Tumor-Associated Macrophages Promote Malignant Progression of Breast Phyllodes Tumors by Inducing Myofibroblast Differentiation

Breast phyllodes tumor is a biphasic breast tumor composed of cellular spindle stroma with epithelial elements. It constitutes 0.3% to 1% of all breast tumors and 2.5% of fibroepithelial lesions of the breast (1). Currently, phyllodes tumors are histologically classified as benign, borderline, or malignant on the basis of stromal cellularity, mitotic activity of stromal cells, stromal nuclear atypia, stromal overgrowth, and types of border (infiltrating or pushing). The clinical outcome of phyllodes tumors is hard to predict, with frequent local relapse and sometimes distant metastasis. Adjuvant chemotherapy or radiotherapy is not effective against phyllodes tumors (2). The potentially recurring and metastatic behavior of phyllodes tumors is attributed to the characteristics of stromal cells, mainly fibroblasts (3). The normal fibroblast can acquire an “activated” phenotype, which expresses the α-smooth muscle actin (α-SMA) as a hallmark and is so-called as myofibroblasts. Our previous studies have indicated myofibroblasts were the major malignant component of phyllodes tumors. The increased myofibroblast population drives the tumorigenicity of phyllodes tumors. In addition, α-SMA can serve as an independent prognostic factor for phyllodes tumors with better predictive values than histologic classification (4). We further demonstrated that the fibroblasts–myofibroblasts transition in phyllodes tumors is driven by the elevated miR-21 (4), whereas the mechanism of miR-21 upregulation and how it drives tumorigenicity of phyllodes tumors remain unknown.

It is well established that tumor-associated macrophages (TAM) are one of the most abundant cell type in tumor microenvironment (5), which are involved in tumor metastasis and progression (6). Clinical and epidemiologic studies have shown a strong correlation between the increased TAMs density and poor prognosis in several types of cancer (6, 7), including breast cancer (8, 9). However, how TAMs impact on the malignant progression of phyllodes tumor and whether TAMs density can be a prognostic factor for phyllodes tumors are still unknown.

Macrophages are usually found in close proximity with collagen-producing myofibroblasts (9). Macrophages produce profibrotic mediators that directly activate fibroblasts, including TGFβ1 and platelet-derived growth factor (PDGF; ref. 10). Macrophages also provide insulin-like growth factor-1, which stimulates the proliferation and survival of myofibroblasts and promotes collagen synthesis by these cells (11). In this study, we found that TAMs are essential for driving myofibroblast differentiation (fibroblasts–myofibroblasts transition) in the malignant progression of phyllodes tumors via the CCL18/NF-κB/miR-21/PTEN/AKT axis and targeting CCL18 is a promising strategy for treating phyllodes tumors.

Breast phyllodes tumor samples were obtained from 268 female patients with 167 benign, 36 borderline, and 65 malignant phyllodes tumors in the Breast Tumor Center, Sun Yat-Sen Memorial Hospital, Sun Yat-Sen University, from January 2000 to June 2011. The patients were followed up for 8–148 months (median follow-up is 112 months). Pathologic diagnosis, as well as mitoses and stromal overgrowth status, was confirmed by two pathologists independently. All human samples were collected with informed consent from the donors according to the International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS). The study was performed after approval by the institutional review board (IRB) of Sun Yat-Sen Memorial Hospital.

Phyllodes tumor cells were isolated from benign and malignant phyllodes tumors as previously described (4). TAMs were isolated from eight fresh breast malignant phyllodes tumor samples as previously described (12, 13), with slight modifications. Briefly, the tissues were minced into small (1–2 mm) pieces and digested with 5% FBS DMEM containing 2 mg/mL collagenase I and 2 mg/mL hyaluronidase (Sigma) at 37°C for 2 hours. The cells were sequentially filtered through 500 μm mesh, 100 μm, and 70 μm cell strainer. The cells were then centrifuged in a Beckman Allegra X-15R centrifuge at 2,500 rpm for 20 minutes with 1 mL cell suspension above 5 mL 45% Percoll (GE Healthcare) in the middle and 5 mL 60% Percoll at the bottom in a 15 mL tube. Mononuclear cells were collected from the cell layer in the interphase between 45% and 60% Percoll. CD14⁺ monocytes and macrophages were isolated by a magnetic-activated cell sorting using direct CD14 Isolation Kit (Miltenyi Biotec) according to the manufacturer's instructions.

