Targeting the C-terminus of galectin-9 induces mesothelioma apoptosis and M2 macrophage depletion

ABSTRACT

Galectin-9 has emerged as a promising biological target for cancer immunotherapy due to its role as a regulator of macrophage and T-cell differentiation. In addition, its expression in tumor cells modulates tumor cell adhesion, metastasis, and apoptosis. Malignant mesothelioma (MM) is an aggressive neoplasm of the mesothelial cells lining the pleural and peritoneal cavities, and in this study, we found that both human MM tissues and mouse MM cells express high levels of galectin-9. Using a novel monoclonal antibody (mAb) (Clone P4D2) that binds the C-terminal carbohydrate recognition domain (CRD) of galectin-9, we demonstrate unique agonistic properties resulting in MM cell apoptosis. Furthermore, the P4D2 mAb reduced tumor-associated macrophages differentiation toward a protumor phenotype. Importantly, these effects exerted by the P4D2 mAb were observed in both human and mouse in vitro experiments and not observed with another antigalectin-9 specific mAb (clone P1D9) that engages the N-terminus CRD of galectin-9. In syngeneic murine models of MM, P4D2 mAb treatment inhibited tumor growth and improved survival, with tumors from P4D2-treated mice exhibited reduced infiltration of tumor-associated M2 macrophages. This was consistent with an increased production of inducible nitric oxide synthase, which is a major enzyme-regulating macrophage inflammatory response to cancer. These data suggest that using an antigalectin 9 mAb with agonistic properties similar to those exerted by galectin-9 may provide a novel multitargeted strategy for the treatment of mesothelioma and possibly other galectin-9 expressing tumors.

Introduction

Galectin-9 is a tandem-repeat-type galectin with two carbohydrate recognition domain (CRD) with distinct physiologic activities, that has recently emerged as a novel candidate for cancer treatment.^1–4 Both in vitro and in vivo studies showed that recombinant galectin-9 induces apoptosis of tumor cells, such as hematologic malignant cells,^5,6 melanoma,⁷ and gastrointestinal tumors.^8–11 Studies with immune cells suggest that galectin-9 could also modulate cells of the tumor microenvironment as T cells,¹² B cells, and macrophages,^13,14 although it is unclear if this modulation leads to an antitumor or protumor effect.^15,16

MM is a lethal cancer linked to asbestos that is increasing in incidence worldwide.¹⁷ Macrophages were demonstrated to have a crucial role MM carcinogenesis as well as for its development.¹⁸ Tumor-associated macrophages (TAMs) are abundantly present in the MM microenvironment and play an important role in inducing T-cell suppression.¹⁹ It has been demonstrated that pleural effusions from MM patients induce recruitment of monocytes and influence their differentiation into M2 macrophages.²⁰ These macrophages promote the development and metastatic capacity of tumors due to the production of protumor factors like the enzyme arginase^1,21 and a larger M2 component of the total macrophage count is inversely correlated with survival.^22–24

The role of galectin-9 in MM remains uncharacterized. In this study, we evaluated the expression of galectin-9 in murine and human MM cells and developed several antigalectin-9 targeted monoclonal antibodies with the goal of modulating the activity of galectin-9 and evaluating the effects on both cancer and immune cells. We provide evidence that immunotherapies utilizing a unique antigalectin 9 mAb exhibiting agonist activity to galectin-9 represents a promising new approach in cancer treatment.

Results

Human MM tumors express galectin-9

Galectin-9 is expressed in several human tumors and has been shown to modulate tumor progression, metastasis, and apoptosis as well as predict cancer patient survival.^5–7 The expression of galectin-9 in MM remains unknown. Therefore, we performed immunohistochemistry galectin-9 staining of 16 human MM biopsies and three normal human mesothelial lining samples. Staining analysis indicated that 14 out of 16 MM biopsies showed detectable levels of galectin-9 in the tumor biopsies, ranging from focally to diffusely positive. In contrast, galectin-9 expression was very low to undetectable in the normal mesothelial lining samples (Supplementary Table 1). Galectin-9 staining was localized in both nucleus and cytoplasm of cells (Figure 1).

Figure 1.

Profiling of galectin-9 tissue expression in MM tumors. (a–c) Galectin-9 staining on three representative MM samples; (d) Gal9 staining on a representative normal mesothelial lining. Original magnification 200×.

Novel antigalectin mAbs bind to both human and mouse galectin-9

To further evaluate the significance of galectin-9 in MM, we generated a series of antigalectin-9 mAb clones, and evaluated their binding to human and mouse galectin-9. We identified 8 mAbs that bound to human galectin-9, with only P4D2 and P1D9 clones binding to both human and murine galectin-9 (Figure 2a). We evaluated the binding of these two mAb clones to two versions of human galectin-9, with (hGalectin-9M) or without (hG9NC) the linker peptide. Both mAbs showed binding to both versions of galectin-9 (Figure 2b).

Figure 2.

