Osteopontin Shapes Immunosuppression in the Metastatic Niche

Despite the advances in tumor prevention, early diagnosis, and therapeutic treatments, breast cancer remains one of the leading causes of death among women, mostly because of its predisposition to metastasize to distant sites, such as lungs, liver, brain, and bones. Much effort has been dedicated to the identification of molecular mechanisms conferring metastatic advantage to specific neoplastic clones so as to target the relevant pathways. Most of these pathways are also necessary for the local invasiveness of the primary tumor, and therefore their targeting may represent an appealing approach to contrast both aspects at the same time.

Besides genetically altered cancer cells, a tumor mass comprises a variety of normal cells that actively participate in tumor progression, including the metastatic process. Tumor and stromal cells produce the extracellular matrix (ECM), composed of structural proteins, such as collagens and fibronectin, and proteins that regulate cell function and their interaction, such as osteopontin (OPN, Spp-1), thrombospondin, SPARC (secreted protein acidic and rich in cysteine), and others, collectively referred to as matricellular proteins. These molecules are generally highly expressed during development, poorly in steady-state condition of adult tissues, whereas readily upregulated in wound healing and tissue remodeling. Matricellular proteins modulate several cellular processes, such as cell adhesion and migration, ECM deposition, cell survival, and proliferation (1). All these processes are also required for primary tumor growth and metastasis, in which matricellular proteins are often aberrantly expressed.

Among matricellular proteins, elevated OPN expression has been correlated with poor survival of patients with cancer with different tumor histotypes (2–4). Accordingly, clinical studies have shown high OPN plasma concentration in patients with metastatic tumors compared with normal samples, especially in breast cancer (5, 6). OPN protumoral and prometastatic activities have been demonstrated in mice and humans (1, 7, 8) through the modulation of tumor cell migration, adhesion, and invasion (9–12).

Nevertheless, the comprehension of molecular mechanisms behind OPN functions in cancer remains far from complete (13). Most studies were centered on OPN produced by tumor cells, whereas only few have focused on OPN produced by the host (14–16), moreover without considering the metastatic setting.

This study, designed to dissect the role of OPN produced by either the tumor or the host in the context of breast cancer, provided the first evidence of OPN tuning the immunosuppressive microenvironment of the metastatic niche.

The 4T1 cell line (CRL-2539; LGC-Promochem) was derived from a murine spontaneous mammary carcinoma arising in a BALB/cfC3H mouse (17). BJMC3879 is a mammary adenocarcinoma developed in a BALB/c female upon inoculation into the inguinal mammary glands of mouse mammary tumor virus (MMTV; refs. 18, 19) and was kindly provided by Dr. Shibata (Osaka Medical College, Osaka, Japan). Female BALB/c mice, 8-week-olds, were purchased from Charles River Laboratories. C57BL/6 spp1^−/− mice (B6.129S6(Cg)-Spp1^tm1Blh/J) were obtained from The Jackson Laboratory and backcrossed on BALB/c background for 10 generations. Mice were maintained at the Fondazione IRCCS Istituto Nazionale Tumori (Milan, Italy) under standard conditions according to the institutional guidelines. All in vivo procedures were approved by the Institute Ethical Committee for animal use.

Stable OPN gene silencing was obtained with the BLOCK-iT Pol II miR RNAi Expression Vector Kit from Invitrogen (see Supplementary Data for details and silencing sequences); Western blot analysis and ELISA on cell supernatants confirmed protein downmodulation (see Supplementary Data), whereas immunohistochemistry was performed on paraffin-embedded sections of primary tumor samples to confirm in vivo downmodulation.

The motility of the 4T1 parental and silenced cells was evaluated in Transwell chambers as previously described (for more details, see Supplementary Data; ref. 20).

Anchorage-independent growth was determined in 0.3% agarose (SeaPlaque; FMC BioProducts) with a 0.6% agarose underlay. Cells (3 × 10³) were plated in semisolid medium (DMEM 10% FBS plus agar 0.3%) and incubated at 37°C in a humidified 5% CO₂ atmosphere. Colonies were counted under a microscope, blindly by two operators, after 8 to 12 days.

For an in vitro anoikis assay, 5 × 10⁵ cells were plated into 6-well tissue-treated plates previously coated or not with polyhydroxyethylmethacrylate (poly-HEMA; Sigma-Aldrich), in DMEM supplemented with 10% FBS or 2% FBS. After 20-hour incubation at 37°C, cells were collected, and anoikis was assessed counting viable cells:dead cells ratio by Trypan blue staining, blindly by two operators.