Chromatin immunoprecipitation (ChIP) assay was performed using Pierce Agarose ChIP Kit (26156, Thermo Scientific) according to the manufacturer's instructions. The specific sequences from immunoprecipitated and input DNA were determined by PCR primers for miR-21 promoter upstream regions:

• miR-21 promoter forward, 5′ -TCCCCTCTGGGAAGTTTC-3′,

• reverse, 5′ -TTGGCTCTACCCTTGTTT-3′.

The negative control primer:

• miR-21-NC-1: forward, 5′-TTCCTCATTTCCCTAAACAACAA-3′

• reverse, 5′ -AATCTACCAGGGATAGCCATAGTC-3′.

• miR-21-NC-2: forward, 5′-GAGGACTTCCCCAACTTAACTATG -3′

• reverse, 5′ -TTATTCTCAAGCAGCAGACCAG -3′.

All animal work was conducted in accordance with a protocol approved by the Institutional Animal Care and Use Committee at the Medical School of Sun Yat-Sen University. Breast phyllodes tumor cells (1 × 10⁷) mixed with MDMs (2 × 10⁶) were inoculated into the mammary fat pads of 6-week-old female nude mice. Mice were from Beijing Vital River Laboratory Animal Technology Co., Ltd. For antibody treatment and the mice were injected with CCL18-specific neutralizing antibody (catalog no. ab9849, Abcam) or isotype-matched IgG (Abcam) via the tail vein at 1 mg/kg twice weekly after the xenografts became palpable (around 0.5 cm in diameter). In some groups, intratumoral injection of M-CSF at 0.2 μg/kg was performed twice weekly after inoculating and 48 hours prior to antibody injection. Tumor growth was evaluated by monitoring tumor volume (TV = length × width² × 0.5) every 3 days for 8 weeks. The animals were sacrificed when the xenografts reached 1.5 cm in diameter. Tumor xenografts and the livers and lungs of mice were harvested for further evaluation. Cryosections (4 μm) of the harvested organs were hematoxylin and eosin (H&E) for histologic assessment, and total RNA was extracted for qRT-PCR analysis of human hypoxanthine phosphoribosyltransferase (HPRT) mRNA expression.

The in vitro data were depicted as mean ± SD of three independent experiments performed in triplicate. All statistical analyses were performed using SPSS 16.0 statistical software package (SPSS). Student t test and one-way ANOVA was used to compare CCL18 expression levels between the phyllodes tumors with different tumor grades, whereas χ² test was used to analyze the relationship between CCL18 expression and clinicopathologic status. Kaplan–Meier curves and log-rank test were used to compare overall survival (OS) and disease-free survival (DFS) in different patient groups. In all cases, P < 0.05 was considered statistically significant.

Evidence from clinical and epidemiologic studies has shown a strong correlation between TAMs density and poor prognosis in several types of cancer (6, 7). To investigate whether malignant progression of phyllodes tumor is correlated with TAMs density, we examined the presence of TAMs in 268 phyllodes tumor samples, including 167 benign, 36 borderline, and 65 malignant phyllodes tumors. Normal breast tissues were used as control. We tested the expression of CCL18, the hallmark of TAM in paraffin-embedded phyllodes tumor samples by IHC. The expression of CCL18 level in malignant phyllodes tumors is significantly higher than that in benign and borderline phyllodes tumors (Fig. 1A). Previously, we have found PITPNM3 was the receptor of CCL18 on the surface of breast tumor cells (12). CCR8 was also reported as the CCL18 receptor (14). We then examined the level of PITPNM3 or CCR8 in phyllodes tumors and found that PIPTNM3 was increased in malignant phyllodes tumors (Fig. 1A), whereas CCR8 was not (Supplementary Fig. S1A). CCL18 is expressed in both TAMs and dendritic cells (DC; 14). To confirm whether CCL18⁺ cells were macrophages, two macrophages markers CD68 and CD163 were used. CD68 is specially expressed by macrophages and in tolerogenic DCs (15). CD163 is a marker of M2 macrophages. Microscopy revealed that both CD68 (Supplementary Fig. S1B) and CD163 (Fig. 1B) immunostaining were colocalized with CCL18⁺ cells in breast phyllodes tumor tissues, showing the CCL18⁺ cells were M2-like macrophages, which in fact were TAMs in the phyllodes tumors.