Generation of antigalectin-9 mAb and corresponding specificity and cross-reactivity. (a) Binding of generated galectin-9 mAbs was evaluated via ELISA plates coated with human or mouse recombinant galectin-9. Mouse serum was used as positive control. Averages of optical densities (OD) are shown as an index of binding. (b) Binding of P4D2 and P1D9 mAbs was compared among human recombinant galectin-9 (hGalectin-9M) and a more stable version of human recombinant galectin-9 missing the linker peptide between N- and C-CRD (hG9NC). Commercially available galectin-9 mAb, clone 9M1-3, was used as a control. (c) hG9G8 (Gal-9 N-terminus CRD only) or hG8G9 (Gal-9 C-terminus CRD only) fusion proteins were used to evaluate P4D2 and P1D9 mAb CRD binding specificity. Commercially available galectin-9 mAb, clone 9S2-3, was used as a control. (d) P4D2 mAb Fv sequencing and modeling of its interactions with galectin-9 was structurally analyzed with SAbPred modeling software. Briefly, RNA was extracted from the P4D2 Hybridomas using TRIzol® Reagent (Thermo Fisher Scientific) and converted to cDNA with SuperScript™ III First-Strand Synthesis System (Invitrogen). Contaminating V_L cDNA from the P3X63Ag8.653 cells were labeled using 5ʹ biotinylated P3 CDR-L3 primers (5’-CAGCACATTAGGGAGCTTACACG-3’) (IDT) and then removed using Streptavidin-linked Dynabeads. The enriched cDNA was amplified using primers K6b and revCk for V_L, and primers H2 and IgG2a for V_H. Both V_L and V_H amplicons were sequenced using Sanger sequencing and annotated in NCBI igBLAST with an IMGT number.²⁵ All primers were designed as previously described.²⁶ IMGT numbers for V_L and V_H were entered into the SAbPred modeling software to generate a protein data bank (PDB) file. The PDB for P4D2 V_L/V_H was entered, along with the galectin-9 crystal structure (PDB ID: 3WV6) into the SAbPred epitope modeling software to identify their binding sites.²⁷ Interactions between the predicted P4D2 Fv structure and galectin-9 crystal structure were then analyzed and results showed binding exclusively to the C-terminal CRD of galectin-9. Galectin-9 amino acids that interact with P4D2 Fv identified with SAbPred are listed.

Differential binding of antigalectin mAbs to the C-terminal (P4D2) and N-terminal (P1D9) CRD of galectin-9

Galectin-9 contains two CRDs in the N- and C-terminal regions (N-CRD and C-CRD, respectively) that display different activities, with the former involved in the regulation of innate immune cells and the latter more effective in inducing T-cell apoptosis.⁴ To identify the CRD recognized by each mAbs P4D2 and P1D9, we generated two fusion proteins, hG9G8 and hG8G9, which has one of the CRD from galectin-9 substituted with the CRD from galectin-8. The fusion protein hG9G8 comprises the N-terminal CRD from human galectin-9, but has the C-terminal CRD of galectin-8, while hG8G9 includes the N-terminal from human galectin-8 and the C-terminal from galectin-9. ELISA plates were coated with these two fusion proteins, and binding of either of the two (P4D2 or P1D9) mAbs evaluated. In these experiments, P4D2 showed strong binding with hG8G9, containing the C-terminal region of galectin-9. Binding of the P4D2 clone to hG9G8 was significantly reduced compared with binding to hG8G9. In contrast, the P1D9 mAb showed stronger binding with hG9G8 when compared with hG8G9 (Figure 2c). To further characterize the interaction between the mAbs and galectin-9, we sequenced the variable domain (Fv) of the P4D2 clone and used this information to design a digital model of this mAb. We then performed simulation modeling between the digital prototype of P4D2 clone and the crystal-structure of galectin-9 using SAbPred. This analysis confirmed binding of P4D2 mAb to the C-terminal of galectin-9. The galectin-9 amino acids involved in the interaction with the Fv of P4D2 mAb that were identified with SAbPred are listed (Figure 2d).

P4D2 has agonistic properties and induces apoptosis of MM cells

It has been demonstrated that endogenous galectin-9, by binding carbohydrates on the cell membrane, forms two- or three-dimensional lattice structures that perform important functions including organizing cell membrane domains, determining thresholds of cell signaling and regulating receptor turnover on the cell surface.^28–30 Treatment with recombinant galectin-9 has been demonstrated to interact with these lattice structures and promote apoptosis in different types of tumor cells.^5–11 In our preliminary experiments, we observed that both human and mouse MM cells express surface galectin-9 (Supplementary Figure 1a–c) and therefore investigated if P4D2 and P1D9 mAbs could mimic the apoptotic effect induced by recombinant galectin-9. We assayed two human MM cell lines (Mill and ROB) and two mouse MM cell lines (CRH5 and EOH6). In Mill cells, we observed reduced viability during P4D2 treatment compared with untreated controls at day 2, 3, 4, and 5. hG9NC treatment also reduced cell viability of Mill cells from day 2 until the end of the assay (day 5) (Figure 3a). Both P4D2 and hG9NC reduced viability significantly of ROB cells, from day 2 until the end of the assay (day 5). In CRH5 cells, P4D2 significantly reduced cell viability for the entire duration of the culture from day 1 to day 5. Mouse galectin-9 (mG9NC) also decreased cell viability but only at day 5. In EOH6 cells, both P4D2 and mG9NC reduced viability at day 3 through the end of the assay. We also evaluated MM cell viability following combined treatment using recombinant galectin-9 and P4D2. In these assays, we observed decreased cell viability using the two agents in combination compared with the single treatments. Importantly, we did not observe reduced MM cell viability following treatment with other lectins such as galectin 3 or erythrina cristagalli (Supplementary Figure 2a,b).

Figure 3.

P4D2 mAb exerts agonist effects in inducing apoptosis in MM cells. (a) Human MM cells (ROB and Mill) were treated with P4D2 or P1D9 mAbs, and viability assessed with an MTT assay. Controls included the human stable recombinant galectin-9 (hG9NC) and no treatment (Ctrl). Differences between Ctrl and P4D2 as well as Ctrl and hG9NC were statistically significant with P ≤ 0.01; n = 3 (*). A viability assay was used to evaluate the effects of P4D2 and P1D9 mAbs on mouse MM cells (CRH5 and EOH6). Mouse stable recombinant galectin-9 (mG9NC) was included for comparison. Differences between Ctrl and P4D2 as well as Ctrl and mG9NC were statistical significant with P ≤ 0.01; n = 3 (*). (b) Analysis of apoptosis for human (ROB and Mill) and mouse (CRH5 and EOH6) MM cells after P4D2 or recombinant galectin-9 (hG9NC or mG9NC) treatment as evaluated by flow cytometry. Percentages of PI- Annexin V+ (early apoptotic) cells and PI+ Annexin V+ (late apoptotic) cells are shown. Statistically significant differences between treatment and no treatment (Ctrl) were assessed with two-way ANOVA followed by the Bonferroni test and indicated with *, P ≤ 0.01, n = 3.

