M2 tumour-associated macrophages contribute to tumour progression via legumain remodelling the extracellular matrix in diffuse large B cell lymphoma

Effects of M2 tumour-associated macrophages on the pathogenesis of diffuse large B cell lymphoma (DLBCL) are still controversial. Our data showed that the number of CD163-positive M2 macrophages correlated negatively with DLBCL prognosis. Macrophage depletion by clodronate liposomes significantly suppressed tumour growth in a xenograft mouse model of DLBCL using OCI-Ly3 cells. Moreover, M2 polarization of macrophages induced legumain expression in U937 cells. Exogenous legumain promoted degradation of fibronectin and collagen I, which was abolished by administration of a legumain inhibitor RR-11a. Overexpression of legumain in Raw 264.7 cells also induced tube formation of endothelial cells in matrigel. In the xenograft mouse model of DLBCL, decreased fibronectin and collagen I, as well as increased legumain expression and angiogenesis were found at the late stage tumours compared with early stage tumours. Co-localization of legumain and fibronectin was observed in the extracellular matrix of tumour tissues. Administration of the legumain inhibitor to the xenograft DLBCL model suppressed tumour growth, angiogenesis and collagen deposition compared with the control. Taken together, our results suggest that M2 tumour-associated macrophages affect degradation of the extracellular matrix and angiogenesis via overexpression of legumain, and therefore play an active role in the progression of DLBCL.

value in DLBCL 11 . Further study on the correlation between the clinical stage as well as other prognostic indexes and M2 TAMs is urgently needed.
Legumain is an asparaginyl endopeptidase that is classified as a member of the C13 family of cysteine proteases 12 . Accumulating data have identified legumain expression in a variety of tumour types, suggesting a role in tumour progression 13 . Furthermore, studies have verified the effect of legumain on modulation of tumourand stroma-derived components of the cancer degradome 14,15 . A recent study on the crystal structure revealed multi-branched and context-dependent activation processes of legumain, suggesting broad roles in addition to its primary role as an endolysosomal cysteine proteinase 16 . TAMs have been found to express abundant amounts of legumain. A vaccine or inhibitor against legumain significantly eliminates M2 TAMs and consequently inhibits tumour growth and metastases in murine tumour models 17,18 .
Here, we provide evidence that a high percentage of M2 TAMs predicts poor outcomes in DLBCL patients. M2 TAMs in the lymphoma microenvironment overexpressed and secreted legumain. In addition, legumain was involved in angiogenesis and tumour progression via extracellular matrix (ECM) remodelling. Therefore, we have identified a novel therapeutic target as well as a diagnostic index for the treatment of DLBCL.

Results
A high density of CD163-positive M2 TAMs is a poor prognostic factor in DLBCL. Whether a higher ratio of CD163-or CD68-positive macrophages can serve as a predictive index for a poorer prognosis of DLBCL is still controversial. Here, we collected clinical data from 139 DLBCL patients with various stages of disease and calculated their international prognostic index (IPI). In the tumour samples, immunohistochemical staining of CD163 was performed to visualize the CD163-positive macrophages. Our data showed various ratios of CD163-positive macrophages in DLBCL tumour tissues. The mean ratio of CD163-positive M2 macrophages was 13.23% (Fig. 1A, range: 7-50.5%). Compared with stage I, II and III patients, more stage IV patients showed a high percentage of CD163-positive M2 macrophage infiltration (Fig. 1B, 60% vs 27%, 28%, and 17%). Moreover, the percentage of M2 TAMs in low risk groups was significantly lower than that in the high risk group as evaluated by the IPI (Fig. 1C, 9.459 ± 1.014% vs 14.34 ± 1.634, p = 0.0099). These findings suggest that the ratio of CD163-positive M2 TAMs correlates positively with the disease progression and prognosis of DLBCL.
M2 TAMs overexpress legumain together with changes in ECM deposition and angiogenesis in the OCI-Ly3 xenograft mouse model of DLBCL. To explore the mechanism underlying the effect of M2 TAMs on the pathogenesis of DLBCL, especially the role of legumain, we established an OCI-Ly3 xenograft mouse model of DLBCL. The tumour growth curve is shown in Fig. 2A. Animals were sacrificed and primary tumour samples were collected when the tumour sizes reached 800 mm 3 or 1500 mm 3 (n = 5, separately). Flow cytometry analysis suggested that the ratio of CD206-positive M2 TAMs increased with tumour progression (Fig. 2B). Furthermore, we determined the expression of legumain mRNA in the homogenates of tumour tissues. Legumain mRNA levels in the early stage group were significantly lower than those in the late stage group (Fig. 2C). Immunohistochemical staining of legumain and CD206 showed co-localization of legumain and CD206, suggesting legumain expression in M2 TAMs (Fig. 2D). In addition, we identified a change in the ECM contents with tumour progression in this model. As shown in Fig. 2E, immunofluorescence staining of CD31 suggested an increase in blood vessel density with tumour progression. Sirius red staining was performed to identify the collagen content in tumour tissues. Figure 2F shows that the percentage of the collagen area was higher at the early stage than at the late stage. These data suggested remodelling of the ECM, including angiogenesis, and less deposition of collagen with the progression of DLBCL. Furthermore, the increased number of M2 TAMs overexpressed the protease legumain.