Furthermore, we measured the CCL18 mRNA levels in fresh frozen phyllodes tumor tissues and primary phyllodes tumor cells isolated from benign, borderline, or malignant phyllodes tumors, using normal breast tissues as a control. Using qRT-PCR, we found that the mRNA levels of CCL18 from fresh frozen tissues were progressively increased from normal breast tissue to benign, borderline, and malignant phyllodes tumors (Fig. 1C). The mRNA levels of CCL18 from primary phyllodes tumor cells did not show an increasing tendency from benign, borderline to malignant phyllodes tumor cells. Although TAMs isolated from the malignant phyllodes tumors expressed extremely high levels of CCL18 mRNA (Supplementary Fig. S1C). ELISA assay was used to measure the serum CCL18 level of phyllodes tumor patients or the secreted CCL18 level in the culture suspension of primary cells isolated from phyllodes tumor tumors. We found that the CCL18 protein levels in the serum of patients with malignant phyllodes tumors were significantly higher than in those with benign and borderline phyllodes tumors (Fig. 1D). However, primary phyllodes tumor cells secreted few CCL18 into the culture medium. Similar to the mRNA level, the secreted CCL18 level in the culture supernatant of the primary TAMs isolated from malignant phyllodes tumor tumors was much higher than in those from phyllodes tumor cells (Supplementary Fig. S1D). Together, these results indicate that TAM density is associated with the malignant progression of phyllodes tumors and CCL18 is mostly produced by TAMs.

Next, we tested whether CCL18 had clinical prognostic value for phyllodes tumor patients. The 268 phyllodes tumor patients were followed up for 8–148 months (median follow-up time is 112 months). During the follow-up, 49 cases were diagnosed with recurrence, including 18 in the benign group, 9 in the borderline group, and 22 in the malignant group. In addition, 31 cases were diagnosed with metastasis, with 3 in the borderline group and 28 in the malignant group.

We then analyzed the association of CCL18 expression with the clinicopathologic status of phyllodes tumors (Supplementary Table S1). The expression of CCL18 increased with higher tumor grade, mitotic activity, and stromal overgrowth (P < 0.001), but was not associated with age and the size of tumor (Supplementary Table S1). The expression of CCL18 was also more abundant in the phyllodes tumors with local recurrence and distal metastasis (P < 0.001; Supplementary Table S1). Furthermore, Kaplan–Meier survival curve demonstrated that patients with low CCL18 expression (staining index, SI ≤ 4, refers to the Supplementary Methods and ref. 4) have a longer overall survival (OS) and disease-free survival (DFS) than those with high CCL18 expression (P < 0.001, Fig. 1E and F). We also used ROC curve to evaluate the efficacy of serum CCL18 in phyllodes tumor patients as a diagnostic marker. It showed that CCL18 serum level could distinguish aggressive cases from benign phyllodes tumors (Supplementary Fig. S1E and S1F). In addition, multivariate Cox regression analyses demonstrated that CCL18 (P < 0.001), stromal overgrowth (P < 0.001), and tumor grading (P = 0.006) were independent prognostic predictors for local recurrence-free survival (LRFS; Supplementary Table S2). These results suggest that the TAMs density is associated with malignant progression of phyllodes tumor, and the levels of CCL18 can be used to predict the outcome of phyllodes tumor patients.