We next investigated the degree of apoptosis induced by the antigalectin-9 mAb treatment of MM cells using a combination of Propidium Iodide (PI) + Annexin V costaining. In Mill cells, hG9NC and P4D2 induced higher percentages of PI⁺/Annexin V⁺ cells representing late apoptotic cells, compared with untreated MM cells. Also, an increase in early apoptotic cells (PI⁻/Annexin V⁺) was detected with P4D2 (Figure 3b). In ROB cells, we observed an increase in both early and late apoptotic cells for P4D2, while hG9NC treatment induced only an increase in late apoptotic cells. In CRH5 cells, both P4D2 and mG9NC induced higher percentages of late apoptotic cells compared with untreated controls. In these cells, a significant increase in the numbers of early apoptotic cells was assessed also for both treatments (P4D2 and mG9NC) when compared with controls. In EOH6 cells, P4D2 was the only treatment that induced higher percentages of both early and late apoptotic cells when compared with untreated controls.

P4D2 mAb modulates human monocyte differentiation with reduced formation of CCR5⁺ tumor macrophages

Next, we investigated whether the antigalectin-9 mAbs would alter monocyte differentiation or reduce formation of protumor myeloid cells. Human blood-derived monocytes (hBDM), which express galetin-9 (Supplementary Figure 3) were differentiated with human AB serum in the presence of either P4D2 or P1D9 and compared to those differentiated in the presence of hG9NC. After 1 week of culture, cell surface levels of CD68 and CCR5 were assessed. CD68 is a marker highly expressed by cells of the monocyte lineage, including circulating and tissue macrophages,³¹ while CCR5 increases during monocyte-macrophage differentiation and is responsible for the recruitment of TAMs in the tumor microenvironment.³² We observed a significant reduction in CD68⁺ CCR5⁺ mature macrophages after treatment with either P4D2 or hG9NC, compared with control cells differentiated in AB serum without any treatment. Interestingly, the number of live cells significantly increased with P4D2, suggesting that this mAb acts by blocking monocyte differentiation rather than killing mature macrophages (Supplementary Figure 4). In contrast, P1D9 displayed the same percentage of CD68⁺ CCR5⁺ cells compared to controls (Figure 4a,b). We also induced monocyte-macrophage differentiation using supernatants from human ROB MM cells untreated or treated with P4D2 (Figure 4c). In these assays, the number of CD68⁺ and CCR5⁺ cells was sharply reduced when the supernatant from P4D2-treated ROB MM cells was used, compared to controls cultured with supernatant of untreated ROB MM cells (Figure 4d,e).

Figure 4.

Treatment with a P4D2 mAb shifts primary human monocyte differentiation away from protumor phenotype. Human primary monocytes were differentiated with human AB serum in presence of P4D2 or P1D9 mAbs. Two other conditions included: (1) human serum and (2) human stable galectin-9 (hG9NC). (a) Representative figures from the flow cytometric analysis of CD68+ and CCR5+ mature macrophages. (b) Percentages of CD68+ and CCR5+ cells are showed for the different treatments. Statistical differences among treatments (n = 3/group) were assessed with one-way ANOVA followed by the Bonferroni test and indicated with *P ≤ 0.01. (c) Illustrative schematic of monocyte-macrophage differentiation experiment. (d) Monocyte-macrophage differentiation was evaluated using supernatants from ROB MM cells (conditioned media) or from ROB MM cells treated with P4D2 mAb (conditioned media + P4D2). P4D2-treated cells were used as control. Flow cytometry representative images show the reduction of CD68+ and CCR5+ mature macrophages induced by P4D2-treated ROB media. (e) Differences in the percentage of CD68+ and CCR5+ cells in ROB media compared to P4D2 mAb-treated ROB media, or P4D2 (n = 3/group) were evaluated using one-way ANOVA followed by the Bonferroni test and indicated with *P ≤ 0.001. (f) Maturation of primary monocytes was also measured with real-time PCR using primers for the M2 marker, MARCO. Differences in ROB media compared to P4D2 mAb-treated ROB media for MARCO were defined using one-tailed paired Student’s t test (n = 3/group) and indicated with *P ≤ 0.001.

Real-time PCR was used to assess mRNA levels for CD68 and the macrophage receptor with collagenous structure (MARCO) in cells differentiated with supernatants from either untreated human ROB MM cells or treated with P4D2. MARCO is a marker expressed by immune-suppressive TAMs that has been linked to poor prognosis in cancer patients.³³ Macrophages differentiated with the supernatant from P4D2-treated ROB MM cells showed decreased MARCO mRNA compared with cells differentiated with the supernatant from untreated ROB MM cells. Levels of CD68 did not change between the two conditions (Figure 4f). In all the experiments listed here, we also evaluated transferrin receptor CD71 expression in mature macrophages as index of metabolic activity.³⁴ Interestingly, only P4D2 increased the number of CD71⁺ cells, compared with all the other treatments (Supplementary Figure 5a–e). We observed additional phenotypic differences between P4D2, P1D9, and hG9NC in macrophages matured with human serum. In these experiments, macrophages matured with hG9NC showed higher expression of CD11b and CD206, while expression of CD11c and CD14 was reduced. P4D2 did not alter the expression of these markers except for CD14 that was significantly increased when compared with all the other conditions (Supplementary Figure 6).

CCR2 is a chemokine receptor that mediates the recruitment of myeloid cells to the tumor microenvironment.³⁵ In our analysis, we observed a reduction of CD68⁺ CCR2⁺ macrophages following incubation with either P4D2 or hG9NC, even if statistical significance was obtained only for hG9NC. Surprisingly, P1D9 dramatically increased CCR2 expression compared with controls (Supplementary Figure 7). Further experiments were also conducted to evaluate migration and cytokine secretion of macrophages differentiated with AB serum with or without P4D2. In these experiments, cells incubated with serum + P4D2 showed increased migration and IL-8 production when compared with macrophages matured only with AB serum (Supplementary Figure 8).