An adhesion assay was performed in 6-well plates in which 1G11 endothelial cells (a kind gift of Dr. Vecchi, Humanitas, Milan, Italy) were previously seeded 48 hours before (24 + 24 hours in the presence or absence of TNF, at 5 ng/mL to induce cell activation). 4T1GFP and 4T1silOPN GFP cells (5 × 10⁴) were added and let to adhere for 30 minutes; after PBS1x wash, cells were collected by trypsinization and GFP⁺ cells were evaluated by flow cytometry.

Parental, silenced, or control cell lines were inoculated orthotopically in the mammary fat pad (i.f.p.) at a dose of 5 × 10⁴ cells per mouse or intravenously (i.v.) at a dose of 10⁴ cells per mouse. For BJMC3879, 5 × 10⁶ cells per mouse were used. Primary tumor growth was measured once a week with a caliper, and the volume was calculated using the formula d² × D/2, where d and D are the short and long diameters, respectively.

For micrometastasis evaluation, lungs were removed from each mouse 28 days after tumor injection, and a clonogenic assay was performed (see Supplementary Data). For macrometastasis evaluation, lungs were removed after intratracheal perfusion with 15% China ink and fixed with Fekete solution (10% formaldehyde; 5% acetic acid in Et-OH 70%). Metastases, identified as white spots, were counted blindly by two operators. Alternatively, metastases were counted on H&E (hematoxylin and eosin)-stained lung sessions under the microscope.

For a lung colonization assay, 4T1 cells or 4T1 silOPN527, both expressing GFP, were injected i.v. (2 × 10⁶ cells/mouse) in BALB/c mice. Peripheral blood leukocytes (PBL) were drawn at 2, 4, 8, 16, 24, and 48 hours and analyzed by flow cytometry for the presence of GFP⁺ CD45.2⁻ cells. For lung evalution, 1, 3, 4, and 5 days after cells injection, mice were sacrificed, their lungs minced, digested with collagenase/hyaluronidase solution (STEMCELL Technologies), and analyzed by flow cytometry.

Samples, fixed in 10% neutral buffered formalin, were embedded in paraffin. Serial sections were cut at 4 μm from each block and stained with H&E. For immunohistochemical staining, the sections were dewaxed and rehydrated. After heat-mediated antigenic retrieval using a pH 9.0 Tris–EDTA buffer for 30 minutes in thermostatic water bath, the sections were incubated with 3% hydrogen peroxide for 5 minutes to quench endogenous peroxidase activity, and for 10 minutes with protein block (Novocastra, Ltd.). The sections were subsequently incubated with primary antibodies: anti-human/mouse OPN (dilution, 1:100; R&D Systems, Inc.), anti-P-Stat3 (dilution, 1:200; Acris Antibodies, GmbH), and anti-human CD68 (dilution, 1:100; Dako). For negative controls, mouse or rabbit immune sera were used instead of primary antibodies. Staining was revealed with a polymeric system (Novolink, Max Polymer Detection System; Novocastra Ltd.) and visualized by using AEC⁺ High Sensitivity Substrate Chromogen Ready-to-use (Dako Cytomation Liquid AEC Substrate Chromogen System; Dako). Slides were counterstained with hematoxylin and evaluated under a Leica DM2000 optical microscope.

For double immunofluorescence, two sequential rounds of single-marker immunofluorescence were performed using CD68/OPN and CD68/arginase-1 markers. The sections were incubated with the first primary antibody (OPN or arginase-1) overnight at 4°C in a humidified chamber, then washed in PBS, and incubated for 1 hour with an Alexa Fluor 488–conjugated anti-rabbit antibody (dilution, 1:100) for arginase-1 and with an Alexa Fluor 488–conjugated anti-goat secondary antibody (dilution, 1:100) for OPN. Then, the slides were washed in PBS and incubated with CD68 for 1 hour, which was revealed using an Alexa Fluor 568–conjugated anti-mouse Ab (dilution 1:200). The sections were counterstained with DAPI (DNA stain) for 1 minute and mounted using coverslip with a drop of mounting medium.