Macrophages are frequently found in close proximity with collagen-producing myofibroblasts (9), and there is strong evidence that this interaction is reciprocal (16). Macrophages produce profibrotic mediators that directly activate fibroblasts, including TGFβ1, platelet-derived growth factor (PDGF; ref. 10). Macrophages also produce insulin-like growth factor-I, which stimulates the proliferation and survival of myofibroblasts and promotes collagen synthesis by these cells (11). As we have found TAM density is correlated with malignant progression of phyllodes tumors and our previous studies have reported that myofibroblast differentiation is associated with the malignant progression of phyllodes tumors (4), we speculated that TAMs may play an important role in the myofibroblast differentiation of breast phyllodes tumors. To demonstrate this hypothesis, we cultured primary benign phyllodes tumor cells with primary TAMs isolated from malignant phyllodes tumor patients. TAMs in breast cancers are primarily M2 cells (17). Because M2 macrophages, which can be induced from monocyte-derived macrophages (MDM) by M-CSF (18), have been shown to induce myofibroblast transition (9), we also cultured benign phyllodes tumor cells with M-CSF pretreated MDMs by transwell assay. Benign phyllodes tumor cells were seeded into the bottom chamber. MDMs or TAMs were seeded into the top chamber of a six-well transwell apparatus. The mRNA and protein level of α-SMA, the hallmark of myofibroblasts, was detected to evaluate the phenotype change of phyllodes tumor cells. Compared with untreated phyllodes tumor cells, cells cocultured with primary TAMs or M-CSF–activated MDMs had a significant higher level of α-SMA (Fig. 2A and B). The expression level of α-SMA in phyllodes tumor cells treated with M-CSF alone or cocultured with untreated MDMs had no obviously changes (Fig. 2A and B). Immunofluorescent staining (Fig. 2C) also showed the increased levels of α-SMA and fibroblast activation protein (FAP), which is also a marker of myofibroblast.

Because myofibroblasts are known to have an increased ability to induce collagen gel contraction (19), collagen contraction assay was used to test whether TAMs-treated phyllodes tumor cells function as myofibroblast. We observed that benign phyllodes tumor cells cocultured with TAMs or M-CSF–pretreated MDMs contracted collagen gels to a much greater extent than phyllodes tumor cells treated with or without M-CSF or phyllodes tumor cells cocultured with untreated MDMs (Fig. 2D). Together, these findings suggest that TAMs induce myofibroblast function in the phyllodes tumor cells.

Previous studies have reported that myofibroblasts in epithelial tumors have an increased proliferative activity and can promote cancer invasion and metastasis (20, 21). We then examined the effects of TAMs on the proliferation, migration, and invasion of primary phyllodes tumor cells. Cell viability and clonogenic assays showed that TAMs or M-CSF–pretreated MDMs increased the foci formation of phyllodes tumor cells (Fig. 2E), indicating that TAMs promoted the growth of phyllodes tumor cells. Boyden chamber assays also showed that TAMs or M-CSF–pretreated MDMs significantly increased the number of migrated and invaded phyllodes tumor cells (Fig. 2F and G; P < 0.01). These data suggest that TAMs not only induces the myofibroblast differentiation, but also promotes their malignancy.

Previously, we have reported that CCL18 released by TAMs promotes breast cancer metastasis, causing poor survival of breast cancer patients (22). It has been shown above that elevated CCL18 expression is also associated with poor outcome of breast phyllodes tumor patients, and that myofibroblast differentiation is progressively increased during the malignant progression of phyllodes tumor in breast. Therefore, we examined whether CCL18 induced the myofibroblast differentiation. We found that breast phyllodes tumor cells treated with 20 ng/mL CCL18 had a significant increase in the mRNA and the protein levels of α-SMA (Fig. 3A–C) and enhanced activity of collagen gel contraction, proliferation (Fig. 3D and E), migration, and invasiveness (Fig. 3F; Supplementary Fig. S2A and S2B), suggesting that CCL18 induced myofibroblast differentiation and promoted the proliferation and invasion of the phyllodes tumor cells. Consistent with the results described above, when cocultured with primary TAMs and control IgG, the phenotype and function changes of phyllodes tumor cells were similar to CCL18 treated alone. However, anti-CCL18–neutralizing antibody inhibited myofibroblast differentiation, proliferation, and invasion induced by TAMs (Fig. 3A–F), indicating that the TAM-released CCL18 is responsible for the myofibroblast differentiation of phyllodes tumor cells.

Our previous study has shown that miR-21 expression levels are upregulated in the malignant phyllodes tumors. miR-21 not only induces the myofibroblast differentiation of phyllodes tumor cells, but also promotes their malignant properties, including proliferation and invasion (4). Thus, we speculate that CCL18 may regulate the myofibroblast differentiation through upregulating miR-21. To further demonstrate this hypothesis, we first analyzed the expression correlation between miR-21 (detected by ISH as in ref. 4) and CCL18, and revealed that the percentage of miR-21⁺ cells was positively correlated with that of CCL18-producing cells in the 268 phyllodes tumor samples (Fig. 4A, r = 0.752, P < 0.001). Then we used qRT-PCR to test the miR-21 level of phyllodes tumor cells in response to CCL18 treatment. Compared with the untreated group, the RNA level of miR-21 of CCL18-treated phyllodes tumor cells increased by 3.9-fold (Fig. 4B).