P4D2 shifts mouse monocyte differentiation toward an M1 phenotype

We demonstrated above that P4D2 hinders human monocyte maturation to macrophages with tumor-promoting characteristics, and next investigated mouse bone marrow-derived monocytes (mBMMs). In these assays, we used four different conditions to induce mBMM differentiation to macrophages: (1) M-CSF; (2) M-CSF plus P4D2; (3) supernatants from mouse MM CRH5 cells; and (4) supernatants from CRH5 cells treated with the P4D2. Flow cytometry was used to measure F480⁺ cells as a marker that identifies mouse macrophages. Results showed a significant increase of F480⁺ macrophages in cultures using the supernatant from P4D2-treated MM cells when compared with cells cultured with the supernatant from untreated MM cells. No differences were observed between cells differentiated with M-CSF with or without P4D2 (Figure 5a,b). We further characterized F480⁺ cells using markers for M1 antitumor (CD38^hi) and M2 protumor (Egr2+) macrophages.³⁶ With both M-CSF and CRH5 media used as maturation stimuli, we observed a significant increase in F480⁺CD38⁺ macrophages when P4D2 was employed. Regarding F480⁺Egr2⁺ macrophages, we observed a complete lack of differentiation into these cells in the presence of P4D2, either when mBMMs were differentiated with M-CSF or CRH5 media (Figure 5c,d). When we calculated M1 (F480⁺CD38⁺) and M2 (F480⁺Egr2⁺) ratios, we observed an increase toward the M1 phenotype for cells treated with P4D2 mAb and M-CSF compared with cells treated with M-CSF, and for mBMMs cultured with the supernatant from P4D2-treated MM cells compared with those cultured with supernatant from untreated CRH5 MM cells (Figure 5e). For all of these conditions, we did not detect differences in the number of live cells (Figure 5f). We also performed experiments to evaluate the effects of P4D2 during mBMM differentiation induced with higher dosages of M-CSF, to mimic the tumor microenvironment in which higher concentrations of this cytokines were found compared with normal tissues.³⁷ As observed for lower doses of M-CSF, P4D2 strongly reduced the percentages of F480⁺ cells, but the number of live cells was also dramatically reduced (Supplementary Figure 9). Surprisingly, mG9NC also decreased the number of F480⁺ cells, but viability was not altered, even at high concentrations of M-CSF.

Figure 5.

P4D2 mAb shifts mouse monocyte differentiation toward an M1 phenotype. Mouse bone marrow-derived monocytes were differentiated to mature macrophages using either M-CSF or supernatant from mouse MM CRH5 cells. During M-CSF-driven differentiation, cells were also treated with P4D2 mAb (M-CSF + P4D2) while controls only received m-CSF. Differentiation with MM supernatant was instead performed using media from CRH5 mouse MM cells untreated (conditioned media) or treated for 24 h with P4D2 mAb (conditioned media +P4D2). (a and b) The left panels contain representative flow cytometry images showing the effects of P4D2 mAb on F480+ cells. On the right, percentages of F480+ macrophages are shown for the different treatments. Differences between CRH5 media compared to media from P4D2 mAb-treated CRH5 were evaluated using one-tailed paired Student’s t test (n = 3/group) and indicated with *P ≤ 0.001. (c and d) Phenotype of the differentiated macrophages was also investigated using markers for M1 (CD38^hi) and 2 (Egr2+). On the top left, representative flow cytometry images show the effects of P4D2 mAb on F480+ CD38^hi M1. On the top right, percentages of F480+ CD38^hi cells are indicated for the different treatments. On the bottom left, representative flow cytometry images show the effects of P4D2 mAb on F480+ Egr2+ M2. On the bottom right, percentages of F480+ CD38^hi cells are displayed for the different treatments. (e) M1/M2 ratios were calculated and indicated for the different conditions. (f) Percentages of live cells were showed for the different treatments. For (b–d) statistical significance for experiments was evaluated using one-tailed paired Student’s t test (n = 3/group) and indicated with *P ≤ 0.01.

In vivo P4D2 treatment hinders tumor growth and improves survival in MM animal models

The in vitro findings with murine MM cells and macrophage differentiation prompted us to assess the potential antitumor effect of P4D2 in animal models of MM. BALB/c mice were inoculated subcutaneous (s.c.) with either CRH5 or EOH6 MM cells. When tumors reached 3–4 mm in diameter, mice were i.p. injected with 400 μg of P4D2 mAb followed by another injection of the same dose 7 d later. Control mice were left untreated or injected with P1D9. Treatment with P4D2 resulted in reduced tumor growth compared to controls for both MM cells (Figures 6a and Supplementary Figure 10). Survival analyses revealed that P4D2 injected mice, carrying CRH5 tumors, also exhibited prolonged median survival compared with untreated controls and P1D9-treated mice (Figure 6b). Since MM develops from mesothelial cells lining internal body cavities, we developed a clinically relevant peritoneal MM model for testing the therapeutic efficacy of P4D2. In these experiments, we injected CRH5 MM cells into the peritoneum of two groups of BALB/c mice. Seven days later, when tumors started growing and spreading within the peritoneal cavity, mice were treated with P4D2 or left untreated. Survival analyses showed a significant increase in median survival of P4D2-treated mice compared to controls (Figure 6c). In all animals treated with P4D2 no adverse events were observed such as acute effects, distress, or weight loss, and gross tissue examination failed to indicate any toxicity in the organs (kidney, brain, spleen, liver, and lungs).

Figure 6.