For confocal microscopy analysis, myeloid cells or 4T1 cells were seeded onto poly-d-lysine–coated glasses, let to adhere, and 2 hours later, fixed with PFA 4%. For intracellular staining of OPN, glasses were sequentially incubated with 0.1% Triton X-100, 1% BSA, polyclonal rabbit anti-OPN (Abcam), in 0.1% BSA solution and Alexa Fluor 546 anti-rabbit secondary antibody. For staining of the endoplasmic reticulum (ER) concanavalin A (Alexa Fluor 488–conjugated; Life Technologies) was added as the final step. Glasses were analyzed with a Microradiance 2000 (Bio-Rad Laboratories) confocal microscope equipped with Ar (488 nm), HeNe (543 nm), and red laser diode (638 nm) lasers. Confocal images (512 × 512 pixels) were obtained using a 20×, 0.5 NA Plan Fluor DIC or 60×, 1.4 NA oil immersion lens and analyzed using ImagePro 7.0.1 software.

For FACS analysis, primary tumors or lungs were collected and perfused with DMEM plus 10% FBS, minced, and then filtered through a 40-μm-pore cell strainer (BD Biosciences). Red blood cells were removed using ACK lysis buffer (ammonium chloride potassium). Cells were Fc-blocked using CD16/32 antibody (eBioscience) before staining. The following antibodies were used: CD45.2, Gr-1, CD11b, F4/80, Ly6C, CD11c, CD4 and CD8, CD44, CD69, and CD62L (all from eBioscience). For regulatory T cells (Treg) detection, after surface staining for CD4, cells were fixed, permeabilized, and stained with FoxP3 Ab, following the manufacturer's instructions (eBioscience).

Samples were acquired using a BD LSR II Fortessa instrument and analyzed with FlowJo software (TreeStar). All samples are analyzed in single; in each experiment, at least three to four samples were analyzed for each group.

Myeloid-derived suppressor cells (MDSC) were purified using CD11b-conjugated microbeads (for overall population) and the Myeloid-Derived Suppressor Cell Isolation Kit (for separation of the two subsets; Miltenyi Biotec) following the manufacturer's instructions.

For an in vitro suppression assay, 4 × 10⁵ naïve BALB/c splenocytes were labeled with carboxyfluorescein succinimidyl ester (CFSE; Sigma-Aldrich) and cocultured with the different MDSC population at different ratios in the presence of 2 μg/mL of soluble anti-CD3 and 1 μg/mL of anti-CD28 to activate lymphocytes. Each sample was seeded in triplicate. Proliferation of CD4 and CD8 T cells was assessed 2 and 3 days later, by flow cytometry evaluating CFSE dilution in the CD4⁺- and CD8⁺-gated populations. Results are shown as the percentage of proliferated cells.

For qPCR analysis, MDSC subsets and other leukocyte populations were purified from the lungs of tumor-bearing mice, pooling at least 5 mice per group; lungs were minced and digested for 45 minutes at 37°C in collagenase/hyaluronidase solution, washed with complete medium, and stained with the specific Abs before cell sorting with a FACSAria instrument (BD Biosciences). Purity of the different population was >99%.

RNA was purified from TRIzol (Invitrogen)-disrupted cells by phenol/chloroform extraction followed by loading onto an RNeasy MINI or MICRO kit (Qiagen). On-column DNAse treatment was routinely performed. RNA purity and yield was assessed using NanoDrop. RNA was reverse transcribed using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). qPCR reaction was prepared using TaqMan Fast Universal PCR Master Mix and run on a 7900 HT Fast Real-time PCR System (Applied Biosystems). The following TaqMan probes were used: Gapdh (Mm99999915_g1), Spp1 (Mm00436767_m1), Arg-1 (Mm00475988_m1), Nos2 (Mm00440502_m1), Il6 (Mm00446190_m1), Stat3 (Mm01219775_m1), Vegf (Mm01281449_m1), and (Mm01178820_m1).

For evaluation of lung metastases, data were represented singularly, and median was shown in the graph; statistical significance was evaluated with the Mann–Whitney test using Prism 5 software. For all the other experiments, data were represented as mean ± SD, and an unpaired two-tailed Student t test was used for statistical analysis.

RNA interference was used to generate stable OPN-silenced 4T1 mammary cancer cells using three different sequences, 215, 527, and 602 (see Materials and Methods and Supplementary Data). Efficiency of gene silencing was evaluated by Western blot analysis on cell lysates and by ELISA on cell supernatants (Supplementary Fig. S1A and S1B); as a control we used cells transduced with nontarget sequence (4T1 NT). Initial in vitro and in vivo experiments were conducted with all the three silenced clones, and with clone 527 thereafter (4T1silOPN527).