A recent study showed NF-κB could bind to the promoter of miR-21 and regulate pancreatic β cell death through miR-21 (23, 24). As we have shown that CCL18 activated NF-κB pathway to induce epithelial–mesenchymal transition in breast cancer (25), we proposed that CCL18 might upregulate miR-21 expression via activating NF-κB. To clarify the roles of NF-κB in CCL18-induced miR-21 expression, the luciferase reporter assay was applied in phyllodes tumor cells. We found that NF-κB activity was escalated to a 6.2-fold increase in CCL18-treated cells. The TNFα-treated phyllodes tumor cells were used as positive control (Fig. 4C). Constantly, we revealed that CCL18-induced p65 nuclear translocation could be reversed by both NF-κB pathway inhibitors and siPITPNM3, suggesting CCL18 induced activation of NF-κB pathway through membrane-associated phosphatidylinositol transfer protein 3 (PITPNM3; Supplementary Fig. S3A and S3B). The PITPNM3 RNAi efficiency was detected by Western blot analysis (Supplementary Fig. S3C). Again, we confirmed that CCL18 promoted NF-κB pathway via PITPNM3, as siPITPNM3 eliminated the phosphorylation of IKK and IKB in the presence of CCL18 (Supplementary Fig. S3D). In addition, CCL18 upregulated the mRNA levels of several NF-κB target genes, including IL8, IL4, Twist, etc. (Fig. 4D). Collectively, these data suggest that CCL18 can enhance NF-κB transcriptional activity in phyllodes tumor cells.

To investigate whether NF-κB can directly regulate miR-21 expression, inhibitors of NF-κB pathway [BAY-117082, an IKK inhibitor that also interferes with the ubiquitin conjugating enzymes (26), and JSH-23, an inhibitor of NF-κB nuclear translocation] were used. Both BAY-117082 and JSH-23 reversed CCL18-induced miR-21 upregulation in phyllodes tumor cells (Fig. 4E), indicating that NF-κB activation is involved in CCL18-mediated upregulation of miR-21. A ChIP-qPCR assay with antibodies against NF-κB p65 was used to determine whether miR-21 is the direct target of NF-κB p65. CCL18 increased the binding of NF-κB p65 to the promoters of miR-21 by 12.7-fold (Fig. 4F). We next examined the functional relationships among CCL18, NF-κB, and miR-21 in breast phyllodes tumor cells. When phyllodes tumor cells were treated by CCL18 for 24 hours, miR-21 promoter activity was increased by 8.6-fold (Fig. 4G). Taken together, our results indicate that CCL18 upregulates miR-21 expression via activating NF-κB.

To investigate whether CCL18 upregulates miR-21 expression and thus induces myofibroblasts' differentiation via activating NF-κB, phyllodes tumor cells were pretreated with BAY-117082 or JSH-23 for 1 hour or transfected with miR-21 antisense oligos (ASO) for 48 hours and then exposed to 20 ng/mL CCL18 or cocultured with TAMs for 48 hours. We found that CCL18 treatment increased the mRNA (Fig. 5A and C) and protein level of α-SMA (Fig. 5B and D). However, the NF-κB inhibitor BAY-117082 or JSH-23 or the miR-21 inhibitor blocked the CCL18 or TAM-induced increase of α-SMA in the phyllodes tumor cells, indicating that myofibroblast differentiation was prevented under these conditions. Similarly, collagen contraction assay showed that CCL18 contracted collagen gels to a much greater extent than the untreated cells (P < 0.01), whereas BAY, JSH, or miR-21 ASO suppressed the contractile ability (P < 0.01; Fig. 5E and Supplementary Fig. S4A). We then evaluated the effect of NF-κB and miR-21 on the invasive ability of phyllodes tumor cells treated with CCL18 or cocultured with TAMs. Consistent with previous results, BAY, JSH, and miR-21 ASO also abrogated the CCL18 or TAM-promoted migration, invasion (Fig. 5E; Supplementary Fig. S4B) and proliferation (Fig. 5F; Supplementary Fig. S4C) of phyllodes tumor cells. These data suggest that CCL18 upregulates miR-21 expression and thus induces myofibroblast differentiation and promotes their tumorigenicity via activating NF-κB.