Legumain catabolizes ECM components including fibronectin and collagen I.
To further identify the role of legumain in ECM remodelling, we tested the effect of legumain on the major components of the ECM, fibronectin and collagen I, in vitro. Cultured human U937 monocytic cells were treated for various periods with homogenates of tumour samples from the OCI-Ly3 xenograft mouse model of DLBCL. Our results showed that 6 days of treatment with tumour homogenates induced M2 macrophage makers, CD163, IL-10 and MMP9 expression, as well as the expression of legumain in U937 cells (Fig. 3A,B). Next, we treated isolated primary mouse peritoneal macrophages with LPS or IL-4 to stimulate M1 or M2 polarization, respectively. With the M1 polarization induced by LPS (increased expression of IL-6, iNOS, IL-1β , TNF-α and CXCL-9; decreased expression of CD206 and IL-10), no change on the mRNA level of legumain was detected compared with control ( Fig. 3C). However, M2 polarized macrophages induced by IL-4 (increased expression of CD206, Arg-1, CCL2 and IL-10; decreased expression of TNF-α ) express more legumain in the mRNA level (Fig. 3D). Also, data from the western blot assay suggested that M2 polarization condition of macrophages induced expression of legumain (Fig. 3E). Furthermore, treatment with legumain catabolized collagen I and fibronectin. These effects were rescued by administration of a synthetic legumain inhibitor, RR-11a 19,20 (Fig. 3F,G). In vivo, results of immunofluorescence staining showed co-localization of legumain and fibronectin in the xenografted tumour tissues, suggesting extracellular localisation of legumain in the tumour microenvironment (Fig. 3H). Thus, our data demonstrated that legumain was overexpressed by M2 polarized macrophages in the tumour microenvironment and the capacity of legumain to catabolize ECM components outside of the cell membrane.

Overexpression of legumain in macrophages stimulates angiogenesis in vitro.
Neovascularisation is a critical process during tumour progression. To elucidate the role of legumain overexpressed by M2 TAMs in the angiogenesis of DLBCL, we established a stable legumain-overexpressing RAW 264.7 macrophage cell line. Overexpressed legumain was found in both the cell lysates and conditioned medium of legumain-overexpressing RAW 264.7 cells (Fig. 4A). Next, we examined the effect of legumain overexpressed by macrophages on angiogenesis using endothelial cell tube formation assays. The results showed that pretreatment with conditioned medium from legumain-overexpressing RAW 264.7 cells stimulated the tube formation of HUVECs, and the effect was abolished by administration of the legumain inhibitor RR-11a (Fig. 4B,C).

Inhibition of legumain suppresses tumour progression in the mouse model of DLBCL.
To further explore the effect of legumain expressed by M2 TAMs on the tumour pathogenesis of DLBCL, we established the OCI-Ly3 xenograft mouse model of DLBCL and divided the animals into three groups (n = 5). Clodrolip (25 mg/kg), RR-11a (0.2 mg/kg), or the vehicle only (control liposome) was used to treat mice. As shown in Fig. 5B, there existed significant suppression of tumour growth in groups with macrophage depletion (clodrolip) or legumain inhibition compared with the control group. Immunofluorescence staining identified low infiltration of F4/80-and CD206-positive M2 macrophages in the clodrolip group (Fig. 5C). Similarly, RT-PCR also showed decreased mRNA levels of F4/80 and CD206, as well as suppression of legumain (Fig. 5D). CD31 staining revealed less angiogenesis in clodrolip and RR-11a groups compared with the control (Fig. 5E). Sirus red and immunofluorescence staining of collagen I showed more deposition of collagen in the tumour tissue of clodrolip and RR-11a groups compared with the control (Fig. 5F,G). Taken together, these data suggested inhibitory effects of both macrophage depletion and suppression of legumain on angiogenesis and a protective role against deposition of collagen in the ECM.

