 5F) or introduction of a dominant negative form of c-JUN (Fig. 5G and H).

To examine whether the RB–CCL2 axis is involved in carcinogenesis induced by RB deficiency, we analyzed mouse mammary carcinogenesis in vivo using MMTV-Cre;Rb^flox/flox mice with various Ccr2 genetic backgrounds, including Ccr2^+/+, Ccr2^+/−, and Ccr2^−/−. Previously, several groups have reported that MMTV-Cre;Rb^flox/flox female mice develop focal hyperplastic lesions in the mammary glands. (29, 30). As expected, the percentage of PCNA-positive cells in the mammary glands of MMTV-Cre;Rb^flox/flox; Ccr2^+/+ nulliparous mice (45.3 ± 13.3%) at an average age of examination of 429 ± 13.2 days was dramatically higher than that of wild-type C57BL/6 mice at a similar age (6.5 ± 3.0%). Importantly, the percentage of PCNA-positive cells in mammary glands was lower in Ccr2^± (17.8 ± 2.0%) and Ccr2^−/− (4.5 ± 5.0%) backgrounds at a similar age (Fig. 6A; Supplementary Fig. S2E). Overall survival did not show statistically significant differences among Ccr2^+/+, Ccr2^±, and Ccr2^−/− backgrounds (Supplementary Fig. S2F). We did not observe palpable mammary tumors in mice with any background. Upon autopsy, we frequently observed lymphoma (Supplementary Fig. S2G), which may explain relatively shorter survival of MMTV-Cre;Rb^flox/flox mice (around 16 months). Although not frequently, we observed hepatic and thyroid tumors (Supplementary Fig. S2H and SI). These findings are consistent with nonmammary tissue-specific activation of the MMTV promoter (31).

Consistent with higher PCNA signal, MMTV-Cre;Rb^flox/flox;Ccr2^+/+ nulliparous mice mammary glands examined at 436 ± 40.2 days frequently exhibited hyperplastic features upon whole-mount carmine alum staining and hematoxylin and eosin staining, whereas MMTV-Cre;Rb^flox/flox;Ccr2^−/− mammary glands at a similar age exhibited morphologies reminiscent of those of wild-type C57BL/6 mice at a similar age (Fig. 6B; Supplementary Fig. S2J). More accurately, we examined mammary gland of 14 MMTV-Cre;Rb^flox/flox;Ccr2^+/+ mice, and 11 of them (78.6%) showed hyperplastic phenotype. However, in 22 MMTV-Cre;Rb^flox/flox;Ccr2^−/− mice, we observed no (0%) hyperplastic phenotype. Furthermore, the infiltration of F4/80⁺ macrophages into mammary glands was significantly upregulated in MMTV-Cre;Rb^flox/flox;Ccr2^+/+ but not in the Ccr2^−/− background, even though CCL2 expression in mammary glands was upregulated following RB deletion both in Ccr2^+/+ and Ccr2^−/− backgrounds (Fig. 6C and D). Taken together, these data suggest that activation of the Ccl2–Ccr2 pathway in mammary glands via enhanced Ccl2 secretion is required for carcinogenesis induced by Rb deficiency.

Here, we provide novel evidence indicating that RB inactivation in tumor cells results in the formation of a protumoral microenvironment by promoting angiogenesis and recruitment of TAMs or immunosuppressive cells such as MDSCs into the TME. By using a syngeneic and orthotropic murine soft-tissue sarcoma model, and confirming these findings in a second murine mammary carcinogenesis model, we demonstrated that CCL2 induction following RB inactivation in tumor cells and subsequent activation of the CCL2–CCR2 axis in the TME accelerates tumor progression. CCL2 has been shown to play a critical role in tumor progression in various cancer types including breast cancer via macrophage recruitment into the TME (32, 33). Macrophages in the TME support tumor growth via multiple mechanisms, including the secretion of growth factors and the promotion of angiogenesis (34). In addition to the proliferative advantage imparted to tumors, elevated CCL2 secretion contributes to the formation of an immunosuppressive TME via MDSCs and Tregs recruitment, and results in the evasion of cytotoxic T cells (35). Despite many studies that identify CCL2 as a protumorigenic chemokine, the therapeutic effect of blocking the CCL2–CCR2 axis via treatment with a neutralizing antibody, for example, has been disappointing in clinical trials (36–38). Thus, to achieve the therapeutic benefits of CCL2-CCR2 blockade, we must understand the regulatory mechanisms of CCL2 and determine which patient subtypes respond to this therapy (i.e., which tumor mutations and gene expression signatures are susceptible to CCL2 blockade).