Migratory ability of silenced clones assessed across Transwell membranes was significantly inhibited in comparison with that of parental and NT control (Supplementary Fig. S1C). Consistently, OPN neutralization by an anti-OPN antibody markedly diminished the migration of 4T1 wild-type (WT) and control cells (Supplementary Fig. S1D). Anchorage-independent growth, evaluated in soft agar, showed no difference in the number of colonies (Supplementary Fig. S1E, top), despite a reduction in size of 4T1silOPN527 colonies (bottom). No difference was detected in proliferation rate in vitro (not shown).

OPN-silenced 4T1 cells and their WT and NT control counterparts injected into the mammary fat pad formed primary tumors with similar size and onset (Fig. 1A). Immunohistochemical analysis of OPN expression on tumor nodules confirmed efficient OPN downmodulation in the case of silencing (Fig. 1B). On the other hand, the evaluation of spontaneous metastases showed that OPN-silenced 4T1 clones have significantly reduced metastatic capacity (Fig. 1C and D and Supplementary Fig. S1F).

Injection of tumor cells directly into the blood stream bypasses the initial steps of the metastatic process such as tumor detachment and cell intravasation. Intravenous injection of 4T1silOPN527 cells induced significantly less metastases to the lungs than OPN-proficient controls, indicating that tumor-derived OPN is required for a late step of metastasis, from circulation to the seeding into the target organ (Fig. 1E). Consistently, other OPN-silenced clones were also less metastatic upon i.v. injection (Supplementary Fig. S1G).

Tumor cell–produced OPN was not required for adherence to resting or activated endothelial cells as tested in vitro (Fig. 1F). However, the ability to survive in the bloodstream was different, in agreement with the known activity of OPN-protecting cancer cells from apoptosis (21): 4T1silOPN527 cells showed increased cell death than parental 4T1 cells when cultured in nonadherent condition in vitro (Fig. 1G) and at early time points after in vivo i.v. inoculation. Indeed, when we i.v. injected GFP-expressing 4T1 cells and evaluated their numbers in PBL and lungs at early time points after injection, we found that GFP⁺ 4T1silOPN527 cells were fewer in number than GFP⁺ parental 4T1 cells in the lungs starting 24 hours after injection and thereafter. We excluded an uneven number of cells injected because the amount of GFP⁺ cells, either OPN-proficient or -deficient, in PBL was similar at earlier time points (2, 4, 8, and 16 hours; Fig. 1H).

Once demonstrated that tumor-derived OPN works mainly autocrinally, conferring survival advantage in anoikis conditions, we tested whether or not host-produced OPN contributed to metastasis. This possibility stems from the reported activity of OPN in skewing macrophages to a M1 phenotype through downmodulation of IL10 and upregulation of IL12 production (22), and could, theoretically, negatively affect tumor growth and metastasis. Therefore, we injected 4T1 cells in WT and Spp1^−/− mice and, to exclude any masking effect exerted by the high endogenous expression of OPN by 4T1 parental cells, 4T1silOPN527 cells were also included.

4T1 cells formed significantly lower numbers of metastases in Spp1^−/− mice than WT counterparts (Fig. 2A), despite similar primary tumor growth (not shown), demonstrating that host-derived OPN also contributes to metastasis. The lower metastatic ability of 4T1 cells in Spp1^−/− mice was confirmed in the case of i.v. injection, indicating that host-derived OPN is required at later stages of metastasis directly at the seeding organ (Fig. 2B). The paucity of 4T1silOPN527 metastasis in WT mice did not allow us to detect further decrease in metastasis in Spp1^−/− animals (Fig. 2A).

Whether accessory cells involved in breast cancer metastasis, such as macrophages (23), MDSC (24–26), and T cells (27, 28), might use OPN to favor the metastatic process remains unknown.