Our previous work has revealed that mRNA of PTEN can be targeted by miR-21 in phyllodes tumor cells (4). As PTEN is the major negative regulator of PI3K–AKT pathway, which is most frequently mutated and inactivated in malignant tumors, we hypothesize CCL18 drives tumorigenicity of phyllodes tumor through activating PI3K–AKT pathway. Therefore, the phosphorylation of AKT was examined to explore whether this pathway is activated. We found that CCL18 treatment or TAM coculture decreased PTEN level and increased AKT phosphorylation followed by the elevated phosphorylated p65 (Fig. 6A), suggesting the activation of PI3K–AKT pathway. When CCL18 antibody was added, the PI3K–AKT pathway kept inactive. Consistently, the NF-κB inhibitor (BAY-117082 or JSH-23) or the miR-21 ASO could completely reverse the CCL18-enhanced AKT phosphorylation and increase PTEN level in phyllodes tumor cells (Fig. 6B). These results suggested that the CCL18/NF-κB/miR-21 axis was necessary and sufficient for TAMs to activate PI3K–AKT pathway in phyllodes tumor cells. Then, we detected PTEN and p-AKT in human phyllodes tumor tumors by IHC and confirmed that PTEN level decreased and p-AKT increased from benign to malignant phyllodes tumors (Fig. 6C). Together, these data suggested that TAMs induced myofibroblast differentiation and promoted phyllodes tumor tumorigenicity through secreting CCL18 to activate PI3K–AKT pathway via NF-κB/miR-21/PTEN axis (Fig. 6D). We also explored which factor in the microenvironment promoted macrophages producing CCL18. We compared the cytokine expressed by the phyllodes tumor cells and found that hepatocyte growth factor (HGF) was highly secreted by malignant phyllodes tumor cells comparing with benign phyllodes tumor cells (Supplementary Fig. S5A and S5B). Then a HGF-neutralized antibody was used to block the phyllodes tumor cell–secreted HGF in a transwell culture system with malignant phyllodes tumor cells and corresponding TAMs. We found that TAMs produced less CCL18 upon HGF-neutralized antibody treatment (Supplementary Fig. S5C).

To investigate the role of CCL18 on tumor formation and progression in vivo, we inoculated benign phyllodes tumor cells and MDMs into the mammary fat pads of athymic nude mice (n = 10 per group), and evaluated tumor metastasis to the lungs and livers. When the xenografts became palpable, CCL18-specific neutralizing antibody (CCL18-Ab, 1 mg/kg twice weekly) or isotype IgG (1 mg/kg twice weekly) were injected via the tail vein twice weekly. Intratumoral injection of M-CSF (0.2 μg/kg twice weekly) was performed twice weekly after inoculating and 48 hours prior to antibody injection. We found that compared with MDMs inoculated alone, additional injection of M-CSF significantly increased the tumor formation efficiency (P < 0.01, Fig. 7A) and accelerated tumor growth of xenografts (P < 0.01, Fig. 7B). CCL18-specific neutralizing antibody markedly inhibited the tumor formation and lowered tumor growth of xenografts (P < 0.01, Fig. 7A and B). These data suggest that CCL18 plays an important role in the malignant transformation of breast phyllodes tumors.