Discussion
As the largest cell population in the tumour microenvironment, the effects of macrophages on tumour pathogenesis have been widely studied in various tumour models. Several lines of evidence support that a high ratio of M2 polarized macrophages (M2 TAMs) leads to a poor outcome of DLBCL. Marchesi and colleagues suggested high M2 TAM levels at diagnosis as an unfavourable prognostic factor in patients with DLBCL 20 . In this study, we evaluated correlations between CD163-positive M2 TAMs and other prognostic indexes of DLBCL. Our data show that different ratio of CD163-positive M2 TAMs exist in tumour tissue samples of DLBCL patients with various disease stages. Moreover, the number of M2 TAMs correlated positively with the IPI. Although tight correlation has been established in Hodgkin's lymphoma and several other types of haematological tumours, role of TAMs in DLBCL is still controversial till now 21,22 . Our data support the density of CD163-positive M2 TAMs as a potential prognostic index in DLBCL. Obviously, further research including a well-defined cut off value, as well as enrolling more patients is needed before its application in clinical practice. ECM remodelling, including ECM cleavage, modulating angiogenesis and immune responses provides essential signals for tumour progression 23 . Cleavage of ECM components is performed by various families of proteases 24 . Recent studies have revealed that, in addition to matrix metalloproteinases, cathepsins also target many ECM proteins. Our previous study revealed overexpression of a member of the cysteine proteases, legumain, in the tumour microenvironment 25 . Interestingly, not only tumour cells undergoing hypoxia, but also TAMs were the sources of legumain. In this study, co-localization of the M2 TAM marker CD206 and legumain supported that legumain is overexpressed in the tumour microenvironment of DLBCL. Moreover, the expression level of legumain increased with expansion of the tumour, cleavage of the ECM and angiogenesis. In vitro, experiments showed that conditioned medium from M2 TAMs induced cleavage of ECM and formation of vessel tubes. Moreover, administration of the legumain inhibitor indicated that these effects were mediated by legumain. At this time, it is still hard to distinguish the major source of legumain in the tumour microenvironment since both tumour cells themselves and tumour associated macrophage have been proven to overexpress this protein.
Lin et al. revealed that stress stimuli induced TRAF6 ubiquitination of pro-legumain by K63 polyubiquitin thus enhanced both the intracellular stability and secretion of legumain in the tumour cells 26 . Notably, legumain is synthesized as inactive zymogen in the tumour cells while mainly exists in the active form in macrophages, therefore imply a context-dependent action pattern of legumain in different cellular types. Additional investigation is obviously needed to elucidate the exact role and underlying mechanism of action for legumain.
Studies on cathepsins show that they either degrade ECM proteins via being secreted into the extracellular area or internalize ECM components such as collagen through endocytosis and degrade them in lysosomes 27 . As one of the conserved cathepsins, the functional pattern of legumain during the ECM remodelling is still unknown. It is well accepted that legumain is activated in lysosomes intracellularly 28 . However, our study revealed co-staining of legumain and fibronectin in the tumour microenvironment. Moreover, administration of legumain induced cleavage of fibronectin and collagen I. These data support that legumain performs its function outside of the cell membrane. It seems like conflicted with the maintenance of proteolytic activity and stability of legumain. Interestingly, a recent study on the crystal structure revealed different activation routes for legumain, which explained the pH stability of legumain and its extracellular activities to some extent 16 . In addition, recent studies also suggest that exosomes, 30-100 nm-sized membrane vesicles that originate from multivesicular bodies inside of cells, also mediate the extracellular proteolysis by proteases 29,30 . Hakulinen reported that matrix metalloproteinase-14 is secreted from exosomes into the extracellular space and degrades type I collagen 31 . Whether legumain is also transported via exosomes, thus reaching the extracellular space, is worthy of further study. Besides degrading ECM components, our data showed that legumain promoted tube formation by endothelial cells on the surface of matrigel in vitro. It has been reported that pericellular proteases contribute to cleavage of the ECM and the release of angiogenic factors such as vascular endothelial growth factor. Meanwhile, the degradation products might also directly promote angiogenesis.
Considering the critical role of TAMs, especial M2 TAMs in the progression of tumours, diverse strategies aimed to delete M2 TAMs have been designed and achieved encouraging results on limiting tumour progression 21 . Several studies have reported the effects of cathepsin inhibitors on tumour characteristics. Blocking cathepsin activity with multiple small molecule inhibitors results in suppression of the bone metastasis of breast cancer 32 . In this study, we deleted TAMs using clodronate liposomes or blocked legumain activity using a legumain inhibitor in a xenograft mouse model of DLBCL. Both methods delayed xenograft progression significantly with more ECM deposition and less angiogenesis.
In this study, we identified the M2 macrophage marker CD163 as a potential prognostic index for DLBCL. Moreover, legumain secreted by M2 TAMs contributes to cleavage of the ECM as well as angiogenesis in tumour microenvironments. Administration of a legumain inhibitor or deletion of M2 TAMs limits tumour expansion, ECM cleavage and angiogenesis. Therefore, our study suggests a novel potential method to optimize the treatment of DLBCL. Lentivirus production was performed as described previously 33 . To establish a stable legumain-overexpressing cell line, Raw 264.7 cells were transfected with the pLV-EF1α -legumain-IRES-Bsd plasmid. A control cell line containing the empty vector was established by the same methods described previously 33 .