First, we analyzed the MDSC population, identified as CD11b⁺/Gr-1⁺ cells in the mouse, because of their reported role in 4T1 breast cancer metastasis (25, 29, 30) and of their huge expansion in the lungs of tumor-bearing mice, where they represent 70% to 80% of the infiltrating leukocytes. Overall, the number of CD11b⁺/Gr-1⁺ cells was not different in lungs of Spp1^−/− and WT tumor-bearing mice, whereas splitting the myeloid population into the two subsets of Gr-1^high (granulocytic) and Gr-1^int/low (monocytic) cells highlighted a significant change in their ratio: expanded Gr-1^high subset in Spp1^−/− mice at the expense of the Gr-1^int/low, more suppressive, one (Fig. 2C). Such difference was tumor induced because lungs and spleens of tumor-free mice showed a similar ratio between the two subsets regardless of the Spp1 genotype (Supplementary Fig. S2A). As OPN may have a role in protection from apoptosis, we determined whether lung MDSC from Spp1^−/− tumor bearers had a different apoptotic rate than their WT counterparts. AnnexinV/PI flow cytometry did not show strain difference for either subset (data not shown).

Notably, the Gr-1^int/low subset expressed Spp1 at a much higher level than the Gr-1^high population, as tested by qPCR and confirmed by double immunofluorescence, showing OPN expression mainly in CD68⁺ cells with monocytic appearance (Fig. 2D). This result raised the question of whether OPN plays a role in MDSC immunosuppression.

An in vitro suppressive assay, performed using the purified Gr-1^high and Gr-1^int/low subsets, showed that Gr-1^int/low cells from WT mice were significantly more suppressive than their Spp1^−/− counterparts against CD3/CD28-activated CD8 and CD4 T cells (Fig. 3A and Supplementary Fig. S2B). As expected, the suppressive ability of Gr-1^int/low cells was higher than that of the Gr-1^high subset.

To better characterize the Gr1^high and Gr-1^int/low subsets from metastatic lungs of Spp1^−/− and WT mice, qPCR was performed on genes relevant for MDSC function. Arginase-1 (Arg-1), Tgfb1, Il6, and Stat3 were higher in the Gr-1^int/low than in the Gr1^high subsets; vice versa, Nos2 was higher in the Gr1^high than in the Gr-1^int/low cells (Fig. 3B). Strikingly, Arg-1 expression in the Gr-1^int/low subset was repressed in Spp1^−/− mice, supporting the idea of a less suppressive phenotype we observed in the absence of OPN (Fig. 3A). A similar trend, although not statistically significant, was observed for Stat3, Il6, Tgfb1, and Vegf, with their expression reduced in Spp1^−/− mice. In metastatic lungs, double immunofluorescence analysis confirmed the colocalization of OPN and arginase-1 (Fig. 3C), whereas IHC for phospho (P)-STAT3 stained only cells with a monocytic appearance, which was represented more in WT than Spp1^−/− mice (Supplementary Fig. S3).

To test whether OPN from MDSC was truly relevant for the metastatic spread of 4T1 cells, coinjection experiments were performed. WT, but not Spp1^−/−, MDSC isolated from the spleen of tumor-bearing mice increased lung metastases when coinjected with 4T1 cells (Fig. 3D), and even more significantly when coinjected into Spp1^−/− hosts. No effect was observed when bone marrow–derived macrophages from either Spp1^−/− or WT mice were partners of the coinjection (not shown).

These results indicate that myeloid cell–produced OPN favors the metastatic spread of 4T1 carcinoma.

To test whether host OPN deficiency affected other leukocyte populations infiltrating the lungs of tumor-bearing mice, we performed multiparametric FACS analysis comparing Spp1^−/− and WT mice. No substantial difference in total CD11b⁺ cells (confirming the previous data) and overall CD4⁺ T cells and CD8⁺ T cells was detected. Nevertheless, intracellular staining for the FoxP3 transcription factor, identifying regulatory T cells (Treg), showed a reduced frequency of CD4⁺FoxP3⁺ cells in Spp1^−/− mice (Fig. 4A), a result confirmed by IHC analysis on lung sections (Fig. 4B). A reduced Treg number was associated with increased frequency of CD4⁺-activated T cells (detected as CD44⁺CD69⁺ and CD62L^lowCD69⁺ cells) in Spp1^−/− mice (Supplementary Fig. S4A and S4B). On the same leukocyte populations, sorted from the lungs of tumor-bearing mice, we evaluated by qPCR the level of Spp1 expression to assess whether other cells, besides MDSC, express it. We confirmed a high Spp1 level in the Gr-1^int/low MDSC subset, whereas other leukocytes express it at a much lower level (Supplementary Fig. S4C).