Treatment with P4D2 mAb hinders tumor growth and improves survival in MM animal models. BALB/c mice with subcutaneous CRH5 MM tumors were treated with P4D2 or P1D9 mAbs. Control mice were left untreated. (a) Tumor size is shown for the different treatments. Differences between control and P4D2 mAb groups were compared with two-way ANOVA followed by the Bonferroni multiple comparison test (n = 5/group) and indicated with *P ≤ 0.01. (b) Mice treated with P4D2 mAb (n = 5/group) show increased survival compared to controls *P ≤ 0.01. (c) Survival curves for mice carrying intraperitoneal CRH5 MM tumors, treated or not-treated with P4D2 mAb. Differences in P4D2 mAb treated vs. controls were evaluated using Kaplan-Meier curves with log-rank test and indicated with *P ≤ 0.01 and n = 5.

Treatment with P4D2 induces mesothelioma apoptosis in vivo and alters intratumor macrophages M1/M2 ratio with increased production of iNOS

We next characterized tumors from mice treated with P4D2 and untreated. MM cells were identified using mesothelin mAbs and the frequency of apoptosis revealed using annexin V staining methods. These flow cytometry assays confirmed that P4D2 induce MM apoptosis also in animal models (Supplementary Figure 11). When we analyzed TAMs using F480 antibodies in immunofluorescence we observed a reduced number of these immune cells in tumors from mice treated with P4D2 compared with controls (Figure 7a). Flow cytometry results confirmed these data showing lower percentages of F480⁺ macrophages in tumors from P4D2-treated mice compared with controls (Figure 7b). Analysis of TAMs for markers for M1 (CD38^hi) and M2 (Erg2) macrophages showed that mice injected with P4D2 had reduced percentages of F480⁺ CD38^hi M1 cells and F480⁺Egr2⁺ M2 macrophages compared with controls (Figure 7c). Analysis of iNOS and arginase-1 mRNA produced by M1 and M2 macrophages, respectively, showed higher production of iNOS mRNA for P4D2-treated mice compared to controls, but no differences in arginase-1 mRNA levels. To evaluate if P4D2 skews macrophage differentiation toward the M1 phenotype in vivo, we calculated ratios between M1 and M2 macrophages using data from flow cytometry and real-time PCR. In both cases, M1:M2 ratios were significantly higher for P4D2-treated mice indicating a prevalence of M1 iNOS-secreting TAMs in these animals (Figure 7d,e).

Figure 7.

P4D2 mAb alters intratumor M1/M2 ratio while increasing iNOS and reducing intratumor cytokines. CRH5 MM tumors from mice treated with P4D2 mAbs and untreated controls were excised and intratumor macrophages characterized. (a) Immunofluorescence using anti-F480-FITC was performed to quantify numbers of TAMs. The left panel shows representative images of stained cells and the right panel shows percentages of F480+ macrophages (n = 5/group), *P ≤ 0.001. (b) Percentages of TAMs were also evaluated using flow cytometry analysis. (c) TAM phenotypes were then characterized using markers as CD38 for M1 and Egr2 for M2. M1 and M2 phenotypes were also evaluated using real-time PCR for M1 (iNOS) and M2 (Arg1). On the left, representative figures from the flow cytometry analysis of F480+ CD38^hi M1 (top) and F480+ Egr2+ M2 (bottom). Percentages of F480+ CD38^hi (top) and F480+ Egr2+ (bottom) cells are shown. On the right, data from RT-PCR are shown for iNOS (top) and Arg1 (bottom). M1/M2 and iNOS/Arg1 ratios are shown in (d) and (e), respectively. For (b–e), differences between the two groups were compared using one-tailed paired Student’s t test (n = 5/group) and indicated by *P ≤ 0.01.

Importantly, T-cell frequency, proliferation, and granzyme B secretion were analyzed in tumors from either P4D2-treated or untreated mice. No differences were recorded for any of these parameters. In addition, markers for T regulatory cells (CD25 and FoxP3) were analyzed and no differences were detected between the two groups (Supplementary Figure 12). Similarly, in our analysis of other immune cell types that included dendritic cells, neutrophils, or monocytic myeloid-derived suppressor cells (MDSC), no differences were observed between P4D2-treated and untreated mice (Supplementary Figure 13).

Discussion

MM is a devastating cancer related to asbestos exposure. Clinical symptoms arise late in the course of the disease, 30–40 years after asbestos exposure, when the efficacy of therapeutic interventions is limited. In this study, we found that human MM tumors express high levels of galectin-9, a lectin that recently emerged as a promising target for cancer immunotherapy due to its dual role in cancer cell apoptosis^5-11 and TAMs differentiation.^14,38 We generated and tested a novel P4D2 mAb that binds galectin-9 C-terminal CRD and shows unique agonistic properties with recombinant galectin-9 in inducing cancer cell apoptosis and in modulating macrophage polarization. In contrast, the P1D9 mAb that binds galectin-9 N-terminal CRD exhibited negligible biological activity.

The agonistic properties of P4D2 are supported by data showing that MM cells treated with either P4D2 mAb or with recombinant galectin-9 undergo apoptosis and show reduced viability. It has been already demonstrated in different types of tumors that recombinant galectin-9, through its interaction with the endogenous galectin-9, induces cancer cell apoptosis. Here, we show for the first time that a unique antigalectin-9 mAb can be used as a substitute to recombinant galectin-9 or even used in combination to further decrease viability of cancer cells.

Similarities between P4D2 and recombinant galectin-9 were not limited to the induction of apoptosis as both treatments also modulated monocyte-macrophage differentiation in a similar way by depleting human macrophages expressing CCR5 a receptor crucial for the migration of M2 TAMs in the tumor microenvironment.^32,35 In contrast, when we evaluated CD71 expression in differentiated macrophages, we observed a significant increase only following P4D2 treatment. Significant differences between P4D2 and recombinant galectin-9 were also observed for other well-established macrophage markers including CD11b, CD11c, CD14, and CD206. These data, together with those obtained in mouse macrophages differentiated with higher concentrations of M-CSF, indicate that P4D2 and recombinant galectin-9 have very similar but not identical mechanisms of action.