To further evaluate whether CCL18 regulates myofibroblast differentiation in vivo, we examined the protein levels of α-SMA, PTEN, p-Akt, and Ki67 in the xenografts using IHC (Fig. 7C). Similar to the results obtained in vitro, injection of M-CSF enhanced the levels of α-SMA, p-Akt, and Ki67 in the xenografts, but decreased the expression of PTEN, whereas CCL18-specific neutralizing antibody attenuated the upregulation of α-SMA, p-Akt, or Ki67 in the xenografts, but increased the expression of PTEN (Fig. 7C). CD163⁺ cells were significantly increased in M-CSF–treated MDMs group, indicating that M-CSF induced MDMs to M2 macrophages in the xenografts. As increased migration and invasion of tumor cells are linked with metastasis, we evaluated whether CCL18 promoted the metastasis of breast phyllodes tumor xenografts. Consistent with the findings in vitro, injection of M-CSF into the fat pads of the athymic nude mice activated MDMs to M2 macrophages, induced myofibroblast differentiation, promoted migration and invasion of breast phyllodes tumor, and significantly increased the number of metastatic nodules and human hypoxanthine-guanine-phos-phoribosyltransferase (HPRT) mRNA in the livers (Fig. 7D, E, and G). However, there was no significant lung metastasis observed in the mice according to histologic examination (data not shown), wet weight of lung (Fig. 7F), and human HPRT mRNA level (Fig. 7G). In addition to the M2 polarization of macrophages in the tumors, we also found that the peritoneal macrophage of the M-CSF–treated mice, especially of those from the group with the fastest tumor growth, showed M2 polarization (Supplementary Fig. S6).

Myofibroblasts are activated mesenchymal stromal cells that are differentiated from fibroblasts upon tissue injury. They contribute to tissue repair during wound healing by their strong contractility and extracellular matrix (ECM) secretory function. At the end stage of tissue repairing, myofibroblasts disappear by massive apoptosis (27). However, the excessive contraction and ECM protein secretion by myofibroblasts severely impair tissue repairing, such as in hypertrophic scars, scleroderma, and Dupuytren's disease (28). In addition, continuous stimulation by toxic, infectious, and metabolic agents or chronic inflammation leads to continuous myofibroblast differentiation and induces fibrosis in liver, kidney, and heart (29, 30). Myofibroblasts were also found to promote invasion and metastasis of breast, pancreatic, and colorectal cancer (31). In phyllodes tumors, myofibroblasts derive from stromal fibroblasts and constitute the major malignant component of tumor mass. Transition of the mesenchymal fibroblasts to myofibroblasts (FMT) has been suggested as a key process in the tumorigenesis of phyllodes tumors (4). TGFβ1 was shown to be able to induce FMT in fibrosis and thus played an important role in cancer progression (32, 33). In this study, we found that the infiltration of TAMs increased dramatically along with the transformation from benign to malignant phyllodes tumors. TAMs in breast cancer are skewed to M2 phenotype by secreting high levels of IL10, CCL18, and CCL22 (25), and our previous studies have shown that the chemokine CCL18 released by TAMs induces epithelial–mesenchymal transition and promotes metastasis of breast cancer (12). Here we went one step forward by showing that CCL18 from the TAMs of phyllodes tumors is also a key regulator of FMT in myofibroblast differentiation. Blocking CCL18 by anti-CCL18–neutralizing antibody dramatically reverses the myofibroblast differentiation and thus inhibits phyllodes tumor growth in vivo. According to our results, the FMT in phyllodes tumors is driven by CCL18 from TAMs, suggesting an essential role of TAMs in the process of FMT of phyllodes tumor malignant development.

It has been well documented that TAMs secrete CCL18 to promote cancer progression and metastasis, such as breast cancer (34), pancreatic cancer (35), and lung cancer (36), by activating PI3K/AKT (37), Lin 28 (38), and Nir1-ELMO1/DOC180 signaling pathways (39, 40). Our previous work revealed that CCL18 could serve as an NF-κB activator to promote breast cancer progression (25). In this study, we showed that TAMs induced myofibroblast differentiation and promoted the malignant progression of breast phyllodes tumors via secreting CCL18. Furthermore, CCL18 level can serve as an independent prognostic factor of phyllodes tumors. The NF-κB pathway has been identified as a key regulator of fibroblast function and matrix remodeling (41). Activated NF-κB pathway is involved in maintaining myofibroblast phonotype (42, 43). However, whether NF-κB pathway is involved in the FMT process and promotes tumor progression, especially in phyllodes tumors, remains obscure. In this study, we have revealed that the NF-κB pathway was activated by CCL18 in myofibroblast. Subsequently, the nuclear translocated p65 binds to the promoter of miR-21 gene and increases its transcription in the myofibroblast. We have previously illustrated that miR-21 induces myofibroblast differentiation and promotes the malignant progression of phyllodes tumors (4). Thus, miR-21 is a key downstream mediator of p65 in TAM-induced FMT of myofibroblast differentiation.