Material and Methods
Scientific RepoRts | 6:30347 | DOI: 10.1038/srep30347 Animal models of tumours. Six-to eight-week-old male NOD/SCID mice were purchased from Beijing HFK Bio-Technology (Peking, China) and housed at the Animal Centre of Collagen of Life Science, Nankai University. All animal experiments were approved by the Institutional Animal Care and Use Committee at Nankai Univeristy, and were performed in accordance with relevant guidelines and regulations. To establish the animal model, 100 μ l of OCI-Ly3 cells (1 × 10 8 cells/ml) was subcutaneously injected at the forth rib of mice. For tumour staging, mice were divided randomly into two groups (n = 5) and sacrificed when the lymphoma volume reached 800 mm 3 (early stage) or 1,500 mm 3 (late stage). To analyze the effect of M2 TAMs and legumain on tumour progression, mice were divided randomly into three groups (n = 5) and intravenously injected with PBS (control) or clodronate liposomes (25 mg/kg) every 4 days (clodronate liposomes and PBS liposomes were provided by Nicovan Rooijen, Amsterdam, Netherlands); or intraperitoneally injected with legumain inhibitor RR-11a (0.2 mg/kg) every day for 12 days. Tumour growth curves were recorded and the mice were sacrificed at the end of the experiments. Primary tumours were collected for tissue sectioning and homogenate preparation.
Tissue processing and Sirius red staining. Tumour tissues were fixed with 3.7% paraformaldehyde, embedded in paraffin, and sectioned. For Sirius red staining, nuclei were stained with Weigert's haematoxylin for 15 2 min, followed by washing under running water for 10 min. Collagen was stained with picrosirius red (Abcam, Cambridge, MA) for 1 h, washed with acidified water, dehydrated, and mounted in a resinous medium. Images were captured by a BX53 research microscope (Olympus, Tokyo, Japan).
Immunohistochemical and immunofluorescence analyses. The detailed procedures of immunohistochemical and immunofluorescence analyses have been described previously 34 . Protein expression was detected in OCI-Ly3 xenograft tissue sections with antibodies against F4/80 (Santa Cruz Biotechnology, Santa Cruz, CA), CD206 (eBioscience, San Diego, CA), CD31 (BD Biosciences), collagen I (Abcam), and legumain (Santa Cruz Biotechnology). Specimens from DLBCL patients were fixed and immunohistochemically stained for CD163 (BD Biosciences). Images were captured by an IX51 research microscope. (Olympus) Tube formation assay. HUVECs were used in the tube formation assay. Briefly, matrigel was coated onto 48-wells plates and then incubated at 37 °C with 5% CO 2 for 30 min to solidify the matrigel. Once the matrigel had set, 200 μ l Raw 264.7 cell-derived conditioned medium was mixed with 50 μ l resuspended HUVEC cells (approximately 2 × 10 4 cells) and then added to the plates. After 6 h, the wells were washed with PBS and the tube network was imaged using the IX51 research microscope.
ECM degradation and western blot analysis. Fibronectin and collagen I were coated to the bottom of 96-well plates and treated with or without legumain for 2 h. Then, protein samples were collected and subjected to western blotting with antibodies against fibronectin (Santa Cruz Biotechnology) and collagen I (Abcam). To detect proteins extracted from cells, equivalent amounts of total protein from each lysate were loaded for western blotting to analyze the target protein level with the indicated antibody including anti-legumain and anti-β -actin (Santa Cruz Biotechnology) antibodies. An ECL chemiluminescence kit (Millipore, Billerica, MA) was used to detect the proteins.