The capacity of P4D2 to modulate macrophage differentiation was also evaluated by culturing hBDMs with supernatants from MM cells treated or untreated with the antigalactin-9 mAb. In these experiments, we hypothesized that P4D2 would interfere with the cytokines secreted by the tumor cells and indirectly reduce CCR5+ macrophage formation. The results supported this hypothesis, with the supernatant from P4D2-treated MM cells showing a lower number of CCR5+ macrophages compared with the supernatant from untreated MM cells. To definitively demonstrate that P4D2 does not have a direct effect on monocytes differentiation, these experiments should be repeated using MM cell supernatants in which the antigalectin-9 antibodies have been removed. However, even if it is unclear that P4D2 has an indirect effect on monocyte differentiation, this antibody clearly inhibits the formation of protumor macrophages as demonstrated by the reduction in MARCO mRNA levels. MARCO expression defines a subtype of TAMs with an M2-like immunosuppressive gene signature. Targeting these TAMs with MARCO mAbs have been shown to induce antitumor activity in carcinoma and melanoma tumor models,³³ suggesting that treatment with P4D2 may achieve similar results. In our experiments, macrophages differentiated with P4D2 also showed increased motility and IL-8 secretion when compared with hBDMs differentiated with only AB serum. This is another clue that P4D2 decreases the differentiation of hBDMs to protumor macrophages, which are known to have reduced motility and lower production of IL-8.^39,40

We further confirmed in animal models of MM our findings that P4D2 induces two different beneficial effects: tumor cell apoptosis and polarization of tumor macrophages toward M1-like myeloid cells. In these animal experiments, P4D2 treatment induced MM cell apoptosis, reduced tumor growth, and improved survival, with tumors from P4D2-treated mice exhibiting reduced infiltration of TAMs. Further characterization of tumors from P4D2-treated mice showed complete depletion of M2-like (F480⁺Egr2⁺) cells, as observed in experiments of monocyte-macrophage differentiation in vitro. Polarization toward an M1 phenotype was also confirmed by an increased production of the antitumor enzyme, iNOS.

Our data collectively suggest that targeting galectin-9 using antibodies against its C-terminal CRD could be an attractive strategy for cancer immunotherapy. However, there are still important limitations to consider in targeting galectin-9 in MM as animals do eventually progress based on our survival experiments. Indeed, the complex network of immunosuppressive pathways present in tumors is unlikely to be overcome by intervention with a single immune-modulatory agent.⁴¹ Since our data suggest that P4D2 acts on macrophage polarization but does not interfere with T-cell activity, a synergistic increase in the antitumor efficacy of this treatment is likely to be achieved by T-cell checkpoint blockade using PD-1 or PDL-1 mAbs. Recent findings revealed that macrophages are essential when targeting the PD1-PDL1 axis. In one study, it was shown that macrophages can remove anti-PD1 antibodies from T cells, blunting their response. A second study demonstrated that macrophages also express PD1 on their surface, which impairs their phagocytic activity.^42,43 Moreover, another recent report suggested that durable regression of established tumors requires concurrent immunotherapy with four distinct agents, which target complementary aspects of innate and adaptive immunity.^14,41 Thus, the development of novel therapies that include antigalectin 9 P4D2 mAb to modulate TAMs will provide a new path toward development of more effective immunotherapeutic options for the treatment of human cancers.

Materials and methods

Immunohistochemistry (IHC)

IHC was performed on paraffin-embedded tissue sections from human MM tumor biopsies and normal human peritoneal mesothelium (a kind gift from Dr. Harvey Pass, New York University, NY). Assessment of tumor content was based on hematoxylin-eosin staining, combined with immunohistochemical features (WilmTumor-1, Calretinin, Cytokeratin5/6 stains). Expert pathologists in pleural pathology, independently evaluated the biopsies (Dr. Pass and M.C.). Immunohistochemistry for galectin-9 (LS-B6275 mAb, LS Bio) was performed as previously described.⁴⁴ Presence of galectin-9 positive cells was evaluated on 10 fields/slide at 200× magnification with a BX43 microscope (Olympus, PA, ISA).

Mice

Female 6–8 weeks old BALB/c mice, were obtained from the Jackson Laboratory. Galectin-9 KO BALB/c mice were provided by GalPharma, Co. (Takamatsu, Kagawa, Japan). Animal experiments were performed in accordance with institutional guidelines and approved by the University of Hawaii IACUC (#16-2355).

Recombinant proteins

Recombinant human galectin-8, stable galectin-9 (hG9NC),⁴⁵ and mouse stable galectin-9 (mG9NC) were obtained from GalPharma. The recombinant hG9NC consists of an N- and C-terminal CRDs linked by His-Met residues, where N-CRD and C-CRD correspond to the 1st–148th and 178th–323rd amino acids respectively of the human galectin-9 sequence (GenPept-BAB83624.1). mG9NC is composed of N-CRD and C-CRD linked by Gly-Ser residues where N-CRD and C-CRD correspond to 1st–147th and 177th–322nd amino acids, respectively of the mouse galectin-9 sequence (GenPept-AAH03754.1). hG8G9 and hG9G8 are artificial structures generated by replacing the N-CRD and C-CRD of galectin-8 and -9. Briefly, the open reading frames of hG9NC and hG8NC cloned in pET11a vector were digested by NdeI to cut off DNA fragments coding for N-CRDs of human galectin-8 and human galectin-9, respectively, then ligated with the rest of the DNA fragments in alternate combinations. All of these molecules were purified from endotoxin using Cellufine ETclean (Chisso). Wild-type galectin-9 was purchased from R&D Systems (Minneapolis, MN, USA).

Production, purification, and titration of antigalectin-9 mAbs

Galectin-9 knock out (KO) female BALB/c mice were immunized with hG9NC or mG9NC. ELISA was performed on sera to determine hG9NC and mG9NC reactivity with preimmunization sera used as controls. The mouse responding best to galectin-9 was boosted for 5 d, then sacrificed and its splenocytes fused with P3X63Ag8.653 cells at a ratio of 5:1 as previously described.⁴⁶ Hybridoma supernatants were assessed via ELISA for binding to hG9NC and mG9NC. Hybridomas of interest were subcloned twice and expanded. Purification was carried out on Protein G-Sepharose columns (HiTrap Protein G HP, GE) according to the manufacturer’s recommendations and purity confirmed by SDS-PAGE. Purified mAbs were subsequently dialyzed and concentrated against 0.05 mM PBS, pH 7.4, and were tested to have endotoxin < 1 EU/mg. mAbs titer was determined by ELISA, utilizing twofold serial dilutions.