It is well established that miR-21 is involved in cellular survival, invasiveness, and apoptosis by suppressing specific target genes as PTEN, PDCD4, RECK, TIMP3, Smad3, etc (44–46). Our previous data suggested that Smad7 and PTEN are targets of miR-21 in phyllodes tumors (4). PI3K–AKT pathway is most frequently activated in breast cancer. A case report based on whole genome sequencing revealed N-RAS mutation with concomitant activation of PI3K/Akt/mTOR in phyllodes tumor (47), suggesting the PI3K–AKT pathway may contribute to myofibroblast transition in phyllodes tumors. Indeed, CCL18/NF-κB/miR-21 axis between TAMs and phyllodes tumor cells decreases PTEN level and further activates AKT. Therefore, an intercellular communication, where TAMs secreted CCL18 to activate the NF-κB/miR-21/PTEN–AKT pathway in myofibroblasts, is established (Fig. 6D). Our previous data and this study have demonstrated that CCL18 and miR-21 play major roles in myofibroblast transition of phyllodes tumors, suggesting that the above intercellular communication pathway from TAMs to myofibroblasts may serve as a major driver of tumorigenicity in phyllodes tumors.

Even with surgical resection, local recurrence rate of phyllodes tumors is still as high as 8% to 36% (48). Moreover, recurrent phyllodes tumors are frequently more aggressive in phenotype (49). Our findings demonstrate that CCL18 drives phyllodes tumor progression by inducing FMT of the stromal cells. Higher CCL18 level dramatically correlates with local or distal recurrence of the malignancy and is a better prediction factor for recurrence compared with other clinical and pathologic markers of phyllodes tumors (Supplementary Tables S1 and S2). Given that chemotherapy or radiotherapy is not effective against phyllodes tumors (2), there is a pressing requirement to develop new therapeutic strategy. In this study, our in vivo findings show that blocking CCL18 with neutralizing antibody effectively shrinks phyllodes tumors in mouse xenograft models, which suggests that antagonizing the CCL18 signaling may emerge as a promising strategy to treat phyllodes tumors. Together, our data suggest that the intercellular communication between TAMs and myofibroblasts via CCL18/NF-κB/miR-21/PTEN/AKT axis plays a central role in the tumorigenesis of phyllodes tumors. Monitoring CCL18 level and targeting this pathway raise the possibility of precision diagnosis and treatment for breast phyllodes tumors.

No potential conflicts of interest were disclosed.

Conception and design: E. Song

Development of methodology: Y. Nie, J. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Nie, J. Chen, D. Huang, Y. Yao, J. Zeng, X. Chao, H. Yao

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Nie, Y. Yao, X. Chao

Writing, review, and/or revision of the manuscript: Y. Nie, S. Su, F. Su, H. Hu, E. Song

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Nie, Y. Yao, J. Chen, L. Ding, E. Song

Study supervision: H. Hu, E. Song

E.W. Song is supported by grants from the National Key R&D Program (2016YFC1302301), Natural Science Foundation of China (81490750, 81230060, 81442009), Science Foundation of Guangdong Province (S2012030006287), Guangzhou Science Technology and Innovation Commission (201508020008, 201508020249), Guangdong Science and Technology Department (2015B050501004), Translational Medicine Public Platform of Guangdong Province (4202037), and Guangdong Department of Science & Technology Translational Medicine Center grant (2011A080300002), grant KLB09001 from the Key Laboratory of Malignant Tumor Gene Regulation and Target Therapy of Guangdong Higher Education Institutes, Sun Yat-Sen University and grant [2013]163 from the Key Laboratory of Malignant Tumor Molecular Mechanism and Translational Medicine of Guangzhou Bureau of Science and Information Technology. H. Hu is supported by Natural Science Foundation of China (81672738). Y. Nie is supported by the Natural Science Foundation of China (81502301). H.R. Yao is supported by the Natural Science Foundation of China (81372819, 81572596, U1601223), Specialized Research Fund for the Doctoral Program of Higher Education (20120171110075), and received funding from Guangzhou Science and Technology Bureau (2014J4100170).

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.