According to our findings that demonstrate the functional relevance between the RB and CCL2–CCR2 axis, aberrant RB expression in certain tumor types might act as a possible marker for the development of an effective therapy by CCL2-CCR2 blockade. Currently, it is thought that the CCL2-CCR2 blockade exhibits tumor-suppressive function, at least in part, by enhancing the antitumoral function of the TME via inhibition of the infiltration of immunosuppressive cells (1). TNBC is a highly heterogeneous subtype compared with others such as ER-positive or HER2-positive breast cancer (39). RB inactivation by genetic and epigenetic factors and significantly higher CCL2 levels is frequently found in this aggressive subtype (40, 41). Although TNBC is sensitive to chemotherapy, the overall outcomes of TNBC are worse in patients with breast cancer because of the lack of targetable molecules such as ER and HER2 (39). Thus, immunotherapy including ICB treatment is emerging as a promising new option for patients with TNBC, but there are no established prognosis markers to estimate its efficacy. Our current study revealed that RB inactivation via genetic mutation or transcriptional suppression via DNA hypermethylation in the RB gene promoter in patients with TNBC could be a potential marker for both CCL2-CCR2 blockade efficacy and poor immunogenicity in TNBC, and anti-CCL2 treatment might enhance therapeutic effect of ICB treatment. However, it is now becoming clear that the accumulation of DNA damage in tumors resulting from defects in the DNA repair pathway, DNA-damaging chemotherapy, and/or radiotherapy is associated with immunogenic cell death and neoantigen production, promoting an antitumor immune response (42–44). It is possible that increased genomic instability and subsequent responses to DNA damage in RB-inactivated cancer cells might potentiate the efficacy of immunotherapy, but further research is needed to uncover how RB inactivation alters tumor immunogenicity in both cell-autonomous and non–cell-autonomous manners.

Tumor angiogenesis in the TME is crucial for tumor progression (26). By using a syngeneic and orthotropic murine soft-tissue sarcoma model with a Ccr2-null background, we demonstrated that Rb inactivation in cancer cells promotes aberrant angiogenesis through the activation of the Ccl2–Ccr2 axis in the TME. In particular, RB inactivation–dependent infiltration of TAMs into the TME is clearly suppressed in Ccr2 KO mice. TAMs secrete high levels of angiogenic factors including Vegfα, leading to neovascularization in the TME. Because Rb inactivation also induces Vegfα secretion from tumor cells themselves (24), RB might regulate tumor angiogenesis via both cell-intrinsic and cell-extrinsic mechanisms.

In our previous study (17), we partially clarified the mechanism whereby RB inactivation leads to increased production of IL6 via increased FAO activity and MS production. We disclosed that RB inactivation increased the expression of a number of genes involved in FAO in an E2F-dependent manner, therefore increased oxidative metabolism leads to increased MS production and JNK activation. JNK activation is critical for IL6 secretion. In addition, we demonstrated that malonyl-CoA level drops following RB inactivation. This finding is the most consistent with increased FAO, because malonyl-CoA is the strongest suppressor of FAO (17). However, in the previous work, we did not clarify why malonyl-CoA level drops following RB inactivation. In this work, we provided evidence that increased CCL2 secretion following RB inactivation is due to elevated FAO activity and MS production similar to in the case of IL6. Moreover, we linked RB inactivation to downregulation of malonyl-CoA by AMPK and ACC. RB inactivation dramatically increased AMPK phosphorylation (Fig. 5A). Phosphorylated AMPK phosphorylates ACC. Phosphorylated ACC loses its activity to synthesize malonyl-CoA from acetyl-CoA. Decrease in malonyl-CoA allows CPT1 to transport long-chain fatty acids into mitochondria for FAO. Why RB loss increases AMPK phosphorylation is currently under investigation. We are determining ATP/AMP and NAD⁺/NADH ratio in cells before and after RB depletion (45). In addition, RB has been suggested to be involved in various facets of cellular metabolism (46). Further study would be necessary to thoroughly determine the mechanism of RB–AMPK axis.

No potential conflicts of interest were disclosed.

Conception and design: F. Li, S. Kitajima, C. Takahashi

Development of methodology: F. Li, S. Kohno, N. Okada, C. Takahashi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F. Li, S. Kitajima, S. Kohno, A. Yoshida, S. Tange, S. Sasaki, Y. Nishimoto, H. Muranaka, N. Nagatani, M. Suzuki, T. Nishiuchi, T. Tanaka, N. Mukaida

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F. Li, S. Kitajima, S. Kohno, S. Tange, T. Tanaka, C. Takahashi

Writing, review, and/or revision of the manuscript: F. Li, S. Kitajima, H. Muranaka, T. Tanaka, D.A. Barbie, N. Mukaida, C. Takahashi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Kitajima, N. Okada, H. Muranaka, S. Masuda, T.C. Thai, C. Takahashi

Study supervision: S. Kitajima, C. Takahashi

We thank Dr. T. Baba for technical instruction and useful discussion, Dr. N. Mahadevan for pathologic diagnosis, and Mr. S. Sundararaman for critical reading of the manuscript. This work was supported by Funding Program for Next Generation World-Leading Researchers LS049 (to C. Takahashi), Grant-in-Aid for Scientific Research on Innovative Areas 15H01487 and 17H05615 (to C. Takahashi), Grant-in-Aid for Scientific Research 17H03576 (to C. Takahashi) and 25830077 (to S. Kitajima), Hokuriku Bank Research Grant for Young Scientist (to S Kitajima), the Uehara Memorial Foundation Post-Doctoral Fellowship (to S. Kitajima), the Strategic Young Researcher Overseas Visit Program for Accelerating Brain Circulation (to S. Kitajima), and JSPS Postdoctoral Fellowship for Research Abroad (to S. Kitajima).

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.