Cells

Murine AB12 cells derived from asbestos-induced tumors in a BALB/c mouse were kindly provided by Dr. B. Robinson (University of Western Australia, Nedlands, Australia).⁴⁷ Human REN cells were kindly provided by Dr. A. Albelda (University of Pennsylvania, PA, USA). Human Mill and ROB cells were characterized and provided by Dr. H. Pass.⁴⁸ Human Hmeso cells were from ATCC. Murine CRH5, EOH6, and EOH9 cells were isolated from peritoneal ascites developed in asbestos- or erionite-injected mice in carcinogenesis experiments previously described.⁴⁴ All cells were cultured in Ham’s F12 medium (Corning) containing 10% FBS and antibiotics. All the MM cells used in this study were provided to our laboratories or purchased in years between 2004 and 2007.

Flow cytometry analysis of antigalectin-9 mAb binding to MM cells or monocytes

MM cells or monocytes were washed and stained for 1 h at 4°C with either P4D2 mAb or IgG2 isotype control clone MG2a-53 antibodies conjugated with FITC or PE using the lighting link labeling kit (Innova Bioscience). Cells were then analyzed using LSRFortessa Flow Cytometer (BD Biosciences) and analyzed with FlowJo software (BD).

Evaluation of MM cell viability and apoptosis

Mouse and human MM cell viability was assessed by MTT assay. Briefly, 2,000 cells were plated in each well of 96 well plates in Ham’s F12 culture medium. After 24 h, cells were treated in reduced FBS (2%) with 20 µg/ml of either P4D2 or P1D9 mAb. In experiments with mouse cells, mG9NC was also used at 1 µg/ml. In experiments with human cells, hG9NC was used at 2 µg/ml. Controls were left untreated. In these assays, the lowest mAb that induced significant effects in our titration experiments were used (Supplementary Figure 14). Viability was evaluated in triplicate for each condition every 24 h. Fold increases of viability were calculated by dividing the value of each day with the viability measured at day 1, before mAb treatment. Apoptosis vs. necrosis was evaluated in MM cells using flow cytometric analysis of 106 MM cells cultured as above described. These cells were collected after 48 h using Cell Stripper (Corning) and stained with V500 conjugated Annexin V (BD Biosciences) and Propidium Iodide cell viability dye (Biolegend). Percentages of early (PI⁻/Annexin V⁺) and late (PI⁺/Annexin V⁺) apoptotic cells were assessed using LSRFortessa (BD Biosciences) Flow Cytometer and FlowJo software.

Monocyte-to-macrophage differentiation

The effects of P4D2 and P1D9 mAbs as well as those of recombinant galectin-9 on monocyte-to-macrophage differentiation were evaluated on both mouse bone marrow-derived monocytes (mBMM) and human blood-derived monocytes (hBDM). For mBMM, marrow was flushed from femurs and tibiae with HBSS, using a syringe with a 25-gauge needle, and cell suspensions were then passed through a 40 μm pore cell strainer to remove tissue debris. mBMMs were plated in DMEM (Thermo Fisher), containing 10% FBS and antibiotics. Differentiation of mBMMs was induced with either 20 g/ml or 40 ng/ml M-CSF (Bio Legend). In these experiments, cells were also incubated for 7 d with either 20 µg/ml P4D2, 20 µg/ml P1D9, or 1 µg/ml mG9NC. Control cells were untreated and media replaced at day 4. In other experiments, differentiation of mBMMs were induced using 50% of media from CRH5 cells treated with 20 µg/ml P4D2 mAb for 48 h. Media from untreated CRH5 cells was used as control. After 7 d, differentiated mBMMs were incubated with Fc-block (BD Biosciences) for 15 m on ice, followed by incubation with anti-F480-FITC clone BM8 and anti-CD38-PE clone 90 (BioLegend) for M1 macrophage staining. A Fixation/Permeabilization kit was used in combination with anti-Egr2-APC clone erongr2 (Thermo Fisher) to stain M2 macrophages. Live cells were distinguished from debris using ZOMBIE® violet cell viability dye (Biolegend). Cells were analyzed on an LSRFortessa (BD Biosciences) and analyzed with FlowJo software (BD). Blood for hBDM was obtained from healthy volunteers under an approved protocol (CHS#19442). Monocytes were isolated using Histopaque-1077 (Sigma-Aldrich) and cultured in X-Vivo-10 media (BioWhittaker) containing 5% AB serum (Sigma-Aldrich) with either P4D2 or P1D9 mAbs (20 µg/ml), hG9NC (2 µg/ml), or no antibody for untreated control cells. Cells were incubated with fresh media on day 3 of culture, and fully differentiated human macrophages were analyzed by flow cytometry on day 6 of culture. In other experiments, differentiation of hBDMs were induced using 50% of media from ROB cells treated with 20 µg/ml P4D2 mAb for 48 h. Media from untreated ROB cells was used as control. Anti-CD71 clone CY1G4, anti-CCR5 clone J418F1, anti-CCR2 clone K036C2, anti-CD68 clone FA-11, anti-CD11b clone ICRF44, anti-CD11c clone 3.9, anti-CD206 clone 15-2, and anti-CD14 clone M5E2 antibodies (all from Biolegend) were used to characterize mature human macrophages. Surface/intracellular staining and flow cytometer analysis were performed as above described. Migration of hBDMs differentiated with AB serum with or without P4D2 was measured using 8 μm Transwell inserts (Greiner). In these assays, 3,000 cells were seeded in serum-free X-Vivo-10 media in the upper chamber, while media with 5% AB serum was used in the lower chamber. Following 24 h incubation at 37°C, the inserts were removed and the nonmigrating cells wiped away with a cotton swab, while cells attached to the under-surface were methanol-fixed and stained. Migrated cells were counted from four randomly selected fields using a ×20 objective. Cytokine secretion of hBDMs differentiated with AB serum with or without P4D2 was assessed by cytometric bead array using the human inflammatory cytokine kit (BD).