Altogether, these data point to a role of host-derived OPN, mostly from monocytic MDSC, in tuning the immunosuppressive microenvironment of the metastatic lung niche.

We tested whether tumor-produced OPN contributed to molding the local immunosuppression in lung metastases. No significant differences were observed in leukocytes distribution and ratio in 4T1silOPN527 versus 4T1 metastatic lungs; MDSC number and ratio, as well as Treg number, were similar in lungs of mice bearing either parental or OPN-silenced 4T1 tumors (not shown). However, the analysis of in vitro suppressive activity of MDSC isolated from the lungs of mice bearing 4T1 cells or their OPN-silenced counterpart showed less suppression of T-cell proliferation in case Gr1^int/low cells were from mice bearing 4T1silOPN527 tumors (Fig. 5A). Accordingly, this subset expressed lower level of Arg1 and Nos2 than its counterpart isolated from 4T1 WT tumor-bearing mice (Fig. 5B). No different expression for Il6, Vegf, Tgfb1, and Stat3 was detected (not shown).

Considering that 4T1silOPN cells produce less metastases than the 4T1 counterpart, their contribution in terms of immunomodulatory factors should be reduced, explaining the lower immunosuppressive capacity of myeloid cells infiltrating these metastatic lungs. Either alternatively or concomitantly, the downmodulation of OPN expression in tumor cells may alter their secretion of immunosuppressive factors. This was indeed the case, at least for Vegf and Il6 that were less expressed in OPN-silenced tumor cells (Fig. 5C).

Although tumor-derived OPN contributes to shaping MDSC immunosuppressive activity, 4T1 cells, fully competent for OPN production, are not able to compensate for the absence of OPN in the host, and when injected into Spp1^−/− mice, these cells show severely impaired metastatic ability.

To explain such observation, we considered the recently described intracellular form of OPN (iOPN), which has been found enriched in cells of the monocytic lineage (31), and we hypothesized that tumor cells produce OPN mainly in the secreted form (sOPN), whereas myeloid cells in the intracellular form.

We evaluated the amount of sOPN in the supernatant from tumor and myeloid cells in relation to their OPN expression by qPCR. One million 4T1 tumor cells secreted roughly 1 μg/mL/24 hours of OPN, which was almost undetectable in the supernatant from myeloid cells (Fig. 6A). Although gene expression data by qPCR confirmed the trend of a much higher level in tumor cells than in myeloid cells, the ratio is not so striking as in the ELISA assay (300× vs. 10,000× in qPCR and ELISA, respectively; Fig. 6B). In addition, considering that qPCR data account for both sOPN and iOPN, it is likely that most OPN from tumor cells is secreted, whereas that from myeloid cells is retained in the cell. This interpretation was confirmed by confocal analysis showing that in tumor cells OPN was mainly cytoplasmic and colocalized with concanavalin A (staining ER), whereas in monocytic myeloid cells, OPN remained confined to specific spots under the cellular membrane, as reported in ref. 31, and not localized in the ER (Fig. 6C).

To assess whether host-derived OPN favors metastatic dissemination of mammary carcinomas other than the 4T1 model, the BJMC3879 adenocarcinoma (18, 19), another spontaneously metastasizing BALB/c breast tumor, was tested. Along with tumor progression, BJMC3879 expanded CD11b⁺/Gr-1⁺ cells in PBL, spleen, and lungs (data not shown) and, phenocopying 4T1 cells, produced similar primary tumors but less lung metastases in Spp1^−/− than WT mice (Fig. 7A–C). In addition, similar to the 4T1 model, the CD11b⁺/Gr-1⁺ cell population showed an increase in the Gr-1^high fraction in the absence of OPN (Fig. 7D). These data indicate that the effect of host-produced OPN in breast cancer lung metastasis is not restricted to the 4T1 tumor model.

To assess whether our finding in the mouse models has any clinical correlation, we evaluated the expression of OPN in lung metastases from 5 patients with breast cancer undergoing metastasectomy (see Supplementary Table S1 for clinical parameters). In the cases evaluated, OPN expression was detected in cells with monocytic morphology infiltrating metastatic cancer foci (Fig. 7E). Double immunofluorescence analysis using Abs for human OPN and the monocytic/macrophagic marker CD68 confirmed the myeloid/monocytic nature of OPN-expressing cells infiltrating breast carcinoma lung metastases (Fig. 7F).