Murine therapeutic experiments

To evaluate tumor dimensions and survival, s.c. mouse models of MM were employed. In these experiments, 105 CRH5 or EOH6 cells were injected in the hind flank in cohorts of five BALB/c mice. When tumors became palpable on day 7 (3–4 mm in maximal diameter), mice received an i.p. injection of 400 µg of either P4D2 or P1D9 mAbs. A second dose of the mAbs was given 7 d later. This treatment protocol was proven to be efficacious in our preliminary experiments, in which lower doses of mAbs were ineffective (Supplementary Figure 15). Tumor size was measured weekly using a digital caliper until the first death was recorded. Survival was then followed until tumors reached volumes > 1,000 mm³ for CRH5 tumors and > 200 mm³ for slow growing EOH6 cells. Survival was also evaluated in an i.p. model of MM. In these experiments, CRH5 cells (105 in 100 μl PBS) were injected i.p. in cohorts of 5 BALB/c mice. P4D2 mAb injections i.p. started 7 d later when, in preliminary experiments, tumor nodules of 0.5 mm of diameter were detectable in the peritoneal cavity. Mice received a total of four injections with 200 µg of P4D2 mAb, at day 7, 10, 13, and 16. Animals were monitored weekly and euthanized as soon as they appeared moribund according to IACUC guidelines.

Isolation and analysis of mesothelioma and tumor-infiltrating immune cells from mice

CRH5 cells (10⁵) were injected s.c. in cohorts of 5 BALB/c mice. When tumors reached 50 mm in maximal diameter, mice received an i.p. injection of 400 µg of P4D2. A second dose of mAb was given 7 d later and tumors excised after another 7 d. Tissues were washed with PBS, minced and incubated for 1 h at 37°C in digestion buffer consisting of 1 mg/ml collagenase IV, 100 µl/ml hyalurodinase, and 15 mg/ml DNAse I (all from Roche Applied Sciences) in PBS. After digestion, tumors were forced through a 40 μm cell strainer. A total of 106 cells were stained for flow cytometer analysis with macrophages were characterized as described above. In experiments to characterize mesothelioma cells and other tumor-infiltrating immune cells, the following antibodies were instead used: anti-CD3-APC/Cy7 clone OKT3, anti-CD4-PE/Dazzle 594 clone RM4-5, anti-CD8-PE/Cy5 clone 53-6.7, anti-Ki67-FITC clone 16A8, anti-CD25-APC clone PC61, anti-FoxP3-PE clone MF-14, anti-Gr1-APC clone RB6-8C5, anti-MHCII-FITC clone M5/114.15.2, anti-CD11b-PE clone M1/70, anti-Ly6C-AF700 clone HK1.4, anti-CD11c-APC clone N418 (all from Biolegend), anti-mesothelin-PE clone 295D (MBL international), and Annexin V-V500 (BD Biosciences).

Analysis of RNA expression

RNA was extracted using the RNeasy Mini kit and treated with RNase-free DNaseI (all from Qiagen). Synthesis of cDNA was performed using Superscript III (Invitrogen) and oligo dT primers. For real-time PCR, 1 μL of cDNA was used in 10 μL reactions using Platinum SYBR Green qPCR SuperMix (Invitrogen) carried out in a LightCycler 480 II thermal cycler (Roche). Oligonucleotides used for PCR included primers specific for the housekeeping gene β-actin, murine iNOS, and arginase-1. For the analysis of hBDM maturation, specific primers were used for the housekeeping gene ubiquitin c (UBC), human CD68 and MARCO. Cycling conditions were used as suggested in the SYBR Green kit instructions and results analyzed using Relative Quantification Software (Roche).

Immunofluorescence

Frozen sections of CRH5 tumors were fixed with 4% paraformaldehyde, blocked for endogenous biotin activity and then incubated overnight at 4°C with 1:300 anti-F480-FITC clone BM8 (Biolegend). Slides were counterstained and mounted using DAPI mounting medium (VectarLab). Expression of F480 was evaluated with an Axioskop2+ fluorescent microscope (Zeiss). Percentage of F480⁺ cells was evaluated on 10 fields with at least 100 cells in the same slide using ImageJ.

Statistical methods

All statistical tests were performed using GraphPad Prism (GraphPad5.0). Means of two groups were compared using one-tailed paired Student’s t test. When more than two groups were compared, two-way ANOVA followed by the Bonferroni multiple comparison test was conducted. For survival, differences were evaluated using Kaplan-Meier curves with log-rank test. Data are represented as mean ± s.e.m. with statistical significance values indicated in the figure legends together with the n values used to calculate the statistics. All in vitro experiments with MM cells have been repeated at least three times using samples from the same source as technical replicates. In vivo studies as well as experiments with primary cells were also repeated at least three times using different sources as biological replicates.

Funding Statement

This research was supported by core facilities supported in part by National Institutes of Health (NIH) grants P30GM114737, P30GM103341, G12RR003061, G12MD007601, P20GM103466, U54MD007584, 5P30GM114737, P30 CA071789-17, and 2U54MD007601. L.C.N. receives support from the National Institutes of Health (NIH) grant R01MH112457.

Acknowledgments

We thank Ann Hashimoto for blood collection and the volunteers that donated their plasma. We also acknowledge Dr. Owen Chan and Dr. Hugh Luk from the University of Hawaii Cancer Center Pathology Shared Resource.

Disclosure of Interest

The authors have declared that no conflict of interest exists.

Supplementary material

Supplemental data for this article can be accessed on the publisher’s website.