Tumor development and progression require a suitable microenvironment in which tumor cells conscript and corrupt normal resident cells, such as fibroblasts and endothelial cells, while recruiting accessory cells from the bone marrow toward their needs (32). Besides cellular elements, tumor stroma comprises the ECM, a complex network of macromolecules with different physical, biochemical, and biomechanical properties, also relevant for tumor progression and metastasis (33). Among matricellular proteins, OPN has been extensively studied as a tumor cell product conferring migratory, adhesive, and growth capacity. Whether OPN produced by stromal cells contributes to tumor progression and metastasis has not been defined. Here, we have dissected the role of OPN in breast cancer primary tumor growth and metastasis when its production was either tumor or host derived. Whereas our data substantially agree with the available literature in finding tumor-derived OPN prometastatic, the effect on primary tumor was negligible in our models, while rather relevant in other transplantable tumors (34–37). The effects of OPN inhibition in tumor cells have been investigated mainly in vitro testing cell adhesion, migration, invasion through Matrigel (34, 35, 37) or cell proliferation (38). Silencing OPN in 4T1 tumor cells, we obtained concordant inhibition of cell migration and invasion but no effect in cell proliferation and adhesion to different substrates (data not shown). Such in vitro effects, however, found little correlation with the in vivo metastatic activity we observed, in which tumor-produced OPN renders the cells intrinsically more resistant to apoptosis while in the blood stream, rather than favoring detachment from the primary site or intravasation. Accordingly, we found a similar antimetastatic effect of OPN silencing, using both intra-mammary fat pad and intravenous injection routes and did not observe any difference in the epithelial-to-mesenchymal transition (EMT) phenotype between OPN-silenced or WT primary tumors (not shown). Nevertheless, data showing smaller soft-agar colonies (Fig. 1D), and smaller metastatic foci in the lungs of mice injected with OPN-silenced tumor cells, suggest that OPN might also contribute to lung parenchyma colonization.

Here, we provide novel findings showing that host-derived OPN, in particular from myeloid cells, molds the immunosuppressive microenvironment at the metastatic niche. The metastatic potential of mammary tumors was greatly reduced when the recipient mice were OPN deficient, despite the fact that tumor was fully competent for OPN production, and the primary tumor grew unmodified. The analysis of the immune microenvironment at the metastatic site (i.e., lungs) showed that OPN favors the presence of immunosuppressive leukocytes, and that such leukocytes were less suppressive in Spp1^−/− mice. Myeloid cells endowed with suppressive activity, generally termed MDSC, comprise two main subsets in the mouse, which can be distinguished by the different expression of the Gr-1 marker: highly expressed on the most abundant, and poorly suppressive, granulocytic subpopulation (Gr-1^high), whereas expressed at low to intermediate level in the highly suppressive monocytic subset (Gr-1^int/low). Such MDSC are massively recruited to both primary tumor and metastatic lungs in the 4T1 mammary carcinoma (39, 40). We show here that in Spp1^−/− mice, the ratio of the two subsets is significantly different, with a less abundant Gr-1^int/low-suppressive population (Fig. 2C). Interestingly, OPN expression seems to be mainly confined to the monocytic subset, suggesting that OPN may directly contribute to their immunosuppressive phenotype. Indeed, when tested for their ability to suppress T-cell proliferation in vitro, Gr-1^int/low cells from lungs of 4T1-bearing Spp1^−/− mice were less suppressive than their WT counterparts (Fig. 3A). We hypothesize that the different ratio of the two MDSC subsets in favor of the Gr1^high population and the reduced suppressive activity of the remaining Gr-1^int/low subset create a lung microenvironment less permissive to metastasis. The reduced suppression of spp1^−/− Gr-1^int/low cells correlated with a lower expression of Arg-1, IL6, and Stat3, all marks of immunosuppression highly expressed in OPN-proficient MDSC (Fig. 3B). In vivo, coinjection of tumor cells with MDSC from either Spp1^−/− or WT mice confirmed that only those from WT mice promoted metastasis of 4T1 cells, regardless of the Spp1 genotype of recipient mice (Fig. 3D). Whether MDSC are produced or recruited differently from the bone marrow or whether they can transdifferentiate locally in relation to Spp1 genotype is currently unknown.

The absence of host OPN seems to affect other aspects of local immune response in the lung metastatic niche, that is, it associates with a decrease in the number of Treg and with enhanced activation of CD4 T cells. Although OPN influences MDSC-suppressive activity, it does not modify Treg–mediated immunosuppression, at least in vitro, according to our unpublished data. Whether the effect on Treg number depends directly on the different MDSC population in Spp1^−/− versus WT mice deserves further investigation.

The MDSC compartment was affected not only by host-derived but also by tumor-produced OPN (Fig. 5), likely because OPN-silenced tumors form less metastases and contribute less in producing immunosuppressive factors, such as VEGF and IL6.

Despite such contribution from tumor-derived OPN, 4T1 cells, fully competent for OPN production, are not able to compensate their absence in the host, and when injected into Spp1^−/− mice show severely impaired metastatic ability. To explain this datum, we investigated OPN localization in tumor and myeloid cells. Indeed, other than the soluble OPN, an intracellular form (iOPN) has been recently identified in leukocytes, particularly in cells of the monocytic lineage (31). Combining dosage of secreted OPN by ELISA with qPCR expression analysis and confocal microscopy (Fig. 6), we demonstrate that tumor cells produce OPN in high amount in the secreted form, localized within the Golgi apparatus and ER. Differently, in myeloid cells, OPN is mainly condensed in specific spots localized in perimembrane areas, phenocopying the described iOPN localization (33).

The newly described role of host-derived OPN in metastasis was corroborated using a different spontaneously metastatic breast carcinoma model, the BJMC3879 cell line (41), that formed less spontaneous lung metastases when injected into Spp1^−/− mice (Fig. 7A–D). Very few reports analyzed the role of host-produced OPN in tumor progression (14, 15, 35). Two reports (14, 15) studied carcinogen-induced papilloma and cutaneous squamous cell carcinoma (SCC), induced directly into Spp1^−/− mice, therefore in complete absence of OPN in both host and tumors. The study from Chakraborty and colleagues (35) was the only one on breast cancer and showed a protumorigenic activity of host-derived OPN through the induction of oncogenic molecules, such as MMPs and uPA, and of angiogenesis, without considering the effect on metastasis.

Our demonstration that myeloid-derived OPN contributes to the immunosuppressive environment required for metastasis opens the question of whether other matricellular proteins may regulate tumor-associated immune responses and adds a whole new pathway to those involved in MDSC fitness. Moreover, this study further enforces the concept that matricellular proteins may produce different effects at discrete phases of tumor progression, according to their tumor-intrinsic or microenvironment-associated expression (42).

Notably, our data find clinical correlation in the presence of myeloid cells producing OPN in lung metastases of human ductal mammary carcinomas. In these samples, OPN expression was detected in scattered reactive cells with monocytoid morphology that were mostly localized within the stroma at the interface between the metastatic foci and the surrounding lung parenchyma, suggesting a role in the local events underlying constitution of the metastatic niche. The in-depth characterization of OPN-expressing cancer-associated myeloid populations and their potential biologic/prognostic relevance in cancer patient cohorts will represent an appealing issue of investigation.

In conclusion, our work is the first to show a role for both tumor- and host-derived OPN toward metastatic dissemination through the modulation of local immunosuppression in the lung metastatic niche.

No potential conflicts of interest were disclosed.

Conception and design: M.P. Colombo, C. Chiodoni

Development of methodology: S. Sandri, C. Chiodoni

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Sangaletti, C. Tripodo, S. Sandri, I. Torselli, C. Vitali, C. Ratti, L. Botti, A. Burocchi, A. Tomirotti

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Sangaletti, C. Tripodo, S. Sandri, C. Vitali, C. Chiodoni

Writing, review, and/or revision of the manuscript: S. Sangaletti, C. Tripodo, M.P. Colombo, C. Chiodoni

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Porcasi

Study supervision: M.P. Colombo, C. Chiodoni

The authors thank Dr. Shibata (Osaka, Japan) for providing the BJMC3879 cell line and Ivano Arioli and Barbara Cappetti for technical assistance. The authors also thank the Conventional and Confocal Microscopy Facility for confocal acquisition, and the Cell Sorting Facility for MDSC subsets cell sorting of the institute.

This work was supported by the Associazione Italiana per la Ricerca sul Cancro (AIRC: Investigator Grants 10137 to M.P. Colombo and Program Innovative Tools for Cancer Risk Assessment and Diagnosis, 5 per mille n°12162 to M.P. Colombo and C. Tripodo; MFAG n° 12810 to S. Sangaletti) and the Italian Ministry of Health.

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.